Case-study on the application of building HVAC
Transcription
Case-study on the application of building HVAC
Case-study on the application of building HVAC performance analysis and fault detection using ABCAT Master Thesis T.M.K. Hissel June 2012 Technical University Eindhoven Faculty of Mechanical Engineering Sustainable Energy Technology Master Program Unit Building Physics and Systems In association with Strukton Worksphere Supervisors: prof.dr.ir. Jan Hensen (TU/e) ir. Gert Boxem (TU/e) MSc. Barry Tuip (Worksphere) i Abstract In many cases buildings do not performs as intended. Shortly after completion, the energy performance of a building starts to degrade and energy consumption increases. Conventional building management and monitoring systems do not detect and alarm decreasing performance, neither is there adequate diagnosis of the cause of this performance loss. Monitoring, fault detection and smart diagnosis of the energy use of a building can give a better insight in the loss of performance of building heating and cooling systems and provide guidelines for improvement. In this study, which is part of Strukton Worksphere’s development of building management tools, a case-study is carried out on the application of monitoring, fault detection and diagnosis on an existing building. A measurement setup is applied to support the monitoring process and the ABCAT software (Automated Building Commissioning and Analysis Tool) is used for heating and cooling energy consumption predictions. Fault detection and diagnosis (FDD) is performed by comparison of the predicted consumption with measured values. The applicability of these specific methods on the case-study building has been assessed. Both simulation and FDD were unsuccessful. Within the scope of the project it was not possible to realize a calibrated model with sufficiently accurate simulation results, due to the following reasons: 1. The simulation software (ABCAT) is a prototype which requires more development into a comprehensive and robust product. Additional testing and expansion on more diverse buildings and systems is required to exclude compatibility issues. 2. Calibration is applied on a clearly underperforming HVAC system. It is observed and concluded that various aspects of the HVAC configuration and its control contain flaws. Simulations represent a correctly working system, while measurements represent a malfunctioning system with an obvious mismatch in results as a consequence. Although the goal of simulation-based FDD has not been reached, several faults in the HVAC operation are detected. Careful monitoring and detailed analysis of measurement data by experts, without a simulation reference, has strong FDD possibilities as well. Based on the case-study results. Conclusions and recommendations on the compatibility of methods and essential conditions for successful FDD are presented. ii Acknowledgements This report is the result of my graduation project for the Sustainable Energy Technology master program. Within the SET program I chose to specialize in the Built Environment and Building Performance Simulation direction because I was very interested in energy reduction and optimization in the built environment and the transition into a sustainable society in general. For my research internship and thesis subject I found a pleasant working environment at Strukton Worksphere, where I did my research while at the same time exploring the corporate side of the application of building energy optimizations. First I would like to thank professor Jan Hensen for his overall guidance, good advice and criticism. I am very glad with the existence of his monthly progress meetings which forced regular recaps and status inquiry of my work, as well as presenting feedback moments and presentation practice. I would also like to thank my direct supervisor Gert Boxem for his guidance throughout the whole graduation process and his expert review and suggestions to improve this work. I respect Gert for his knowledge in report writing, being very critical and handing me the best of tips and recommendations. Special thanks to Barry Tuip, my supervisor at Worksphere, for showing me all of the company and assisting me with a lot of practical issues and challenges during the project. I want to thank Barry for always being very inspiring, motivational and a great colleague. I also want to thank all my other colleagues at the Worksphere engineering and consultancy group for their helpfulness and support. Last but not least my friends and family for being interested in my work, supplying some good ideas and motivation, proofreading my work (especially my dad who evaluated the whole thesis in one weekend) and the support through the heavy weeks of thesis-writing. iii Contents Abstract ..................................................................................................................................... ii Acknowledgements ...................................................................................................................iii 1 2 Introduction ....................................................................................................................... 1 1.1 Problem definition..................................................................................................... 1 1.2 Main goals and objectives ......................................................................................... 1 1.3 Introduction to ABCAT .............................................................................................. 2 1.4 Methodology ............................................................................................................. 3 Monitoring ......................................................................................................................... 5 2.1 2.1.1 Gas-powered Heat Pumps ................................................................................. 6 2.1.2 Air Handling Units .............................................................................................. 7 2.1.3 Induction Circuit ................................................................................................ 8 2.2 Commissioning ........................................................................................................ 10 2.3 Measurement setup ................................................................................................ 11 2.4 Calculations ............................................................................................................. 14 2.5 Data processing ....................................................................................................... 15 2.5.1 Data acquisition, transfer and storage ............................................................ 15 2.5.2 Data processing into useful information ......................................................... 17 2.6 3 Case description ........................................................................................................ 5 Initial building energy performance analysis .......................................................... 19 2.6.1 Energy performance results ............................................................................ 19 2.6.2 Improvements to HVAC system control .......................................................... 20 2.6.3 Energy performance after improvements ....................................................... 20 Simulation and fault detection ........................................................................................ 23 3.1 Background .............................................................................................................. 23 3.2 Simulation with ABCAT ............................................................................................ 24 3.3 ABCAT diagnosis functionalities .............................................................................. 28 3.4 ABCAT compatibility ................................................................................................ 33 3.5 The calibration process of ABCAT............................................................................ 35 3.5.1 The need for Calibration .................................................................................. 36 3.5.2 Required amounts of data ............................................................................... 36 3.5.3 Calibration method.......................................................................................... 37 3.5.4 Calibration evaluation and statistics ............................................................... 41 iv 4 3.5.5 Evaluation of the supplied characteristic signatures....................................... 43 3.5.6 Definition of calibration period and test period.............................................. 45 3.5.7 Application of the calibration process on the case-study building ................. 47 3.5.8 Remaining calibration problems and challenges............................................. 51 3.5.9 Conclusions on the Calibration process .......................................................... 51 Case-study results............................................................................................................ 53 4.1 Building heating and cooling energy consumption ................................................. 53 4.2 ABCAT calibration .................................................................................................... 54 4.3 Fault detection results ............................................................................................. 55 4.4 Diagnosis results ...................................................................................................... 56 5 Discussion ........................................................................................................................ 57 5.1 ABCAT assessment .................................................................................................. 57 5.2 Building energy performance assessment .............................................................. 59 5.3 Expert view versus automation ............................................................................... 62 6 Conclusions and recommendations ................................................................................ 65 References ............................................................................................................................... 68 Appendices .............................................................................................................................. 70 A Abbreviations .............................................................................................................. 70 B Detailed description of the ABCAT interface............................................................... 71 C ABCAT output figures .................................................................................................. 74 D Sources for ABCAT inputs and building parameters ................................................... 76 E Characteristic Signatures SWS Son Case-Study Building ............................................. 77 F Details of subsystem measurements .......................................................................... 83 G Measurement accuracy, reliability and validation ...................................................... 88 H Subsystem results ........................................................................................................ 93 I Details on the HVAC control improvements ............................................................... 95 J M-files for data processing .......................................................................................... 98 v vi 1 Introduction The global energy problem is one of the most important challenges of our time. Depletion of fossil fuels, increasing energy costs and climate change due to CO2 emissions are predominant subjects in the media and politics. Reduction of the current and future energy consumption is one of the ways to face this challenge. A big share of the worldwide energy consumption is due to the heating, cooling and electricity demand of existing buildings. The built environment contributes for about 40% in the worldwide energy use. An obvious approach to attack this energy problem is to focus on the energy use of climate systems in large buildings and offices. (Elkhuizen & Rooijakkers, 2008) states that about 30% of all buildings uses more energy than necessary and that more than 70% of all buildings does not function as intended. Decrease of heating and cooling energy use in buildings can be realized by the implementation of new, more efficient techniques and devices and furthermore by better management and control of existing systems. Optimization of climate system configuration and settings, continuous monitoring of performance and timely, well-directed maintenance can contribute greatly in the reduction of the global energy use. Building services contractor Strukton Worksphere participates in this energy reduction trend by various projects focused on building energy optimization. Worksphere has several current developments on building management tools, which can be of great help in building energy monitoring, the discovery of inefficiencies in building energy performance and detection and diagnosis of faults in the heating, ventilation and air conditioning (HVAC) systems. For this project a case-study is carried out on the application of monitoring, fault detection and diagnosis on an existing building. A measurement setup is applied to support the monitoring process and simulation and fault detection software is used for heating and cooling energy consumption predictions as well as fault detection and diagnosis (FDD). 1.1 Problem definition Building HVAC systems malfunction causes increase of energy use. HVAC system underperformance is often not caused by abrupt faults but by constant, time-scaling performance degradation (Morisot & Marchio, 1999). Degradation of system performance leads to increased energy consumption and often also to a decrease of indoor comfort, according to (Zeiler & van der Velden, 2011). Building services contractors like Strukton Worksphere make use of PPP (Public Private Partnership) contract forms in which they are responsible for both energy consumption and indoor comfort of certain buildings. Not achieving the energy and comfort goals will result in financial penalties so detection of system malfunction in an early stage is a very desirable development in the building maintenance field. 1.2 Main goals and objectives To counter excessive cooling and heating energy use, occurring faults in the HVAC systems should be detected in an early stage. The assignment for Worksphere is to come with a 1 continuous fault detection and diagnosis (FDD) tool, with proper diagnosis functionality, which can operate with minimal input of the user. Malfunction of the HVAC systems should be noticed in an early stage and the cause of problems should be pinpointed and reported. Wear and tear of parts of the system should be reported so they can be replaced before real problems or malfunction occur. Objectives to fulfill this goal: Detection of faults in the continuous operation of a building HVAC system Diagnosis of these faults including: o Identification: distinguish faults from random anomalies and rate the fault’s impact/urgency o Isolation: find out in which subsystem or component the fault occurs Reportage: Automate the process and make sure the right information (develop key FDD figures) is supplied to building managers and Strukton mechanics by automated email/SMS alerts. In this study a measurement setup is assembled and applied for continuous monitoring of the HVAC system behavior. A method for fault detection and diagnosis is chosen and its applicability on the case-study building is tested. Research questions for this test procedure are: 1. 2. 3. 4. 5. 6. 1.3 How is the energy performance of the case-study building? Is this a suitable case to apply FDD on? Can we accurately simulate the building behavior? Can we successfully detect and isolate faults in the case-study system? Can we make the required measurements setup to do this? Can we reproduce FDD results from the past (available in literature) on this casestudy building? Introduction to ABCAT For simulation and FDD purposes the program ABCAT is chosen (Automated Building Commissioning Analysis Tool). It is a tool made by the Texas A&M University (Curtin, 2007) and is especially designed for building energy use diagnosis. The ABCAT can operate with a limited amount of input data. A simplified physical model of the building and its climate systems will be calibrated by using measurement data from the building heating and cooling energy consumption. Advantages of ABCAT over other simulation software is that: A top down approach is used which requires less inputs and detail while the influence of the fault on the system as a whole will be more clear. The data traffic and processing is kept manageable by basing simulation on three quantities: the total energy use for cooling, the total energy use for heating and the electricity consumption. A simplified model is used which uses less parameters causing it to be easier to calibrate compared to conventional simulation models. 2 The simplified ABCAT approach makes simulations and FDD more affordable, automated, and easier applicable. The energy use predicted by ABCAT is compared to the measured energy use and faults are detected based on significant deviations between the two. An important feature of ABCAT is to identify faults that, on the long run, will have a negative impact on the performance of the HVAC system. These faults can be found by analyzing their cumulative effects. Additional figures and features are available for further identification and isolation of malfunction. In this project the ABCAT prototype is applied on a case-study building, the Strukton Worksphere office building in Son. ABCAT is a relatively new tool which is hardly applied outside of its own development environment, the Texas A&M University. Case-study tests of this FDD method, especially focused on compatibility with various HVAC systems and additional requirements for successful applicability are interesting to research. The applicability of the ABCAT FDD methods on the case-study building and its conditions is assessed. 1.4 Methodology Fault detection will be realized by performing energy use measurements in the case study building alongside energy use simulations, which will be carried out by an appropriate simulation program. In a previous literature study (Hissel, 2012) various simulation and fault detection methods are investigated and compared and the program ABCAT (Automated Building Commissioning and Analysis Tool) was selected for simulation and FDD applications. The basic FDD process is summarized in figure 1. It consists of the following steps: 1.) Monitoring of the real-time heating (HW) and cooling (CHW) energy consumption in the building. 2.) Physical model-based simulation of the HW and CHW energy use as they would be in a ‘correctly functioning building’. 3.) Detection, which lies in the comparison of the measured and simulated behavior and observations of the residual and anomalies occurring in this residual. ABCAT supplies additional tools and figures for the detection and diagnosis process which will be explained later. Figure 1: Schematic representation of the basic 1) monitoring, 2) simulation and 3) fault detection process 3 Throughout the whole study the following approach with sub questions has been used: 1. Monitoring of the case-study building and visualization of the energy use o Build and configure a measurement setup and data processing treatment that can support the FDD process o Inquiry of which parameters are required for FDD and which are already measured in the building with the existing Building Management System (BMS) o Complement BMS measurements with additional sensors o Combine measurement data of different sources o Validation of measurements and assessment of accuracy o Process measurements into Input data for simulations and calibration of the simulation model 2. Simulation of the total heating and cooling energy consumption using ABCAT o Validation and assessment of ABCAT o Is it applicable on the case-study building or are there compatibility issues? o What are shortcomings of and possible improvements for ABCAT? o How robust is ABCAT and how flexible in the use on multiple buildings? o Produce an accurately calibrated ABCAT model of the case-study building and HVAC systems 3. Application of fault detection o ABCAT will generate time-energy use plots and time-cumulative error plots o In contrast with the ABCAT baseline and calibration step mentioned before, ABCAT will now be applied on a different ‘test period’ in which fault detection is performed o A proper definition of a fault must be formed. Statistically significant deviations between measurement and simulation indicate a fault o Small but persistent deviations, which are not cancelled out, are detected by their cumulative effect. These are the faults which should be detected because they indicate with a high probability the malfunctioning of the HVAC system which should be resolved as soon as possible to prevent excessive energy use and system break down o What is the performance of the fault detection of ABCAT? o Explore and test additional detection and diagnosis utilities provided by ABCAT o Are faults detected correctly? 4. After fault detection the faults will be diagnosed o Which diagnosis tools are available in ABCAT? o Application of the ABCAT ‘expert rule’ tables o Do the ABCAT ‘expert rules’ work for the case-study building? o Can the detected faults be isolated to a subsystem? o Can the faults be pinpointed to a specific component? o Let Matlab automatically report the likely causes of malfunction. Send this report to the building operators or Worksphere mechanics 5. Thorough analysis of the achievements and problems regarding the FDD process and conclusions and recommendations for future projects in this area. The focus lies on system and software compatibility and essential conditions for successful FDD application. 4 2 Monitoring 2.1 Case description The monitoring and fault detection methods are applied on a building in practice. The office building of Strukton Worksphere in Son, Netherlands was facilitated as the test-case building. The main advantages of the building are the easy availability and the unlimited access to all climate and control systems. Because of its state-of-the art climate systems and short lifetime (no degradation) the building HVAC performance was assumed to be quite optimal. The monitoring and analysis of the behavior will point out the validity of this assumption. The characteristics of the building are: 7 floors of office space Attached magazine/logistics hall About 6700 m2 floor area 400 kg/m2 thermal mass Built in 2010 Figure 2: Test-case building in Son Figure 3: Floor plan of the case-study building Figure 4: Design drawing of the building front view 5 Characteristics of the climate systems: Four gas-engine powered heat pumps for both heating and cooling purposes, source: ambient air, regular heat pump cycle. 71KW cooling, 84 KW heating each Additional hot water boiler for peak heating demands One CAV AHU for office ventilation of about 35.000m3/h air flow capacity, 235 KW cooling, 150 KW heating capacity and with 60% heat wheel energy recovery One CAV AHU for restaurant ventilation of about 4.000 m3/h air flow capacity, 40 KW cooling, 40 KW heating capacity and with heat exchanger heat recovery Heating and cooling distribution by four pipe induction system The different components of this HVAC system will be discussed in more detail below: 2.1.1 Gas-powered Heat Pumps The building HVAC system’s main heating and cooling plant consists of four gas-powered heat pumps from ‘Gasengineering/ Aisin Toyota Group (GasEngineering, 2011)’. A running heat pump which is either heating or cooling also delivers free engine heat on the side. A cooling heat pump is able to produce heat at the same time as cooling. Parameters of the heating and cooling system are found in the ‘building delivery report (Utiliteitsgebouwen, 2009)’: Gas engine driven heat pumps Source: outside air Heating temperature range: 35°C < Tsupply < 45°C Backup energy from gas boiler for peak demand Generation efficiency heating = 1.5 System efficiency heating (system losses) = 0.84 Generation efficiency cooling = 1.95 System efficiency cooling = 0.85 Maximum heating capacity = 336 kW Maximum cooling capacity = 284 kW Figure 5: Impression of the Heat pumps and Gas boiler present in the case-study building (Hissel, 2011) 6 Free engine heat HW buffer Regular output Figure 6: Two heat pumps in heating mode (GasEngineering, 2011) It can be seen that the heat pump delivers an amount of heat equal to Energy from gas * PER to the heat storage tank. For even better performance also the heat produced by operation of the engines is delivered to the hot water tank as well. If a specific heat pump is cooling it can still contribute to the total heat production by supplying the engine heat to the hot circuit. Part of the energy is lost due to heat loss and the remainder is the heat delivered to the building. 2.1.2 Air Handling Units Figure 7: Impression of the AHU's present in the case-study building (Hissel, 2011) On the building roof there are two large Air Handling Units present from manufacturer Carrier(Carrier, 2009). The larger one of the AHU’s is dedicated to the office floors of the building. The other one supplies air to the restaurant and to two meeting rooms. AHU office: Heating and cooling coil with water from central heating/cooling plant Temperature and relative humidity sensors Heating capacity 150kW Cooling capacity 235kW Heat recovery (sensible and latent) by energy wheel, Hygroscopic 310kW Filtering: synthetic bag filters type F5 and F7 7 Rated air flow 10,3 m3/s Constant air flow control with two settings (Dahlander Fans, 1500/750rpm) AHU restaurant: Heating capacity 43kW Cooling capacity 37kW Only sensible heat recovery by heat exchanger, 32kW Filtering: synthetic bag filters type F5 and F7 Rated air flow 1,6 m3/s Variable frequency fans, but used as constant air flow 2.1.3 Induction Circuit The heating in the building is mainly done by an induction heating and cooling system. Figure 8: Example of an induction segment of an induction heating and cooling system (Trane, 2012) There is an extensive, branched induction circuit throughout the roofs of all the office area. All the branches have local valves which make sure the induction segment is supplied by the right mix of cold and hot water to ensure the correct supplied temperature. Every few meters a thermostat is present, which directly controls the nearest induction circuit valve, so the indoor temperature can be adjusted for every single segment and section of the office. The induction system is a so called four pipe system. There are two supply pipes and two return pipes for each induction unit. One of both contains hot water, the other one of both contains cold water. The temperature, or choice between heating or cooling can be made for each section of the office individually. The indoor climate can be adjusted to the occupants wishes. 8 Figure 9: The overview of the plant side of the HVAC system, derived from (Priva, 2011) with: The four heat pumps + engines (which can produce free engine heat), the hot water boiler (red square), the hot and cold water storage tanks and the hot and cold circuit piping. Temperature sensors indicate the temperature at various points used for the heat pump control. On the right the water distribution system is connected. Figure 10: Example of the Priva interface showing the hot water plant, buffer and distribution (Priva, 2011). A similar distribution schematic holds for the cold water distribution. A measurement plan is made for the monitoring of the energy consumptions in the building climate system. The primary purpose of the setup is to get measurement results for the total Heating (HW) and cooling (CHW) energy use. This is the basis for the top-down FDD approach which is applied by the use of ABCAT. The total HW and CHW energy consumption are the main measurement inputs for ABCAT. Secondary objectives are additional measurements that are assumed to be interesting for the general energy performance assessment of the building and for the diagnosis & isolation part of the FDD process. These can be measurements on building level or on subsystem/component level. 9 2.2 Commissioning Before being able to apply FDD, a commissioning process is essential, as stated by (Zeiler & van der Velden, 2011). Initial improvements are not permanent, continuous commissioning can be a solution. Good commissioning is necessary before setting up a fault detection system. The continuous checking of a certain level of performance and detection of deviation from this performance level only make sense when an initial assessment of this performance has been carried out. Maintaining inefficient behavior would have absolutely no added value. The chosen case-study building is the new office building of Strukton Worksphere and contains a, state of the art, supposedly well performing HVAC system. A thorough commissioning process has not taken place, though. During the initial phases of the project it was assumed that the case-study HVAC building would show (close to) optimal energy behavior because performance degradation over several years of exploitation has not been the case. Later on in this project this assumption has been invalidated and the understanding of the need of commissioning has become ever more clear. In general, commissioning can be seen as an in depth examination and validation of a building state and its performance (EPA, 2012). The monitoring of a building and detection of malfunction are no use if the initial state of the building is not assessed by a commissioning process. Another important aspect of commissioning is that it can bridge the gap in view and expectations of the building designer, building contractors, building systems operator and building owner. Because the commissioning process can have a role in all the phases of a building’s lifetime the visions of all these involved actors can be combined. There are several types of commissioning that can be distinguished: Initial commissioning: Is carried out in the design and production phase of the building. Before the building is accepted by the new owner the configuration and working of the climate systems is thoroughly investigated and fine-tuned. Retro-commissioning: The commissioning of an existing building that never had an initial commissioning. From a lot of buildings the performance is never properly assessed. By retro-commissioning usually a lot of potential energy savings are found. Re-commissioning: Commissioning carried out on a building that has had an initialor retro- commissioning in the past. With re-commissioning the building owner can verify, improve and document the performance of the building and its systems. Continuous-commissioning: After commissioning and improving a building’s systems it is another challenge to avoid the decrease of this new performance. By continuously monitoring and immediate solving of occurring problems the performance of the building can be maintained. Fault detection and diagnosis can have an important role in continuous commissioning. One aspect of commissioning is the comparison of the target building behavior with references (Van Kleef, 2011). The following references are often used for comparison: Design aspects (Simulation results from building design software) 10 Status report from building completion phase, which should include measurements from building systems tuning and testing Measurements in exploitation phase Reference buildings with comparable specifications After these comparisons and in combination with basic inspection of the building and its systems a team of experienced building systems engineers or consultants can bring out a verdict on the building performance. Is it functioning properly or not? A commissioning report can also contain advice on which measures have the most potential for improvements to the building. When assessing the current performance of the building and systems certain aspects have to be addressed: Energy efficiency Thermal comfort for occupants Efficient use of equipment (conservation of lifetime) All these three aspects are important but there is an interaction and often also a contradiction. For instance if the building’s energy consumption is minimized the thermal output of the HVAC system can be insufficient resulting in discomfort, or it can lead to harmful equipment use resulting in early breakdown of system components. A large number of heat pump startups and shutdowns are disastrous for its lifetime but can on the other hand be an improvement measure for energy efficiency or guarantee of indoor comfort. It is important to find the right balance for these and to find an optimum for all three simultaneously. Experienced maintenance engineers, energy consultants or other similar professionals can accurately assess the processes in a specific building case and can also foresee problems occurring in other aspects of building performance. Where monitoring and fault detection are processes which can be automated by use of software, this is much less the case for commissioning. If this is to be automated the software must be able to correctly combine different aspects of performance which is challenging. Computer intelligence is not expected to be suitable for all the aspects of building commissioning. 2.3 Measurement setup The installation and optimization of this measurement system is performed as part of this graduation project. The Priva building management system is already installed in the building and inhabits various sensors measuring the climate system status and room conditions throughout the building. In addition to the measurements from the BMS additional measurements are carried out with the WiSensys wireless measurement system. There are about 70 measured quantities available in the Priva BMS. Only the ones necessary to achieve the project goals are selected and further processed. The choice of measurement parameters, acquisition and processing of data are steps in the monitoring process carried out for this project. 11 Acquisition of the total heating and cooling energy use Ideally the total heating energy use (HW) and cooling energy use (CHW) would be directly measured at the source (Gas-powered heat pumps) or at the main supply circuits. Various methods for measuring these quantities were addressed but they all presented some major obstacles: Because both the heating and cooling are generated by the same plant (four heatpumps) they are not measurable individually Climate installation equipment does not allow direct extraction of activity information (heat pump energy production, circulation pump activity (rpm/flow)) The main supply circuit lacks measurement options. Valves with Δp connections are occupied by sensors for pump control, alternative (ultrasonic) flow measurements are expensive and overdone for the scope of this project. Because of this an alternative was chosen where total HW and CHW energy flows are determined using individual measurements on the three supply subsystems (figure 11): Figure 11: Overview of the embranchment of the building systems plant side into the three supply subsystems: the induction circuit, AHU office and AHU restaurant. This embranchment is similar for the HW and CHW net. Because the heat flows or energy consumption is not measurable on the plant side these are obtained by the combination of subsystem measurements Energy supply to the induction circuit. By measurement of the mass flow and supply + return temperatures the supplied HW and CHW energy can be determined. Energy supply to the heating and cooling coils of the two AHUs can be determined by temperature and relative humidity measurements in combination with the produced air flow rate. Special attention is required for latent heat recovery by the energy wheel in the office AHU. 12 The measurement plan is not the most accurate solution but because of the technical difficulties of the more direct measurement approaches this is the best achievable alternative. Table 1: Parameters measured in the case-study building and the respective measurement systems used. (Priva BMS = existing measurements from building management system, WiSensys and Maestro are additional wireless measurement systems, KNMI is meteorological data from nearby weather station) Parameter Outside temperature Outside humidity Indoor climate Electricity use Gas use ΔP Induction T Induction Tsupply, Treturn AHU Texhaust AHU Air flow rate AHU RH AHU Priva BMS X WiSensys X X Maestro KNMI X X X X X X X X X X X X Gas Use The building’s total gas consumption is already measured for energy monitoring purposes. In this project the gas use can be very useful for validation of energy use results. Energy use can be calculated by temperatures and energy flows, but also by gas use, energy content and system efficiency. The comparison of a result obtained by two different approaches can give very useful information about the correlation and so validate the results .From the gas use quantities no division between heating and cooling quantities can be made, it can only be compared to the combined HW and CHW energy use. Gas consumption measurements come from a meter on the main gas line in of the building and are available in the Maestro energy management database. Electricity Use Electricity consumption is measured both on the main meter and on electricity group level. The building’s total electricity use is used as an input for ABCAT and is an essential quantity to measure. Electricity use of individual groups can be useful for fault isolation purposes and is measured and stored as well using the KNX system and Honeywell data acquisition equipment. The overview of the applied measurement setup and all the quantities monitored can be seen in table 5. Measurements are obtained from different sources and all the information is gathered and combined using Matlab. Measurement data is stored in a database on a Worksphere server. The measurement accuracy and the validation of measurements is discussed in appendix G. The following paragraph treats the mentioned subsystems in more detail and later on the calculations and processing required to determine the heating and cooling totals are explained. Details on the subsystem measurement methods can be found in appendix F. 13 2.4 Calculations First of all the equation for total heating and cooling energy use Induction The measurements supporting these calculations are treated in more detail in appendix F. AHU office QAHU is actually the energy supplied to the heating or cooling coil (Qin) which can be derived from the energy balance of the heat transfer processes in the AHU. Due to its measurement complexity and assumed insignificance heat losses in the AHU are neglected. In figure 12 a schematic overview of the AHU is given and important Temperatures and energy flows are indicated. Figure 12: Schematic overview of AHU office, as presented in the Building Management System, showing key parameters for the energy balance and energy flow calculations (Priva, 2011) AHU energy balance: The thermodynamics and corresponding formulas of an AHU are treated in (Howell, Coad, & Sauer, 2009). In this balance Qin can be either positive (heating) or negative (cooling). While processing the calculations this is taken into account accordingly. 14 The energy recovery using the energy wheel is obtained by the following relation: Sensible heat recovery by common flat plate heat exchanger: Energy recovery by energy wheel: Using the enthalpy given by: The results of these calculations are presented in the results section (4.2) and the validation and accuracy of results are treated in appendix G. 2.5 Data processing 2.5.1 Data acquisition, transfer and storage Figure 13: An illustration of the Matlab program written to automatically gather and transfer the necessary measurement variables from the building management system to the measurement database on the Worksphere server Measurements from the WiSensys sensors are uploaded directly to the server on which the web-based database utility is running which can be seen in figure 14. The WiSensys sensors are wireless, battery powered packages of one or more sensors with a transmitter circuit. Each sensor should have a base-station nearby. A base station has line-power, continuously receives data packages from one or more sensor units and continuously transmits the gathered measurement data to the server. The connection to the server is realized through a mobile phone network. Measurements from the PRIVA building management system are stored in a MS Access database present at the PRIVA server. Data transfer from the local database to the WiSensys database is done by a Matlab program especially written to support this measurement setup. The steps carried out in this acquisition program are: - Connection to MS Access database and inquiry of available parameters 15 - Automatic selection of required parameters Comparison already acquired data range with the data range on PRIVA database Import of only the new data Connection to Worksphere server via php Upload of the just imported data to the right places in the Worksphere database Scheduled to automatically run every night. Error report via email if transfer fails For data storage, the following storage functionalities have been developed and installed: - Automatic storage of WiSensys measurements in Worksphere database Automatic Matlab uploads of Priva data to Worksphere database All data is combined and made available via web browser as can be seen in figure 14 Data downloads by Matlab via php scripts and database login authentication key Figure 14: The combined measurement data stored in a database on Worksphere server. Directly available to Matlab for data processing and also available in a web interface for manual monitoring purposes The required data is now available in the WiSensys database. The m-file for processing (Appendix J) downloads the required data columns, applies data synchronization and format processing and executes the calculations given in the calculations paragraph for each of the subsystems. 16 2.5.2 Data processing into useful information Figure 15: Overview of measured parameters and their processing into required input parameters for ABCAT Data processing tasks as carried out with Matlab include: Cutting of all vectors to the same length, making all vectors have the same starting and end point. Vectors lengths are automatically checked and adjusted to match the shortest one Removal of NaN’s (Not a Number) and filling of gaps in measurement data by either average values ore by interpolation of the data in between Values must be converted to Imperial units Vectors with different time intervals (3min, 5min, 15min) must be converted to hourly values and daily averages Time vectors must be converted from datenum to Unix and dd/mm/yy format Electricity consumption must be summed to 24hour values Separate columns must be produced with the correct ABCAT input values (as can be seen in figure 15) for preoccupied, occupied and post-occupied periods The processed data can either be copy-pasted, exported to .mat or .txt or directly written to Excel via the xlswrite functionality. After this process the ABCAT input data is present in the ABCAT Input tab and simulations can start. 17 Figure 16: Overview graph of the processed measurement data showing daily values for Heating energy consumption, Cooling energy consumption, Gas consumption and the combined heating and cooling energy consumption. This figure is used for basic validation and comparison purposes 18 2.6 Initial building energy performance analysis During the validation and processing of measurement data some observations could be made. Later on this evolved into a more precise energy performance analysis of the casestudy building involving the assistance of energy consultants and control technicians. The involvement of these experienced people can be interpreted as being the expert view in the case-study HVAC system performance analysis. Diagnosis of the building energy consumption results pointed out inefficiencies in the system operation and plans for improvements were made and put into use. 2.6.1 Energy performance results Figure 17: The first energy consumption results yielded interesting observations on the building energy perfomance: unexpected high amounts of simultaneous heating and cooling are reported in the summer months indicating energy inefficient behavior High amounts of simultaneous heating and cooling can be observed in the months august and September. Simultaneous heating and cooling can occur because it is a four pipe system for which this is an intended functionality. The amounts of simultaneous heating and cooling are out of proportions though, resulting in significant energy waste. The cause of this inefficient behavior is found in the set-points for hot water and cold water circuits, which have relatively high values throughout the whole year. Hot water (45oC) and cold water (10oC) are continuously transported through the whole building without being used. Heat loss and heat transfer between hot and cold pipelines are a direct waste of energy. The weather conditions in these months were very moderate and the amount of heating or cooling required to produce a comfortable indoor climate appear to be just fractions of the actual heating and cooling energy used in this period. The expert verdict on this matter: ’The building shows very poor energy performance and actions are required to improve the HVAC control software.’ After these initial observations on system malfunction and the reportage of severe underperformance, Worksphere acknowledged the energy performance issues and measures for improvements have been taken. Several improvements to the HVAC control have been carried out which are described in the following paragraph. Further monitoring and FDD applications are delayed until these improvements are implemented. 19 2.6.2 Improvements to HVAC system control Initial measurements Observations of HVAC system underperformance Improvements on HVAC control software Measurements on improved system and application of FDD Figure 18: Overview of the different steps in the building performance assessment + improvement process A proposal for improvements on the building control system (RTO, (Meulman, 2011) has been written and is implemented into a new version of the Priva control software. The new control software is put into use on dec8th 2011 and fine-tuned in the subsequent weeks. Figure 19: Illustration of the HW and CHW water flows in the induction circuit, that shows the moment of implementation of the new control software which immediately deactivated the unnecessary cooling which was previously present in the building. A short overview of the improvements in the new software: Summer and winter block Cyclic switching of the heat pumps AHU heating curve optimization by use of return temperatures Heat pump & engine heat - release and set point optimization The contents of the existing HVAC control strategies and the applied improvements are explained in more detail in appendix I. 2.6.3 Energy performance after improvements After the optimizations in the control software the FDD process is continued. As can be seen in the calibration chapter the period after improvements has been defined and used as the calibration period. FDD has been applied on the test period, the period before improvements, in which there is guaranteed underperformance of the HVAC system. Although many improvement have been carried out, the HVAC energy performance still appears to be not optimal. During the continuous operation after the optimization efforts more suboptimal performance and faults in behavior are observed, including: Slacking outside temperature indicator Deviations from design in climate installation configuration 20 Buffer tank layering problems and short circuiting Valve leakage in induction circuits Moreover results are obtained from a research on the energy efficiency of the same building but carried out independently (Daelmans, 2012). From this research, energy consumption simulation results are obtained which are produced using the original design models of the building and the simulation program VABI. The results are assumed to be correct and are not further validated. A comparison is made between the simulated annual gas consumption and the actually measured annual gas consumption. Because the design is likely not the same as the actually built systems and because the simulations are carried out for a different year (simulations for VABI reference year, measurements for 2011) it is not expected that the results show a close match. It is interesting though to validate if the simulations are anywhere near the measured quantities. This is not the case, the building consumes much more gas than is simulated in the building design phase. This is an extra indication that the building is currently underperforming. Gas consumption for heating and cooling purposes [m3/year] 150000 100000 50000 0 Measured VABI 117 VABI Elements Figure 20: Measured heating and cooling gas consumption compared to simulations with VABI. Carried out to assess the current energy performance in relation with the intended performance. The calibration of ABCAT has been performed on this period with suboptimal behavior and the applicability and performance of ABCAT is analyzed. There was no time to delay the calibration and FDD application a second time. The observation of these additional flaws in the system is relatively new, no actions for improvements have been initiated until now. The observed flaws in HVAC system operation are: Delay in outside temperature indicator Cold nights and sunny days in march. Cooling is controlled using a damped version of the outside temperature. It takes the average outside temperature over the past few hours. When this delayed outside temperatures increases over the set-point outside temperature the building cooling system is activated. In the morning the slacking temperature will rely heavily on the cold night temperature. On a sunny day, where high solar irradiation heats up the office space the temperature for cooling control will fall behind. The cooling will not activated despite of high indoor temperatures. In the afternoon people start to go home and the occupancy decreases. The delayed control temperature now inhabits a large share of the warmer temperatures from during the day and will have a high value. The cooling increases while there is no more demand and use for it. 21 Deviations from design In the current HVAC configuration wrong pipeline connections are reported, resulting in hydraulic unbalance (Daelmans, 2012). The engine heat supply pipeline is connected on the wrong place which causes counteracting flows, as is shown with blue arrows in figure 21. Buffer tank short circuiting For the heating circuit the supply temperature (Tsupply) should always be higher than the return temperature (Treturn) because the water in the heating circuit delivers heat to the building. It is observed that sometimes the return temperature after the buffer tank is higher than the return temperature before the Figure 21: Schematic of hot water distribution and buffer tank showing: Hydraulic unbalance in heat pump tank (Treturn). This indicates there is a direct connection and supply and return temperature indications, related to the buffer tank short circuiting flow (indicated in figure 21 with a purple (Priva, 2011). arrow) from the buffer tank into the return water stream, which is actually the buffer tank short circuiting. Valve leakage After the application of the new control software on December 8th the circulation of induction cooling circuit is shut down. In figure 22 can be seen that after this change the temperature of the cooling circuit rises quickly to a temperature similar to the heating circuit average temperature. Off course the temperature of stationary water will adapt to the ambient temperature and this ambient temperature can be influence by the heating circuit pipelines which are situated alongside the cooling circuit ones. According to HVAC technicians the rate of change of this temperature rise does indicate that there is a leakage between the HW and CHW circuit, most likely caused by leaking vales. This leakage is a direct transfer of heat and could be a huge waste of energy, which is very unwanted. Figure 22: Measurements from the HW & CHW supply and return temperatures showing a high rate of change of the CHW circuit warming up after circulation shutdown. 22 3 Simulation and fault detection 3.1 Background Before any kind of FDD can be applied, first the correct or normal behavior of a system or component must be known. If incorrect behavior is observed the necessary improvements must be carried out to achieve correct behavior and define a base case situation. In order to guard and maintain the energy improvements and system performance realized by the commissioning process it is required to continuously monitor the building performance and detect faults in the HVAC functioning. In the literature study report (Hissel, 2012) several FDD methods are compared and their possibilities are assessed. A promising and interesting FDD tool is ABCAT. ABCAT is chosen for FDD applications in this study. Some more detail on a suitable simulation and FDD approach is given below: FDD approach In the analysis of the behavior of a system two approaches can be distinguished: The bottom up and top down approach. Both approaches are investigated and compared . For an affordable, easy to implement and automated FDD system the top down approach is the better option to get to a total building energy analysis and to keep the data processing manageable. ABCAT makes use of the top down approach (Curtin, 2007). In the top down approach the detection of faults is carried out using information from only the highest level of the target system, in this case the total heating and cooling energy consumption. For further analysis it is possible to address lower levels of the system. Method for simulation and FDD Various model-based and statistical regression methods are compared in the research internship (Hissel, 2012). For the model-based methods both the first principle (white box, physical model based) and data driven (black box) approaches are compared. All these approaches have their advantages and disadvantages but still all of them are viable for building FDD and deserve and require additional research in the development of FDD as a whole. As sated before, for this project a physical model-based simulation method is chosen in the form of ABCAT (Curtin, 2007). In the model-based FDD process, first simulation results of the target building are obtained. Fault detection is then carried out by comparison of simulated results with measured results and the analysis of the simulations-measurement residual (also called analytic redundancy). Not every anomaly detected in the system behavior is necessarily an indication of system malfunction. Real faults in the system must be separated from randomly occurring anomalies by means of identification. After fault detection and identification the question remains in what part of the system the fault occurs. For this step in the FDD process the total heating and cooling energy consumption measurements are not sufficient and additional measurements on component level have to be included. Analysis of component level measurements by an expert rule system will be used for the isolation purposes in the case-study project. 23 3.2 Simulation with ABCAT By physical model-based simulations, ABCAT will supply predictions for the energy use of the building for heating and cooling purposes. With ABCAT, faults can be detected making use of analytic redundancy, which is the analysis of the residual between the measured energy use and simulated energy use. Significant deviation (exceeding a certain threshold) will be an indication for an anomaly, which is a possible fault in system operation. Detection of a one-time deviation between measurement and simulation does not necessarily indicate a real fault. One way to identify a real fault is by adding the analysis of cumulative deviations. Without speaking of abrupt faults, a system can show a constant time-scaling performance degradation. Real faults (malfunction) are often not a onetime occurring deviation but are persistent for a longer time. With the proper tools and figures presented by ABCAT, real faults can be identified. ABCAT is an Excel-based program. The interface (input specification and output numbers and figures) consists of various sheets in which inputs and parameters are supplied and numerous output figures are presented. The physical model behind de simulations is present in several Visual Basic macros and will not be visible or of any use for the common ABCAT user. The physical model For simulation ABCAT makes use of physical models of the HVAC system and its components. The models are based on the ‘Simplified Energy Analysis Procedure’ (SEAP) (Knebel, 1983) and can predict the cooling and heating energy consumptions using simplified procedures and only a limited amount of input. Additional characteristics of the SEAP include: Simulations are run on a daily average basis, applied on both the input variables and the output data. This is sufficient for the FDD purpose of the program. The model allows the specification of a fraction of the measured electricity consumption that is converted into heat gain within the building. The building is represented as a two zone model: the exterior zone (subjected to building envelope loads) and the interior zone, which has shown to be an adequate simplification in modeling (Curtin, 2007). Systems with similar functionality can be consolidated into a single system. To be able to simulate various kinds of HVAC system setups ABCAT presents four basic ‘system types’ available for modeling (1 in the schematic). The basic system types are: Single Duct Reheat (SDRH) Single Zone Heating and Cooling (SZHC) Dual Duct Variable Air Volume/Constant Volume (DDVAV/DDCV) Dual Fan Dual Duct (DFDD) The specific HVAC system configuration can be modeled by choosing the most appropriate system type and modify it to represent the system. For the case-study building the Single 24 Duct Reheat model is the most suitable, modifications include: no economizer, 100% OA and heat recovery. Building zones In ABCAT the building is represented as a two zone model, consisting of the interior zone and exterior zone (all areas subjected to the building envelope). The model is divided this way because of the physical differences in building envelope loads (solar transmission and conduction losses). The two zone building model shows to be an adequate simplification in modeling (Wei, Liu, & Claridge, 1998). The interior zone area (%) is one of the ABCAT simulation parameters. As can be seen in figure 3 in chapter 2.1 about all of the building area is subjected to the envelope so the interior zone area is estimated at 10%. The calibration process of the ABCAT model later on shows that alteration of the interior zone area hardly influences the simulation results. This appears to be a very insensitive building parameter. The operation of ABCAT, its required inputs and the produced output are explained in the schematic overview below: 25 26 ABCAT operation details Resolution All the time-dependent inputs and outputs used in ABCAT are treated on a daily basis. Temperatures and humidities are used as 24 hour averages. Energy consumptions are used as total consumption over 24 hours. The use of daily values instead of hourly values (or even smaller) makes sure that the data handling remains manageable. Also the ABCAT developers point out that daily consumption data at the whole building level can provide enough detail to identify and diagnose faults that have a significant effect on energy consumption in buildings. Units Because ABCAT originates from the U.S.A. imperial units are used instead of SI units. To streamline the ABCAT use and calibration processes the units are not changed during this project, the required conversions are made accordingly. For FDD purposes the output data is treated in a relative way, the units are not important for detection and diagnosis results. In case SI units are important for the presentation and comparison of results a conversion will be made. The following units are present in the ABCAT: Table 2: overview of the SI and Imperial units for important ABCAT quantities Quantity Temperature Humidity Thermal energy Length Mass Volume flow Unit SI [°C] Relative humidity [%] Joule Meters Kilograms 3 m /s Unit ‘Imperial’ (ABCAT) [°F] Dewpoint temperature [°F] MMBtu(British thermal unit) Feet Pounds 2 cfm/ft All essential inputs have been discussed. With these inserted in the program, simulations can be run. Simulation produces two columns of output data: HW and CHW energy use. Automatically all the ABCAT output figures will be generated and updated with the new data for the desired time-range. In appendix B the full ABCAT interface and all its inputs and outputs, with their sources, are treated in more detail. The complete set of output figures is treated in appendix C. The most important output figures and their use in the FDD process are discussed below. Before starting to work with ABCAT it should be checked whether the ABCAT models are compatible with the case-study climate system (3.4) and a calibration process has to be carried out to optimize the ABCAT results to represent the building systems behavior accordingly (3.5). 27 3.3 ABCAT diagnosis functionalities The diagnosis process consists of three parts: detection, identification and isolation. In this paragraph the measured HW and CHW energy use (the total energy use for heating and cooling) is mentioned regularly. These are produced from measurements on the case-study building. Detection ABCAT produces energy consumption predictions and presents them in various figures. The following output figures have the specific purpose of anomaly detection and analysis and are expected to be useful for the current project: #1: Time – Energy consumption Figure 23: Example of the regular time-energy consumption plots of both the measured and the simulated energy consumption used for detection. Similar for both HW and CHW. Significant deviations are detected but are not necessarily faults (Curtin, 2007). ABCAT produces a variant for HW and for CHW energy use. Both the measured and predicted energy consumptions are presented. Statistically significant deviations between measurement and prediction indicate a possible fault but a single-time occurring deviation is not a guaranteed fault. Random anomalies can also cause a large deviation, which in case of an anomaly usually disappears just as sudden as it appears. The single time occurring faults are hard to keep apart from single time occurring anomalies. Additional diagnosis is necessary to identify the real faults from other anomalies. Small, persistent deviations will not stand out in the normal time-energy plot. Other figures are more suitable to point these out. 28 #2: Cumulative residual Figure 24: Example of the cumulative residual plot used to detect small but persistent deviations. Small, persistent deviations are a strong indicator for real faults in system performance (Curtin, 2007). In this figure the cumulative value of the residual (measurement – prediction) is plotted as a function of time. If the residual alters between positive and negative the effect will be cancelled out and there will not be a clear, overall trend in the cumulative figure. A real fault, which can be indicated as a constant too high energy use (positive residual) will stand out in these figures. The cumulative residual appears to be a good indicator of the significance of a fault. The value of ABCAT does not lie in daily short-term observations but rather in observations on the order of weeks to months. However, the cumulate plot has plenty of use for identification. #3: Consecutive days of excess consumption levels Figure 25: Example of a consecutive days of excess consumption plot which indicates the persistence of deviations by showing the number of days a fault persists (Curtin, 2007). A slight variation of the previous cumulative plots in which the days of excess energy consumption (anomalies) are counted and plotted as bars. Anomalies lasting for a short time are cancelled out and disappear after a short amount of time. Anomalies persisting for multiple consecutive days will be emphasized in these plots. Lasting anomalies are an indication for faults. The cumulate plot has plenty of use for identification. 29 #4: Outside temperature – energy consumption Figure 26: Example of a plot showing the energy consumption related to the outside temperature making it time-independent. These figures are used to detect and identify structural fault patterns, for example schedule related faults. The red circle indicates HVAC operation in weekends, which is less than on working days (Peitsman, Aker, Claridge, & Bynum, 2010). These figures show a different, time-independent relation between measurements and simulations. Single deviations from the main trend of temperature-energy use will in most cases be random anomalies happening. Usually there is a second trend, being lower than the main trend shown. This second ‘branch’ originates from measurements and simulations in the weekend. During the weekend the HVAC equipment is used less. The outside temperature dependent plots show different relations than the time dependent plots do. Different faults can be found using these plots. #5: Temperature – energy consumption for subsequent periods Figure 27: Example of the energy consumption for subsequent periods plot used to identify degradation of performance over time. Higher energy consumption levels indicate malfunction (Peitsman et al., 2010). Translation of the graph along the y-axis is an indication for change of the system performance over time. An increase shows degradation of performance. A decrease indicates an overall improvement of the system performance, most likely caused by improvement measures taken. Translation and variation along the x-axis is due to variation in weather conditions and overall scatter in results. It does not give information on the buildings energy use behavior. In most simulations one or more of these visualizations of the simulation-measurement relation will definitely show anomalies. The next step is to identify the anomaly. Is it a regular inaccuracy of disturbance or is it system malfunction? 30 Identification Identification is all about separating real faults in the system behavior from regular disturbances or spread in simulation results. ABCAT supplies some figures to aid in this process (as shown in the detection part) but the actual identification should happen by user observation. There is no specific functionality for identification in ABCAT. With the combination of the normal threshold exceeding + cumulative identified faults the indication for malfunction can be quite good. Randomly occurring disturbances will not be present in the cumulative analysis. Logs of maintenance, replacement of components, power blackouts and other clearly observable events influencing the HVAC operation are an important aid in testing the detection and identification capabilities in the testing and assessment step of an FDD application. Isolation The purpose of isolation is to point out in which section of the system the fault takes place. For this step usually additional, subsystem or component level measurements are required. The total building energy use, as simulated by ABCAT does not give any information on subsystem level so the ABCAT output figures have no direct functionality in the isolation process. Isolation is mainly carried out manually and based on analyst experience. The diagnosis method used together with ABCAT (Curtin, 2007) explains how partial isolation can be carried out using the Expert rule method, which is based on the field experience of climate system experts. For the expert rule method the total HW and CHW energy consumption is used in combination with the outside temperature, supply air temperature and economizer deactivation temperature. The economizer is a control part in the AHU which toggles between use of indoor or outdoor air in relation with the ambient and required supply temperatures. 31 Table 3: The expert rule table, as presented in (Curtin, 2007). The expert rule method uses the total HW and CHW energy consumption in combination with the outside temperature, supply air temperature and economizer deactivation temperature in order to determine the category in which the fault is likely to be. The simplified expert rule diagnosis method, used with ABCAT, brings down the complexity of an extensive fault list approach by focusing on a few categories in which the faults are most likely to happen. The seven fault type categories can be seen in the expert rule table (table 2). In the left column of the expert rule table, the likely to occur types of faults are presented. On the right side of the table, columns are present for different temperature ranges: - Outside temperature < Supply air temperature Supply air temperature < Outside temperature < Economizer shutdown temperature Economizer shutdown temperature < Outside temperature For each temperature range the simulated and measured energy use are compared. Whether the residual is strongly positive, strongly negative or within the expected limits gives an indication towards one or more of the fault categories. If a fault with a specific pattern of heating and cooling is observed in only one temperature range, three or four of the fault categories may apply. As more data is gathered from one or more of the temperature ranges, the possibilities for a more deterministic diagnosis improve. Apart from the expert rule method, for this case-study project there are several subsystem measurement results that can be used to point out the malfunctioning subsystem. Detailed isolation on component level is significantly harder because there is less information available and the search gets extended and more complicated. The isolation capabilities with the approaches mentioned will be investigated . 32 3.4 ABCAT compatibility All the required predictions, figures and tools for diagnosis will be produced with ABCAT. The methods available for detection and isolation of faults are investigated and can be applied after accurate simulation results are obtained. Whether the simulations carried out with ABCAT yield accurate results is strongly dependent on the compatibility between the simulations models and the actual HVAC system being simulated. Are the correct physical models available in ABCAT? The compatibility of ABCAT is addressed by investigation of restrictions and requirements presented in the documentation and by comparison with reference (successful) projects. The system restrictions stated in the ABCAT documentation give some basic requirements that the building HVAC system must meet. The HVAC system basics show a close match: AHUs (Single duct, hot and cold coil, no economizer, heat recovery) Single duct CAV (which is one of the options in ABCAT) Heat pumps + boiler as central plant (the HW, CHW plant is not discussed in the documentation) Available System Types in ABCAT: 1. 2. 3. 4. SDRH: Single Duct Re Heat (VAV and CV) SZHC: Single Zone Heating and Cooling DD: Dual Duct (VAV and CV) DFDD: Dual Fan Dual Duct The building shows the best match to system type 1, which will be used for further simulation. The case study HVAC setup does meet the system type requirements. Next to requirements for the core HVAC system type there are additional system configuration assumptions that have been confirmed as well, as can be seen in table 3. Table 4: ABCAT system configuration assumptions with their validity for ABCAT System configuration assumptions The conditioned area is served by single duct AHUs Both the supply and return fan have a VFD Economizer mode with interlinked relief air, return air and outside damper AHU coils receive HW and CHW water from a central plant The CHW and HW circuits have VFD circulation pumps Preheat coil and humidifier are not used in this system Compatibility + ± ± + + + Remarks Supply fan = fixed but ABCAT does accept fixed frequency after all Economizer is absent and can be toggled off in ABCAT Most of the configuration assumptions are met without a doubt. Two of them show a slight difference in the exact definition, but the ABCAT parameter sheet shows that these deviating components can be excluded from simulation, which makes these configuration incompatibilities irrelevant. The case-study HVAC system does meet the configuration assumptions. 33 Comparison current HVAC system with systems treated in literature In literature (Claridge & Lin, 2009; Curtin, 2007; Peitsman et al., 2010) a lot of examples are available of FDD projects successfully carried out using ABCAT. By comparing the HVAC setup and building characteristics of these successful projects with the setup in the casestudy building more insight can be obtained on the likely compatibility. Test cases of this type of FDD in real buildings provides the most realistic and valuable feedback on successful application and compatibility with building systems. Compatible HVAC setups found: Sbisi Dining Hall (College Station, TX) Dining facility, 19 AHUs for all heating and cooling purposes, SDVAV and SDCV with terminal reheat, operation only during working days and hours, thermal energy (chilled water (CHW) and hot water (HW)) supply from central utility plant. Computing Services Building (Austin, TX) Computer test laboratories and offices, SDVAV units with economizers and mixed air preheat for interior zones, DFDD units for exterior zones, energy intense computer laboratories are complemented with extra cooling fan coils. DASNY Corporate Headquarters (Albany, NY) Office space, two chillers and five boilers for CHW and HW generation, twelve SDVAV AHUs with mixed air preheat, economizers and terminal reheat and hot water radiant heating. OPPD Energy Plaza East Building (Omaha, NE) Office space, SZHC AHU with economizer and four SDVAV AHUs with terminal reheat and economizers. Vertigo Building TU/e (Eindhoven, NL) Offices and general study space, one heat pump and two boilers as CHW and HW plant, four AHUs with heat recovery, convective radiators and a four pipe climate roof for heating and cooling distribution. The Vertigo building represents the case-study setup best. Successful simulation and FDD on this building strengthens the assumption that there will be no compatibility problems between ABCAT and the Strukton case-study building. The fact that gas-powered heat pumps are used as heating and cooling source in the case-study does not present any issues. In the FDD projects from literature the CWH and HW source is always mentioned as ‘Thermal energy from central plant’. The source and type of device generating the heat and cold is irrelevant. Most of the presented HVAC systems make use of only AHUs for heating and cooling of the building. A wide variation of AHU types are successfully simulated in the past so it can be presumed that the modeling of the case-study AHUs will not likely be a problem. However, 34 additional devices for heating and cooling distribution, like radiators, induction heating and floor heating are not treated in these projects. This raises question on the extend of the compatibility of ABCAT with these systems. Because of the correspondence of systems in the case-study building with the systems available in ABCAT and the close match with the Vertigo case-study, which was successful, it is expected that ABCAT can be applied without direct compatibility issues. On the other hand compatibility must not be taken for granted and incompatibility could still cause deviating results in the further steps of the simulation and FDD process. The data acquisition and insertion in ABCAT has been successful. ABCAT simulations can be run without much trouble. The results from the initial simulations, however, do not correlate well with the measured CHW and HW energy use. As can be seen in (Curtin, 2007), this is very common for the first simulation rounds of a new building in ABCAT. The ABCAT model must be calibrated to adjust the simulation results to match better with the measurements. 3.5 The calibration process of ABCAT In order to be able to successfully simulate the cooling and heating energy use of a building the calibration to actual measured heating and cooling energy consumption has shown to be an essential step. While reading some ABCAT oriented reports (Curtin, 2007; Peitsman et al., 2010), the unique calibration functionality of the program is mentioned numerous times but how this function works or how it is applied is never explained. Concerning the functioning of ABCAT, the program is filled with various Macros that contain all of the simulation processes of the program. Initially this calibration functionality was expected to be present in these Macros as well. Because of the lack of Excel macro experience, it was not possible to find and use the calibration function within these macros. The calibration functionality remained a mystery for a long time and only after sending out various help requests to experienced ABCAT users an answer was found: Calibration is carried out manually instead of automated with Macros Calibration happens by alteration of building parameters, there are no specific calibration parameters Manual calibration happens to be a process by altering values for several iterations and switching to simulation result figures to see the improvements and make further adjustments Statistics will guide the calibration process, statistic metrics comparing the simulation and measurement results are available in the ABCAT interface A manual for calibration was obtained (Claridge, Bensouda, Lee, & Wei, 2003) The manual calibration process to achieve this agreement between simulations and measurements appears to be time-consuming. The calibration process can be optimized and speeded up using a proofed methodology for rapid calibration explained in the “Manual of procedures for calibrating simulations of building systems“ (Claridge et al., 2003). This manual introduces the use of calibration signatures, that characterize the difference between measured and simulated performance. Specific calibration signatures are available 35 which are suitable for the four different system types: SDVAV, SDCAV, DDVAV, DDCAV. These system types are not the same as the system types present as ABCAT physical models (Chapter 3.2), although they have strong similarities and can be used together. Individual sets of calibration signatures are available for different climate types. The method for calibration and its application will be further explained in this chapter. 3.5.1 The need for Calibration Conventionally the inputs for energy simulations of buildings have been based on design data. From the experience of the authors, who have carried out hundreds of energy simulations, it can be seen that differences of 50% or more between simulations and measurements are not unusual. These differences are not thought to be due to errors in the simulation software itself, but more likely due to errors in the assumptions of input parameter values for a particular building, due to misunderstanding of the building’s design or to the differences between the design and as-build conditions and parameter values. Various techniques for calibration exist, which either measure or infer the characteristics of individual buildings as they were built and operated and identify candidate changes in model inputs that may resolve the differences. The improvements in the correspondence between the simulated and measured consumptions realized by calibration found by the authors are deviations of less than 5% on a yearly basis and 5-10% on a monthly basis. Once a probable error in a simulation input has been identified, the analyst must typically assess whether the change makes physical and intuitive sense. Calibration processes in general are time consuming and require a great deal of specialized expertise. A procedure that can quickly produce a reliable calibration for simulations of large commercial building HVAC systems would be very valuable. Uses for these calibrated simulations are: References for energy audits of existing buildings Energy savings determination after a retrofit Estimation of potential energy savings for different virtual scenarios Continuous fault detection and diagnostics 3.5.2 Required amounts of data Regular model based simulation programs can require large amounts of input parameters of which a few dozen are crucial to the simulation results. In case that simulations are carried out on a short time period and/or with large increments (for instance monthly values) it could happen that the number of parameters that can be varied exceeds the amount of data points that are being fit. In this case the model for this simulation will be ‘mathematically over determined’. There will be more equations than unknowns. Even if a strong correspondence between measurements and simulation results is achieved by calibration, this will not necessarily be the case for future data as well. Calibrations carried out on problems with insufficient amounts of training data are not guaranteed to be reliable for further predictions. It is advised to use at least several months of daily consumption data for the calibration process to eliminate this problem. For the calibration of ABCAT in this project only a period of four months was available. This four 36 months of daily energy consumption data is not guaranteed to be sufficient. Ideally a larger amount of data should be used for calibration but this was not possible in the scope of the project. Depending on the calibration and simulation results it can be concluded whether the data set is sufficient or not. Another key requirement for the calibration data set is that the simulation period covers enough of the annual ambient temperature range. Preferably the whole year is covered and multiple years is even better to have more extreme weather situations covered. For the calibration in the case-study period (January – April as will be explained later) this variation in temperature is not fully covered. The use of hourly data (or even smaller increments) is another possibility, although it has a major downside. Dynamic effects of thermal and other fluctuations that happen over the course of the day will become more evident and can cause problems in the calibration process. Using daily consumption data these effects will be averaged out and will have no effect, which makes the use of daily data the most common choice. Hourly data can still be used for calibration fine tuning but it is not addressed during this project. The quality of the measured data is also an essential factor to the success of calibration. The data quality is assessed and addressed in appendix G. Identification and fixing of erroneous and missing data is an important step to take before starting a calibration procedure. In some cases it is appropriate to interpolate in order to fill gaps in the data range, while in other cases it is better to remove erroneous parts from the measurement range. In the measurement data validation chapter (Appendix G) and the data processing chapter (2.5) the quality of available measurement data is assessed. 3.5.3 Calibration method Calibration is carried out using the calibration signature method, it enables experienced user to calibrate a building with large built-up systems in 10-40 hours. This calibration approach has been successfully demonstrated in test cases with ABCAT and other simulation programs, carried out by people from Texas A&M. No doubt this will also be a suitable method for calibration of the ABCAT simulations in this project. The calibration of ABCAT is explained in the schematic overview. 37 Figure 28: Schematic overview of the calibration process for ABCAT. All steps in this overview are treated in more detail in the text. 2) Calibration Signature The calibration is based on a specific graphical representation of the difference between simulation results and energy use measurement, which is called the ‘calibration signature’. (Wei et al., 1998) found out that calculating the difference between measured heating and cooling consumption and the consumption predicted by uncalibrated simulation, normalizing them and plotting them as function of the ambient temperature, provides useful information about the required changes in input variables. Energy uses are represented against the outside temperature in order to make most of the calibration a time-independent process. The only aspects of calibration in which time stays an important factor is the calibration of the differences between workday and weekend HVAC operation. These differences are mainly dependent on: building occupancy schedules, HVAC weekend schedules and weekend electric load (Internal gains). Calibration of the weekend behavior should be carried out using time-based figures. For a given system type and climate the graph of this simulation-measurement difference against the outside temperature has a typical shape which is directly related to the cause of the difference. In order to make the characteristics more distinctive a second order polynomial fit of the simulation-measurement residual is used to eliminate a wide spread and to give a clear overview of the main trend. The actual calibration signature is obtained as follows: 38 The figure below shows an example of the use of the calibration signature. The measurement-simulation graph shows that the simulated cooling consumption for low outside temperatures is much too high. The calibration signature on the right shows the same information but then in a more distinct way and also tells which profile of change is required in order to get a closer agreement, as will be explained below. The next step is to find out which alterations to the simulation will result in a large CHW energy use decrease for low temperatures while not changing the energy use for high outside temperatures. Figure 29: CHW Measurement-simulation relations for the same situation. Left: individual measured and simulated consumptions. Right: Edited variant showing the CHW simulation-measurement residual as a calibration signature. Note that this calibration signature always has positive values, decreases with increasing outside air temperature, is about 60% at 35°F and reaches zero at about 75°F. These are characteristics of the occurring deviations that will be useful to determine what errors are present in the simulation inputs. Calibration signatures are typically calculated on a daily average basis, but other time steps are possible as well. The energy consumption values can be whole building or specific system consumption. They can originate from electric or thermal data and will end up being dimensionless. Because ABCAT is used for the simulations it will be the total HW and CHW energy use in MMBtu/day. 3) Characteristic signatures Next to the calibration signature, which illustrates the current measurement-simulation deviations, the calibration method makes use of ‘characteristic signatures’. These are again normalized plots, but this time they reflect the changes in simulated energy consumption as a result of the alteration of a specific input parameter. The characteristic signature is defined as: By comparison of the calibration signature of an uncalibrated simulation with a variety of published characteristic signatures an indication can be found on which input parameters to change in order to improve the simulation results. A library consisting of several characteristic signatures is included with the calibration manual but it is also possible to calculate new characteristic signatures using any simulation program, so also with ABCAT. This is done by simulating the building with one value for an input parameter, the baseline run. Then changing that same input parameter by a given amount an rerunning the simulation. The change in simulated energy consumption is the basis of the characteristic 39 signature for that specific parameter change. The characteristic signature shows all changes in terms of the percent change relative to the maximum value of the energy consumption in the baseline case. Again the characteristic signature is displayed as a function of the outside temperature. The characteristic signatures can also be seen as a parametric sensitivity analysis of the building and HVAC system of interest. CHW (%) 0 -10 20 40 60 80 100 Outside Temperature (F) V_eDsqft - 0,01cfmft 10 HW (%) V_eDsqft - 0,01cfmft 10 0 -10 20 40 60 80 100 Outside Temperature (F) Figure 30: Example of a characteristic signature. By decreasing the exterior zone air flow the chilled water energy use will remain unaltered. The hot water energy use will increase slightly though. The signature has a negative slope so The figures below show an example of a characteristic signature. By decreasing the exterior zone air flow the chilled water energy use will remain unaltered. The hot water energy use will increase slightly though. The signature has a negative slope so the hot water energy use will increase more for low outside temperatures than for high outside temperatures. Also the percentage of change is given which is a good indication for with which magnitude the specific input parameter should be changed in order to match the simulation calibration signature. The clues provided by the characteristic calibration signatures are much clearer when CHW and HW signatures are combined. These two will typically show different trends, and the combination is a powerful indicator of the input parameter that need to be changed in order to improve the simulation results. 3) Weather signatures The characteristic signatures are not only dependent on the specific input parameter and the type of HVAC system being simulated, but also on the weather conditions for the building-system location. The signatures clearly depend on the outside temperature. But also, although not explicitly shown, they depend on the ambient humidity levels when these are high enough to induce latent cooling loads. Because of this the signatures will be different for sites with different weather conditions. This is treated by categorizing different sets of characteristic signatures for different weather characteristics. Characteristic signatures can be reused from weather signatures that match the weather signature of the site in question. 40 Eindhoven, NL 20 30 40 50 60 70 80 90 100 110 Outside Temperature (F) Figure 31: The weather signatures used in the manual (Claridge et al., 2003) and the weather signature for Eindhoven. The weather characteristics are found by plotting the relative humidity with respect to the outside temperature for Eindhoven for a long time period. This characteristic gives a clear representation of the present climate type. Using the best fit regression the big cloud of individual data points is processed to an overall, single line, trend. Weather data is derived from the KNMI database spanning 10 years of measurements in Eindhoven. The Eindhoven weather signature seems to match the most with the one from Oakland, CA. So, when addressing the characteristic signatures given in the manual the signatures from Oakland, CA to would be the best set to be used. But still the weather signatures do not form a perfect match and so the exact calibration changes performed by using the Oakland characteristic signatures will be slightly different. When it appears to be necessary for proper calculation, it can be chosen to adapt the characteristic signatures for the specific weather situation in Eindhoven. 3.5.4 Calibration evaluation and statistics In order to evaluate the quality of calibration of the simulations 1) several metrices can be used. The two metrices suggested in the calibration manual, and also available in ABCAT, are the Root Mean Square Error (RMSE) and the Mean Bias Error (MBE). The RMSE is defined as: The RMSE is a good measure of the overall magnitude of the deviation. It represents the size of errors and the amount of scatter, but it gives no information about the overall bias in the data. If large errors are distributed both above and below zero the RMSE would have a high value. Similarly if all the errors would be positive, the RMSE would have the same value. The RMSE is stated as being an useful metric of how ‘good’ the simulation results are in comparison with the measured results. The manual also says that in general a RMSE between 5 and 10% of the mean value of energy consumption is a very respectable result. It will be difficult to achieve a lower RMSE in a regular building energy use calibration process. 41 The MBE is defines as: In contrast with the RMSE, for MBE the positive and negative errors will cancel each other out. The MBE is an overall measure of how biased the data is and is also a good indicator of how much error will likely be introduced in the annual energy consumption estimates, since here positive and negative errors are cancelled out as well. A simulation with a small RMSE, but with significantly larger MBE, indicates a possible error in simulation inputs. The residual is not significantly large but it is constantly either positive or negative. The simulations can usually be improved by altering the responsible parameter for a specific amount. A simulation with a large RMSE, but with a small MBE, might have no errors in input parameters, but the building performance is likely to show some unexpected, unmodeled behavior that is difficult to simulate. Minimizing both the RMSE and MBE is essential if a calibrated simulation is to be used as a baseline for estimation of energy savings or fault detection and diagnosis. The basic calibration process will start out by carefully monitoring the RMSE for both cooling and heating after each input parameter alteration carried out. Also the mean RMSE is important because often a change to inputs will increase the heating RMSE while decreasing the cooling RMSE, or the other way around. If the mean RMSE decreases, it was a useful alteration after all. The residuals should be scattered around zero and the calibration signature has to be flat and should show no trend with the outside air temperature (Claridge et al., 2003) in order to come to an accurate calibration result. In the evaluation process (step 4) in the calibration overview) given pairs of heating and cooling calibration signatures are compared to the characteristic signature pairs, especially paying attention to characteristic intersects, slopes and bulges. It also has a high priority to pick a characteristic signature that results in improvement of both the HW as CHW results. Input parameters should be identified that, when changed, are most likely to minimize the residuals over the targeted range of outside air temperature. If multiple pairs of characteristic signatures have similar shapes, it has to be estimated which one is most likely to be inaccurate in the initial simulation. If the calibration signatures do not show a strong resemblance with any of the characteristic signatures, take an input/signature that will counter the residual with the highest magnitude or one that will remove any strong irregular shapes in one of the calibration signatures. The amount of change should be estimated by comparing the magnitudes of the normalized residual and the normalized result change in the characteristic signature. Parameter alteration (step 5) is followed by new simulations (step 1), new calibration signatures (step 2) and the evaluation (step 4) of the new RMSE, residuals and cooling and heating calibration signatures. Step 1, 2, 4 and 5 (in the calibration overview) have to be repeated until the total RMSE is minimal and the calibration results are satisfactory. 42 3.5.5 Evaluation of the supplied characteristic signatures Four HVAC systems are provided: SDCV, single-duct constant-air-volume SDVAV, single-duct variable-air-volume DDCV, dual-duct constant-air-volume DDVAV, dual-duct variable-air-volume The HVAC system in the case study building is of the SDCV type. The air flow produced by the air handling unit is constant and the amount of heating and cooling of the air stream is constantly adapted to the user’s demand. Only characteristic signatures of the SDCV system type set are addressed in this study. The following parameters can be found in the characteristic signature libraries: Cold deck temperature Hot deck temperature (DD systems) Supply air flow rate (CV systems) Minimum air flow rate (VAV systems) Floor area Preheat temperature Internal gains Outside air flow rate Room temperature Envelope U-value Economizer These parameters have been chosen by the authors because they are frequently containing errors causing inaccurate simulations (Claridge et al., 2003). In order to find out the significance of these given parameters for the case study project the Characteristic signatures supplied with the manual are compared with the creation of new signatures: 1. The Eindhoven weather characteristics do not exactly match the given location weather signatures, which is one incentive to choose for making new ones. 2. The given characteristic signatures include a lot of parameters that are either absent in the set of ABCAT input parameters or later on appear to be very insignificant parameters with little influence on the simulation results. Examples are: Hot deck temperature (N/A for SDCV systems), Minimum air flow rate (N/A for SDCV systems), Preheat temperature (N/A) and Economizer parameters (N/A). 3. ABCAT has a lot of input parameters that are not covered in the signature library. Parameters like Heating & Cooling (+ setback) set-points, Wall and window areas and U-values, Heating and Cooling curve set-points, Various Air Flow rates, Thermal Mass and Area per occupant are likely to be significant input values. The complete 43 list with ABCAT inputs is covered elsewhere in the report. This availability of plenty of additional input parameters is another good reason to create new signatures. 4. General characteristic signatures are not necessarily valid and precise for any kind of building and simulation program. Custom made signatures for a specific situation are expected to be better than generic signatures. Because of these reasons, it was chosen to create a new set of characteristic signatures for the specific building for the locations Eindhoven, Netherlands. How the signatures are made can be seen in the section about characteristic signatures earlier on. The produced characteristic signatures can be found in appendix D. 5) Input parameters to adjust A selection has been made of parameters that: - Make sense to change - Are not strictly defined for the case study building. Certain obvious and very specific parameters such as: system-type, solar irradiation, outside temperatures, absence of economizer and the measured electric energy use are strictly defined and not suitable for alteration - Actually do have an impact on the simulated energy uses. Parameters that need a large, unrealistic alteration to achieve significant changes are discarded Area and mass: - Floor Area: Is not present in the supplied library but clearly has a significant impact on heating and cooling energy consumption. - Area per Occupant: The increase of area per occupant brings very interesting changes in the form of a strong negatively sloped profile. Slight increase of the area per occupant decreases heating and cooling for low temperatures and increase them for high temperatures. - Thermal Mass: An increase does not change anything. A decrease has a positive influence on the cooling demand. AHU volumetric flow rates: - Exterior Zone Air Flow - Interior zone Air Flow - Exterior Zone Outside Air Flow - Interior Zone Outside Air Flow - Exterior Zone Minimum Air Flow - Interior Zone Minimum Air Flow All of these inputs have an influence on the heating and cooling energy consumption and are therefore added to the set of characteristic signatures. Envelope: Wall area and window area: These parameters need large alterations in order to achieve significant changes to the energy use. Large increase of these areas produce a small cooling 44 increase for low temperatures and a small heating decrease for high temperatures. Because such large alterations are required it might be better not to use these parameters. Wall U-value and window U-value: A decrease of the U values results in a decrease of cooling and an increase of heating requirement, both with a positive slope. This makes it an interesting profile for simulation manipulation. Heating and cooling setpoints: These are interpreted as the temperatures at which the hot and cold circuits are maintained. Water from these circuits will in turn be used to provide the heating and cooling coils in the air handling units. The heating setpoint only influences the heating consumption for low outside temperatures. The cooling setpoint does influence both the heating as the cooling energy uses. 30 25 20 15 10 5 0 Heating set point (°C) Cooling set point (°C) Heating and cooling curves: In ABCAT specified as ‘Coil set-point temperature schedules’. 0 30 25 20 15 10 5 0 0 20 25 30 Outside temperature (°C) 20 25 30 Outside temperature (°C) Figure 32: The heating and cooling curves of the case-study Air Handling Units (AHUs), which are important parameters for the ABCAT calibration procedure. A low outside temperatures require less cooling, hence the setpoint for cooling has a higher value for low outside temperatures and a lower setpoint (more cooling) for higher temperatures. Two of these setpoints are used in combination with linear behavior between them as can be seen in the cooling cure in figure 32. Similarly more heating is required for low outside temperature and less heating is required for high temperatures, resulting in the heating curve shown above. Alteration of both the outside temperature markers and the corresponding control set points result in significant and a wide variety of changes to the simulated energy uses and is an important addition to the signature set from literature. The corresponding characteristic signatures can be found in appendix E. 3.5.6 Definition of calibration period and test period Before calibration can be started it is important to make a substantiated decision on which time period is to be used for calibration purposes and which period will be the target test period for the case-study system. A calibration period must show uniform, consistent behavior and the time range should be sufficient. A test period has less requirements but it is advantageous if there is a period with changes to the system, for which both the time and the description of the change is documented. 45 Two periods of measurements of uniform HVAC system operation are available: 1 August – 30 November 2011 (before the HVAC control improvement) 1 January 2012 still running (after the HVAC control improvement) The first period is not suitable because the system has not been optimized yet, which greatly decreases the value of simulating this period with inferior functioning. The second period is much shorter, which can be problematic for calibration but here the system is optimized and will yield much more useful results for building optimization and FDD. It is chosen to use the period of the 1st of January to the 1st of April 2012 for calibration of the ABCAT model. During the month December several modifications and optimizations have been carried out on the HVAC system. The modifications have been properly documented. This period is suitable for testing the calibrated model on its fault detection abilities. Figure 33: Indication of different periods within the total measurement period, respectively: 1) Period of initial measurements, pointed out to represent suboptimal system behavior 2) Period with major system changes and improvements making it a good test period 3) Calibration period, representing continuous, optimized climate system behavior The measured quantities (mass flow and heat flow in the induction circuit) are not important in this figure. It is the characteristics of the behavior shown (chaotic with much deviations or stable and consistent) which is important in this figure. 46 3.5.7 Application of the calibration process on the case-study building Trough the analysis of strange calibration results caused by certain parameter alterations some incorrect relations in the interface-model connection (section 5.1) were found and corrected before calibration procedures could start. The calibration process is initiated according to the step-by-step calibration overview presented in 3.5.3. The firs ABCAT simulations are applied on the calibration period with the best estimates of input data and settings, yielding the following results: ABCAT simulation results using best estimates for input parameters: a b c d Figure 34: ABCAT simulation results next to measured data during the calibration period and for simulations using the best estimates of building parameters. a) CHW measurement and simulation as function of time b) HW measurement and simulation as function of time c) CHW measurement and simulation as function of ambient temperature d) HW measurement and simulation as function of ambient temperature Observations: - - 34a and c: Overestimation of the cooling energy consumption. ABCAT does not ‘understand’ the cooling should be off for outside temperatures below 5°C. This can be corrected by applying a manual winter cooling block in the ABCAT calibration process. 34b: The simulations show a strong weekend dependency. For weekends the model overestimates the energy consumption. Instead of higher energy use in weekends, this should be lower than for working days. 47 Calibration signature generated for the best estimates / baseline simulation case: For the base case simulation with best estimates the calibration signatures are generated in an extra ABCAT output sheet especially designed for the calibration process of this project. The calculation of a calibration signature is treated in 3.5.3. This calibration signature represents the deviations between simulations and measurements in the current state of the ABCAT model as can be seen in figure 35 and is used for the selection of further calibration steps. a b c d Figure 35: ABCAT calibration signature showing simulation - measurement residuals related to the outside temperature for the ‘best estimates’ model situation. RMSE values are added for model accuracy evaluation. These calibration signatures are the baseline for the characteristic signature calibration approach. a) CHW simulation-measurement residuals for each individual measurement point b) CHW calibration signature, including normalized and polyfit representation c) HW simulation-measurement residuals for each individual measurement point d) HW calibration signature, including normalized and polyfit representation In the baseline simulation some specific, large deviations between simulations and measurements stand out (as is seen in the observations of figure 34). These are the Winter Block issue and the Weekend Behavior issue. These problems are not caused by wrong estimates and easily solvable by calibration but are more related to flaws or shortcomings in the ABCAT model. A lot of improvements are expected from the application of winter block and weekend fix before further application of the calibration method. The order of improvements/calibration steps is as follows: 48 Initial 'Best Estimates' simulation 2) Application of the calibration method 1) Correction obvious flaws in model Calibrated model Figure 36: The order of steps taken in the calibration process 1) Correction for model shortcomings Winter block: The case-study building is outfitted with new control software which includes a blockade for cooling in winter and heating in summer, as further explained in 2.6.2. For outside temperatures below 5oC the cooling is shut down completely. This is some behavior that is not present in the ABCAT building models and it is logical that this is not simulated correctly. An easy way to fix this is to add a manual Winter Block. The manual Winter Block is applied by changing the values in the ABCAT output cells by a simple filter loop: IF outside temperature < 5 simulated value CHW consumption = 0 The manual Winter Block works with satisfying results, as can be seen in figure 37. Because of practical reasons (every simulation run removes the filter again, and the filter does overrule the exiting simulation results) this solution will be ordered to be the last applied step of processing. Figure 37: Application of the winter block for cooling, related to time. Weekend Fix: To correct the overestimation of energy consumption in weekend, several occupancy and HVAC operation schedules are altered to find a parameter which can correct this behavior. The weekend fix is an adjustment that is related to the weekday-weekend behavior schedules and is in no way compatible with the calibration signatures and characteristic signatures methods(temperature related). Schedule alteration has to be applied separately from the calibrations signatures method to prevent negative influences. The best correction for weekend behavior was found by decreasing the HVAC Operation Fractions for weekends from 1 to 0.5. 49 Figure 38: Application of corrections of the weekend schedule and behavior shown on the HW energy consumption figure, related to time. 2) Calibration using the characteristic signatures method Making use of the calibration signatures method calibration is carried out to find the best possible match between measurements and simulations. The calibration signature above (figure 38) was used in combination with the characteristic signatures from appendix E. The order of calibration steps is based on the best shape matches of the signatures: The internal gains factor was optimized to 0.6. This gave the best tradeoff of improvements to both heating and cooling simulation. The Air flow (OA, exterior zone) decrease from 0.1cfm/ft2 to 0.03cfm/ft2 was a major improvement for the cooling simulations. In order to increase the heating simulation results the heating set-point was increased from 21oC to 24oC which gave the best matching results. In table 5 the Root Mean Square Error for the cooling, heating and total simulation accuracy are presented for the chosen steps in the calibration process. The RMSE is a performance indicator for the quality of calibration results. The average values of measured cooling and heating energy consumption are: CHW about 3 MMBtu/day and HW about 14 MMBtu/day. As a percentage of the mean consumption values the final RMSE values are respectively 50% and 35% resulting in a total relative RMSE of 43%. According to the calibration manual a RMSE between 5 and 10% of the mean value of energy consumption is a very respectable result for a FDD oriented building model. The use of RMSE is further explained in 3.5.4. Table 5: Showing the resulting RMSE after several steps in the calibration process. The RMSE indicates the quality of calibration, lower RMSE indicates a better result. Calibration state Initial situation Internal gains factor correction Air flow correction Heating set-point correction Winter block & Weekend fix Cooling RMSE 1 [MMBTU/day ] 10.71 (360%) 10.66 (360%) 3.29 (110%) 3.29 (110%) 1.5 (50%) 1 Heating RMSE [MMBTU/day] 4.58 (33%) 4.48 (33%) 7.70 (55%) 5.51 (40%) 4.9 (35%) Total RMSE [MMBTU/day] 7.64 (200%) 7.60 (200%) 5.50 (80%) 4.39 (70%) 3.3 (43%) Million British Thermal Units (1 BTU = 1,055.06 joules) The unit is not important though, the results are treated in a relative way. 50 By the combination of the two simulation improving strategies mentioned above the best simulation calibration result is obtained. The best results are presented in the results section (Chapter 4). 3.5.8 Remaining calibration problems and challenges Nearly all the parameter changes influence both the heating and cooling energy consumption. Improving one of them often results in the other one getting significantly worse. The ideal combination of parameter changes to have them both optimized has not been found. As can be seen in figure 39a (emphasized with red lines) the produced HW energy consumption simulations all appear to be very flat figures. Correct simulations would show more sloped figures, caused by the outside temperature dependence of the heating demand. This temperature dependence sloped relation with the outside temperature is not present. There is an unexpected and unexplained boundary at 55°F outside temperature as can be seen in figure 39a (emphasized with the blue vertical line). The simulation results for heating energy use are completely different on the two sides of this boundary. This phenomenon is expected to be related to the applied physical model but no explanation of this phenomenon is found. The 55°F value is NOT present as an ABCAT input and did not change during the calibration process and alteration of many of the ABCAT input parameters. Sometimes with a small alteration of a parameter, ABCAT suddenly comes up with an extreme energy use estimation shoot out (figure 39b) that cannot be explained. All the environment data inputs (electricity use, temperatures and humidity’s) of the days around these shoot outs are completely in line with those from the days with predicted shoot outs. Figure 39: The following figures illustrate some of the mentioned problems. They are not from the same simulation and should not be seen as being related in some way. ABCAT simulation figures giving examples of: o Absence of temperature dependence, the 55 F boundary and unexplained shoot-outs. 3.5.9 Conclusions on the Calibration process Calibration insufficient as can be seen in the large deviations between measured and simulated energy use during the calibration period and the high values for RMSE. The calibration does not lead to respectable results by far. Calibration failure is assumed to be caused by both ABCAT issues (prototype and unguaranteed compatibility) and unwanted behavior of the HVAC systems in the case-study building. 51 52 4 Case-study results 4.1 Building heating and cooling energy consumption Figure 40: Showing a bar plot representation of the daily energy use for heating and cooling over time. The average outside temperature is added to emphasize the temperature dependency of the energy consumptions. Through the measurement setup applied on the case-study building and the proper processing of individual measurements into building total values, the following daily heating and cooling energy consumption quantities are obtained. The measurements and calculated results are validated (Appendix G) and are expected to provide a close representation of the real energy consumptions in the building. The accuracy is not perfect but assumed high enough for building level FDD application. 53 4.2 ABCAT calibration As already mentioned in chapter 3.5.8, the calibration of ABCAT was not successful. Making use of the calibration signatures method, figure 41 shows the best measurement – simulation match that could be achieved. Based on visual comparison of the calibration and characteristic signatures there still seems to be room for parameter alterations which improve the measurement – simulation match. In practice though, no more alterations have been found which improve both the heating and cooling energy use simulations simultaneously. Improvements to CHW results are accompanied by decreasing correspondence for HW results and vice versa. Figure 41: ABCAT output figures for simulations with most accurate calibration results. Because of the poor correlation between measurements and simulations this calibrated model not suitable for FDD purposes. Observations: 41a &c: Due to the manually applied Winter Block the cooling for low outside temperature shows a satisfactory match. The small differences are explained by the daily averages that ABCAT uses. If the average outside temperature is < 5oC the cooling for the whole day will be 0. In reality, on cold days (mostly < 5oC ) for the majority of hours during the day there will be no cooling, but for the hours during the day that the temperature > 5oC there can be some cooling. So small amounts of cooling can still be measured. By the calibration method it was possible to have the cooling predictions in the same range as the measurements, as can be seen in the figure. However, the separate measurement and simulation data points do not properly match though due to improper simulation results. The heating results shows even less correlation. 54 4.3 Fault detection results As previously stated, no accurate simulations of the building energy consumption have been achieved. The fault detection process, making use of measurement-simulation comparison, is strongly dependent on the quality of simulation of the target system. Consequently, the intended level of fault detection is not possible with this poor simulation quality. However, it has been investigated to which extend it is possible to detect certain faulty behavior with these poor results. As can be seen in figure 42 there are certain large deviations that even stand out to the overall residual that is present. Figure 42: ABCAT results with the poorly calibrated model for the 'test period'. Although calibration is considered insufficiently accurate for FDD purposes it is possible to observe and analyze some of the large and obvious deviations in these figures. Observations and provisional conclusions: 42a&c: The measured cooling consumption during warm days is significantly higher than simulated. Even though the simulation quality is poor and the deviations can just as well be caused by wrong simulations, observation of these figures can be an incentive to take another look at cooling set-points in the building for optimization possibilities. 42b&d: A large deviation between simulated and measured heating is observed for higher outside temperatures. The simulated values are rather low and unrealistic. The measurements show the actual behavior of the system. The simulation quality is indeed too low to detect faults in the system behavior. Observed deviations are due to simulation flaws and not to anomalies in HVAC system operation. 55 4.4 Diagnosis results Because simulation quality was insufficient , the ‘Expert rule’ method could not be used. No conclusions on the capabilities of this method could be drawn. Although the simulation and fault detection using ABCAT FDD functionalities is proven to be unsuccessful there are other ways to diagnose and locate faults in the system behavior. Analysis of measurements carried out in subsystems or on individual components of the building HVAC system, without the application of any simulations can also yield interesting observations. Measurements on some well chosen, sub-system level locations can give valuable information on the performance of this specific subsystem. Measured anomalies in behavior can be directly related to failure or unwanted behavior of the subsystem. An expert is required to analyze the measurements in order to correctly identify the fault. Analysis of component level measurements is not so strong for detection and identification purposes. There are a large number of individual measurements to take into account in contrast with the top down approach, in which faults are detected by analysis if total energy consumption only, which supplies a much better overview on the building climate system status. Individual component measurements also show much more fluctuations than the total energy use, making the identification more difficult. On the other hand will component level analysis yield direct isolation possibilities. If a fault is detected then it is directly known in which component this is caused. Although it was not a direct goal or intention for this project, possibilities for the mentioned diagnosis methods occurred. As part of the building energy performance assessment several subsystem measurements were observed during regular use of the BMS interface and during the data processing applications with Matlab. From these subsystem measurements various conclusions are drawn which can are definitely interesting applications of diagnosis without use of simulations software. Examples (treated in more detail in chapter 2.6.3: building energy performance after improvements): Temperature measurement of supply water, return water and return water after buffer tank outlet will give insight on the buffer tank layering capabilities. Analysis of the rate of temperature change in sections of HW or CHW piping after being (temporarily) shut down yields information on possible leakage between the two. By having a control technician (expert) analyze the control schemes present in the BMS various control faults will definitely be found: o There is no winter/summer block resulting in simultaneous heating/cooling o There is no heat pump cycle implemented o The buffer tank setup does not contain the essential sensors to apply proper buffer tank layering 56 5 Discussion 5.1 ABCAT assessment For the consideration of the use of ABCAT for current and future FDD applications by TU/e and Worksphere, the experiences in the use and applicability of ABCAT are documented. Is ABCAT any good as simulation program? The calibration process of ABCAT was focused on a time period which is assumed to represent a correctly functioning building, also referred to as the baseline period. For a correctly functioning building the measurements and simulations must show a strong correlation. If not, there is not a valid basis to apply an FDD method on. Past projects show that ABCAT is able to produce accurate simulation results of various buildings and systems. In this case-study it was not possible to reproduce this level of simulation accuracy. The quality of simulation depends on the compatibility between model and system. Although ABCAT is expected to be a good simulation program it didn’t turn out to be appropriate for the case-study building because of compatibility issues and underperformance of the building. Is ABCAT simulating the case-study building behavior correctly? ABCAT is simulating (sub-) systems as they should be functioning while the actual systems are not functioning as intended. Mismatch between results of the calibrated model and actual measurements are caused by both faults in building HVAC system and general underperformance and shortcomings of ABCAT prototype and compatibility issues. In the starting phase of this graduation project it was assumed that ABCAT would be compatible with the case-study system. Although ABCAT was mostly used to simulate US buildings in the past, there is also a good example of successful simulation of the Vertigo building in the Netherlands (Peitsman et al., 2010). The differences between US and EU building styles should be kept in mind when using ABCAT and testing its compatibility with the case-study system. In the ABCAT documentation some requirements for the target HVAC system are given. There is no direct exclusion of the systems in the case-study building but also not a direct guarantee that it would work. In practice, calibration of ABCAT turned out to be unsuccessful, which is an indication for possible compatibility issues after all. ABCAT interface and ease of use The interface and the documentation coming with ABCAT are not complete and could be more detailed on various aspects of use. A lot of inputs and steps to take were unclear in the first sessions using ABCAT, leaving multiple obstacles which made it a time consuming process just to get to the completion of parameter input and realization of simulations. ABCAT is a tool which is in development and the version used in this study is still a prototype. This explains the encountered unclarities and obstacles. The current version is not made for retail use and there is no customer support. 57 Additional complications with the interface presented themselves in the form of incorrect connections between the interface and the physical model. For example the alteration of parameters involving weekend behavior resulted in changes of the simulations for working days and vice versa. Plenty more of these unexpected mismatches in the interface were found by trial and error in the use of ABCAT resulting in a lot of unnecessary hours of delay. ABCAT simulation quality Another thing is to validate and assess the correctness of ABCATs energy use predictions. ABCAT appears to be very reliant on input parameters. During the calibration process it is found out how sensitive the results are to small parameter alterations. Building parameters are often estimates, or based on design values which do not exactly match the real-life parameters. Consequently the calibration becomes a rather difficult and instable process. There is no good balance in the transition from the initial simulations to the calibration process, the physics-based model seems not able to produce good absolute results and results rely heavily on the calibration process and the manipulations of parameters to get to the desired result. Adaptation and expansion of the ABCAT models ABCAT has no options for summer and winter block in its physical model. It is not expected to be able to predict the very sharp threshold of shutting down the cooling for outside temperatures below 5°C, which is happening in the actual building. By manually adding a script for the appliance of this winter block you are actually adapting the existing ABCAT with models for average HVAC systems to fit the specific case-study system. Besides the winter block, which absence was very clear in the ABCAT simulation results, there are definitely more of these specific behaviors of the case-study system that are not implemented in the ABCAT models and thus influencing the calibration results. Incompatibility between ABCAT physical models and the actual behavior of the target HVAC systems is very likely one of the most important causes of the failure of accurate simulation. In the end, not only for this projects ABCAT flaws are noted and recommendations for improvements are done. (Zeiler & van der Velden, 2011) also states that ABCAT is a very promising tool, with good FDD results, but that it requires additional development from a prototype into a real product. Possibilities for generic calibration One of the questions we had on the applicability of ABCAT is about its flexibility: When applying ABCAT on a next building, is FDD possible without a second calibration procedure? Generic calibration is not possible, calibration is carried out by alteration of building inputs. For a second building most of the inputs are different so the previous calibration efforts cannot be transferred to this new building in any way. 58 5.2 Building energy performance assessment After the findings of poor building performance in chapter 2.6 it was chosen to look at the circumstances around this building performance issue from a higher perspective. It is investigated how certain steps in the design, building and exploitation process have taken place and where things went wrong. Also is the development and delivery process for the case-study building compared to the building sector in general. By means of interviews with technicians and engineers participating in the construction, delivery and exploitation of the case-study building (concerning the questions presented below) a better view on the procedures has been formed. Moreover plenty of conversations and discussions with colleagues at Worksphere (Maintenance engineers and energy consultants) have taken place which also supplied much insight in these affairs. Why was the building never commissioned? Commissioning is never included in the price, a detailed commissioning process is expensive. For the building owner/client this has no high priority, he will not be the one paying the energy bill. A building realized in a fast and cheap way has much more value to him. If all the climate systems work, somehow he is satisfied already. Why does a building services contractor like Worksphere accept a building like this? The purchase and acceptance of a newly rented office space are carried out by people from the organizational/financial side of the company. There is no knowhow on climate systems and their performance. The technicians and people with knowledge of climate systems are hardly ever involved in this process so performance issues are not taken into account. Why was assumed that the building and its systems were working correctly? Because there was no knowhow on building performance assessment and issues some wrong assumptions were made. The first logical thought everyone has when encountered with a brand new building is that it would function quite well. It would make sense that every newly delivered building has had some kind of commissioning and performance assessment. Apparently these things just don’t happen. A building and its systems are constructed and put to use and no one checks it. Only after comfort issues and months later someone will come and take a look at the random settings that were applied to an unchecked/tested system. In the world of other consumer products (not being buildings) the inspection of newly produced goods is a routine that is hardly ever skipped. If a new product is bought the quality and initial performance is usually pretty good and this performance only declines throughout the time of use. In general the consumers are very critical and demanding. For buildings this is apparently not the case. Some of the engineers at Worksphere were new and unknowing towards the issues around building delivery and underperformance of climate systems just like me. Others know this phenomenon already for years, it’s called the ‘throw it over the hedge’ mentality. 59 Building companies, and also the process of constructing and exploiting buildings mostly has two separate sides. The design and construction side, and the exploitation/maintenance side. When a building is finished the constructors pack and leave, throw the product over the hedge and don’t look back. There is hardly ever any information or instructions left behind for the people who start to use and maintain the building. The building management and maintenance people receive the newly constructed building without any information of its status and probably just assume it is functioning well. They don’t receive information about how the system is configured exactly and whether it has been properly tested on energy performance or not. Possibly it is assumed that the people who constructed the building and its climate systems actually did some testing/fine-tuning work but it just didn’t happen. Only after some time (months) of experience with the new, not-well-understood, building they start to notice and realize the flaws and malfunctions present. Moreover: Configuration of the BMS control settings are often executed without being thought over and by people with a lack of knowledge of the target building. When comfort issues present themselves set-points are altered to solve the short-term complaints. Hereby the energy use of the systems is not taken into account, which makes that improvements of comfort are often accompanied by a strong decrease of energy efficiency. Why are poorly performing climate systems built and delivered that way? This issue also has a very important financial side. Large building construction projects sometimes tend to exceed budget. In the later stages of construction (where buildings systems get installed) it appears that money is running out. This can lead to various presumed consequences: Installment cheaper systems than intended in the design Leaving out expensive sustainable sources and applications Time pressures for the constructors: the climate installation is placed then there is not time and money for inspection and fine-tuning of the systems A real commissioning/testing phase of the new building being discarded For project managers, an achievement oriented approach is dominating. Often projects should be finished as fast and as cheap as possible. Have the constructors place everything as fast as possible, save in construction time and budget to end up with a higher margin. In general the systems are doing what they should do without malfunction, but will not function very effectively. From a project management perspective these are good results and they will not directly be accounted for building energy performance. Even if some people with knowledge about the functioning of building systems do know about or suspect improper performance of the systems, there is in most cases not much incentive to do something about it. It will involve additional costs (even though the finetuning and improvements of systems save a lot more money) so it is difficult to convince the management of taking action and supply budget for improvement purposes. 60 (Elkhuizen & Rooijakkers, 2008) gives a good indication of the various causes of underperforming buildings. In general 85% of the problems is caused by improper use and management of the building. Only a small share (15%) of the problems is related to faults in the design and construction phase. In practice it is observed that in the BMS various components are not properly controlled and set-points are not tuned to the actual physical properties of the building and the actual use of the building. Project management will sometimes also decide for savings and budget cuts on sustainable energy applications and other HVAC equipment. Cheaper components are chosen to achieve higher margins or to be able to meet budget requirements. The cheaper solution will also work but the energy performance over the subsequent years is likely to be significantly lower. Sustainable applications are important for the process of energy savings in buildings and can also be beneficial in costs for building owners over a longer time period. Because of the high capital costs they are often cut from the initial building design. Who built/assembled the HVAC systems in the Worksphere office building and how was their cooperation? - The owner and client of the building is Van Schijndel Vastgoed The building is rented and occupied by Strukton Worksphere and Centric Van Schijndel Bouwgroep and Strukton Bouw realized the construction of the building Strukton Worksphere – Business Unit Projects was in the role of subcontractor responsible for the building installations There was poor communication and documentation about the work carried out, state of the systems at the time of transfer between parties and the performance of the delivered system. In case the fine-tuning and testing phase of the installment of equipment is skipped for some reason it Why was such a poorly functioning system delivered? Presumed causes for the bad state of equipment and configurations for newly constructed buildings entering the exploitation phase: Lots of contributing parties and no clear, overall responsible party Poor communication and transfer of work between constructing parties and maintaining parties No ‘hours’ available for proper testing and revision of the installed machinery Budget cuts in equipment purchase Often there is only time available (as in hours available to spend on the project according to the budget) for basic placement and connection of equipment. There is a general lack of time and money for testing, tuning and revision of the installed equipment. 61 We know that some previously undiscovered energy problems were present in the casestudy building. Are these things that occur in more buildings? In many cases, just like with the current test-case building, there is no good insight in the functioning of the climate systems is a building. There is a large demand for measurements and visualization of the energy flows present in the building systems. For the case-study building, a clear overview of the total amounts of HW and CHW produced and the division over the different subsystems in the building would be of great help: The amount of energy produced by heat pumps, the amount of energy coming from the heat exchangers, from excessive engine heat and thermal energy produced by a hot water boiler are important to know. Moreover it is interesting to know which fraction of the gas consumed is used for office heating and which fraction is used in the large, less insulated magazine, which is heated by a single hot water boiler. In the current situation these specific flows are not measured. More detail on these individual energy flows will also greatly improve the fault detection and localization process. There is a general lack of knowledge of the building energy performance state and a lot of improvements are possible in both the technical and the organizational aspects of monitoring and commissioning. From talks with various built environment oriented engineers and consultants it seems that the issues presented in this chapter are occurring very often in all kinds of building projects. It is hard to quantify the real extend of these inefficient and unwanted proceedings concerning the construction and delivery of buildings. 5.3 Expert view versus automation In almost every current commissioning and FDD application, experts are used for manual analysis and conclusions. Because of shortage and rising costs for the use of experts there is a large need for automation of this process, as stated by (Zeiler & van der Velden, 2011). The processing of large data amounts, as well as the modeling and simulation of complete buildings are still complex and expensive methods. New technology is required to achieve better results and reduce the costs. One of the advantages of manual assessment by experts compared to automation is that experienced maintenance engineers, energy consultants and other similar professionals can accurately assess the processes in a specific building case and can also foresee problems occurring in other aspects of building performance if actions are taken. If this is to be automated the software must be able to correctly combine different aspects of performance which is currently still a challenge. It is agreed upon that automation and simulation can aid the experts in various ways: ‘Powerful analytics that add “intelligence” to existing building infrastructure have the potential to transform the way in which companies manage energy across their real estate portfolio. Also building engineers can be empowered to take a more targeted, data-driven approach to their work while automation improves their productivity. Substantial cost savings can be achieved with relatively low capital investments”. (Kofmehl, Levine, Falco, & Schmidt, 2011). 62 There is a difference though between automation as an aid and full automation of an assessment process: When there is a model of the building in a simulation program, how do you turn it into automated assessment? This will fully rely on the absolute quality of the simulation program results because it has to be assumed that the results are correct. Consequently, compatibility issues between the modeled behavior and the real behavior of the building will result in an offset that indicates a dysfunctional building. But is this really the case? Engineers will always stay involved for the validation of proper simulation and FDD functionality of software. Also in the comparison between manual and automatic assessment the costs – benefit relation in both time and money should be taken into account. For some situations the expert assessment can have its advantages: If the unassessed building is available to an expert and the expert is free to make changes and analyze the results this is a big advantage for the expert approach. The expert can develop improvements, implement them right away and monitor and report the performance results and side effects of his actions in the subsequent days. A model based approach would have to be developed from scratch, which is costly and time consuming and is not in favor for this situation. In case a model of the target building is already available in any practicable simulation software this will save a lot of time and costs, which is a benefit for the automated approach. Making a detailed model of a building from scratch is a time-consuming and expensive task. Before actually starting this task consider the time required and the use of it. Spending all this time is usually worth it if the model can be used multiple times for diverse purposes. A one-time assessment is not always of enough importance to undergo such a task. Other options might be more efficient. Development of a standardization process for building modeling and FDD could be of great help to overcome these problems. Standardization would mean the use of standard models, parameters, algorithms and control strategies which greatly decrease the work that different parties currently carry out individually. 63 64 6 Conclusions and recommendations During this study the applicability of the ABCAT FDD methods on the case-study building as well as the energy performance of the building itself has been assessed. Building performance In the building energy analysis process (chapter 2.6) it is pointed out that the building HVAC system is not functioning as it should. Various improvements were carried out but still it is not said that the building functions according design specifications. Still there are faults in the machinery setup and operation, such as buffer tank short-circuit, HW-CHW circuit leakage and improper heating and cooling set-points. We showed that the measured building energy use is much higher than designed. In order to set-up a proper FDD system, a properly functioning building is a necessity which must be confirmed by a commissioning process, which did not happen for the case-study building. Presumed reasons for building underperformance (as discussed in 5.2): Budget cuts and time constraints during building and delivery Poor communication and transfer of work between involved actors Disinterest of building performance by owners and management Overall absence of incentive and knowhow of energy performance assessment Simulation and fault detection results For the application of monitoring and FDD on the case-study building additional information, not provided by the building management system, had to be collected. A capable measurement setup is applied (chapter 2) and measurement data of sufficient quality for building level FDD is acquired. On the other hand ABCAT calibration and simulation results (3.5) are unsatisfactory. There is no good correlation found between the measured and the simulated energy consumptions during the calibration period which is essential for further FDD procedures. This does not mean that these imperfect simulations are useless. The fact that large deviations between simulation and measurements are found are a clear indication that something is wrong with the building operation and additional research is required. It emphasized the need for actions on the climate system performance. Causes of calibration mismatch ABCAT is simulating (sub-) systems as they should be functioning while actual systems are not functioning as intended. Mismatch between results of the calibrated model and actual measurements are caused by: Faults in building HVAC system and general underperformance (4.4) Shortcomings of ABCAT prototype and compatibility issues (5.1) Shortcomings and possible improvements of ABCAT: ABCAT is a promising tool with good FDD results, but it requires additional development from a prototype into a real product before it should be used any 65 further. More functionality is required for use on more diverse (European) buildings so HVAC systems and compatibility can be guaranteed. ABCAT requires the look of an experienced building engineer. The engineer has to inspect output figures manually with the expert rule table in hand for detection and isolation of problems. More automation and intelligence in the software are expected to be a great improvement. The inclusion of component level measurements of some significant variables to the expert rules method, as well as cumulative figures of subsystem behavior, could be of great value for isolation purposes. Although the goal of simulation-based FDD has not been reached, several faults in HVAC behavior are detected after all. Careful monitoring and detailed analysis of measurement data by experts, without a simulation reference, has strong FDD possibilities as well. Recommendations The importance of the conditions and the testing environment necessary for successful application of FDD were not fully known or greatly underestimated at the start of the project. During the project much knowledge and understanding of the importance of these conditions has been obtained. As a recommendation the most essential conditions for FDD should be emphasized. Conditions for a good basis to build model-based FDD on when choosing a case-study building are listed. The building must: Have been properly commissioned, it would help when the proper documentation of the commissioning would be available. Have sufficient post-commissioning measured consumption data by which to define the baseline situation and sufficient continuous measured data from that point on. Have had a known significant degradation in energy performance that was detected by for instance sensible changes in indoor thermal conditions. These known and documented faults are important for testing purposes. Have additional information as to the nature or cause of the degradation in energy performance available to compare with the diagnostic results of FDD applications. Not have experienced any major changes or retrofits during the period analyzed. Changes to the setup have a strong negative influence on the continuity of the process and the ability to model it. Recommendations for further FDD use and development for Worksphere FDD is an important development with a lot of potential for energy savings and comfort improvements. Before it is fully applicable, additional development and testing is required. The lessons we learned during this project are an important step for Worksphere in that direction. For further FDD developments it is suggested to aim at: Application of further FDD research on different buildings that are proven to be functioning as they should. The knowhow of setting up a suitable monitoring environment is earned during this graduation project. Application of ABCAT also deserves extra attention. It has to be tested if ABCAT yields better results on a different, better performing building (4.4). Testing the applicability of other simulation software. Many other simulation software is available, plenty of them are even for free. The main goal is to acquire satisfactory simulation results. If valid simulation results are obtained with a 66 different program, it is even possible to apply the ABCAT detection and diagnosis figures discussed in this report (3.3), which can be reproduced outside of ABCAT. Application of simulation-free FDD, only making use of careful monitoring of the climate system behavior. In this report certain component level experiments and observations are mentioned (4.4.3) that aided in the detection of faults. These measurements can be applied and monitored in other buildings as well. For instance: buffer tank layering checkup, test-case for HW – CHW leakage by looking at rates of temperature change and simultaneous heating and cooling checks. Much more of these specific component level measurements and experiments are possible. It is expected to be very beneficial to extend the number of these experiments and further optimize them and put them into use for building performance assessment. Expansion of regular building control into more ‘intelligent’ building control. Predicting functionalities could be beneficial for the efficiency of building climate control. Use the expertise obtained with Matlab data management (2.5) and acquisition of KNMI weather conditions (2.3) to include some prediction functionality to a Building Management System: o Low temperatures predicted? Increase pre-heating in the morning o High solar irradiation predicted or measured? start cooling and ventilation before the indoor temperature gets uncomfortably high o Sunny afternoon predicted? Stop heating the building in the morning o Tomorrow a scheduled meeting? Adjust settings for pre heating/cooling of the meeting room 67 References Andersson, T. & (2010). TA Link Manual. Current. Tour & Andersson. Carrier. (2009). Technical Specifications Systems Strukton Worksphere Eindhoven. Carrier. Claridge, D. E., Bensouda, N., Lee, S. U., & Wei, G. (2003). Manual of procedures for calibrating simulations of building systems. Claridge, D. E., & Lin, G. (2009). Retrospective Testing of an Automated Building Commissioning Analysis Tool ( ABCAT ). Curtin, J. M. (2007). Development and testing of an Automated Building Commissioing Analysis Tool (ABCAT). Simulation. Texas A&M University. Daelmans, A. (2012). Possibilities energy neutral building: Worksphere Son. EPA. (2012). Heating, Ventilation and Air-Conditioning (HVAC) Systems. United States Environmental Protection Agency. Retrieved from http://www.epa.gov/iaq/schooldesign/hvac.html Elkhuizen, B., & Rooijakkers, E. (2008). Visie op ontwikkelingen gebouwbeheersystemen. VV+, mei, 336-338. GasEngineering. (2011). R 410 A specifications. Gebhardt, N. (2011). Centrifugal fans belt driven. Hissel, T. (2011). Pictures HVAC systems SWS Son. Hissel, T. (2012). Monitoring and FDD in building HVAC systems. Research internship Sustainable Energy Technology Master program. Eindhoven. Howell, Coad, & Sauer. (2009). Principles of Heating Ventilation and Air Conditioning (pp. 393-480). Knebel, D. E. (1983). Simplified Energy Analysis Using the Modified Bin Method. Atlanta: ASHRAE. Kofmehl, A., Levine, A., Falco, G., & Schmidt, K. (2011). Energy-Smart Buildings. Demontrating how information technology can cut energy use and costs of real estate portfolios. Meulman, G. (2011). Nieuwbout Kantoor Ekkersrijt, Regeltechnische Omschrijving. Eindhoven. 68 Morisot, O., & Marchio, D. (1999). Fault Detection and Diagnosis on HVAC Variable Air Volume system using Artificial Neural Networks. Proceedings IBPSA Building Simulation. Peitsman, H., Aker, K. van den, Claridge, D. E., & Bynum, J. (2010). Energyanalyse en foutdiagnose Vertigo-gebouw TU/e. TVVL Magazine, 46-51. Priva. (2011). Priva Top Control. Retrieved from http://www.priva.ca/en/products/priva-top-control Siemens. (2011). Symaro – innovative sensors , measurable quality innovative measurement that pays off over the long term. Trane. (2012). Pictures Induction Units. Retrieved from http://myfactoryrep.com/InfoSheets/ROSEMEX/Induction_Units.aspx Utiliteitsgebouwen, E. (2009). EPU Strukton Worksphere Son. Van Kleef, S. (2011). Plan van Aanpak, Gebouw en Installatie Monitoring. Eindhoven. Wei, G., Liu, M., & Claridge, D. E. (1998). Signatures of Heating and Cooling Energy Consumption for Typical AHUs. Proceedings of the Eleventh Symposium on Improving Building Systems in Hot and Humid Climates. WiSensys. (2012). WiSensys ® Wireless Sensor WS-DLTc. Zeiler, W., & van der Velden, J. A. J. (2011, October). Van Commissioning naar Presatieborging. TVVL Magazine, 18-22. 69 Appendices A ABCAT FDD HVAC PPP BMS HW CHW KNMI AHU NaN PHP Abbreviations Automated Building Commissioning and Analysis Tool Fault Detection and Diagnosis Heating Ventilation and Air Conditioning Public Private Partnership Building Management System Hot Water Chilled Water ‘Koninklijk Nederlands Meteorologisch Instituut’ (Royal Dutch Meteorological Institute) Air Handling Unit Not a Number PHP Hypertext Preprocessor 70 B Detailed description of the ABCAT interface ABCAT is an excel based program. The whole interface (input specification and output numbers and figures) is inserted in the various sheets of the excel file. The physical model behind de simulations is present in several visual basic macros and will not be visible or of any use for the common ABCAT user. The interface of ABCAT consists of the following sections: Interface (time domains for calibration and testing, statistics, tool for diagnosis) Data (inputs and outputs including: Temp, RH, CHW&HW energy use, Electricity use) Input (building parameters) Simulation results Interface sheet (tools for calibration and time period selection) On the interface tab the desired x-axis range of the ABCAT output figures can be altered. After standard simulation the start date of the given data-set will be visible together with the number of weeks available. By changing this start date and amount of weeks you can set the figures to a period that is used as calibration period. This way outputs of other time ranges are not interfering with the visual calibration process. ABCAT produces certain output figures (in tab “Periods1&2”) especially for the comparison of results of two separate time periods (the training period (correct behavior) and test period). The time range for these two periods can also be specified in the interface tab. Moreover ABCAT gives some basic statistic values corresponding to the time ranges mentioned above. These statistics guide in the calibration and Diagnosis processes for ABCAT. Figure 43: Data range input and basic statistic output in the ABCAT interface One of ABCAT’s detection methods is the cumulative error figure. This output and detection method makes use of a “day in excess” threshold for fault detection. In the ‘Consecutive Day Chart Excess Consumption Levels’ box this threshold can be defined and altered. The interface tab presents one utility for the diagnosis (localization) process. In the ‘Diagnostic Summary Count’ table all the days of the simulation period are placed in one of three categories. The different categories are separated into three daily average outside air temperature ranges, Toa < Ts,sp (supply air temperature); Ts,sp < Toa < Te,sp (temperature the economizer deactivates); Toa >Te,sp. For each outside air temperature range, the total days where the simulated consumption (CHW and HW) is greater than, less then or within de deviation range of the measured consumptions are counted and fill the table appropriately. This table is the major source of information for the Expert rules diagnostic method, which will be presented later on in the report. 71 Figure 44: Diagnostic Summary Count table, to use with the Expert rule method Besides the previously mentioned contents the interface sheet covers the following, less interesting inputs: ‘costs per energy unit’, ‘consumption totals’, ‘alarm parameters’ and ‘input and output data file locations’. Data Sheet (time-dependent in- and outputs) This sheet has two versions, the ‘Imperial unit’ version and ‘SI’ version. All processes in ABCAT make use of imperial units but the SI variant is there as extra information for SI users. The data in the SI tab is linked to the imperial tab and shows the SI representation of all the data in there. This does not work the other way around. This sheet stores the ambient and indoor conditions, measured and simulated consumption values as well as the results of several calculations performed with this data. For the indoor temperature, indoor dewpoint-temperature and electricity use this tabs makes a distinction in Pre-occupied period average value, Occupied period average value and Post-occupied period average value. This distinction is important for the building occupation related factors in the simulation process. Columns A thru O on this sheet are configured as follows: A-C: Date, Average Daily Outside Air Temperature, Average Daily Dew Point Temperature (or Relative Humidity) D-F: Measured Daily Chilled Water, Hot Water and Whole Building Electricity consumption G-H: Average Preoccupied Period Outside Air Temperature and Whole building Electricity consumption I-K: Average Occupied Period Outside Air Temperature, Dewpoint-Temperature (or Relative Humidity) and Whole Building Electricity consumption L-M: Average Postoccupied Period Outside Air Temperature and Whole Building Electricity consumption N-O: Daily Simulated Chilled Water and Hot Water Consumption The remaining columns are the results of calculations performed upon the data in columns A-O with user specific inputs on the “Interface” sheet. Inputs Sheet (building and environment parameters) This sheet also contains inputs for the model, but this section only contains building and system parameter based inputs, not inputs that are measurement based. The input sheet consists of: In this sheet the physical model system type to use is defined. It is possible to use a set of up to 5 system types and different sets of parameters simultaneously. 72 However for most building simulations one system type will be sufficient. By enabling and disabling individual sets different configurations can be compared Four system types are currently available to simulate. These include 1. SDRH (Single Duct Reheat), 2. SZHC (Single Zone Heating and Cooling), 3. DDVAV or DDCV (Dual Duct Variable Air Volume or Constant Volume) and 4. DFDD (Dual Fan Dual Duct) Tdewpoint_or_RH – specifies whether the ambient humidity measurement in the Occupied Period column of the “Data” sheet on the workbook is Dewpoint temperature (F) – (0) or Relative Humidity (%) – (1) Total Floor Area and % Interior Zone Area(not subjected to envelope gains/losses) SWR – Surface Weight Ratio (lb/ft2), typical values depending on construction type will usually range from (30 – light) to (130 – heavy) Heating and Cooling Set Point and Night Setback Temperatures Humidification – Does an electronic humidifier exist? No – (0); Yes –(1). If so, the electric load required for the latent heat of vaporization of the water is calculated and subtracted from the whole building electric load to avoid considering this load as heat gain. This is complemented with the Minimum RH set point Occupancy Schedules (0-24) hours for Weekdays, Saturdays, Sundays and Vacation periods HVAC Schedules (0-24) hours for Weekdays, Saturdays, Sundays and Vacation periods Volumetric Flow Rates: Total, Outside Air, Minimum, Leakage, Infiltration. Both for Exterior Zone and Interior Zone. OA_Control – controls whether the outside air is controlled at a constant percentage (0) of total flow or a constant flow level (1) Hot Deck Specifications (DFDD only). HD_onoffTemp, Z_iRH (Interior zone reheat (1/0)), dT_Rduct (Return air temperature rise) Economizer: yes/no (1/0), Temperature Set Points, Outside Air Fractions. Envelope Areas and U-Values: A_Wall, A_Win, U_Wall, U_Win Supply Fan Efficiency and Total Pressure Solar Transmission (Btu/hr): A linear interpolation of the given temperatures and corresponding transmissions is used for simulations throughout the year, the input values are: T_Jan, q_Jan, T_July, q_July Area per Occupant for exterior and interior zones (ft2/person) Lighting and Equipment Heat Gain: Either measured electric loads are used as input (0) or electric loads are specified (1) from the inputs below. Electric gains for Exterior and Interior zone for Occupied and Unoccupied periods. Complemented with the fraction of the measured electric load that contributes to heat gain in the system simulated (0-1) and the fraction of electric heat gain contributing to respectively the interior and exterior zone (should sum to 1) AHU Coil Set Point Temperature Schedules: T_Cset1, T_C1, T_Cset2, T_C2 with linear interpolation between the given set-points for cooling. Similar for heating set-points and carried out both for inside air and outside air coils. Coils are disabled in simulation by entering set-point temperatures to well above or below realistic values. Electric Load and HVAC operation Fractions based on day of the week including: Start and end of the measured electric load for the occupied period (see Data tab) and factors to adjust the occupied period electric gains and total heating/cooling by HVAC system (usually always 1) Vacation period definition 73 C ABCAT output figures Figure 45: Set 1 of ABCAT output figures used for basic detection (Peitsman et al., 2010) a) CHW measurement and simulation as function of time b) HW measurement and simulation as function of time c) CHW measurement and simulation as function of ambient temperature d) HW measurement and simulation as function of ambient temperature Figure 46: Set 2 of ABCAT output figures used for additional detection and identification (Peitsman et al., 2010) a) Cumulative (measurement-simulation) energy use difference for both CHW and HW simulation b) Electricity consumption used for quick check of anomalies in weekly pattern c) CHW measurement-simulation difference [%] with 7day moving average d) HW measurement-simulation difference [%] with 7day moving average 74 Figure 47: Set 3 of ABCAT output figures used for detection, identification and isolation (Peitsman et al., 2010) a) Consecutive days of excess consumption levels b) CHW CUSUM alarm, based on cumulative values and RMSE c) Bar plot of the diagnostic day count table d) HW CUSUM alarm, based on cumulative values and RMSE Figure 48: Set 4 of ABCAT output figures used detection, identification and isolation (Peitsman et al., 2010) a) CHW energy consumption related to ambient temperature, comparison of period 1 and 2 b) Residuals of the combined CHW and HW energy consumption related to ambient temperature c) HW energy consumption related to ambient temperature, comparison of period 1 and 2 d) Electricity consumption related to the ambient temperature, comparison of period 1 and 2 75 D Sources for ABCAT inputs and building parameters Table 6: Time-dependent inputs for ABCAT and their source Input Ambient temperature Ambient relative humidity Indoor temperature Indoor relative humidity Electricity consumption Source KNMI (IMPORTALL.M toevoegen) KNMI Priva BMS Priva BMS Utility provider Table 7: Static building parameters for ABCAT and their source Parameter Total floor area Source EPU (Utiliteitsgebou wen, 2009) Estimate Value SI 2 6755 m Value Imperial units 2 72710 ft Comments 0.9 0.9 Almost every room borders the envelope 400kg/m Set-point heating Set-point cooling Night setback EPU (Utiliteitsgebou wen, 2009) BMS BMS BMS 21 °C 17 °C - 69.8 °F 62.6 °F - Humidification BMS - - Occupancy Weekend occ. HVAC schedules Weekend Volumetric flow BMS BMS BMS BMS Carrier documentation (Carrier, 2009) Carrier documentation (Carrier, 2009) Carrier documentation (Carrier, 2009) BMS EPU EPU VABI VABI Carrier doc. Carrier doc. KNMI KNMI 7-19h 0 7-21h 7-21h 3 10 m /s 7-19h 0 7-21h 7-21h 21189 CFM 1 1 Only outside air is used - - Constant % control 0 2 2306 m 2 1981 m 6.6 1.1 0.82 - Not available VABI Estimate 0.6 0 2 24813 ft 2 21316 ft 2 0.06 Btu/hrft °F 2 0.01 Btu/hrft °F 0.82 5 / 0 / 3.9 w.g. 30 °F / 90 °F 135713 Btu/h 348815 Btu/h 2 60,48 ft /person 0.6 Priva - - Not 100% available, some are estimated Interior zone area Thermal mass Outside air ratio AHU control Economizer A_wall A_win U_wall U_win Fan Efficiency AHU pressure Temp Jan & July Solar transmission Area per Occupant heat generation fraction electric gain Coil Set Points 2 81.93lb/ft 76 2 No signs of night setback temperatures No specific humidification system present No weekend and holiday occupancy HVAC also operational in weekends Air flow measured at AHU completion Combination of irradiation/area and building total outside area (jan-jul) E Characteristic Signatures SWS Son Case-Study Building T_h + 2degC 10 T_h + 2degC 10 R² = 0.146 HW (%) CHW (%) R² = #N/A 0 0 20 40 60 80 100 20 40 60 Outside Temperature (°F) Outside Temperature (°F) T_c + 1degC 10 T_c + 1degC 10 R² = 0.6328 HW (%) CHW (%) R² = 0.4601 0 0 20 40 60 80 100 -10 20 40 60 80 100 -10 Outside Temperature (°F) Outside Temperature (°F) T_hsb + 1degC 10 T_hsb + 1degC 10 R² = 0.5851 HW (%) R² = 0.7097 0 0 20 40 60 80 100 -10 20 40 60 80 100 -10 Outside Temperature (°F) Outside Temperature (°F) T_csb + 1degC T_csb + 1degC 10 10 R² = #N/A HW (%) R² = 0.7165 CHW (%) 100 -10 -10 CHW (%) 80 0 0 20 40 60 80 100 -10 20 40 60 80 100 -10 Outside Temperature (°F) Outside Temperature (°F) 77 V_eDsqft - 0,01cfmft 10 V_eDsqft - 0,01cfmft 10 R² = 0.6962 HW (%) CHW (%) R² = 0.0137 0 0 20 -10 40 60 80 100 Outside Temperature (°F) 40 60 V_iDsqft - 0,01cfmft 10 HW (%) R² = 0.6962 0 0 20 40 60 80 100 -10 20 40 60 80 100 -10 Outside Temperature (°F) Outside Temperature (°F) V_OAeDsqft - 0.01cfmft 10 V_OAeDsqft - 0.01cfmft 10 R² = 0.7097 HW (%) R² = 0.8039 CHW (%) 100 Outside Temperature (°F) R² = 0.0086 0 0 20 40 60 80 100 20 40 60 80 100 -10 -10 Outside Temperature (°F) Outside Temperature (°F) V_OAiDsqft - 0,01cfmft 10 V_OAiDsqft - 0,01cfmft 10 R² = 0.7097 HW (%) R² = 0.8039 CHW (%) 80 -10 V_iDsqft - 0,01cfmft 10 CHW (%) 20 0 0 20 40 60 80 100 20 40 60 80 -10 -10 Outside Temperature (°F) Outside Temperature (°F) 78 100 A_Wall + 10000ft2 10 A_Wall + 10000ft2 10 R² = 0.724 HW (%) CHW (%) R² = 0.7106 0 0 20 40 60 80 100 20 40 60 80 -10 -10 Outside Temperature (°F) Outside Temperature (°F) A_Window + 10000ft2 10 A_Window + 10000ft2 10 R² = 0.724 HW (%) CHW (%) R² = 0.7106 0 0 20 40 60 80 100 -10 20 -10 U_wall decrease of 50% R² = 0.7106 0 60 80 100 U_wall decrease of 50% 10 HW (%) CHW (%) 10 40 Outside Temperature (°F) Outside Temperature (°F) R² = 0.724 0 20 40 60 80 100 -10 20 40 60 80 100 -10 Outside Temperature (°F) Outside Temperature (°F) U_window decrease of 50% 10 U_window decrease of 50% 10 R² = 0.7106 HW (%) CHW (%) 100 0 R² = 0.724 0 20 40 60 80 100 -10 -10 Outside Temperature (°F) 79 20 40 60 80 Outside Temperature (°F) 100 T_Cset1 + 5 degC 10 T_Cset1 + 5 degC 10 R² = #N/A HW (%) CHW (%) R² = 0.0729 0 0 20 40 60 80 100 -10 40 60 T_C1 + 5 degC T_C1 + 5 degC 10 HW (%) CHW (%) R² = 0.0779 0 0 20 40 -10 60 80 100 Outside Temperature (F) 20 40 -10 60 100 T_Cset2 + 2 degC 10 R² = 0.0729 HW (%) R² = 0.0195 CHW (%) 80 Outside Temperature (F) T_Cset2 + 2 degC 10 0 0 20 40 60 80 100 -10 20 40 60 80 100 -10 Outside Temperature (F) Outside Temperature (F) T_C2 + 2 degC 10 T_C2 + 2 degC 10 R² = 0.1962 HW (%) R² = 0.0178 CHW (%) 100 Outside Temperature (F) R² = 0.0816 0 -10 80 -10 Outside Temperature (F) 10 20 0 20 40 60 80 100 Outside Temperature (F) -10 80 20 40 60 80 100 Outside Temperature (F) T_Hset1 + 5degC 10 T_Hset1 + 5degC 10 R² = 0.0729 HW (%) CHW (%) R² = 0.0729 0 0 20 40 60 80 100 -10 20 -10 T_H1 + 5 degC 60 HW (%) CHW (%) R² = 0.0785 0 0 20 40 60 80 100 -10 20 40 -10 T_Hset2 + 2 degC 10 HW (%) 0 80 100 T_Hset2 + 2 degC 10 R² = 0.0729 60 Outside Temperature (F) Outside Temperature (F) CHW (%) 100 T_H1 + 5 degC 10 R² = 0.0785 R² = 0.0729 0 20 -10 40 60 80 100 Outside Temperature (F) 20 40 60 80 100 -10 Outside Temperature (F) T_H2 + 2 degC 10 T_H2 + 2 degC 10 R² = 0.1964 HW (%) R² = 0.1964 CHW (%) 80 Outside Temperature (F) Outside Temperature (F) 10 40 0 0 20 40 60 80 -10 100 20 40 60 80 100 -10 Outside Temperature (F) Outside Temperature (F) 81 Floor area + 10% Floor area + 10% 10 R² = 0.8037 HW (%) CHW (%) 10 0 R² = 0.0736 0 20 40 60 80 100 -10 20 40 60 80 -10 Outside Temperature (F) Outside Temperature (F) Themal Mass -30% 10 Themal Mass -30% 10 R² = #N/A HW (%) CHW (%) R² = 0.4562 0 0 20 40 60 80 100 -10 20 40 60 80 100 -10 Outside Temperature (F) Outside Temperature (F) Area per Occupant +10% 10 Area per Occupant +10% 10 R² = 0.7441 R² = 0.6785 HW (%) CHW (%) 100 0 0 20 40 60 80 -10 100 20 40 60 80 -10 Outside Temperature (F) Outside Temperature (F) 82 100 F Details of subsystem measurements Plant/Heat pumps/Gas Use The building HVAC system’s main heating and cooling plant consists of four gas-powered heat pumps from ‘GasEngineering/ Aisin Toyota Group ’(GasEngineering, 2011). A running heat pump which is either in heating or cooling operation mode bus is also able to deliver free engine heat on the side. The HVAC setup also includes a hot water boiler used for peak demand heating situations but it’s activity and share in the overall heating energy use is assumed to be very low and therefore neglected. Gas use measurement data is obtained from the existing energy monitoring activities using the Maestro energy management system. Parameters of the heating and cooling system are found in the ‘building delivery report’ (Utiliteitsgebouwen, 2009) and heat pump specifications (GasEngineering, 2011): Gas engine driven heat pumps Source: outside air Heating temperature range: 35C < Tsupply < 45 Backup energy from gas boiler for peak demand Generation efficiency: Nopw,verw 1.5 System efficiency: Nsys;verw 0.84 Generation efficiency cooling: Nopw;koel 1.95 System efficiency cooling: Nsys;koel 0.84, 0.85, 0.86 From the measured gas consumption quantities, Dutch natural gas energy content, the average rated COP of the heat-pump array and documented system efficiency the total combined HW & CHW energy use can be obtained: Gas Use (m3) * Energy content (MJ/m3) * Heat Pump COP * System efficiency = Combined HW & CHW Energy Use Air Handling Units One idea was to determine the AHU heating /cooling production with the activity and heat flows in the heating and cooling coils. As can be seen in the figure some temperatures in these coils are available. Also the supply pump activity and valve control can be derived. The one parameter that cannot be found or derived is the mass flow through these coils. Because this is an essential factor to determine the heat flows, this approach is discarded. 83 Figure 49: Schematic overview of AHU office. On the building side temperature and humidity measurements are available in the BMS. On the ambient side of the AHU (outside air and exhaust air) additional T and RV measurements using the WiSensys system are applied. This was similar for the AHU restaurant. Air mass flow in the AHUs. This is a category of measurements we initially missed. After looking further into the acquisition of this information the operation rpm of the fans were found in the building management system. They were not present in the History database but were from that moment on included. From rpm to air flow was not very easy to convert. A fan performance curve (Gebhardt, 2011) was obtained but the relations to get to the air flow were too complex and essential information was missing to successfully derive the occurring air flow from fan rpm. 84 Figure 50: Fan performance curves for the fans present in the building AHUs. No direct relation between fan rpm and resulting air flow was found (Gebhardt, 2011) There is no simple, direct relation between fan rpm and the resulting air flow, as can be seen in figure 42. The air flow produced strongly depends on both the static and dynamic pressure the air flow encounters on its trajectory. In order to calculate the energy for heating and cooling in the AHU’s the mass flow through the unit ducts is essential. This mass flow is not logged or measured but can be calculated using the fan operation and parameters. Fan operation is logged in form of percentage of the maximum rotational speed. Using this rotational speed in combination with the resistance pressure the fan has to cope with and other parameters like efficiency and power it should be possible to derive the volume and mass flow of air through the system. The office AHU has two Dahlander belt driven fans as can be found in the Carrier AHU technical specification (Carrier, 2009) that was obtained from the AHU manufacturer. This means that the fan has two different rotational speeds it can operate at. Investigation pointed out that the AHU fans operate at their high rpm level at weekdays 07:00 – 18:00 and at low rpm for the resulting time. The fans are never shut down. Although the restaurant AHU’s are fit out with variable speed fans they appear not to be controlled and used this way. The fan rpm is either fixed on a steady operating rpm of about 75% or shut down. This also happens according a strict time schedule which makes an easy relation for time – rpm possible without any necessary measurements. 85 Other important information derived from the technical specifications is that the occurring air flow at these operation speeds is measured during the building and testing phase of the delivery of the AHU’s. Because these speeds are measured for the exact building and setup they can be assumed to be valid for continuous building HVAC operation. Table 8: Parameters AHU (Carrier AHU restaurant SWS Eindhoven (Carrier, 2009)) Fan 1 Total pressure Power Engine Speed Volume flow Air velocity Fan rpm Efficiency Axis power RZR 11-0280 972 [pa] 3 [kW] 3000 [rpm] (@ 100%) 2250 [rpm] (@ 75%) 1.56 [m3/s] 12.13 [m/s] 3202 [rpm] 70 [%] 2.15 [kW] Fan 2 Total pressure Power Engine speed Volume flow Air velocity Fan rpm Efficiency Axis power RZR 11-0355 918 [pa] 2.2 [kW] 1500 [rpm] (@ 100%) 1125 [rpm] (@ 75%) 1.44 [m3/s] 7.13 [m/s] 2098 [rpm] 77 [%] 1.72 [kW] The calculation of the exact air volume flow brings some problems because of the complex relation (Gebhardt, 2011) between pressure, power, efficiency, fan rotation and volume flow. Therefore, for now, the following approximation of the air volume flow will be applied: Table 9: AHU fan speed to air flow conversions for AHU restaurant variable frequency drive fans Frequency control 100% 75% 0% Fan 1 speed 3202 rpm 2402 rpm 0 rpm Fan 2 speed 2098 rpm 1574 rpm 0 rpm Fan 1 airflow 1.56 m3/s 1.44 m3/s 0 m3/s Fan 2 airflow 1.17 m3/s 1.08 m3/s 0 m3/s The fans in this AHU operate from 6.00 to 21.00 for every day of the workweek. During nighttime and weekends the fans are shut down. For now I assume AHU airflows of 1.44m3/s (fan1) and 1.08m3/s (fan2) for all operation hours. AHU office The same assumptions are used for now, a linear relation of the fan rpm and airflow. The air flow at 750rpm is an assumption, not a given value. Table 10: AHU fan speed to air flow conversions for AHU office Dahlander control fans Dahlander control 2 1 0 Fan 1 speed 1500 rpm 750 rpm 0 rpm Fan 2 speed 1500rpm 750 rpm 0 rpm 86 Fan 1 airflow 9.72 m3/s 4.86 m3/s 0 m3/s Fan 2 airflow 10.28 m3/s 5.14 m3/s 0 m3/s Induction Circuit The pressure difference over the main control valve of the induction circuit pipeline is measured. Using conversion factors given in the valve documentation the pressure difference is converted to the flow in the induction circuit in (m3/s). Figure 51: TA Link pressure sensor & Example of flow measurement application on a control valve (Andersson, 2010) The flow combined with the supply and return temperatures, which are measured in the Priva BMS, yields the energy flows present in the induction circuits, as is displayed in the overview in figure 45. This procedure is carried out for both the HW and CHW circuit. M Figure 52: Overview of the entire HW system for explanation of the induction subsystem measurements. Induction subsystem = blue box, Temperature measurements = T, Flow measurement = M (Priva, 2011) 87 G Measurement accuracy, reliability and validation Most measurement data is obtained from common BMS system. The Priva BMS is supplied with Siemens, industry grade sensors for temperature and relative humidity. The accuracy of these sensors is not thoroughly tested. Sources in literature indicate that common industry BMS sensors are accepted as sufficiently accurate for building FDD purposes. PRIVA sensor accuracy The sensors used are typical industrial sensors, and their accuracies are respectively ±0.5% for temperature sensors and ±5% for relative humidity. These values should be taken into account when determining the threshold for fault detection. The WiSensys sensors were calibrated for a previous project. The actual measurement results were checked for accuracy when setting up the sensors for this project’s measurements. (Siemens, 2011) gives little information about the accuracy of their sensors. However they do mention that the actual sensors are standard pt100 and pt1000 components which means they have the same properties as the sensors used by WiSensys. The accuracy is given in the WiSensys sensor specifications (WiSensys, 2012): Measurement range 10% to 95%RH non-condensing -20°C to +80°C Measurement accuracy ±3%RH from 20% to 80%RH; ±5%RH otherwise ±0.4°C @ 25°C, ±1°C from 0°C to +50°C, ±2°C from -20°C to +80°C Measurement resolution 0.1%RH, 0.1°C Long term stability Drift < 0.5%RH per year TA-Link Δp measurement However for the parameters measured through the additional, self installed, measurement Table 11: conversions and accuracies for the different links in the flow measurement setup Step in conversion Δp Measurement Control valve Logger Description TA Link TA DN100 valve WiSensys voltage logger Conversion 0-100kPa 0-10V Q = √Δp *100*KV 0-10 V Accuracy <± 1 kPa ± 0.1 V These inaccuracies together can be processed into a flow dependent total inaccuracy, which is displayed in the following figure: 88 50.0% 150 1 10 19 28 37 46 55 64 73 82 91 100 0.0% KV factor 200 Error 100.0% -50.0% KV DN100 KV DN80 100 50 0 -100.0% 0.5 1.5 2.5 3.5 4.5 5.5 6.5 7.5 dp measuremnt [kPa] Vale opening setting Figure 53: Figures used for the validation of TA-Link accuracy It can be seen that for low Δp the error becomes gigantic. The higher the Δp, the lower the error gets. In order to achieve accurate water flow measurements, the control valve should be closed so far that the pressure difference over it becomes in the range of 30-100 kPa. Beware; closing the valve too far will influence the building water flow and heating/cooling behavior. Closing the valve ever further will breach the pressure sensor’s maximum range of 100 kPa, resulting in malfunction. Δp over the control valve is measured and using the manufacturer’s formula and ‘valve setting value’, KV, the volume flow can be calculated. The following figure shows KV in relation to the valve opening setting. (8 is fully open, 0 is closed). As can be seen the KV behavior is fairly linear and clear for valve opening settings of 3 and higher. Lower valve settings result in low values for KV and thus a less clear relation and higher accuracy. It is not wise to use a low valve opening. In order to achieve an acceptable Δp the valve opening setting has been decreased to realize measurements in an accurate and reliable range. Measurement reliability Priva server downtime Sometimes for unknown reason the Priva History server was not running. The History service was transferred to a more protected environment to make sure this unwanted downtime was over. Connection problems When the distance between a base-station and sensor package is too high the signal strength becomes low and connection problems can occur. Sometimes a working connection was lost because of irregular weather conditions (rain, wind or something). Also when HVAC maintenance personnel moves sensors because of their activities the connection can get lost. The base-station was moved around several times to find the optimal position with minimal downtime. 89 Broken sensors When comparing the measured RH with KNMI RH data it is easy to detect sensor malfunction. Even if there is a certain error between KNMI and measurement the real faulty operation of a sensor can easily be detected. This can be seen in the following figure where the green graph starts showing completely different and unrealistic readings from the end of august. Figure 54: Relative humidity measurements showing a broken sensor (green), the measurements are continued using data from KNMI (blue). The data from both sources shows a close match. Measurement setup trouble shooting Some of the sensors and logging processes of these sensors are a constant cause of trouble. Continuous watching and troubleshooting are required to keep the setup up and running. List of issues: History being disabled for some reason, detection and manual enabling required. Resulting in data loss: holes like can be seen in the following figure One of the virtual sensors (PRIVA input) had a flaw in the translation from PRIVA to the WiSensys database. WiSensys showed completely different data than the source from that data. One error in Matlab could be found fixing the problem partially. Other causes of this fault are still unknown. Solved by adding an additional virtual sensor and repeating every step of the translation. Manual upload of data. Sensor breakdown because of water damage Disconnection of the base station because it was sent away for software updates. WiSensys informally reassured us the intermediate measurement data would be stored in the sensors itself, and would still be uploaded after reconnection of the base station some days later. This appeared to be the case for most of the sensors and most of the data, but not for everything. This can be seen in following figures. 90 Figure 55: Example of WiSensys system downtime, indicating flaws in system reliability Measurement validation AHU operation measurements The behavior of the air handling unit fans can be assessed by plotting the fan control readings from the building management system for a couple of days. The figures are not a good example because they contain a gap of lost measurement data. Fans AHU office [1500/750/0 rpm] 2.5 Fan 1 2 Fan 2 1.5 1 0.5 0 6-06-11 0:00 8-06-11 0:00 10-06-11 0:0012-06-11 0:0014-06-11 0:0016-06-11 0:0018-06-11 0:00 Figure 56: The AHU office is equipped with fans with Dahlander control, which means they have two fixed speeds they can operate at. Observation of the fan behavior shows they are at full speed (1500 rpm) during occupation and at half speed (750 rpm) while the building is unoccupied 100 Fans AHURestaurant [%] 80 Fan 1 60 Fan 2 40 20 0 6-06-11 0:00 8-06-11 0:00 10-06-11 0:0012-06-11 0:0014-06-11 0:0016-06-11 0:0018-06-11 0:00 Figure 57: The AHU restaurant is equipped with variable frequency fans. The fans though are operated in a strict schedule of: activated at 75% during occupation and off while the building is unoccupied 91 Measurement replacement by KNMI data Figure 58: Temperature measurements from different sources (BMS sensor, WiSensys sensor and KNMI) are compared. They do not show a very strong correlation but it is acceptable for the total building FDD purpose. During sensor downtime the WiSensys measurements are replaced with KNMI data 92 H Subsystem results AHU office & restaurant For the caclulation of the energy recovery in the office AHU both the temperature and the relative humidity are required in three different points. The measurement setup supplies these for the outside air, return air (air subtracted from inside of the builiding) and exhaust air. From the temperature and relative humidity the enthalpy is calculated. From the three obtained enthalply’s the total energy recovery by energy wheel Qrecovery is calculated, which is an important step in the acquisition of the total heating/cooling supplied by the AHU. Figure 59: During the measurement data processing phase subsystem results, including the air handling units, were plotted and observed to validate the expected behavior and absence of anomalies Induction system Figure 60: During the measurement data processing phase subsystem results, including the induction circuit, were plotted and observed to validate the expected behavior and absence of anomalies 93 Figure 61: Induction heating and cooling energy flows in comparison with the outside temperature Temperatures and flow combined results in the actual heat transfer by the induction system which has frequently is the induction system heating and cooling in comparison with the outside temperature. Important observations on the initial energy analysis of the building were done with these figures. 94 I Details on the HVAC control improvements The unlocking of heat pumps In the central controller the four heat pumps are unlocked depending on the outside temperature. One of the hydro modules (heat pump) is always unlocked for cooling purposes. This hydro module is used to maintain the CHW buffer tank set point temperature. The main advantage of this control is that there is always free engine heat available, which can be used in case there is a (slight) heat demand for the HW circuit. In this case it is not necessary to enable a heat pump for heating. Improvement: Summer/winter block. For an outside temperature < 5oC (adjustable) no heat pumps are unlocked for cooling purposes. For an outside temperature > 25oC (adjustable) no heat pumps are unlocked for heating purposes. Heat pump selection The heat pumps are switched on a daily or weekly basis and in case of malfunction another heat pump will automatically take over. The heat pump selection can be applied automatically or manually. Heat pump switching should occur during the night to exclude heating/cooling performance loss. The number of startups/shutdowns and operation time per pump are registered and this information is included in the selection process of the heat pumps. Heat pump startup/shutdown should be kept to a minimum in order to save heat pump lifetime. Improvement: A heat pump is not allowed to be started up more than three times per hour. Multiple heat pumps are enabled/disabled in cascade order. Heat pump heating operation The heat pumps are currently not disabled due to too high/low return temperatures. Improvement: If the return temperature > 40oC (adjustable), one of the working heat pumps is shut down. After a given latency, the process is repeated to see if more pumps can be shut down. For supply temperature > 45oC (adjustable), first the circulation pump of the engine heat is shut down. The superfluous heat is released to the outside air. In case the supply temperature drops below the set point again, the pump is re-enabled. A certain latency (adjustable) is applied to this process. Heat pump cooling operation During operation the buffer tank set point temperature is maintained. The secondary supply temperature is set at 9°C. Based on the residual of the calculated and measured secondary supply temperature one or more heat pumps are enabled for cooling. The return temperature is monitored, if it drops below 11°C (adjustable), one heat pump is shut down. Hot water boiler The hot water boiler is unlocked in case there is a heating demand AND the HW supply temperature becomes significantly lower than the HW supply set point temperature OR when the heat pomp configuration gives an error signal. To avoid repeated enabling and disabling a 5 minute latency is used. 95 Floor heating and radiators This represents a very small share of the total building heating. They work autonomous of the BMS. Heat is obtained from the hot water circuit and the share of these heating subsystems is included in the calculations of the induction circuit. Induction circuit control The amounts of water delivered from the plant to the induction circuits is controlled by valves actuated by a PID controller. The PID controller determines the valve setting from the HW and CHW set-point curves and the actually measured supply temperature. The HW and CHW set-point curves are depending on the outside temperature and are configured as follows: Temperature set-point Induction heating Outside temperature X1 = -10°C (adjustable) Set-point temperature Y1 = 50°C (adjustable) Outside temperature X2 = 15°C (adjustable) Set-point temperature Y2 = 40°C (adjustable) 60 40 HW 20 CHW 0 -20 Induction cooling Outside temperature X1 = -10°C (adjustable) Set-point temperature Y1 = 15°C (adjustable) Outside temperature X2 = 15°C (adjustable) Set-point temperature Y2 = 1°C (adjustable) -10 0 10 Outside temperature 20 30 Figure 62: The heating and cooling curves for the induction system. Temperature set-points are related to the outside temperature These are temperature settings for the water that is produced and made available in the induction circuits. These temperatures do not indicate the heating/cooling actually delivered. The heating/cooling on room level is controlled locally by thermostat readjustment which defines the desired temperature and the mix of HW and CHW required to achieve this supply temperature. AHU control In case the outside temperature during the day exceeds 24oC the night ventilation is unlocked. During the night the ventilation is activated when the room temperature remains above a certain set-point temperature and, on the other hand, the outside temperature is sufficiently low as well. When the room temperature drops below a second set-point temperature the night ventilation is deactivated again. The cooling and heating coil are not active during night ventilation. Delayed startup. When the outside temperature is below some threshold the AHU will operate with a delayed startup. The delay time is configurable. The supply air temp set-point will start at a significantly higher temperature. After the delayed startup the supply air will be decreased with 0.1oC per time unit until it reaches the original set-point. Temperature control The AHU supply temperature is controlled depending on the calculated and measured supply air temperature. A supply temperature curve specifies the desired supply temperature depending on the outside temperature. A heating and cooling coil fed with hot/cold water from the plant are used to realize the demanded temperature. 96 The AHU supply air temperature is not very determining for the actual heating or cooling supplied to the indoor environment. The AHU does not have a significant capacity for heavy heating or cooling. The induction system supplies of the majority of heating and cooling. 97 Temperature set-point AHU temperature curve Outside temperature X1 = 0°C (adjustable) Set-point temperature Y1 = 21°C (adjustable) Outside temperature X2 = 20°C (adjustable) Set-point temperature Y2 = 16°C (adjustable) 25 20 15 10 5 0 -10 0 10 20 Outside temperature Figure 63: AHU temperature curve 30 J M-files for data processing %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Version 3.0 of the PRIVA History Uploader % % This uploader - imports his instructions from settings.mat % % - checks last uploaded data from lastupload.mat % % Latest improvements 16/08/2011 Fix decimals in Wysensys upload % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %load settings auth='944b9aa7b823b3caf26ce169d37990c0'; % required for server authentication load settings.mat % which includes the definition of which parameters to upload enoughdata = 1; % variable which is used when checking if there is new data to process %start log, an errolog is written and if necessary it is send by email delete errorlog.txt diary errorlog.txt error = 0; %check activity, there is a button to temporarily deactivate the program without having %to remove it if settings.on == 1 disp('uploads are scheduled, program running') disp(' ') else disp('nothing is scheduled, program closing') disp(' ') close end %check upload tasks if settings.numberofuploads > 1 disp(sprintf('%d wisensys uploads are planned',settings.numberofuploads)) disp(' ') elseif settings.numberofuploads ==1 disp(sprintf('%d wisensys upload is planned',settings.numberofuploads)) disp(' ') else %I started on a email upload as well but it is never put into use disp(sprintf('no wisensys uploads planned, moving on to emails')) end %% open and format starttime % try to find lastupload.mat try % open lastupload.mat load lastupload.mat % tells the program until which timestep the data is already transfered lastuploadd = 1; disp('last upload found, only uploading new data') catch disp('no previous upload found, uploading all data') disp(' ') % if there is no 'lastupload.mat' the program will upload all the data it can find lastuploadd = 0; end if lastuploadd ==1 % the MSAccess query system uses time notations which are much different % from Matlab time syntax. Conversion is required. [a,b,c,d,e,f]=datevec(lastupload); b = formatt(b); 98 c = formatt(c); d = formatt(d); e = formatt(e); f = 0; f = formatt(f); %format the start time for query syntax starttime = sprintf('%s%s%s%s%s%d %s%s%s%s%s%s','#',b,'/',c,'/',a,d,':',e,':',f,'#'); else end %% Set db connection preferences with setdbprefs. % standard script for database connection s.DataReturnFormat = 'cellarray'; s.ErrorHandling = 'store'; s.NullNumberRead = 'NaN'; s.NullNumberWrite = 'NaN'; s.NullStringRead = 'null'; s.NullStringWrite = 'null'; s.JDBCDataSourceFile = ''; s.UseRegistryForSources = 'yes'; s.TempDirForRegistryOutput = ''; setdbprefs(s) % Make connection to database. Note that the password has been omitted. % connect... database name, login, password conn = database('History','',''); %% start loop of imports and uploads for i = 1:settings.numberofuploads % the parameter name in MSAccess database and ID for worksphere % database are extracted from the settings file if settings.numberofuploads == 1 upload = settings.upload{i}; ID = settings.ID{i}; else upload = cell2mat(settings.upload{i}); ID = cell2mat(settings.ID{i}); end sdtid=ID; % one by one the parameters are addressed with onscreen notifications disp(sprintf('uploading %s to ID: %s',upload,ID)) % let op. cell2mat gebruikt! try %% extract specified data from the access db % one version for defined time range, one version for upload without time % boundary if lastuploadd == 1 Time = fetch(conn, sprintf('SELECT ALL Systeemtijd FROM "%s" WHERE Systeemtijd > %s',upload,starttime)); Value = fetch(conn, sprintf('SELECT ALL Waarde FROM "%s" WHERE Systeemtijd > %s',upload,starttime)); else Time = fetch(conn, sprintf('SELECT ALL Systeemtijd FROM "%s" WHERE Systeemtijd',upload)); Value = fetch(conn, sprintf('SELECT ALL Waarde FROM "%s" WHERE Systeemtijd',upload)); end catch %detection of failure: error message + assignment to send error email error = 1; disp('something failed in database communications, sending error log'); end 99 % If the newly extracted data is very little (less than 10 entries) it is % not worth it to continue, about all the data is on the worksphere % server already, the user is advised to try again later if length(Time) < 10 disp('not enough data to upload, try again later') enoughdata = 0; else end % in the settings file it is stated if a parameter needs an additional % devision in order to get to the correct value (a lot of temperatures are % a factor 10 too high and must be corrected) try divided = settings.divide{i,1}; divided = cell2mat(divided); catch divided = settings.divide{i}; end divided = str2num(divided); if enoughdata ==1 % application of the division correction if divided == 1 valuee = cell2mat(Value); valuee = valuee * 0.1; Value = valuee; disp('values divided by ten') else disp('no division required') Value = cell2mat(Value); end %% The actual upload data=''; %empty vector which will be filled with data in the loop below % the upload script for the worksphere database required again a different % syntax for the time vector + data points. it will be converted here. for j = 1:length(Time) Time2 = Time{j,1}; aa = str2num(Time2(1:4)); bb = str2num(Time2(6:7)); cc = str2num(Time2(9:10)); dd = str2num(Time2(12:13)); ee = str2num(Time2(15:16)); ff = str2num(Time2(18:19)); Timedatenum(j)=datenum(aa,bb,cc,dd,ee,ff); % the actual syntax for data upload to worksphere database data = strcat(data, sprintf('%04d %02d %02d,%02d %02d %02d,%3.1f;', aa,bb,cc,dd,ee,ff , Value(j))); end % the upload to the server.. using urlread with the targed ID, % authentication key and data string as formatted above try params={'sdtid',sdtid,'auth',auth,'data',data}; %parameters string [S,status] = urlread('http://www.gebouwinbeeld.nl/import.matlab.php','post',params); %url write command catch error = 1; % if this script fails there is an error report for server connection problems disp('something failed in server communications, sending error log') end else end 100 end % uploads are finished disp('scheduled upload finished. see you later') %% Close database connection. close(conn) clear conn s if enoughdata ==1 %store last uploaded time value, this value will be used the next time the %program is run to define the time range lastupload=datenum(Time(length(Time))); save lastupload.mat else end %close log diary off %message if error ==0; report = 'The PRIVA export was successfull'; else report = 'The PRIVA export has failed'; end % email script % Define these variables appropriately: mail = '[email protected]'; %Your GMail email address password = 'privaexport1'; %Your GMail password % Then this code will set up the preferences properly: setpref('Internet','E_mail',mail); setpref('Internet','SMTP_Server','smtp.gmail.com'); setpref('Internet','SMTP_Username',mail); setpref('Internet','SMTP_Password',password); props = java.lang.System.getProperties; props.setProperty('mail.smtp.auth','true'); props.setProperty('mail.smtp.socketFactory.class', 'javax.net.ssl.SSLSocketFactory'); props.setProperty('mail.smtp.socketFactory.port','465'); % Send the email. Note that the first input is the address you are sending the email to file = 'errorlog.txt'; sendmail('[email protected]',report,report,file) %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Version 2.1 of the Import Tool % % This importer retrieves all the new data from the server/database % and puts them in a local .mat file for the calculations in the % following m-files % % Latest improvements 6/11/2011 % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% clear all close all clc % % % % % % Choose which data should be imported 1 = wysensys sensors (T, RV etc.) 2 = KNMI 3 = AHU office 4 = AHU office FANS 5 = AHU restaurant 101 % % % % % w q 6 = AHU restaurant FANS 7 = Room conditions 8 = dp induction circuit 9 = temperatures induction circuits 10 = Energy date from excel = 7; = 2; %for excel only. 1 = new, else = addition % choose time range % Take extra care in choosing these values % overlap in time range results in overwriting of values begint = datenum(2012, 3, 1, 00, 00, 00); endt = datenum(2012, 5, 1, 00, 00, 00); disp('importer started') starttime = datevec(begint) %display begin and endtime for validation endtime = datevec(endt) % authentication key for database access auth='944b9aa7b823b3caf26ce169d37990c0'; % load excisting data load alldata % format timerange values [s_year, s_month, s_day, s_hour, s_min, s_sec] = datevec(begint); [e_year, e_month, e_day, e_hour, e_min, e_sec] = datevec(endt); a=formatt(s_year); b=formatt(s_month); c=formatt(s_day); d=formatt(s_hour); e=formatt(s_min); f=formatt(s_sec); g=formatt(e_year); h=formatt(e_month); i=formatt(e_day); j=formatt(e_hour); k=formatt(e_min); l=formatt(e_sec); %% DATA IMPORTS % Step 1: Wysensys sensors if w == 1 %% temp sensor 041 SDTID=637; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); 102 % Add up previous data with new data % after that filter and sort unique data alldata.timet041 = [alldata.timet041; data(:,1)]; alldata.t041 = [alldata.t041; data(:,2)]; [alldata.timet041,I,J]=unique(alldata.timet041); alldata.t041 = alldata.t041(I); % show the last uploaded timepoint for validation purposes lastuploaded = datevec(alldata.timet041(length(alldata.timet041))) %% temp sensor 042 SDTID=638; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); % splitsen tijd en data olddata = length(alldata.t042) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.timet042 = [alldata.timet042; data(:,1)]; alldata.t042 = [alldata.t042; data(:,2)]; [alldata.timet042,I,J]=unique(alldata.timet042); alldata.t042 = alldata.t042(I); %% temp sensor 043 SDTID=640; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); % splitsen tijd en data olddata = length(alldata.t043) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.timet043 = [alldata.timet043; data(:,1)]; alldata.t043 = [alldata.t043; data(:,2)]; [alldata.timet043,I,J]=unique(alldata.timet043); 103 alldata.t043 = alldata.t043(I); %% temp sensor 053 SDTID=642; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); % splitsen tijd en data olddata = length(alldata.t043) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.timet053 = [alldata.timet053; data(:,1)]; alldata.t053 = [alldata.t053; data(:,2)]; [alldata.timet053,I,J]=unique(alldata.timet053); alldata.t053 = alldata.t053(I); %% RH sensor 043 SDTID=639; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); % splitsen tijd en data olddata = length(alldata.t043) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.time043 = [alldata.time043; data(:,1)]; alldata.RH043 = [alldata.RH043; data(:,2)]; [alldata.time043,I,J]=unique(alldata.time043); alldata.RH043 = alldata.RH043(I); %% RH sensor 053 SDTID=641; m1=num2str(SDTID); % download data from website 104 params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); % splitsen tijd en data olddata = length(alldata.t043) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.time053 = [alldata.time053; data(:,1)]; alldata.RH053 = [alldata.RH053; data(:,2)]; [alldata.time053,I,J]=unique(alldata.time053); alldata.RH053 = alldata.RH053(I); % confirmation of last stored timestamp laststoreddate = datevec(alldata.time053(length(alldata.time053))) % % % % % % % %% pyranometer % % % % % % % SDTID=62; % % % % % % % m1=num2str(SDTID); % % % % % % % % % % % % % % clear data % % % % % % % % % % % % % % % download data from website % % % % % % % params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; % % % % % % % data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % % % % % % % % % % % % % % % standaard tijd omzetten naar datenum tijd % % % % % % % length(data) % % % % % % % % % % % % % % data=datahandle(data); % % % % % % % % % % % % % % olddata = length(alldata.t043) % % % % % % % newdata = length(data) % % % % % % % % % % % % % % % Add up previous data with new data % % % % % % % % after that filter and sort unique data % % % % % % % alldata.timepyro = [alldata.timepyro; data(:,1)]; % % % % % % % alldata.pyro = [alldata.pyro; data(:,2)]; % % % % % % % % % % % % % % [alldata.timepyro,I,J]=unique(alldata.timepyro); % % % % % % % alldata.pyro = alldata.pyro(I); % Step 2: KNMI data elseif w == 2 knmidata = urlwrite('http://www.knmi.nl/klimatologie/uurgegevens/datafiles/370/u urgeg_370_2011-2020.zip','knmidata.zip'); unzip('knmidata.zip'); newData1 = importdata('uurgeg_370_2011-2020.txt'); 105 % Create new variables in the base workspace from those fields. vars = fieldnames(newData1); for i = 1:length(vars) assignin('base', vars{i}, newData1.(vars{i})); end date = num2str(data(:,2)); hour = data(:,3); %produce the real date in datenum for i = 1:length(date) a = str2num(date(i,1:4)); b = str2num(date(i,5:6)); c = str2num(date(i,7:8)); date2(i) = datenum(a,b,c,hour(i),0,0); end Time = date2; Temperatuur = data(:,8)./10; RH = data(:,18); alldata.timeKNMI = Time'; alldata.TKNMI = Temperatuur; alldata.RHKNMI = RH; laststoreddate = datevec(alldata.timeKNMI(length(alldata.timeKNMI))) % Step 3: AHU OFFICE elseif w == 3 %% T1 SDTID=637; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); size(data); % conversion of website time to datenum data=datahandle(data); % splitsen tijd en data olddata = length(alldata.t043) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.time1 = [alldata.time1; data(:,1)]; alldata.T1 = [alldata.T1; data(:,2)]; [alldata.time1,I,J]=unique(alldata.time1); alldata.T1 = alldata.T1(I); %% RH1 SDTID=641; 106 m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); % splitsen tijd en data olddata = length(alldata.t043) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.timeRH1 = [alldata.timeRH1; data(:,1)]; alldata.RH1 = [alldata.RH1; data(:,2)]; [alldata.timeRH1,I,J]=unique(alldata.timeRH1); alldata.RH1 = alldata.RH1(I); % datevec(data(length(data),1)) %% T2 SDTID=640; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.t043) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.time2 = [alldata.time2; data(:,1)]; alldata.T2 = [alldata.T2; data(:,2)]; [alldata.time2,I,J]=unique(alldata.time2); alldata.T2 = alldata.T2(I); %% RH2 SDTID=639; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); 107 % conversion of website time to datenum data=datahandle(data); % Add up previous data with new data % after that filter and sort unique data alldata.timeRH2 = [alldata.timeRH2; data(:,1)]; alldata.RH2 = [alldata.RH2; data(:,2)]; [alldata.timeRH2,I,J]=unique(alldata.timeRH2); alldata.RH2 = alldata.RH2(I); %% T3 SDTID=643; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.T3) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.time3 = [alldata.time3; data(:,1)]; alldata.T3 = [alldata.T3; data(:,2)]; [alldata.time3,I,J]=unique(alldata.time3); alldata.T3 = alldata.T3(I); %% T4 SDTID=645; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.T4) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.time4 = [alldata.time4; data(:,1)]; alldata.T4 = [alldata.T4; data(:,2)]; [alldata.time4,I,J]=unique(alldata.time4); alldata.T4 = alldata.T4(I); 108 % laststoreddate = datevec(alldata.time4(length(alldata.time4))) % Step 4: AHU office fans elseif w == 4 %% fan1 SDTID=646; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); alldata.timefan1 = data(:,1); alldata.fan1=data(:,2); %% fan2 SDTID=647; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); alldata.timefan2 = data(:,1); alldata.fan2=data(:,2); %% AHU RESTAURANT elseif w == 5 %% rest T1 SDTID=637; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); 109 olddata = length(alldata.T1rest) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.time1rest = [alldata.time1rest; data(:,1)]; alldata.T1rest = [alldata.T1rest; data(:,2)]; [alldata.time1rest,I,J]=unique(alldata.time1rest); alldata.T1rest = alldata.T1rest(I); %% rest T2 SDTID=648; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.T1rest) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.time2rest = [alldata.time2rest; data(:,1)]; alldata.T2rest = [alldata.T2rest; data(:,2)]; [alldata.time2rest,I,J]=unique(alldata.time2rest); alldata.T2rest = alldata.T2rest(I); %% rest T3 SDTID=649; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.T1rest) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.time3rest = [alldata.time3rest; data(:,1)]; alldata.T3rest = [alldata.T3rest; data(:,2)]; 110 [alldata.time3rest,I,J]=unique(alldata.time3rest); alldata.T3rest = alldata.T3rest(I); %% rest T4 SDTID=650; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.T1rest) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.time4rest = [alldata.time4rest; data(:,1)]; alldata.T4rest = [alldata.T4rest; data(:,2)]; [alldata.time4rest,I,J]=unique(alldata.time4rest); alldata.T4rest = alldata.T4rest(I); laststoreddate = datevec(alldata.time4rest(length(alldata.time4rest))) %% fan1 rest elseif w == 6 SDTID=652; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); alldata.timefan1rest = data(:,1); alldata.fan1rest=data(:,2); %% fan2 SDTID=653; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; 111 data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); alldata.timefan2rest = data(:,1); alldata.fan2rest=data(:,2); %% RUIMTE CONDITIES elseif w == 7 %% floor1 SDTID=658; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.Tfloor1) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.timefloor1 = [alldata.timefloor1; data(:,1)]; alldata.Tfloor1 = [alldata.Tfloor1; data(:,2)]; [alldata.timefloor1,I,J]=unique(alldata.timefloor1); alldata.Tfloor1 = alldata.Tfloor1(I); % floor2 SDTID=659; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.Tfloor2) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.timefloor2 = [alldata.timefloor2; data(:,1)]; 112 alldata.Tfloor2 = [alldata.Tfloor2; data(:,2)]; [alldata.timefloor2,I,J]=unique(alldata.timefloor2); alldata.Tfloor2 = alldata.Tfloor2(I); %% floor3 SDTID=660; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.Tfloor3) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.timefloor3 = [alldata.timefloor3; data(:,1)]; alldata.Tfloor3 = [alldata.Tfloor3; data(:,2)]; [alldata.timefloor3,I,J]=unique(alldata.timefloor3); alldata.Tfloor3 = alldata.Tfloor3(I); %% floor4 SDTID=661; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.Tfloor4) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.timefloor4 = [alldata.timefloor4; data(:,1)]; alldata.Tfloor4 = [alldata.Tfloor4; data(:,2)]; [alldata.timefloor4,I,J]=unique(alldata.timefloor4); alldata.Tfloor4 = alldata.Tfloor4(I); laststoreddate = datevec(alldata.timefloor4(length(alldata.timefloor4))) %% RH3 SDTID=644; m1=num2str(SDTID); % download data from website 113 params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.RH3) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.timeRH3 = [alldata.timeRH3; data(:,1)]; alldata.RH3 = [alldata.RH3; data(:,2)]; [alldata.timeRH3,I,J]=unique(alldata.timeRH3); alldata.RH3 = alldata.RH3(I); %% INDUCTIE CIRCUITS elseif w ==8 %% dp hot SDTID=664; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.dphot) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.timedphot = [alldata.timedphot; data(:,1)]; alldata.dphot = [alldata.dphot; data(:,2)]; [alldata.timedphot,I,J]=unique(alldata.timedphot); alldata.dphot = alldata.dphot(I); clear data % dp cold SDTID=665; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); 114 % conversion of website time to datenum data=datahandle(data); olddata = length(alldata.dpcold) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.timedpcold = [alldata.timedpcold; data(:,1)]; alldata.dpcold = [alldata.dpcold; data(:,2)]; [alldata.timedpcold,I,J]=unique(alldata.timedpcold); alldata.dpcold = alldata.dpcold(I); laststoreddate = datevec(alldata.timedpcold(length(alldata.timedpcold))) %% heating Tin elseif w == 9 SDTID=654; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); % Add up previous data with new data % after that filter and sort unique data alldata.timeTHWin = [alldata.timeTHWin; data(:,1)]; alldata.THWin = [alldata.THWin; data(:,2)]; [alldata.timeTHWin,I,J]=unique(alldata.timeTHWin); alldata.THWin = alldata.THWin(I); %% heating Tout SDTID=655; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); % Add up previous data with new data % after that filter and sort unique data alldata.timeTHWout = [alldata.timeTHWout; data(:,1)]; alldata.THWout = [alldata.THWout; data(:,2)]; [alldata.timeTHWout,I,J]=unique(alldata.timeTHWout); 115 alldata.THWout = alldata.THWout(I); %% cooling Tin SDTID=656; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); % Add up previous data with new data % after that filter and sort unique data alldata.timeTCHWin = [alldata.timeTCHWin; data(:,1)]; alldata.TCHWin = [alldata.TCHWin; data(:,2)]; [alldata.timeTCHWin,I,J]=unique(alldata.timeTCHWin); alldata.TCHWin = alldata.TCHWin(I); %% cooling Tout SDTID=688; m1=num2str(SDTID); % download data from website params1 = {'sdtid',m1,'start',[a '-' b '-' c ' ' d ':' e ':' f],'end',[g '-' h '-' i ' ' j ':' k ':' l],'auth',auth}; data = str2num(urlread('http://www.gebouwinbeeld.nl/export.matlab.php','GET' ,params1)); % conversion of website time to datenum data=datahandle(data); % splitsen tijd en data olddata = length(alldata.TCHWout) newdata = length(data) % Add up previous data with new data % after that filter and sort unique data alldata.timeTCHWout = [alldata.timeTCHWout; data(:,1)]; alldata.TCHWout = [alldata.TCHWout; data(:,2)]; [alldata.timeTCHWout,I,J]=unique(alldata.timeTCHWout); alldata.TCHWout = alldata.TCHWout(I); elseif w ==10 % q = 1; % 1 for fresh excel import, 2 for addition of some new data, with unique function if q == 1 %% Elektricity Use 116 newData1 = importdata('Electricityson.xls'); % version with UNIX timestamp time = newData1.data; [a b c d e f]=datevec(time); a = a+1900; c = c-1; time=datenum(a,b,c,d,e,f); value = newData1.textdata(2:length(newData1.textdata),4); % operation to remove the 'KWh' string from the input values for i = 1:length(value) value2 = value{i}; b = str2num(value2(1:length(value2)-3)); valuenum(i)=b; end value = valuenum'; % removal of double entries value = value(find(unique(time))); time = unique(time); alldata.timeElec = time; alldata.Elec = value; clear time value newData1 %% Gas Use newData1 = importdata('Gasson.xls'); time = newData1.data; [a b c d e f]=datevec(time); a = a+1900; c = c-1; time=datenum(a,b,c,d,e,f); value = newData1.textdata(2:length(newData1.textdata),4); for i = 1:length(value) value2 = value{i}; b = str2num(value2(1:length(value2)-4)); valuenum(i)=b; end value = valuenum'; % removal of double entries value = value(find(unique(time))); time = unique(time); alldata.timeGas = time; alldata.Gas = value; % laatstewaardesExcel = datevec(alldata.timeGas(length(alldata.timeGas))) else % here is the addition part %% Elektricity Use newData1 = importdata('Electricityson.xls'); % versie met timestamp (data als nummer) time = newData1.data; 117 [a b c d e f]=datevec(time); a = a+1900; c = c-1; time=datenum(a,b,c,d,e,f); value = newData1.textdata(2:length(newData1.textdata),4); for i = 1:length(value) value2 = value{i}; b = str2num(value2(1:length(value2)-3)); valuenum(i)=b; end value = valuenum'; timeold = alldata.timeElec; elecold = alldata.Elec; timenew = [timeold;time]; elecnew = [elecold;value]; value = elecnew(find(unique(timenew))); time = unique(timenew); alldata.timeElec = time; alldata.Elec = value; clear time value newData1 %% Gas Use newData1 = importdata('Gasson.xls'); clear time time = newData1.data; [a b c d e f]=datevec(time); a = a+1900; c = c-1; time=datenum(a,b,c,d,e,f); value = newData1.textdata(2:length(newData1.textdata),4); for i = 1:length(value) value2 = value{i}; b = str2num(value2(1:length(value2)-4)); valuenum(i)=b; end value = valuenum'; clear timeold timeold = alldata.timeGas; gasold = alldata.Gas; timenew = [timeold;time]; gasnew = [gasold;value]; value = gasnew(find(unique(timenew))); time = unique(timenew); alldata.timeGas = time; alldata.Gas = value; 118 laatstewaardesExcel = datevec(alldata.timeGas(length(alldata.timeGas))) end end disp('finished') save alldata alldata %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Calculations for Heating and Cooling Energy AHU office % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % regular initiation clear all close all clc % loading and definition of the used parameters load alldata % .mat file with all measurement data load CP % table with CP values time1 = alldata.time1; time2 = alldata.time2; time3 = alldata.time3; time4 = alldata.time4; Tout = alldata.T1; Tsup = alldata.T4; Tret = alldata.T3; Texh = alldata.T2; T1 = Tout; T2 = Texh; T3 = Tret; T4 = Tsup; timeRH1 = alldata.timeRH1; timeRH2 = alldata.timeRH2; timeRH3 = alldata.timeRH3; RH1 = alldata.RH1; RH2 = alldata.RH2; RH3 = alldata.RH3; %outside %exhaust %return % the vectors are cut to the same lenght.. starting at the maximum value of % the first entry of eacht vector and ending at the minimum value of the % last entry of each vector startt = max([time1(1),time2(1),time3(1),time4(1),timeRH1(1),timeRH2(1),timeRH3(1)]); endd = min([time1(length(time1)),time2(length(time2)),time3(length(time3)),time4(le ngth(time4)),timeRH1(length(timeRH1)),timeRH2(length(timeRH2)),timeRH3(lengt h(timeRH3))]); % display of the start and endtime of the synchronized vectors starttime = datevec(startt) endtime = datevec(endd) % a time vector is created to represent the vector of the correct length % and making use of a step size of 8 minutes Time = startt:datenum(0,0,0,0,8,0):endd; Time = Time'; %all vectors are interpolated to fit this vector... and giving all %parameter vectors the same step size T1 = interp1(time1,T1,Timefixed); T2 = interp1(time2,T2,Timefixed); T3 = interp1(time3,T3,Timefixed); 119 T4 = interp1(time4,T4,Timefixed); RH1 = interp1(timeRH1,RH1,Timefixed); RH2 = interp1(timeRH2,RH2,Timefixed); RH3 = interp1(timeRH3,RH3,Timefixed); % old values are cleared clear Timefixed starttime endtime time1 time2 time3 time4 timeRH1 timeRH2 timeRH3 % calculation of the enthalpy at the three measured points in the air % handling unit outside, exhaust and return position for i = 1:length(Time) H1(i) = interp1(CP.Tdry,CP.CPdry,T1(i)+273)*(T1(i)+273) + RH1(i)*interp1(CP.Tvap,CP.CPvap,T1(i)+273)*(T1(i)+273); H2(i) = interp1(CP.Tdry,CP.CPdry,T2(i)+273)*(T2(i)+273) + RH2(i)*interp1(CP.Tvap,CP.CPvap,T2(i)+273)*(T2(i)+273); H3(i) = interp1(CP.Tdry,CP.CPdry,T3(i)+273)*(T3(i)+273) + RH3(i)*interp1(CP.Tvap,CP.CPvap,T3(i)+273)*(T3(i)+273); end H1=H1'; H2=H2'; H3=H3'; % figure of intermediate results figure plot(Time,H1,Time,H2,Time,H3) legend('H outside','H exhaust','H return') title('Enthalpys AHU office') xlabel('Time') ylabel('Enthalpy [kj/kgK]') datetick % % % % % % save the image filetype = '-dtiff'; resolution = '-r750'; title2 = 'Enthalpy AHUoffice2'; print(filetype,resolution,title2) % dahlander 1500/750 rpm % the air flow is determined from the operation schedules (measurement was % not possible) timevector = datevec(Time); for i = 1:length(Time) if weekday(Time(i)) >1 & weekday(Time(i)) < 7 % selection of the day of the week (workday and weekend operation) if timevector(i,4) >= 9 & timevector(i,4) <21 % selection of the time of the day (AHU operation hours 9-21) v1(i) = 9.72; v2(i) = 10.28; else v1(i) = 4.86; v2(i) = 5.14; end else v1(i) = 4.86; v2(i) = 5.14; end end % all values in m3/s % convesion to Kelvins for further processing T1 = T1+273; T2 = T2+273; T3 = T3+273; T4 = T4+273; % table with air density values in degC Trho =[-40 120 -20 0 20 40 60 80 100]'; Trho = Trho+273; rho = [1.52 %kg/m3 1.40 1.293 1.205 1.127 1.067 1.000 0.946]'; % just to test wether there are still NaN's present NaNcount = 0; % Energy calculations for i = 1:length(Time) % q[m3/s] rho[kg/m3] Cp[kj/kgK] T[K] --> Q[kW] Qout(i) = v1(i) * interp1(Trho,rho,T1(i)) * interp1(CP.Tdry,CP.CPdry,T1(i)) * T1(i); %energy flow entering the AHU from outside air test1(i) = interp1(Trho,rho,T1(i)); %partial calculation to validate if data and calculations are correct if isnan(test1(i)) NaNcount = count+1; % count Not a Numbers end test2(i) = interp1(CP.Tdry,CP.CPdry,T1(i)); %partial calculation to validate if data and calculations are correct Qsup(i) = v1(i) * interp1(Trho,rho,T4(i)) * interp1(CP.Tdry,CP.CPdry,T4(i)) * T4(i); %energy flow supplied to the building Qrec(i) = v2(i) * interp1(Trho,rho,(T4(i)+T2(i))/2) * interp1(CP.Tdry,CP.CPdry,(T4(i)+T2(i))/2) * (T2(i)-T4(i)); %energy flow from heat recovery end Qin = Qsup-Qout-Qrec; %the energy supplied to the AHU via the heating and cooling coil.. which is the thermal energy consumption Qout = Qout'; NaNcount %for calculation and data validation Timeplot = Time; % time vector for plots Qinplot = interp1(Time,Qin,Timeplot); % the energy flow inserted to the AHU must be devided in heating and % cooling for i = 1:length(Qinplot) if Qinplot(i)>= 0 heat(i)=Qinplot(i); cool(i)=NaN; else cool(i)=Qinplot(i); heat(i)=NaN; end end % figure of AHU heating and cooling results figure plot(Timeplot,heat,'r',Timeplot,cool,'b','linewidth',2) legend('Qin heating','Qin cooling') xlabel('Time') ylabel('Heat flow [kW]') datetick 121 title('Illustration heating/cooling power AHU office') % the same but with 0 instead of NaN's which is required for ABCAT for i = 1:length(Qinplot) if Qinplot(i)>= 0 heat(i)=Qinplot(i); cool(i)=0; else cool(i)=Qinplot(i); heat(i)=0; end end % calculations results are stored in the alldata.mat alldata.timeoffice = Timeplot; alldata.qinoffice= Qinplot; alldata.qheatoffice = heat'; alldata.qcooloffice = -cool'; save alldata alldata %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Calculations for Heating and Cooling Energy AHU Restaurant % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % regular initiation close all clear all clc % loading and definition of the used parameters load alldata % .mat file with all measurement data load CP % table with CP values T1 T2 T3 T4 = = = = alldata.T1rest; alldata.T2rest; alldata.T3rest; alldata.T4rest; length(alldata.T1rest) length(alldata.T2rest) length(alldata.T3rest) % the vectors are cut to the same lenght.. starting at the maximum value of % the first entry of eacht vector and ending at the minimum value of the % last entry of each vector startt = max([alldata.time1rest(1),alldata.time2rest(1),alldata.time3rest(1)]) endd = min([max(alldata.time1rest),max(alldata.time2rest),max(alldata.time3rest)]) % a time vector is created to represent the vector of the correct length % and making use of a step size of 8 minutes Timefixed = startt:datenum(0,0,0,0,8,0):endd; Timefixed = Timefixed'; Time2=Timefixed; %all vectors are interpolated to fit this vector... and giving all %parameter vectors the same step size T1 = interp1(alldata.time1rest,T1,Timefixed); T2 = interp1(alldata.time2rest,T2,Timefixed); T3 = interp1(alldata.time3rest,T3,Timefixed); T4 = interp1(alldata.time4rest,T4,Timefixed); % old values are cleared 122 clear Time end1 end2 starttime endtime % table with air density values in degC Trho =[-40 -20 0 20 40 60 80 100]'; Trho = Trho+273; rho = [1.52 %kg/m3 1.40 1.293 1.205 1.127 1.067 1.000 0.946]'; % change the temperatures to kelvin T1 = T1+273; T2 = T2+273; T3 = T3+273; T4 = T4+273; %% get air mass flow % the air flows are determined from the operation schedules %(measurement was not possible) timevector = datevec(Time2); for i = 1:length(Time2) if weekday(Time2(i)) >1 & weekday(Time2(i)) < 7 if timevector(i,4) >= 9 & timevector(i,4) <21 %only active on working days between 9 and 21h v1(i) = 1.44; v2(i) = 1.08; else v1(i) = 0; v2(i) = 0; end else v1(i) = 0; v2(i) = 0; end end % all values in m3/sec % we have m[m3/s], rho[kg/m3], Cp[kj/kgK], and T[K] --> Q in [kW] for i = 1:length(T1) Qout(i) = v1(i) * interp1(Trho,rho,T1(i)) * interp1(CP.Tdry,CP.CPdry,T1(i)) * T1(i); Qsup(i) = v1(i) * interp1(Trho,rho,T4(i)) * interp1(CP.Tdry,CP.CPdry,T4(i)) * T4(i); Qrec(i) = v2(i) * interp1(Trho,rho,(T4(i)+T2(i))/2) * interp1(CP.Tdry,CP.CPdry,(T4(i)+T2(i))/2) * (T2(i)-T4(i)); end Qin = Qsup-Qout-Qrec; %the energy supplied to the AHU via the heating and cooling coil.. which is the thermal energy consumption Timeplot = Time2; %time vector for plots Qinplot = Qin; 123 % the energy flow inserted to the AHU must be devided in heating and % cooling for i = 1:length(Qinplot) if Qinplot(i)>= 0 heat(i)=Qinplot(i); cool(i)=NaN; else cool(i)=Qinplot(i); heat(i)=NaN; end end % figure of AHU heating and cooling results figure plot(Timeplot,heat,'r',Timeplot,cool,'b','linewidth',2) legend('Qin heating','Qin cooling') xlabel('Time') ylabel('Heat flow [kW]') datetick title('Illustration heating/cooling power AHU restaurant') % the same but with 0 instead of NaN's which is required for ABCAT for i = 1:length(Qinplot) if Qinplot(i)>= 0 heat(i)=Qinplot(i); cool(i)=0; else cool(i)=Qinplot(i); heat(i)=0; end end % calculations results are stored in the alldata.mat alldata.timerest = Timeplot; alldata.qinrest = Qinplot; alldata.qheatrest = heat'; alldata.qcoolrest = -cool'; save alldata alldata %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Calculations for Heating and Cooling Energy AHU office % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % regular initiation close all clear all clc % loading and definition of the used parameters load alldata % .mat file with all measurement data load CP % table with CP values dphot = alldata.dphot; dpcold = alldata.dpcold; THWin = alldata.THWin; THWout = alldata.THWout; TCHWin = alldata.TCHWin; TCHWout = alldata.TCHWout; timedphot = alldata.timedphot; timedpcold = alldata.timedpcold; timeTHWin = alldata.timeTHWin; timeTHWout = alldata.timeTHWout; timeTCHWin = alldata.timeTCHWin; 124 timeTCHWout = alldata.timeTCHWout; % the vectors are cut to the same lenght.. starting at the maximum value of % the first entry of eacht vector and ending at the minimum value of the % last entry of each vector startt = max([timedphot(1),timedpcold(1),timeTHWin(1),timeTHWout(1),timeTCHWin(1),tim eTCHWout(1)]) endd = min([timedphot(length(timedphot)),timedpcold(length(timedpcold)),timeTHWin(l ength(timeTHWin)),timeTHWout(length(timeTHWout)),timeTCHWin(length(timeTCHWi n)),timeTCHWout(length(timeTCHWout))]); % a time vector is created to represent the vector of the correct length % and making use of a step size of 8 minutes starttime = datevec(startt) endtime = datevec(endd) % maak de juiste tijdvector aan Timefixed = startt:datenum(0,0,0,1,0,0):endd; Time = Timefixed; %all vectors are interpolated to fit this vector... and giving all %parameter vectors the same step size dphot = interp1(timedphot,dphot,Timefixed); dpcold = interp1(timedpcold,dpcold,Timefixed); THWin = interp1(timeTHWin,THWin,Timefixed); THWout = interp1(timeTHWout,THWout,Timefixed); TCHWin = interp1(timeTCHWin,TCHWin,Timefixed); TCHWout = interp1(timeTCHWout,TCHWout,Timefixed); % old values are cleared clear Timefixed starttime endtime time1 time2 time3 time4 timeRH1 timeRH2 timeRH3 % table with water density values in degC % T [C] TRHO = [100 80 60 40 30 25 22 20 15 10 4 0 -10 -20]; TRHO = TRHO+273; %in K % RHO water [kg/m3] RHO= [958.4 971.8 983.2 992.2 995.6502 997.0479 997.7735 998.2071 999.1026 999.7026 999.972 999.8395 1000 1000]; 125 %% calculations % flow(formulas from TAlink manufacturer) % the control valve opening settings are changed twice in the meaurement % period. For both the changes a factor in the equations had to be altered, % resulting in three individual calculations for t = 1:length(Time) if Time(t) <= datenum(2011,9,28,16,0,0) flowhot(t) = 100.*80.*dphot(t).^0.5./1000; elseif Time(t) <= datenum(2011,10,3,13,40,0) flowhot(t) = 100.*44.*dphot(t).^0.5./1000; else flowhot(t) = 100.*16.*dphot(t).^0.5./1000; end flowcold(t) = 100.*41.*dpcold(t).^0.5./1000; end % in m3/h time=Time; %% PLOTS % Plot, temperatures in Hot water circuit and chilled water circuit % for a clearer view temperatures are converted to degC figure(1) subplot(2,1,1) plot(time,THWin(1:length(time)),time,THWout(1:length(time))) legend('THWin','THWout','location','SouthWest') title('Hot water Circuit'); xlabel('time'); ylabel('temperature [^oC]'); datetick axis tight subplot(2,1,2) plot(time,TCHWin(1:length(time)),time,TCHWout(1:length(time))) legend('TCHWin','TCHWout','location','NorthWest') title('Chilled water Circuit'); xlabel('time'); ylabel('temperature [^oC]'); datetick axis tight % flow in m3/h figure(2) plot(time,flowhot(1:length(time)),'r-',time,flowcold(1:length(time)),'b-') legend('Hot water flow','Cold water flow','location','NorthWest') title('water volume flows'); xlabel('time'); ylabel('flow [m3/h]'); datetick axis tight % Temperatures to Kelvin THWin = THWin+273; THWout = THWout+273; TCHWin = TCHWin+273; TCHWout = TCHWout+273; %heat transfer calculation for i = 1:length(time) % here values in m3/h are inserted % calculation flow kg/s = flow m3/s flowhot(i) = flowhot(i) / 60 / 60 * interp1(TRHO,RHO,(THWin(i)+THWout(i))/2); flowcold(i) = flowcold(i) /60 / 60 * interp1(TRHO,RHO,(TCHWin(i)+TCHWout(i))/2); 126 /60 /60 * rho % now they are kg/s % heat flow calculation % Q[kW] = m[kg/s] * Cp[kj/kg K] * dT[K] --> kW Qhot(i) = flowhot(i) * interp1(CP.TCPwater,CP.CPwater,(THWin(i)+THWout(i))/2) * (THWin(i) THWout(i)); Qcold(i) = flowcold(i) * interp1(CP.TCPwater,CP.CPwater,(TCHWin(i)+TCHWout(i))/2) * (TCHWout(i) TCHWin(i)); end % temperature vector from KNMI temperature data TKNMI = interp1(alldata.timeKNMI,alldata.TKNMI,time); % Plot flow in kg/s and heat transfer in KW figure(3) subplot(2,1,1) plot(time,flowhot(1:length(time)),'r-',time,flowcold(1:length(time)),'b-') legend('Hot water flow','Cold water flow','location','NorthWest') title('Water mass flows'); xlabel('time'); ylabel('mass flow [kg/s]'); datetick axis tight subplot(2,1,2) plot(time,Qhot(1,1:length(time)),'r-',time,-Qcold(1,1:length(time)),'b','LineWidth',1) legend('Hot water energy','Cold water energy','location','NorthWest') title('Energy flows'); xlabel('time'); ylabel('Heat Energy flow [kW]'); datetick axis tight % Plot flow in kg/s and heat transfer in KW figure(4) subplot(2,1,1) plot(time,TKNMI,'b-') title('Outside Temperature'); xlabel('time'); ylabel('Temp [^oC]'); datetick axis tight subplot(2,1,2) plot(time,Qhot(1,1:length(time)),'r-',time,-Qcold(1,1:length(time)),'b','LineWidth',1) legend('Hot water energy','Cold water energy','location','NorthWest') title('Energy flows'); xlabel('time'); ylabel('Heat Energy flow [kW]'); datetick axis tight % calculations results are stored in the alldata.mat alldata.tQhot = time; alldata.tQcold = time; alldata.Qhot = Qhot; alldata.Qcold = Qcold; save alldata alldata %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Combination of subsystem results into whole building results % % + preperation of ABCAT inputs % 127 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % regular initiation close all clc clear all % definition of the target time period starttime = datenum(2011,8,1,0,0,0); endtime = datenum(2012,4,8,0,0,0); % import and definition of variables load alldata % heating theatind = alldata.tQhot; theatrest = alldata.timerest; theatoffice = alldata.timeoffice; heatind = alldata.Qhot; heatrest = alldata.qheatrest; heatoffice = alldata.qheatoffice; %cooling tcoolind= alldata.tQcold; tcoolrest = alldata.timerest; tcooloffice = alldata.timeoffice; coolind = alldata.Qcold; coolrest = alldata.qcoolrest; cooloffice = alldata.qcooloffice; % interpolation to 10min values time = starttime:datenum(0,0,0,0,10,0):endtime; timevec = datevec(time); heatind = interp1(theatind,heatind,time); heatrest = interp1(theatrest,heatrest,time); heatoffice = interp1(theatoffice,heatoffice,time); coolind = interp1(tcoolind,coolind,time); coolrest = interp1(tcoolrest,coolrest,time); cooloffice = interp1(tcooloffice,cooloffice,time); % clearign old variables clear theatind theatoffice theatrest tcoolind tcooloffice tcoolrest % the basic summation of subsystems heattotal = heatind + heatrest + heatoffice; cooltotal = coolind + coolrest + cooloffice; % from kW to kWh = interpolation to hourly values timeh = starttime:datenum(0,0,0,1,0,0):endtime; heatkWh = interp1(time,heattotal,timeh); coolkWh = interp1(time,cooltotal,timeh); heatofficekWh = interp1(time,heatoffice,timeh); coolofficekWh = interp1(time,heatoffice,timeh); Tdin = interp1(alldata.timeTdin,alldata.Tdin,timeh)'; Troom = interp1(alldata.timeTroom,alldata.Troom,timeh)'; Elechour = interp1(alldata.timeElec,alldata.Elec,timeh)'; Gashour = interp1(alldata.timeGas,alldata.Gas,timeh)'; % removal of NaN's for i = 1:length(Tdin)-1 if isnan(Tdin(i)) ==1 Tdin(i)=(Tdin(i-1)+Tdin(i+1))/2; %NaN is replaced by the average of the previous and next entry end end 128 % creation of daily values timeday2 = starttime:datenum(0,0,1,0,0,0):endtime; Tdout = interp1(alldata.timeTdout,alldata.Tdout,timeday2)'; Tout = interp1(alldata.time1,alldata.T1,timeday2)'; % voor Tdin alleen waardes voor 7-18 uur % preocc = 0-7 daystart = 7; % occ = 7-18 dayend = 18; % postocc = 18-24 % calculation of daily values and preoccupied - occupied - postoccupied % values i = 1; try while i~=0 heatkWhday(i) = sum(heatkWh((i-1)*24+(1:24))); coolkWhday(i) = sum(coolkWh((i-1)*24+(1:24))); heatofficekWh2(i) = sum(heatofficekWh((i-1)*24+(1:24))); coolofficekWh2(i) = sum(coolofficekWh((i-1)*24+(1:24))); Elecday(i) = sum(Elechour((i-1)*24+(1:24))); timeday(i) = datenum(2011,8,1,0,0,0)+i; Gasday(i) = sum(Gashour((i-1)*24+(1:24))); Troompreocc(i) = sum(Troom((i-1)*24+(1:daystart)))/(daystart-1); Elecpreocc(i) = sum(Elechour((i-1)*24+(1:daystart)))/(daystart-1); Troomocc(i) = sum(Troom((i-1)*24+(daystart:dayend)))/(dayenddaystart); Tdinocc(i) = sum(Tdin((i-1)*24+(daystart:dayend)))/(dayenddaystart); Elecocc(i) = sum(Elechour((i-1)*24+(daystart:dayend)))/(dayenddaystart); Troompostocc(i) = sum(Troom((i-1)*24+(dayend:24)))/(24-dayend); Elecpostocc(i) = sum(Elechour((i-1)*24+(dayend:24)))/(24-dayend); i = i+1; end catch disp('time vector processed, all daily values calculated') end % again some NaN's must be removed for i = 1:length(Tdinocc)-1 if isnan(Tdinocc(i)) ==1 Tdinocc(i)=(Tdinocc(i-1)+Tdinocc(i+1))/2; end end % processing into matrix for correct 'bar plot' representation Troom2 = [Troompreocc',Troomocc',Troompostocc']; Tdew = [Tdout(1:length(Tdout)-1),Tdinocc']; Elec = [Elecpreocc',Elecocc',Elecpostocc']; % make sure there is no negative cooling for i = 1:length(coolkWhday) if coolkWhday(i) < 0 coolkWhday(i)=0; end end timedayfixed = timeday2-1; EleckWh = interp1(alldata.timeElec,alldata.Elec,timeday); Gas = interp1(alldata.timeGas,alldata.Gas,timedayfixed); % figure#1 energy consumptions in relation with the outside temperature 129 figure subplot(3,1,1) bar(timedayfixed(1:length(timedayfixed)-1),heatkWhday,'r') title('Heating Energy Use per day') xlabel('Time') ylabel('Heating Energy [kWh]') datetick axis tight subplot(3,1,2) bar(timedayfixed(1:length(timedayfixed)-1),coolkWhday) title('Cooling Energy Use per day') xlabel('Time') ylabel('Cooling Energy [kWh]') datetick axis tight subplot(3,1,3) bar(timedayfixed,Tout,'g') title('Average outside temperature') xlabel('Time') ylabel('Temperature [^oC]') datetick axis tight % figure#2 energy consumptions in relation with the gas consumption figure subplot(4,1,1) bar(timedayfixed(1:length(timedayfixed)-1),heatkWhday,'r') title('Heating Energy Use per day') xlabel('Time') ylabel('Heating Energy [kWh]') datetick axis tight subplot(4,1,2) bar(timedayfixed(1:length(timedayfixed)-1),coolkWhday) title('Cooling Energy Use per day') xlabel('Time') ylabel('Cooling Energy [kWh]') datetick axis tight subplot(4,1,3) bar(timedayfixed(1:length(timedayfixed)-1),Gasday,'g') title('Gas Use per day') xlabel('Time') ylabel('Gas Consumption [m3/day]') datetick axis tight subplot(4,1,4) bar(timedayfixed(1:length(timedayfixed)-1),heatkWhday+coolkWhday,'y') title('Combined Energy Use per day') xlabel('Time') ylabel('Energy Use [kWh]') datetick axis tight % figures for gas consumption -> energy use correlation figure scatter(Gasday,heatkWhday+coolkWhday,2) figure scatter(heatkWhday+coolkWhday,Gasday) %% formattation to ABCAT columns Tout; Tdout; timeday2 = timeday2'; CHW = coolkWhday'; HW = heatkWhday'; Elecday = Elecday'; 130 Troompreocc = Troompreocc'; ELECpre = Elec(:,1); Troomocc = Troomocc'; Tdewocc = Tdinocc'; ELECocc = Elecocc'; Tpost = Troompostocc'; ELECpost = Elecpostocc'; % one matrix containing all processed data ready for ABCAT implementation toABCAT = [timeday2(1:length(Tout)-1),Tout(1:length(Tout)1),Tdout(1:length(Tout)1),CHW,HW,Elecday,Troompreocc,ELECpre,Troomocc,Tdewocc,ELECocc,Tpost,ELECpos t]; 131