Report on simulation results and evaluation of the

Transcription

INCREASE – Increasing the penetration of renewable energy sources in the distribution grid by
developing control strategies and using ancillary services
D3.3 – Report on simulation results and evaluation of the integrated simulation platform
INCREASE
INCREASING THE PENETRATION OF RENEWABLE
ENERGY SOURCES IN THE DISTRIBUTION GRID BY
DEVELOPING CONTROL STRATEGIES AND USING
ANCILLARY SERVICES
D3.3 – Report on simulation results and evaluation
of the integrated simulation platform
31.08.2015
1
Document info
Project Number
Funding Scheme
Work Programme
Number
Title
608998 – INCREASE
Collaborative Project
Topic ENERGY.2013.7.1.1: Development and validation of methods
and tools for network integration of distributed renewable
resources
D3.3
Report on simulation results and evaluation of the integrated
simulation platform
Dissemination Level
Date
Nature
21.07.2015
Authors
Andreas I. Chrysochos (AUTH), Georgios C. Kryonidis (AUTH),
Eleftherios O. Kontis (AUTH), Matthias Strobbe (UGent), Charis S.
Demoulias (AUTH), Grigoris K. Papagiannis (AUTH)
Contributors
Reviewers
Bart Meersman (UGent), Andrej Gubina (UL)
Document History
Date
Authors
Andreas I. Chrysochos
Georgios C. Kryonidis
Eleftherios O. Kontis
21/07/2015
Matthias Strobbe
Charis S. Demoulias
Grigoris K. Papagiannis
05/08/2015
Bart Meersman
24/08/2015
28/08/2015
Andrej Gubina
31/08/2015
Action
Status
Preparation of the 1st draft
Reviewed first draft
Preparation 2nd draft
Reviewed 2nd draft
Final version
Comments
This deliverable report corresponds to Task 3.3 – ‘Development of a toolset capable to
access jointly the power and the communication in the integrated platform’.
31.08.2015
2
Contents
Document info ..................................................................................................................... 2
Document History ................................................................................................................ 2
Comments............................................................................................................................ 2
List of tables ......................................................................................................................... 5
List of figures........................................................................................................................ 6
1.
2.
Introduction ............................................................................................................... 10
1.1.
Context ............................................................................................................... 10
1.2.
Goals ................................................................................................................... 10
1.3.
Report outline .................................................................................................... 11
Overview of the INCREASE simulation platform ....................................................... 13
2.1.
General architecture .......................................................................................... 14
2.2.
Tool components................................................................................................ 15
2.2.1.
Core platform .............................................................................................. 15
2.2.2.
Draw tool..................................................................................................... 16
2.2.3.
OpenDSS simulator ..................................................................................... 17
2.2.4.
JADE environment....................................................................................... 18
2.2.5.
OMNeT++ simulator .................................................................................... 18
2.3.
3.
Local Control scheme ......................................................................................... 18
Implementation of the Overlaying Control ............................................................... 20
3.1.
OLTC control algorithm ...................................................................................... 20
3.1.1.
Problem formulation................................................................................... 20
3.1.2.
Proposed methodology............................................................................... 21
3.2.
Congestion management algorithm .................................................................. 22
3.2.1.
Introduction ................................................................................................ 22
3.2.2.
Proposed methodology............................................................................... 22
3.3.
FPS control algorithm ......................................................................................... 24
3.4.
Integration with Local Control ........................................................................... 25
3.5.
Actual implementation....................................................................................... 26
31.08.2015
3
3.5.1.
3.6.
4.
6.
Tasks performed by JADE and Core components .............................................. 28
Incorporation of LAN simulator component ............................................................. 30
4.1.
Introduction........................................................................................................ 30
4.2.
Communication network simulation.................................................................. 30
4.2.1.
Network Simulator (ns-2 and ns-3) ............................................................. 30
4.2.2.
OMNeT++ .................................................................................................... 31
4.2.3.
NeSSi ........................................................................................................... 31
4.2.4.
OPNET Modeller.......................................................................................... 32
4.2.5.
Discussion.................................................................................................... 32
4.3.
5.
JADE and Core components ........................................................................ 26
Overview of OMNeT++ and INET ....................................................................... 32
4.3.1.
Introduction ................................................................................................ 32
4.3.2.
OMNeT++/INET simulator configuration .................................................... 33
4.4.
INCREASE modules ............................................................................................. 35
4.5.
Interface with the INCREASE simulation platform ............................................. 36
Simulation results ...................................................................................................... 40
5.1.
Normal Case ....................................................................................................... 43
5.2.
Case of high DRES penetration........................................................................... 51
5.3.
Communication network performance evaluation............................................ 59
5.3.1.
Wired communication network .................................................................. 61
5.3.2.
Wireless communication network .............................................................. 65
Conclusions ................................................................................................................ 70
References ......................................................................................................................... 71
31.08.2015
4
List of tables
Table 5-1: PV units rated power ........................................................................................ 41
Table 5-2: Distribution transformer characteristics .......................................................... 42
Table 5-3: PV units rated power for the case of high DRES penetration .......................... 51
31.08.2015
5
List of figures
Fig. 2.1: Overview of the INCREASE simulation platform. ................................................. 15
Fig. 2.2: Main window of draw library ............................................................................... 16
Fig. 2.3: Indicative GUI of agent......................................................................................... 17
Fig. 2.4: Typical droop curve of controllable DG units ...................................................... 18
Fig. 3.1: LV network with conflict on voltage regulation................................................... 21
Fig. 3.2: The 15-min timeslot concept ............................................................................... 26
Fig. 3.3: Conceptual implementation of Overlaying Control ............................................. 27
Fig. 3.4: Simulation flowchart ............................................................................................ 29
Fig. 4.1: Simple and compound modules .......................................................................... 34
Fig. 4.2: TCP/IP 5-layer protocol stack ............................................................................... 35
Fig. 4.3: High-level view on the implemented INCREASE modules on top of OMNeT++ and
INET .......................................................................................................................................... 36
Fig. 4.4: Interface between power and communication network simulators................... 37
Fig. 4.5: Visual overview of the exchanged messages on application level between a
regular and an aggregator agent.............................................................................................. 38
Fig. 4.6: Visual overview of the exhanged messages on application, TCP and IP level
between a regular and an aggregator agent............................................................................ 38
Fig. 4.7: Example graph for the measured throughput for a setup with 10 regular agents,
one router and an aggregator agent ........................................................................................ 39
Fig. 5.1: Elektro Gorenjska modified network topology.................................................... 41
Fig. 5.2: Active power profile of the aggregated load vs. time ......................................... 42
Fig. 5.3 Reactive power profile of the aggregated load vs. time ....................................... 43
Fig. 5.4: Injected active power of PV units vs. time........................................................... 44
Fig. 5.5: Injected active power of PV units from 10:00 to 16:00 ....................................... 44
Fig. 5.6: Curtailed active power of PV units vs. time ......................................................... 45
Fig. 5.7: Curtailed active power of PV units from 10:00 to 16:00 ..................................... 45
Fig. 5.8: Network losses vs. time ....................................................................................... 46
Fig. 5.9: Network losses vs. time from 10:00 to 16:00 ...................................................... 46
Fig. 5.10: Voltage of the network vs. time......................................................................... 47
Fig. 5.11: Tap setting of the transformer vs. time ............................................................. 48
Fig. 5.12: Daily energy losses ............................................................................................. 48
Fig. 5.13: Daily energy production ..................................................................................... 49
Fig. 5.14: Total injected active power of PV units along a single feeder........................... 50
Fig. 5.15: Positive-sequence voltage profile along a single feeder ................................... 51
Fig. 5.16: Transformer total apparent power vs. time ...................................................... 53
Fig. 5.17: Transformer total apparent power from 10:00 to 16:00 .................................. 53
31.08.2015
6
Fig. 5.18: Injected active power of PV units vs. time......................................................... 54
Fig. 5.19: Injected active power of PV units from 10:00 to 16:00 ..................................... 55
Fig. 5.20: Curtailed active power of PV units vs. time ....................................................... 55
Fig. 5.21: Curtailed active power of PV units from 10:00 to 16:00 ................................... 56
Fig. 5.22: Daily energy production ..................................................................................... 56
Fig. 5.23: Daily energy losses ............................................................................................. 57
Fig. 5.24: Network losses vs. time ..................................................................................... 57
Fig. 5.25: Voltage of the network vs. time......................................................................... 58
Fig. 5.26: Tap setting of the transformer vs. time ............................................................. 59
Fig. 5.27: Overview of the examined grid using the GUI interface of the INCREASE
simulation platform .................................................................................................................. 60
Fig. 5.28: Overview of the simulated wired communication network .............................. 62
Fig. 5.29: Throughput per link and direction ..................................................................... 63
Fig. 5.30: Channel utilization for link between router and aggregator agent ................... 63
Fig. 5.31: End-to-end delays .............................................................................................. 65
Fig. 5.32: Overview of the locations of the simulated agents ........................................... 66
Fig. 5.33: End-to-end delays via TCP measured at 2 regular agents located at the closest
and furthest from the aggregator agent .................................................................................. 67
Fig. 5.34: End-to-end delays via TCP measured at the aggregator agent ......................... 67
Fig. 5.35: End-to-end delays via UDP measured at 2 regular agents located at the closes
and furthest from the aggregator agent .................................................................................. 68
Fig. 5.36: End-to-end delays via UDP measured at the aggregator agent ........................ 68
31.08.2015
7
List of symbols and abbreviations
Symbol
ai
m
c curt
Description
Symbol
Coefficient for fair distribution of PGi
curtailed active power to all DRESs
along the feeder
Curtailment coefficient
PLtot
transf
Pcurt
g1
Curtailed active power of the DRES
i
with the minimum PMPP
of the mth feeder
Change in voltage magnitude of
the i-th connection point
Dead-band of the fixed step width
of tap changer
Fundamental input conductance
gd
Damping conductance
tot
Ploss
Total active power losses
Qtot
Srated
PLi
Absorbed active power of the Load
at the i-th connection point
Maximum power point active
power
∆Ρ
∆ Vgi
|ΔVtap|
PMPP
Pinj
Injected active power
PGtot
Total DRES generation of the
network
m
Pcurt
Active power curtailment of the mth feeder
Total DRES MPP power at the m-th
feeder
Active power curtailment of the mth feeder caused by the Local
Control
Final curtailed active power of the
DRES at the i-th connection point
Reference power of the control
m
PMPP
PLCm,curt
i
Pcurt
Pref
31.08.2015
tot
PMPP
m
PFPS
, curt
m
PCong
, curt
Description
Generated active power of the
DRES at the i-th connection point
Total active power consumption of
Load
Total curtailed active power
according to the congestion
algorithm
Total amount of DRES MPP
generation
FPS actual curtailment of the m-th
feeder
Actual active power curtailment of
the m-th feeder
Transformer reactive power
Apparent rated power of the
transformer
transf
Cumulative apparent flow of the
S
transformer
ʋmin,
Minimum allowable voltage in the
EN
grid as defined
by the
Vlow
corresponding Standards
ʋ0, ʋ1, zero-, positive- and the negativesequence components of the grid
ʋ2
voltage
ʋmax
Maximum allowable voltage in the
grid as defined
by the
corresponding Standards
υcpb
Constant power band voltage
ʋg
Voltage at the connection point
Vg
Voltage at the PCC
Vmin
Minimum voltage along a feeder
w
Coefficient vector
8
i
MPP
P
D
DG
DRES
DSO
DVB
EPRI
FIPA
FPS
GUI
JADE
LF
LV
inverter
Maximum power point active
power of the DRES at the i-th
connection point
Deliverable
Distributed Generation
Distributed Renewable Energy
Source
Distribution System Operator
Desired Voltage-Bandwidth
Electric Power Research Institute
Foundation for Intelligent Physical
Agents
Fair Power Sharing algorithm
Graphical User Interface
JAVA
Agent
Development
Framework
Local and FPS
Low-Voltage
31.08.2015
xij
Elements of the sensitivity matrix
MAS
MV
MPP
Multi-Agent System
Medium-Voltage
Maximum Power Point
OL
OLF
OLFC
OLTC
OLTC and Local
OLTC, Local and FPS
OLTC, Local, FPS and Congestion
On-Load Tap Changer
PCC
PLC
pu
Point of Common Coupling
Power Line Communication
per unit
TSO
WP
Transmission System Operator
Work Package
9
1.
Introduction
1.1. Context
The INCREASE project aims to manage distributed renewable energy sources (DRES) in
low- (LV) and medium-voltage (MV) networks. This is succeeded by providing ancillary
services, namely voltage control and provision of reserve, towards distribution (DSOs) and
transmission system operators (TSOs). The cornerstone of the project activities is the
introduction of three-phase four-wire inverter-interfaced DRES that provides high flexibility
and advanced features, further supported by an intelligent multi-agent-based control system
with enhanced structure and algorithms. INCREASE focuses on providing solutions for
operational problems in power systems, allowing increased penetration of DRES as well as
providing advanced technological solutions and intelligent control strategies for the
prosumers.
Under the INCREASE framework (WP 3), an integrated simulation platform is developed,
enabling the design, analysis, and optimization of the developed solutions. This simulation
platform is a valuable tool for DSOs in order to investigate the performance of DRES in their
distribution grids, implementing either the proposed INCREASE solutions or any other
control scheme. The simulation platform is developed using existing open-source software
and includes the following major features:
•
•
•
•
Simulation of the distribution system (both LV and MV networks) with the presence
of unbalanced loads and generation.
Integration of the locally controlled, inverter-interfaced, distributed generation (DG)
units, i.e. of the Local Control scheme for overvoltage and voltage unbalance
mitigation.
Incorporation of a Multi-Agent System (MAS) taking into account the multi-objective
Overlaying and Scheduling Control algorithms.
Implementation of a communication network simulator for the evaluation of the
existing infrastructure and the communication requirements for the MAS control
system.
The software platform architecture also allows the integration of other external and
independent software modules, such as forecasting algorithms, demand side management
and demand response simulation modules, as well as constraint optimization tools.
1.2. Goals
In this report, the incorporation of the MAS-based Overlaying Control algorithm and of
the communication network simulator in the INCREASE simulation platform is thoroughly
31.08.2015
10
described. The MAS-based multi-objective algorithm is developed to cope with overvoltages
and congestion problems, which are significant technical issues arising when integrating
multiple DRES units in the distribution grid. The LAN simulator component aims to enhance
the features of the INCREASE simulation platform, by focusing on the evaluation of the
communication infrastructure and of the requirements posed by the MAS control system
and the implemented control algorithms.
First, all features of the Overlaying Control are presented, including the transformer onload tap-changer (OLTC) algorithm, the congestion control technique, and the fair power
sharing (FPS) algorithm presented in Deliverable D3.2. The combination of Local and
Overlaying Control is also discussed, presenting the consecutive implementation of both
control schemes and the update of the various control parameters in each simulation time
interval. Furthermore, the actual incorporation of MAS in JADE environment is analyzed,
highlighting the bi-directional communication between the Core and JADE components of
the INCREASE simulation platform.
Next, the LAN simulator component included in the INCREASE simulation platform is
presented. The general framework and the different simulation programs are described,
while the simulation functionalities and features of the open-source OMNeT++ software are
further analyzed. The necessary operational functions and their implementation in the
INCREASE simulation platform are thoroughly discussed, while the actual integration is
presented, focusing on the interface between GUI, Core and LAN components.
The performance of the full Overlaying Control is demonstrated on a selected pilot
installation and is compared to the cases of no DRES power curtailment as well as to the
Local Control scheme. Results show the efficiency of the proposed algorithm in mitigating
overvoltages by changing the OLTC state and enforcing a fair contribution of the curtailed
power among the installed PV inverters, while maintaining low levels of power losses in the
distribution network. Moreover, the efficiency of the congestion control technique is also
examined, assuming a high DRES penetration level. Finally, a performance evaluation of the
communication layer is conducted in the examined pilot installation using the LAN simulator
component, while assuming different wired and wireless communication technologies.
1.3. Report outline
This introductory chapter is followed by:
Chapter 2: Overview of the INCREASE simulation platform. The overall structure and
the individual components of the INCREASE simulation platform are presented. The
overview mainly focuses on the interface between the Core and JADE components as well as
31.08.2015
11
between GUI, Core and LAN components. This section also briefly describes the Local Control
scheme, originally presented in D3.1.
Chapter 3: Implementation of the Overlaying Control. The developed full multiobjective Overlaying Control scheme is incorporated in the INCREASE simulation platform,
featuring the OLTC control algorithm, the congestion management technique and the FPS
among the connected DRES. The integrated model is generalized, introducing proper
weighting factors in order to be also applicable to feeders and PV inverters with arbitrary
characteristics and nameplate ratings.
Chapter 4: Incorporation of LAN simulator component. The integration of the LAN
simulator in the INCREASE simulation platform is described with special emphasis on the
necessary functions to investigate the performance of the communication system,
supporting the MAS-based control. Moreover, the interface between GUI, Core and LAN
components is described.
Chapter 5: Simulation results. Simulation results from a pilot installation are presented,
revealing the effectiveness of the full multi-objective Overlaying Control under normal or
high PV penetration conditions. Results are compared with other control schemes, whereas
a detailed investigation is also performed on the total active power injection and losses of
the distribution network. The communication infrastructure of the same pilot grid is also
simulated to evaluate its performance in supporting the INCREASE distributed control
scheme, while different wired and wireless technologies are investigated.
Chapter 6: Conclusion. General conclusions are summarized and a plan for the next
research steps is proposed.
31.08.2015
12
2.
Overview of the INCREASE simulation platform
The main objective of the INCREASE simulation platform is to simulate and analyze
MV/LV electrical power grids, including inverter-interfaced DRES, different types of loads as
well as the implementation of various control schemes for the DRESs. The INCREASE
simulation platform is able to handle unbalanced AC power flow calculations, using the
phase-domain approach [1], as well as profile-based power flow calculations. Due to the
required analysis over extended observation times, a quasi-dynamic solution has been
selected and implemented instead of a dynamic simulation model. The quasi-dynamic
solution is based on sequential steady-state power flow calculations over a user-defined
time span with a variable time step. This approach is preferred over detailed dynamic
simulations, since it can provide a better insight into system steady-state conditions while it
is also very efficient numerically, requiring significantly smaller execution times [2]. A
detailed description and evaluation of the INCREASE simulation platform can be found in
Deliverable D3.1.
In the INCREASE project framework, three distinct control schemes are proposed and
developed, namely Local, Overlaying and Scheduling Control. Each one of these control
schemes is fully integrated in the INCREASE simulation platform. The incorporated INCREASE
control schemes can be summarized in the following:
•
•
•
The first level control, denoted as Local Control, is a low-level control applied to
controllable DG units via grid-interfaced inverters. Its main objective is to perform
voltage control in low voltage networks by mitigating overvoltages and voltage
unbalances [3]-[7]. The Local Control works continuously, adjusting the total
amount of the inverter active power output and its distribution among the three
phases according to the voltage at the Point of Common Coupling (PCC).
The second level control, characterized as Overlaying Control, is related to the MAS
coordination algorithms for the application of specific control strategies and is
focused on voltage control and transformer congestion management [8], [9]. It is
an event-driven control, interacting with the Local Control system by varying its
operational parameters, such as the slopes of the inverter droop curves and the
corresponding set-points, as well as the MV/LV transformer OLTC settings.
Generally, the Overlaying Control is addressed to all controlled DRES in the
examined network.
The third level control, named as Scheduling Control, addresses problems on a
longer time scale, which varies from minutes to days. This control focuses on the
optimization of the grid performance according to predefined criteria and
31.08.2015
13
constraints, which include load and generation forecasts, demand response, as well
as possible energy market inputs [10]. Scheduling Control is also implemented
through the MAS.
Finally, in the INCREASE simulation platform a discrete event simulator of the
communication infrastructure is employed. This simulator is used to evaluate the
communication performance of the MAS control system and to analyze possible
contingencies of the communications in the operation of the MAS control system.
Furthermore, it can be used to investigate alternative options on the design of the necessary
infrastructure, and to examine the communication system vulnerability and its risks on the
control system performance.
2.1. General architecture
An overview of the INCREASE simulation platform is presented in Fig. 2.1. The developed
software comprises different open-source tool components and their mutual
interconnections. More, specifically, the INCREASE simulation platform includes:
•
•
•
•
•
The Core, which is the base of the simulation platform. The Core controls the
interaction of the different INCREASE platform components. It is also the base for
the implementation of the algorithms of the proposed Local Control.
The Draw tool, which is a graphical pre-processor with design capabilities to allow
the user-friendly input and configuration of the distribution or transmission
network under investigation.
The OpenDSS software [11], which is a phasor-domain grid simulator, capable of
handling unbalanced power flow problems, and also allowing the development and
implementation of the necessary power system component models.
The JADE software [12], which is the tool integrating the MAS and the
corresponding communication in the INCREASE platform for the implementation of
both the Overlaying and Scheduling Control algorithms.
The OMNeT++ simulator [13], which is used for the analysis and the evaluation of
the communication infrastructure of the examined network.
31.08.2015
14
Draw
JADE
Network Designing Tool
Agents’ Environment
1-w
ay
ay
2-w
Core
ay
1 -w
2-w
ay
OMNeT++
OpenDSS
LAN Simulation
Framework
Distribution System
Simulator
Fig. 2.1: Overview of the INCREASE simulation platform.
2.2. Tool components
In this section a brief description of the components incorporated in the INCREASE
simulation platform is presented. A more detailed analysis of all required tools can be found
in D3.1.
2.2.1. Core platform
The core of the platform is developed in MATLAB, which is a high-level language,
suitable for numerical computations, visualization and programming [14]. In the developed
platform the initial GUI windows defining the main simulation parameters as well as the
post-processing tools for the results are implemented within the Core component.
Furthermore, the Core is responsible for the interconnections between the different
components of the developed platform. More specifically, as shown in Fig. 2.1, four distinct
interconnections among the key elements of the INCREASE simulation platform are
implemented. These are:
•
•
A one-way interconnection between the Core and the Draw tool. This interaction is
used for transferring all necessary data from the Draw input pre-processor to the
appropriate variables and structures in the Core.
A two-way interconnection between Core and OpenDSS. This interconnection uses
the well-established COM interface, transmitting all necessary input files (.dss files)
to the OpenDSS simulator. Furthermore, the corresponding results from the power
31.08.2015
15
•
•
flow simulations performed in OpenDSS are transferred back to the Core via the
same connection.
A two-way interconnection between Core and JADE. This link allows the
communication of the MAS components with the core of the INCREASE simulation
platform.
A one-way interconnection between the Core and the OMNeT++. This
interconnection is responsible for the transmission of the necessary data to define
the grid and MAS structure, as well as to create the corresponding input files of the
LAN simulator (.ned and .ini files).
2.2.2. Draw tool
The graphic library for the input of all individual network components is based on
SIMULINK models [14] and contains all components defined in the toolbox of Fig 2.2. This
collection includes all basic power system components, such as loads and generation units,
as well as the more advanced models of controllable inverters and agents, which have been
presented thoroughly in D3.1. All elements work in a drag and drop environment by simply
putting them in the design area, configuring their required parameters, and making the
appropriate connections. Furthermore, all elements are accompanied by brief help
descriptions, when pressing the corresponding button. Finally, the final version of the Draw
component will also include a data conversion tool to provide basic import capabilities from
file formats used in other simulation software packages and design environments to
variables and structures compatible with the INCREASE simulation platform.
Fig. 2.2: Main window of draw library
The GUI for the agent component is shown in Fig. 2.3. By selecting the corresponding
checkboxes, result reports regarding voltages, currents, active and reactive power can be
easily obtained from the power flow simulations performed in OpenDSS. Furthermore,
compared to previous versions of the Draw tool, two new data fields are incorporated in the
31.08.2015
16
GUI of the agent. In the first field, the user is able to define if the connection between the
agents will be wired (power line communication - PLC) or wireless. In the second field, the
relative to the aggregator agent geographical coordinates of each agent are inserted,
allowing the simulation of wireless connections.
Fig. 2.3: Indicative GUI of agent
2.2.3. OpenDSS simulator
The OpenDSS simulation tool [11], [15] has been chosen as the power flow solver within
the INCREASE simulation platform. OpenDSS is a comprehensive, open-source simulation
tool for power distribution systems, developed and distributed by the Electric Power
Research Institute (EPRI). OpenDSS provides highly accurate results, remarkable numerical
performance and vast communication abilities with external programs. Furthermore, the
most common power system analysis algorithms for both steady-state and dynamic analysis
are incorporated in the OpenDSS. It also includes various quasi-dynamic solution modes,
such as snapshot-daily-yearly power flows and harmonic analysis, making it ideal for
sequential time simulations. The time period can be arbitrary selected, whereas the user
may also implement external macros to drive the load models in any arbitrary manner. The
results of a power flow generally include bus voltages, branch currents, grid losses and other
31.08.2015
17
information available for the total system, for each component, and for certain defined
areas.
2.2.4. JADE environment
A detailed presentation of the JADE environment is given in Section 3.
2.2.5. OMNeT++ simulator
A thorough description of OMNeT++ simulator is given in Section 4.
2.3. Local Control scheme
The Local Control scheme is a low-level control applied to controllable DG units via gridinterfaced inverters. Its main objective is to perform voltage control in low voltage networks
by using exclusively local parameters measured at each inverter PCC, such as the grid voltage
and the available power, offering the ability to immediately react on grid disturbances and
ensuring the safe operation of the distribution grid.
The proposed Local Control incorporates two distinct basic control features, namely the
droop control of the injected active power and the voltage unbalance mitigation strategy.
Since, in low voltage distribution grids the R/X ratio is relatively high as distribution lines
have mainly resistive characteristics, voltage control can be accomplished more efficiently by
controlling the active power output of the DRESs connected to the LV network. Thus, the
droop control curtails the active power of DG units, avoiding unacceptable overvoltages
along the distribution feeders. The droop control is based on the voltage at the point of
common coupling (PCC) υg , while a typical droop curve is depicted in Fig. 2.4.
Fig. 2.4: Typical droop curve of controllable DG units
Voltages υcpb and υ max are the thresholds for the activation of active power curtailment
and for the complete cut-off of power injection, respectively. Additionally, υmin is the
inverter minimum operating voltage, whereas Pref is the maximum active power the inverter
31.08.2015
18
can deliver at a specific time instant, defined from the available power of the primary
source.
In the second integrated control scheme of the Local Control, inverters mitigate voltage
unbalance by injecting zero- and negative-sequence currents proportionally to the zero- and
negative-sequence voltages at the PCC, respectively. A proportionality term is introduced,
called damping conductance gd , which results in a resistive behaviour of the inverter
towards the zero- and negative-sequence components of the grid voltage. The injected
currents in symmetrical components are calculated according to the following equation [3][7],
 i0   gd
  
 i1  =  0
 i2   0
0
g1
0
0  υ 0 
0  ⋅ υ1 
gd  υ2 
(2.1)
where υ0 , υ1 , υ2 are the zero-, positive- and negative-sequence components of the voltage
at the PCC, and g1 refers to the injected active power and is the fundamental conductance
of the inverter having an opposite sign of gd in case of generation.
A detailed presentation of the Local Control is available in D2.4. Further results
considering the performance and the effectiveness of the proposed Local Control scheme
are presented in D3.1.
31.08.2015
19
3.
Implementation of the Overlaying Control
A significant drawback of the Local Control is that the active power curtailment is mainly
observed in the DRES units connected close to the end of the radial LV feeders. The
Overlaying Control and especially the OLTC control algorithm acts complementary to the
Local Control, adjusting the voltage level of the network and thus reducing the total curtailed
power of the DRES. Furthermore, the FPS algorithm of the Overlaying Control is responsible
to redistribute the curtailed active power in a fair way among the DRES, while the
Congestion management acts supplementary to the other control schemes in order to
mitigate any transformer congestion.
3.1. OLTC control algorithm
In principle, the MV/LV transformers are equipped with off-load tap changers, which can
only be adjusted off-line during the installation or after a topology change in the network.
However, it is expected that future distribution networks will be equipped with OLTC-based
MV/LV distribution transformers to independently adjust the voltage level of LV networks
during their operation without affecting the corresponding level of the MV network. The
OLTC is capable of adjusting the secondary voltage on-line by controlling a tap changer on
the MV side, thus mitigating voltage fluctuations due to DRES [16], [17]. The incorporated
automatic voltage regulator compares the network voltages with a reference voltage,
adjusting the taps of the transformer when necessary. This is usually done in a predefined
number of steps up and down the rated voltage, and with a predefined voltage change rate
for each step. In the following text the term ‘tap down’ implies a voltage reduction at the LV
side by the pre-defined tap changers step rate, while ‘tap up’ implies a voltage increase at
the LV side, irrelevant of the actual OLTC manufacturer definitions.
3.1.1. Problem formulation
The MV/LV distribution network of Fig. 3.1 is considered, where several feeders are
connected. For simplicity reasons and without limiting the generalized application of the
methodology it is also assumed that feeder k explicitly hosts a number of DRES, while feeder
l contains only loads. In this case, the former feeder is mainly characterized by a voltage rise
when the DRES operate at their peak generation capacities, while in the latter one a voltage
drop is generally observed due to the presence of passive loads. If the OLTC is tapped down
to mitigate the voltage rise problem, e.g. controlling the transformer voltage ratio to lower
the LV side voltage for steady feeding voltage from the MV, a severe undervoltage situation
EN
may be observed in feeder l, violating the low voltage limit Vlow
according to the EN 50160
standard. As a result, a coordinated voltage regulation among DRES and OLTC must be
31.08.2015
20
developed, maintaining the voltage level of all feeders in the desired voltage-bandwidth
(DVB).
Fig. 3.1: LV network with conflict on voltage regulation
3.1.2. Proposed methodology
Since feeder k includes n DRES, it is expected that this feeder experiences the maximum
voltage among the feeders. The DRES voltages at the PCC and the corresponding droopcontrol thresholds can be grouped in vectors Vg and Vcpb , respectively. Meanwhile, feeder l
experiences the minimum voltage Vm in , which is also the lowest among the feeders due to
the sole presence of passive loads. In case of mitigating the voltage rise or drop in these two
feeders, the voltage levels of the remaining feeders definitively remain in the desired DVB,
which is determined by the difference between the maximum element of Vg and Vm in .
The operational procedure of the proposed OLTC control can be summarized in the
following situations, assuming a certain dead-band ∆Vtap due to the fixed step width of the
tap changer:
•
Situation #1a - No tap action:
Vg ≤ Vcpb
and
EN
Vm in ≥ Vlow
•
Situation #1b - No tap action:
Vg > Vcpb
and
EN
Vmin − ∆Vtap < Vlow
•
Situation #2 - Tap up:
EN
Vm in < Vlow
•
Situation #3 - Tap down:
Vg > Vcpb
and
EN
Vmin − ∆Vtap ≥ Vlow
31.08.2015
21
In more detail, if all voltages remain within DVB, then Situation #1a is met and neither
OLTC nor DRES curtailment is activated. In Situation #1b, although there is active power
curtailment due to the activation of the Local Control, no tap action is performed since it
EN
. In case of undervoltage, Situation #2 is met and a tap up
would lead to the violation of Vlow
action is performed, until the problem is mitigated. Finally, the tap down action of
EN
Situation #3 is activated when both the Local Control is activated and no violation of Vlow
is
observed.
In general, the DRES power generation is maximised using the tap down action of the
OLTC, while the DRES power generation is curtailed only when the OLTC cannot tap down
because of a lower voltage in another feeder. This control is activated in an iterative way to
determine the exact required number of tap actions, while the execution is performed after
a user-defined time delay and only after the confirmation that the voltage issue persists.
3.2. Congestion management algorithm
3.2.1. Introduction
In distribution networks with high penetration of DRES, extremely high reverse power
flows may lead to operational points for network elements beyond their physical capacity
limits, and thus congestions may occur. The most congestion vulnerable network elements
in a distribution grid with high DRES penetration at the LV level are the MV/LV transformers
[18], [19]. As a result, DSOs with a high share of DRES in their networks may face challenges
in maintaining the reliability of the network. These challenges are expected to become more
frequent, depending on the different types of connected resources, their geographic
location and the voltage level of the connection.
One aspect of the Overlaying Control aims to deal with the impact of high DRES
generation into the loading conditions of the distribution network and to treat effectively
the resulting congestion issues. For this purpose, the implemented MAS-based active power
curtailment mechanism is modified to also tackle the potential congestion of the MV/LV
transformers in LV networks. This curtailment strategy is thoroughly investigated and
successfully combined with the FPS algorithm of D3.2, aiming to fairly curtail the active
power of the connected DRES in terms of their feed-in capacity in order to solve the
congestion issue.
3.2.2. Proposed methodology
Given that PGi is the generated DRES active power at the i-th connection point, the total
DRES generation PGtot of the network is given by:
31.08.2015
22
N
PGtot = ∑PGi
(3.1)
i =1
where N is the total number of connection points in the network. Similarly, the total load
PLtot is calculated as:
N
P = ∑PLi
tot
L
(3.2)
i =1
tot
whereas the corresponding line and cable total active power losses are denoted by Ploss
.
In each bus of the grid, reverse power flow occurs when the local generation PGi exceeds
the local load PLi . This reverse power flow can cause transformer overload when the
cumulative flow exceeds the transformer rating Srated :
Stransf =
(P
tot
G
− PLtot − Plostots ) + ( Qtot ) > Srated
2
2
(3.3)
tot
where Q corresponds to the transformer reactive power. In this case, congestion occurs
and the necessary amount of total active power curtailment is approximately calculated by:
Pctransf
≅ S transf − Srated
urt
(3.4)
The proposed methodology aims to distribute the required curtailment among all the
feeders on the basis of a fair curtailment mechanism. This is performed by assigning a
m
suitable curtailment coefficient c curt
for each feeder m, based on the individual total feeder
DRES generation on MPP conditions. The MPP value is prefered over the installed capacity,
since this is the maximum power that a PV unit can generate in any moment given the
radiation level. The coefficient is defined as the ratio of the total DRES MPP power at m-th
m
tot
to the total amount of DRES MPP generation in the network PMPP
:
feeder PMPP
ccmurt =
m
PMP
P
tot
PMPP
(3.5)
Thus, the total required active power curtailment can be distributed evenly among the
feeders by multiplying it with the corresponding weighing coefficients. Therefore, the
m
is given by:
curtailment at each feeder Pcurt
31.08.2015
23
m
m
transf
Pcurt
= ccur
t ⋅ Pcurt
(3.6)
Finally, the actual active power curtailment of the m-th feeder, with reference on its MPP
m
conditions PCong
, is calculated in (3.7) by taking into account the corresponding feeder
, curt
curtailment caused by the Local Control on the DRES connected to the m-th feeder PLCm,curt :
m
m
m
PCong
,curt = Pcurt + PLC ,curt
(3.7)
3.3. FPS control algorithm
The last goal of the Overlaying Control is to obtain a fair power curtailment among the
DRES connected to a radial feeder of the LV network, while mitigating the overvoltage in the
feeder. This is thoroughly presented in D3.2, where the sensitivity matrix methodology is
used in order to uniformly curtail the active power of DRES.
According to the FPS formulation, the curtailed active power ∆Ρ of the DRES with the
i
minimum PMPP
on the m-th feeder is given by (3.8), while the active power curtailments of
all the other DRES in this feeder are given by (3.9):
∆Ρ =
∆ Vgi
(3.8)
m
∑w ⋅ x
i
ji
i =1
PFPS ,curt = ∆Ρ ⋅ w
(3.9)
Here, ∆ Vgi is the change in the voltage magnitude of the i-th connection point, x ji are the
corresponding elements of the sensitivity matrix, and w the coefficient vector related to the
MPP conditions. By summing the elements of (3.9), the FPS actual curtailment of the m-th
feeder is acquired with reference on its MPP conditions:
m
; for the m-th feeder
PFPS
,curt = ∑ PFPS ,curt
(3.10)
The combination of the Congestion and FPS control algorithms is achieved by comparing
the actual curtail values of (3.7) and (3.10), which correspond to the m-th feeder. The largest
value is selected as the final power curtailment for the m-th feeder, yielding voltages on the
safe side, while the total amount of curtailed active power is fairly distributed to all DRES
connected to this feeder. This is calculated in (3.11) for the DRES at the i-th connection
point:
31.08.2015
24
i
m
m
Pcurt
= a i ⋅ max ( PCong
, curt , PFPS ,curt )
(3.11)
In (3.11) the weighting factor a i is given by:
ai =
i
PMPP
m
PMPP
(3.12)
The curtailed active power components of (3.11) lead to new power injections for all
inverters, thus a new power flow solution is performed to calculate the corresponding
voltages at the inverter PCCs. From the acquired set-points, the droop characteristics of all
inverters are then reconfigured as proposed in D3.2.
3.4. Integration with Local Control
The integration concept of the Local and Overlaying Control is presented in Fig. 3.2,
where a detail of an arbitrary selected 15-min timeslot is also shown. Since the Local Control
is designed to be embedded at the hardware level of the inverter, it is continuously active
during the 15-min timeslot. After the first 5 min, the MAS component of the INCREASE
platform detects any possible voltage issue and calculates the new tap setting of the
transformer OLTC according to Section 3.1. Then all the data considering the voltages at PCC,
the inverter power injections and the net power flows are monitored and communicated to
the MAS layer. In case of detecting any congestion issue or unfairness in curtailing the DRES
injections, the combined Congestion and FPS algorithm is activated according to Sections 3.2
and 3.3, and the new reference control signals are sent to each inverter. Then, the droop
curves are reconfigured and this state retains until the end of the 15-min timeslot, where the
droop curves reset and the loads as well as the MPP inverter values are updated to new
values. At this point, the OLTC control algorithm executes the tap change after checking
whether the voltage issue persists. The length of the specific time slot is fully flexible and can
be adjusted according to the sampling rates and all other response times of the actual grid
elements and of the monitoring system implemented in each specific distribution grid.
31.08.2015
25
OLTC
t
OLTC
t+15 min
OLTC
t+45 min
t+30 min
Local Control
FPS & Congestion
FPS & Congestion
FPS & Congestion
Detection of Voltage Issue
Calculate New Tap Setting
t+15 min
t+30 min
Local Control
FPS & Congestion
Reset
Load, MPP
Check for OLTC
activation
OLTC
Fig. 3.2: The 15-min timeslot concept
3.5. Actual implementation
3.5.1. JADE and Core components
JADE (Java Agent Development Framework) is a software development framework
aimed at developing MAS and applications conforming to FIPA standards for intelligent
agents. JADE is written in an object-oriented programming language, namely JAVA, because
of the many attractive features it provides.
In the framework of the INCREASE simulation platform, JADE is employed as a MAS
developing environment where the Overlaying Control is implemented. More specifically,
the code referring to the Overlaying Control is actually written in JADE, making it an essential
component of the simulation platform. The link of the Core module, which is implemented in
MATLAB, with JADE is done through TCP/IP communication. Considering a specific time
31.08.2015
26
instant, the Overlaying Control procedure is depicted in Fig. 3.3 comprising 6 steps which are
analyzed as follows:
Fig. 3.3: Conceptual implementation of Overlaying Control
Step 1: The results of the Local Control regarding voltages, injected active power of DRES
etc., are forwarded to the Core module of the INCREASE simulation platform. Then,
the Core module checks whether the Overlaying Control is activated. In the case of
DRES with curtailed active power, the FPS algorithm is launched, whereas the
Congestion and OLTC algorithms are used to address transformer overloading and
voltage violation issues, respectively. Otherwise, the procedure moves to Step 6 and
the Overlaying Control remains deactivated.
Step 2: This step is implemented via a TCP/IP communication channel, established between
Core module and JADE. The data transferred to JADE refer to the network
configuration, the outputs from the Local Control, i.e. voltages, loads, injected active
power of the DG units, etc., and the activated algorithms of the Overlaying Control.
Step 3: Considering the aforementioned Congestion and FPS algorithms, two auxiliary power
flow calculations are necessary. In the first power flow solution, each DG unit is
assumed to inject its maximum available power, i.e. MPP operation, whereas in the
second it injects the active power defined by the Congestion or FPS algorithm. The
power flow calculations are implemented in MATLAB and OpenDSS but are initiated
by the MAS module established in JADE.
Step 4: The results obtained from the power flow calculations are transferred to JADE. Then,
the combined Congestion and FPS algorithm calculates the sensitivity matrix and the
new droop curves based on the first and the second power flow, respectively,
whereas the OLTC algorithm determines the new setting of the transformer on-load
tap changer.
31.08.2015
27
Step 5: The new droop curves of the DG units and the new transformer on-load tap setting
are forwarded back to the Core module via another TCP/IP communication channel.
Step 6: Finally, the droop curves of the DG units and the tap settings are passed to the Local
Control, moving one time step forward and initiating a new iteration.
3.6. Tasks performed by JADE and Core components
The complete simulation flowchart is depicted in Fig. 3.4, including the coordination of
the Local and Overlaying Control, based on the 15-minute timeslot concept of Section 3.4.
Initially, the network data referring to the consumption, generation and network
configuration are loaded. Then, a part of the OLTC algorithm of the Overlaying Control is
executed. An internal Local Control is initiated and the Core module checks whether the tap
setting should change in the new value. Next, the Local Control for the first five minutes is
executed.
The results obtained from the Local Control are forwarded to the Core module, which
first checks whether a new setting of the on-load tap changer is needed, according to the
OLTC algorithm of Section 3.1, which will be forwarded at the next 15-min timeslot. Then,
the Core module compares the calculated apparent power of each transformer with the
corresponding rated one. In case a violation is observed, the Congestion algorithm is
activated and executed in JADE by calculating the required amount of active power to be
curtailed, as described in Section 3.2. Then, depending on the condition that there are DG
units operating in the droop region, the Core module activates the FPS algorithm of Section
3.3, which calculates the new injected active power for each DG unit.
Finally, the new droop curves of the DG units, derived from the combined operation of
Congestion and FPS algorithms are used as inputs to the Local Control, corresponding to the
last ten minutes of the timeslot. The procedure is repeated until the end of the pre-defined
total simulation time.
31.08.2015
28
Start of Simulation
(t = 0)
Overlaying Control
Load network data of the
timeslot (consumption,
generation, etc.)
Internal Local Control
Check for the tap setting
change
Local Control
(first 5 minutes)
t = t + 15
YES
Check for OLTC
activation
OLTC control scheme
(new tap setting for the
next timeslot)
Overlaying Control
NO
YES
Stransf > Srated
Congestion management
scheme
NO
YES
Vg > Vcpb
FPS control scheme
NO
Local Control
(last 10 minutes)
NO
Core (MATLAB)
Last timeslot
YES
End of Simulation
MATLAB/OpenDSS
JADE
Fig. 3.4: Simulation flowchart
31.08.2015
29
4.
Incorporation of LAN simulator component
4.1. Introduction
The goal of the communication network simulator is to simulate the communication
traffic for the control mechanisms developed within INCREASE. The Overlaying Control
scheme is implemented as a MAS with different kinds of agents that communicate with each
other to apply the specific control strategies focused on voltage control and transformer
congestion management.
The communication between these agents can be realized using different
communication technologies. Power-line communication can be used using the existing
feeders, wireless technologies can be also deployed or other deployed telecom solutions like
DSL, coax and fiber. In all cases, it is important to analyse to what extent these technologies
can support the communication traffic associated with the Overlaying Control features in
terms of acceptable delays, needed bandwidth, reliability, etc.
In this section a short overview of the available communication network simulation
platforms is presented and the OMNET++/INET platform used in the INCREASE simulator is
further described. Then, the developed components within the simulation platform for
modeling the different kinds of agents and associated communication patterns are
described, while the interface between the communication network simulator and the
power grid simulator is also investigated.
4.2. Communication network simulation
A short overview of a number of communication network simulators is presented that
are widely used for the development and evaluation of communication architectures and
protocols, and have been used successfully in a smart grid context [20].
4.2.1. Network Simulator (ns-2 and ns-3)
The Network Simulator version 2 (ns-2) is a widely used open-source discrete event
network simulator, created for research and educational purposes. It is targeted at
networking research, with a strong focus on Internet systems. Therefore, it includes a rich
library of network models to support the simulation of e.g. IP-based applications (including
TCP, UDP, etc.), routing, multicast protocols, over wired and/or wireless networks. The ns-2
core is written in the C++ programming language. Users can create new network models or
protocols using the C++ language. Simulation scripts to control the simulation and configure
aspects such as the network topology are created using the OTcl language interface. As a
result, users can create and modify simulations without having to resort to C++
programming and recompiling ns-2.
31.08.2015
30
Development of ns-3, the successor to ns-2, is ongoing. New features include support for
the Python programming language as a scripting interface instead of OTcl, improved
scalability, more attention to realism, better software integration, etc. [21]. However, when
selecting a specific version of ns, it is important to consider that ns-3 is not backwards
compatible with ns-2, i.e. existing ns-2 simulation models must implemented again for ns-3.
Both are widely used for networking research, while both ns-2 and ns-3 are also adopted
in a smart grid context, e.g. a co-simulation approach [22], [23].
4.2.2. OMNeT++
The open-source OMNeT++ discrete event simulation environment has been designed
for the simulation of communication networks (wired and wireless) and distributed systems
in general [13], [24]. The simulation environment has a general design, i.e. it is not limited to
simulating communication networks, and therefore has been used in various domains, such
as wireless network simulations, business process simulation and peer-to-peer networking.
However, OMNeT++ is mostly applied in the domain of communication network simulation.
A comprehensive set of Internet-based protocols is provided by means of the INET
framework extension [25], which includes support for IPv4, IPv6, TCP, UDP, Ethernet, and
many other protocols. Other extensions provide simulation support for mobility scenarios
(e.g. VNS), ad-hoc wireless networks (e.g. INET-MANET), wireless sensor networks (e.g.
MiXiM, Castalia), etc. Distributed parallel simulation is also supported to enable simulation
of large-scale networks. Additionally, federation support based on the High-Level
Architecture (HLA) standard is provided in OMNEST, the commercial version of OMNeT++.
An OMNeT++ simulation model consists of simple modules implemented in C++.
Compound modules consist of other simple or compound modules, and are defined using
the OMNeT++ Network Description Language (NED). Modules communicate by passing
messages via gates, which are the input and output interfaces of the modules that are linked
to each other by so-called connections, forming communication links between modules.
Apart from the networking community, OMNeT++ has also received substantial attention
from the smart grid community for developing smart grid simulators [26], [27], [28].
Examples that focus on the communication aspect of the smart grid include the design and
evaluation of different smart grid communication architectures, the performance of smart
grid protocols, etc.
4.2.3. NeSSi
NeSSi (Network Security Simulator) is an open-source discrete event network simulator
developed at DAI-Labor (Distributed Artificial Intelligence Laboratory) and sponsored by
Deutsche Telekom Laboratories. The primary focus of the tool is on network security related
scenarios in IP networks [29]. Features described to support security related scenarios are
31.08.2015
31
attack modeling, attack detection, security metrics, etc. Distributed simulation is supported
to enable simulation of large-scale networks.
4.2.4. OPNET Modeller
OPNET Modeller is a powerful commercial discrete event network simulator with builtin, validated models including LTE, WIMAX, UMTS, ZigBee, Wi-Fi, etc. It enables modeling of
various kinds of communication networks, incorporating terrain, mobility, and path-loss
characteristics in the simulation models. OPNET Modeller has a visual high-level user
interface, offering access to a large library of C and C++ source code blocks, representing the
different models and functions. It comes with an open interface for integrating external
object files, libraries, other simulators (co-simulation) and even hardware-in-the-loop.
4.2.5. Discussion
The discussed communication network simulators have been used successfully in the
context of smart grid research. OMNeT++ and ns-2/ns-3 are used extensively in academia
due to their open-source nature. In terms of supported simulation models, a wide range of
models is available for each simulator, and the choice mainly depends on prior knowledge
and preferences of the user regarding modeling language and tools, extensibility and
supported programming languages, presence of extensive GUI tools, etc. For example,
OMNeT++ and NeSSi provide an integrated development environment (IDE) that includes
GUIs for building and configuring simulation models, visualization of topologies, result
processing, etc. However, ns-2/ns-3 lacks an extensive set of GUI tools as found in
OMNeT++, making it more complex in its usage. OPNET Modeller on the other hand is a
commercial simulator that has a visual high level interface. Another aspect that may
influence the choice of simulator is the commercial support, which is available for OMNeT++
(i.e., OMNEST) and OPNET. NeSSi, also an open source simulator, distinguishes itself from
the other tools due to its primary focus being network security. Consequently, the OMNeT++
in combination with the INET framework is chosen for the INCREASE simulation platform,
mainly due to prior positive experiences with this simulator.
4.3. Overview of OMNeT++ and INET
4.3.1. Introduction
OMNeT++ is an extensible, modular, component-based C++ simulation library and
framework, primarily for building network simulators [13]. It offers an Eclipse-based IDE, a
graphical runtime environment, and a host of other tools. It runs on Windows, Linux, Mac OS
X, and other Unix-like systems. OMNeT++ provides a component architecture for models.
Components (modules) are programmed in C++, then assembled into larger components and
31.08.2015
32
models using a high-level language (NED) in order to be easily reused. OMNeT++ has
extensive GUI support, and due to its modular architecture, the simulation kernel (and
models) can be embedded into applications.
OMNeT++ provides the basic machinery and tools to write simulations, but it does not
provide itself any components specifically for computer network simulations, queuing
network simulations, system architecture simulations or any other area. Instead, these
application areas are supported by various simulation models and frameworks, such as INET
[25]. INET is an open-source model library for OMNeT++ providing protocols, agents and
other models for researchers and students working with communication networks.
Furthermore, it contains models for the Internet stack (TCP, UDP, IPv4, IPv6, OSPF, BGP,
etc.), wired and wireless link layer protocols (Ethernet, PPP, IEEE 802.11, etc), support for
mobility, MANET protocols, DiffServ, MPLS with LDP and RSVP-TE signaling, several
application models, and many other protocols and components.
INET is built around the concept of modules that communicate by message passing.
Agents and network protocols are represented by components, which can be freely
combined to form hosts, routers, switches, and other networking devices. New components
can be programmed by the user, and existing components have been written so that they
are easy to understand and modify. INET also benefits from the infrastructure provided by
OMNeT++. Beyond making use of the services provided by the OMNeT++ simulation kernel
and library (component model, parameterization, result recording, etc.), this also means that
models may be developed, assembled, parameterized, run, and their results evaluated from
the comfort of the OMNeT++ Simulation IDE, or from the command line.
4.3.2. OMNeT++/INET simulator configuration
An OMNeT++ model consists of the following parts:
31.08.2015
33
•
Modules that communicate with message passing. The active modules are
termed simple modules and they are written in C++, using the simulation class library.
Simple modules can be grouped into compound modules and so forth, thus the
number of hierarchy levels is unlimited. The whole model, called network in
OMNeT++, is itself a compound module. Messages can be sent either via connections
that span modules or directly to other modules.
Fig. 4.1: Simple and compound modules
•
•
•
NED language topology description(s) (.ned files) that describe the module structure
with parameters, gates, etc. NED files can be written using any text editor, but the
OMNeT++ IDE provides support for two-way graphical and text editing.
Message definitions (.msg files). Various message types can be defined and data
fields added to them. OMNeT++ will translate message definitions into full-fledged
C++ classes.
A configuration file (typically called omnetpp.ini). This file contains settings that
control how the simulation is executed, values for model parameters, etc. The
configuration file can also prescribe several simulation runs.
The output of the simulation is written into result files: Output vector files , output scalar
files, and possibly the user's own output files. OMNeT++ contains an Integrated
Development Environment (IDE) that provides an environment for analyzing these files.
Output files are line-oriented text files, which make it possible to process them with a variety
of tools and programming languages as well, including Matlab, GNU R, Perl, Python, and
spreadsheet programs.
31.08.2015
34
4.4. INCREASE modules
In Fig. 4.2 an overview of the 5-layer Internet protocol stack and some typical protocols
used in every layer is given. INET provides modules for most of these protocols. In order to
simulate the typical communication patterns between the INCREASE agents, extra modules
are added in the application layer.
Fig. 4.2: TCP/IP 5-layer protocol stack
The IncreaseAgent and IncreaseAggregator are implemented with small differences in
case TCP or UDP is used as transport protocol on top of OMNeT++ and INET, as shown in Fig.
4.3. The IncreaseAgent contains a number of parameters such as the time period by which
measurement messages are sent to the Aggregator agent (measurementInterval, e.g. 1
minute) and the size of such a measurement message (sendBytesMeasurement). When the
Aggregator agent sends a control message, the regular agent will send a reply message of a
certain
size
(sendBytesControlReply),
possibly
after
a
certain
delay
(replyControlMessageDelay), e.g. to update the settings of the associated PV inverter. A
specific start time of the agent can be specified (startTime) and furthermore a number of
variables are defined to collect statistics on the communication traffic.
31.08.2015
35
Fig. 4.3: High-level view on the implemented INCREASE modules on top of OMNeT++ and
INET
An OMNeT++/INET module consists of a C++ header (.h), a class file (.cc) and a NED file
which contains the different parameters that can be defined in the simulation configuration
file (.ini file) with often a default value, and a number of additional statistics that are
calculated, e.g. the end-to-end delay of the received messages.
4.5. Interface with the INCREASE simulation platform
To perform a simulation with the communication network simulator two files are
essentially needed: A NED file connecting the different modules, i.e. INCREASE agents with
the relevant internet protocols on the different layers, and a configuration file (.ini file) to
provide specific values for the different module and simulation parameters.
In Fig. 4.4 the interface between the INCREASE power simulator and the communication
network simulator is shown. The power simulator contains a GUI representation of the
power grid under study which is converted to Matlab matrices containing information for
the number and names of present Agents in the grid, the type of communication (wired or
wireless), the coordinates of the regular agents on a map and the distances between the
regular agents and the aggregator agent.
31.08.2015
36
Fig. 4.4: Interface between power and communication network simulators
A Matlab parsing script is developed to transform these matrices into NED and INI files,
which can be used by the communication network simulator. Using these files an actual
simulation of the communication aspects can be executed. The results of such a simulation
run can be visually inspected in the OMNeT++ IDE, e.g. the exchanged messages between all
the involved actors and different layers of the network stack.
In Fig. 4.5 an example is provided, showing the exchange of messages on the highest
(application) level between one regular agent and one aggregator agent. In Fig. 4.6 a more
detailed view is given, also showing the TCP and IP layers of the agents and the intermediate
router for the exchange of a measurement message and associated reply. One can now also
see the exchanged TCP ACK messages.
31.08.2015
37
Fig. 4.5: Visual overview of the exchanged messages on application level between a regular
and an aggregator agent
Fig. 4.6: Visual overview of the exhanged messages on application, TCP and IP level
between a regular and an aggregator agent
The OMNeT++ IDE also provides some tools to easily make graphs of the different
measured statistics. Of course, for detailed evaluations any tools can be used to process the
results, but OMNeT++ tools are handy to check if everything works as expected when
defining and testing your simulations. As an example, in Fig. 4.7 the measured throughput
for a setup with 10 regular agents, one aggregator agent and an intermediate router is
31.08.2015
38
shown. As expected, the throughput on the link between router and aggregator is the largest
and could potentially become a bottleneck if for example PLC communication with low
datarates is used.
Fig. 4.7: Example graph for the measured throughput for a setup with 10 regular agents,
one router and an aggregator agent
31.08.2015
39
5.
Simulation results
The combined operation of Local Control with the different versions of Overlaying
Control is demonstrated in the pilot installation of the Slovenian DSO, Elektro Gorenjska. The
examined LV network is depicted in Fig. 5.1 and consists of 79 nodes, where 70 inductive
unbalanced loads and a typical MV/LV distribution transformer are connected. The green
nodes denote the location of the existing controllable PV units.
Since there is no feeder with at least two PV units, this network configuration does not
allow the examination of the FPS feature of the Overlaying Control. Thus, 24 additional PV
units are considered in the installation as shown by the red nodes in Fig. 5.1. The rated
power of the PV units, as well as the transformer data are shown in Table 5-1 and Table 5-2,
respectively. The MV/LV transformer is equipped with an OLTC, while the MV side voltage is
assumed equal to 1.05 pu. The transformer tap range is ±2 % with a voltage dead-band of
2.5 %. Finally, the 4-wire distribution lines have cross sections ranging from 4x16 mm2 to
4x150 mm2, while their lengths vary from 10 m up to 176 m.
In the following simulations 6 different control schemes are assumed and investigated:
1. No Control, where no control is considered and the PV units inject their nominal
active power.
2. Local, where the droop control of the injected active power of PV units is
implemented, according to D3.1.
3. Local & FPS (LF), which is the combined operation of Local Control with the FPS
algorithm of Overlaying Control, according to the 15-minute timeslot concept of
D3.2.
4. OLTC & Local (OL), which is the cooperation of OLTC algorithm with Local Control,
according to Section 3.1.
5. OLTC & Local & FPS (OLF), which is an enhanced version of OL with the incorporation
of the FPS algorithm.
6. OLTC & Local & FPS & Congestion (OLFC), which is the full version of Overlaying
Control that includes the FPS, OLTC, and Congestion algorithms, and it is combined
with Local Control, according to the 15-minute timeslot concept of Section 3.4.
31.08.2015
40
Fig. 5.1: Elektro Gorenjska modified network topology
Table 5-1: PV units rated power
31.08.2015
Name
Node
PV 1
PV 2
PV 3
PV 4
PV 5
PV 6
PV 7
PV 8
PV 9
PV 10
PV 11
PV 12
PV 13
13
14
18
20
22
24
25
4
5
6
7
8
31
Rated
Power
(kWp)
10
12
12
6
10
10
10
6
10
4
10
10
10
Name
Node
PV 16
PV 17
PV 18
PV 19
PV 20
PV 21
PV 22
PV 23
PV 24
PV 25
PV 26
PV 27
PV 28
50
47
52
54
9
57
58
59
60
63
64
70
74
Rated
Power
(kWp)
10
10
10
6
6
10
10
10
10
10
12
4
10
41
PV 14
PV 15
41
45
10
10
PV 29
PV 30
76
78
10
6
Table 5-2: Distribution transformer characteristics
Sn (kVA) Un (kV) Vector group uk (%) No-load losses (W) Load losses (W)
250
20 / 0.4
Dyn5
4
425
3250
This chapter is divided into three sections. In the first one, namely the normal case, a
time-series simulation is presented, where all control schemes (except the OLFC) are
evaluated. Each timeslot is considered equal to 15 min and the simulation covers a period of
24 hours. The active and reactive power profiles of the aggregated load are presented in Fig.
5.2 and Fig. 5.3, respectively, as derived from real-time measurements provided by the DSO.
The generation profile is also acquired by real-time measurements of the current PV
installations and is further applied to all PV units after appropriate scaling. In the second
section, an extreme case with high PV penetration is demonstrated in order to validate the
proposed OLFC and compare it with the OL and OLF control schemes. Finally, in the third
section the different performance metrics of the communication network for both wired and
wireless communication technologies are calculated for the examined LV grid.
60
Phase a
Phase b
Phase c
50
Active Power (kW)
40
30
20
10
0
0
4
8
12
Time (h)
16
20
24
Fig. 5.2: Active power profile of the aggregated load vs. time
31.08.2015
42
30
Phase a
Phase b
Phase c
25
Reactive Power (kVAr)
20
15
10
5
0
0
4
8
12
Time (h)
16
20
24
Fig. 5.3 Reactive power profile of the aggregated load vs. time
5.1. Normal Case
The total injected active power of all PV units with respect to time is presented in Fig.
5.4. In Fig. 5.5 a zoom over a specific time period is shown to clearly highlight the differences
among the proposed control schemes. The corresponding total curtailed active power is
depicted in Fig. 5.6 and Fig. 5.7. The LF control scheme results in a more uniform active
power curtailment among PV units. The main drawback of this method is the further
reduction of the total injected active power compared to the Local control scheme. By
employing the OL control scheme, the total injected power is considerably improved
compared to the Local control scheme. Finally, the integration of the FPS algorithm to the OL
control scheme, i.e. the OLF scheme, reduces the injected power due to the uniform power
curtailment which, however, remains higher compared to the other control schemes.
Therefore, the incorporation of the OLTC algorithm into the Overlaying Control increases the
overall injected active power of PV units.
Considering the network losses, the high PV penetration results in a reverse power flow
during the high generation periods. Thus, the active power losses are approximately
proportional to the square of the generation and present similar trend as shown in Fig. 5.8
and Fig. 5.9.
31.08.2015
43
300
Local
LF
OL
OLF
250
Active Power (kW)
200
150
100
50
0
0
4
8
12
Time (h)
16
20
24
Fig. 5.4: Injected active power of PV units vs. time
300
Local
LF
OL
OLF
Active Power (kW)
250
200
150
100
50
10
12
14
16
Time (h)
Fig. 5.5: Injected active power of PV units from 10:00 to 16:00
31.08.2015
44
90
Local
LF
OL
OLF
80
70
Active Power (kW)
60
50
40
30
20
10
0
0
4
8
12
Time (h)
16
20
24
Fig. 5.6: Curtailed active power of PV units vs. time
90
Local
LF
OL
OLF
80
70
Active Power (kW)
60
50
40
30
20
10
0
10
12
14
16
Time (h)
Fig. 5.7: Curtailed active power of PV units from 10:00 to 16:00
31.08.2015
45
12
Local
LF
OL
OLF
10
Active Power (kW)
8
6
4
2
0
0
4
8
12
Time (h)
16
20
24
Fig. 5.8: Network losses vs. time
11
Local
LF
OL
OLF
10
9
Active Power (kW)
8
7
6
5
4
3
2
1
10
12
14
16
Time (h)
Fig. 5.9: Network losses vs. time from 10:00 to 16:00
In Fig. 5.10, the voltage profile of each network node is depicted for the different control
schemes. It is evident that, since the No Control scheme results in zero active power
curtailment, there are overvoltages, especially during the high generation periods. However,
the use of Local, LF, OL, and OLF control schemes mitigate the overvoltages, while the
31.08.2015
46
voltage is also efficiently controlled. In the cases of OL and OLF, the voltage at several nodes
is less than 1.06 pu due to the OLTC operation, and thus the droop control for a number of
PV units is also avoided.
The tap setting of the transformer is depicted in Fig. 5.11, where only two tap changes
are performed during the examined day. Therefore, the equipment is not stressed, whereas
the injected active power of PV units is further increased. Finally, the daily energy losses and
the produced energy are presented in Fig. 5.12 and Fig. 5.13, respectively, where similar
conclusions can be drawn.
b) Local
a) No Control
1.15
Voltage (pu)
Voltage (pu)
1.15
1.1
1.05
1
0.95
0
4
8
12
Time (h)
c) LF
16
20
24
1
4
8
12
Time (h)
d) OL
16
20
24
0.9
0
4
8
12
Time (h)
16
20
24
16
20
24
Voltage (pu)
1.15
1.1
1.05
1
4
8
12
Time (h)
16
20
24
1.1
1.05
1
0.95
e) OLF
1.15
Voltage (pu)
Voltage (pu)
1.05
0.95
0
1.15
0.95
0
1.1
1.1
1.05
1
0.95
0.9
0
4
8
12
Time (h)
Fig. 5.10: Voltage of the network vs. time
31.08.2015
47
OL
OLF
2
Tap Position
1
0
-1
-2
0
4
8
12
Time (h)
16
20
24
Fig. 5.11: Tap setting of the transformer vs. time
60
50
Energy Losses (kWh)
40
30
20
10
0
Local
LF
OL
OLF
Control Scheme
Fig. 5.12: Daily energy losses
31.08.2015
48
1500
Energy Production (kWh)
1250
1000
750
500
Local
LF
OL
OLF
Control Scheme
Fig. 5.13: Daily energy production
The injected active power of all PV units on the feeder indicated by the blue line in Fig.
5.1 is presented in Fig. 5.14 after the implementation of the different control schemes. The
corresponding positive-sequence voltage profiles are also presented in Fig. 5.15. All data are
shown in a pu scale based on the rated voltage and power of each PV unit and correspond to
the timeslot between 12:00 and 12:15 where the most severe overvoltages occur as shown
in Fig. 5.10.
Considering the No Control scheme, each PV unit injects its nominal active power. Since
no active power curtailment is employed, the voltage at the last nodes of the feeder exceeds
1.1 pu which is the maximum permissible voltage, as defined by the EN 50160 standard. This
overvoltage is avoided by applying the Local control scheme. However, in this case the PV
units located at the end of the feeder suffer from a severe active power curtailment
compared to the ones located at the beginning of the feeder. To overcome this problem, the
LF control scheme is applied, resulting in a uniform active power curtailment among the PV
units of the same feeder, as shown in Fig. 5.14. According to the results of the LF control, the
injected active power of PV units located at the end of the feeder is increased, whereas the
PV units at the beginning of the feeder have now an increased active power curtailment,
compared to the previous Local control case.
The introduction of the OLTC algorithm improves considerably the performance of the
Overlaying Control. More specifically, considering the OL control scheme, only the last three
31.08.2015
49
PV units operate at the droop region, compared to the Local control scheme, due to the
reduction of the positive-sequence voltage at the LV side of the transformer as shown in Fig.
5.15. Furthermore, the OLF control scheme results in a considerable increase of the injected
active power, compared to the LF control scheme, due to the incorporation of the OLTC into
the Overlaying Control.
110
No Control
Local
LF
OL
OLF
100
Active Power (%)
90
80
70
60
50
40
PV 1
PV 2
PV 3
PV 4
PV unit
PV 5
PV 6
PV 7
Fig. 5.14: Total injected active power of PV units along a single feeder
31.08.2015
50
1.14
No Control
Local
LF
OL
OLF
Voltage threshold
1.12
Voltage Magnitude (pu)
1.1
1.08
1.06
1.04
1.02
1
0.98
MV
LV
11
13
14
18
19
Node
20
21
22
23
24
25
Fig. 5.15: Positive-sequence voltage profile along a single feeder
5.2. Case of high DRES penetration
The performance of the various control schemes are validated once again in the pilot
installation of Elektro Gorenjska. As mentioned previously, the network configuration of this
specific installation is not suitable for the implementation of the full Overlaying control. In
fact, despite the assumptions and modifications made in the previous section, the
congestion management control cannot be activated, since for the given load the total
injected power from all installed inverters in the network does not cause high enough
reverse power flow to lead to congestion.
Thus, in order to evaluate the congestion management control, a modified network
topology of Elektro Gorenjska, as shown in Fig. 5.1, is used. However, in this simulation case
the rated power of all DRES is assumed to be considerably higher, compared to the rated
power of all inverters in the previous section. The corresponding data are presented in Table
5-3, while the data of the MV/LV transformer, as well as the unbalanced, time-varying load is
shown in Table 5-2, and in Fig. 5.2 and Fig. 5.3, respectively. Finally, in this case the MV side
voltage of the transformer is considered equal to 1 pu.
Table 5-3: PV units rated power for the case of high DRES penetration
Name
31.08.2015
Node
Rated
Power
(kWp)
Name
Node
Rated
Power
(kWp)
51
PV 1
PV 2
PV 3
PV 4
PV 5
PV 6
PV 7
PV 8
PV 9
PV 10
PV 11
PV 12
PV 13
PV 14
PV 15
13
14
18
20
22
24
25
4
5
6
7
8
31
41
45
15.4
17.6
17.6
8.8
15.4
15.4
15.4
8.8
15.4
6.6
15.4
15.4
15.4
15.4
15.4
PV 16
PV 17
PV 18
PV 19
PV 20
PV 21
PV 22
PV 23
PV 24
PV 25
PV 26
PV 27
PV 28
PV 29
PV 30
50
47
52
54
9
57
58
59
60
63
64
70
74
76
78
15.4
15.4
15.4
8.8
8.8
15.4
15.4
15.4
15.4
15.4
17.6
6.6
15.4
15.4
8.8
The total apparent power which flows during the day through the MV/LV transformer is
depicted in Fig. 5.16. In Fig. 5.17 a zoom over a specific time period from 10:00 to 16:00 is
illustrated, to clearly highlight the congestion events and the necessity of the congestion
management control scheme. By observing Fig. 5.17 it is evident that after 10:00 three
congestion events occur. In the first event the total apparent power is equal to 251.2 kVA,
while both OLF and OLFC can handle efficiently the problem. In fact, the result of both
control strategies is exactly the same, since the activation of the FPS control curtails
adequate amount of active power from the PV units and thus mitigates the problem without
the activation of the congestion management control. However, regarding the second and
the third congestion events, it is clear that the OLF control strategy cannot mitigate
efficiently the congestion issues. Thus, the congestion management control must be applied.
31.08.2015
52
300
250
Apparent Power (kVA)
200
150
100
50
OL
0
0
4
8
12
Time (h)
OLF
16
OLFC
20
Transformer Limit
24
Fig. 5.16: Transformer total apparent power vs. time
300
250
Apparent Power (kVA)
200
150
100
50
OL
0
10
12
OLF
OLFC
14
Transformer Limit
16
Time (h)
Fig. 5.17: Transformer total apparent power from 10:00 to 16:00
The total injected active power of all PV units is presented in Fig. 5.18 and Fig. 5.19.
Furthermore, the corresponding total curtailed active power for all the examined control
strategies is depicted in Fig. 5.20 and Fig. 5.21. Using the OL control scheme, the total
injected active power is maximized compared to the OLF and the OLFC control strategies.
31.08.2015
53
However, the implementation of the OL control cannot ensure a uniform active power
curtailment among the installed PV units. On the other hand, the OLF control results in a
uniform power curtailment, nevertheless reducing the total injected active power. Finally,
the proposed OLFC control scheme curtails higher amount of active power compared to OL
and OLF controls, however, its main advantage is the efficient mitigation of the congestion
issues. Finally, the total daily injected power and energy losses are depicted in Fig. 5.22 and
Fig. 5.23, where similar conclusions can be drawn.
400
OL
OLF
OLFC
350
Active Power (kW)
300
250
200
150
100
50
0
0
4
8
12
Time (h)
16
20
24
Fig. 5.18: Injected active power of PV units vs. time
31.08.2015
54
400
OL
OLF
OLFC
350
Active Power (kW)
300
250
200
150
100
50
10
12
14
16
Time (h)
Fig. 5.19: Injected active power of PV units from 10:00 to 16:00
120
OL
OLF
OLFC
100
Active Power (kW)
80
60
40
20
0
0
4
8
12
Time (h)
16
20
24
Fig. 5.20: Curtailed active power of PV units vs. time
31.08.2015
55
120
OL
OLF
OLFC
100
Active Power (kW)
80
60
40
20
0
10
12
14
16
Time (h)
Fig. 5.21: Curtailed active power of PV units from 10:00 to 16:00
Energy Production (kWh)
2100
2050
2000
OL
OLF
Control Scheme
OLFC
Fig. 5.22: Daily energy production
31.08.2015
56
85
Energy Losses (kWh)
80
75
70
OL
OLF
OLFC
Control Scheme
Fig. 5.23: Daily energy losses
Regarding the network losses, once again the high PV penetration results in a reverse
power flow during high generation periods. Thus, the active power losses are approximately
proportional to the square of the generation and present the same trend as shown in Fig.
5.24.
20
OL
OLF
OLFC
18
16
Active Power (kW)
14
12
10
8
6
4
2
0
0
4
8
12
Time (h)
16
20
24
Fig. 5.24: Network losses vs. time
31.08.2015
57
The voltage profile of each network node is presented in Fig. 5.25. It is clear that all
control strategies can mitigate efficiently any possible overvoltages, ensuring the safe
operation of the network. Furthermore, the incorporation of the OLTC control ensures that
the voltages at the majority of the network nodes are less than 1.06 pu during the day. Thus,
the activation of the droop control, with the Vcpb indicated by the red dotted line, is avoided
and the total injected active power is maximized. The tap setting of the transformer is
illustrated in Fig. 5.26. It is clear that during the day only two tap changes are required for all
the examined control strategies.
b) OLF
1.05
1.05
Voltage (pu)
1.1
1
0.9
0
1
0.95
0.95
4
8
12
Time (h)
16
20
24
0.9
0
4
8
12
Time (h)
16
20
24
c) OLFC
1.1
1.05
Voltage (pu)
Voltage (pu)
a) OL
1.1
1
0.95
0.9
0
4
8
12
Time (h)
16
20
24
Fig. 5.25: Voltage of the network vs. time
31.08.2015
58
3
OL
OLF
OLFC
2
Tap Position
1
0
-1
-2
-3
0
4
8
12
16
20
24
Time (h)
Fig. 5.26: Tap setting of the transformer vs. time
5.3. Communication network performance evaluation
In Fig. 5.27 the different performance metrics of the communication network for both
wired and wireless communication technologies are calculated for a part of the examined LV
distribution grid.
31.08.2015
59
Fig. 5.27: Overview of the examined grid using the GUI interface of the INCREASE
simulation platform
The examined grid contains 30 regular agents that control the PV inverters and 1
aggregator agent located in the transformer LV bus. The following parameters are used in
the simulations:
•
•
Simulated time period: 1 hour
Message size: The agents use ACL messages for their communication with an
envelope size of 200 bytes. On top of that the actual data values need to be added.
o A measurement sample from a regular agent sent towards the aggregator
contains values for the PV injection, the active power exchanged and the
voltage for each of the three phases, thus 12 values in total. Assuming 4 bytes
per measurement value, a total of 48 bytes of data is added to the packet
resulting in a total message of 248 bytes of application data.
o For a control message from the aggregator towards a regular agent one value
per phase is assumed, thus 12 bytes extra and a total message size of 212
bytes.
o Further it is assumed that the reply messages to measurement and control
messages do not contain extra data, thus their size is 200 bytes.
31.08.2015
60
•
Frequency: Every regular agent sends a measurement message every minute, which
is acknowledged by the aggregator with a reply message and is immediately followed
by a control message from the aggregator which is on its turn acknowledged by the
regular agent. As a result, 4 application messages are exchanged every 60 seconds
between the aggregator and each of the regular agents.
For the evaluations presented in this section much details of the physical layer are not
taken into account, thus generic communication channels are assumed, but with realistic
data rates for different technologies. Results are compared for the 2 main transport
protocols TCP and UDP:
•
•
TCP stands for Transmission Control Protocol. It is a connection oriented protocol
where messages are acknowledged when they arrive and are resent in case a
message gets lost. Messages are also rearranged at the receiver side in the order that
they were sent. It is suited for applications that require high reliability (e.g. web
browsing, file transfer, email).
UDP stands for User Datagram Protocol. It is a more lightweight protocol without
guarantee that messages reach their destination and without reordering of packets.
It has a smaller header than TCP (8 bytes vs 20 bytes), so it is especially useful for
applications that require fast and efficient transmission (e.g. games, VoIP).
For the INCREASE control strategies TCP seems the most appropriate protocol, although
UDP could work as well since within the application layer reply messages are sent when
measurement or control messages are received. So even with UDP the communication could
be made reliable on application level, therefore the overhead from TCP to UDP is compared
in the simulations.
5.3.1. Wired communication network
In this case a wired communication network is assumed, where all regular agents and
the aggregator agent are connected to each other via a router, as shown in Fig. 5.28.
31.08.2015
61
Fig. 5.28: Overview of the simulated wired communication network
5.3.1.1. Throughput and channel utilization
The amount of data that is sent per time unit, i.e. the throughput, is first examined using
the settings described above. In Fig. 5.29 the results for TCP and UDP are shown. Logically,
the throughput on the link between the router and the aggregator is much higher than on
the links between the router and the individual regular agents as the former link transports
data from and towards 30 agents. The throughput towards the aggregator is the highest as
the measurement messages are a little larger than the control messages.
Furthermore, it is observed that the throughput for TCP is about 25 % higher than for
UDP. A first reason is because the TCP header is larger than the UDP header and a second
reason is that messages in TCP are acknowledged by additional ACK messages, as shown in
Fig. 4.6.
To calculate the utilization of the communication channel the throughputs in both
directions on a link are aggregated and divided by the maximum data rate of the channel. In
Fig. 5.30 the results for the most used link between router and aggregator are shown for a
few wired technologies: Standard Ethernet with 100 Mbps and 2 types of PLC with their
maximum data rates. As can be seen from the figure there is no problem to transmit all the
data. The maximum channel utilization in case of G3-PLC with TCP is about 15 %.
31.08.2015
62
Fig. 5.29: Throughput per link and direction
Fig. 5.30: Channel utilization for link between router and aggregator agent
In case lower data rates are used for PLC communication to make the communication
more robust, e.g. the most robust transmission rate of PRIME is 5.4 kbps, channel utilizations
close to 100 % are obtained. Thus, in case of larger networks with more agents or more
31.08.2015
63
frequent message exchanges, a pure PLC based network will not support the traffic anymore
and a different, e.g. hybrid communication infrastructure, would be needed. Part of the
communication network could still be based on PLC, e.g. to connect regular agents with the
router, whereas the link between router and aggregator should be replaced by technologies
with faster data rates like Ethernet or fiber. The optimal combination of PLC with other
technologies of course depends on the actual topology of the grid.
5.3.1.2. End-to-end delay
Next, the latencies for the messages exchanged between the agents are compared for
different technologies and the two transport protocols in order to assess if the information
reaches its destination always on time.
In Fig. 5.31 the results for the different communication technologies and transport
protocols are shown. The bars present the average delays for all messages that are received
at a certain agent for a unit of time (in this case within one minute as all messages
(measurement, control and reply messages are sent every minute). For regular agents this is
the average of 2 messages (a control message and a measurement reply message). For the
aggregator agent this is the average of 60 messages (30 measurement messages and 30
control reply messages). The whiskers show the minimum and maximum values.
Note that the y axis is logarithmic to show the values for Ethernet as these are very
small. Furthermore, note that as the channel utilization in all cases is well below 100 %, the
delays are always the same over the whole simulation period of 1 hour.
31.08.2015
64
Fig. 5.31: End-to-end delays
As it can be seen from the graph, the delay is always lower than 1 second. Thus, there is
enough time left for processing and sending back messages within the algorithm frequency
of 1 minute.
During the simulation, messages are simulated in order of number of agent, so that the
first message that arrives at the first agent as well as at the side of the aggregator has a very
low delay, whereas later messages have higher delays due the amount of traffic that is
generated at the same moment and corresponding waiting in queues. Therefore, for the last
regular agent all message experience some delay due to the other traffic which explains the
small difference between the minimum and maximum value.
5.3.2. Wireless communication network
In this section a wireless communication network is examined. In this case, the location
and the distance between the regular agents and aggregator agent is relevant. Since the
exact locations of the agents for the EG network are not known, some typical distances are
assumed in Fig. 5.32, varying from 10 m up to 750 m between the regular agents and the
aggregator agent.
31.08.2015
65
A simple abstracted wireless network is simulated with a transmission range that is high
enough for all regular agents to directly communicate with the aggregator agent and a
relatively limited data rate of 2 Mbps.
Fig. 5.32: Overview of the locations of the simulated agents
In Fig. 5.33 the recorded delays at the closest (Agent 8) and furthest located agent
(Agent 18) from the aggregator are shown assuming TCP communication. Compared to the
frequency of transmission of measurement messages and control messages, i.e. 1 minute, all
delays are very small, but the delays for the furthest agent are clearly higher than for the
closest agent. Similarly, in Fig. 5.34 the delays for all messages arriving at the aggregator
agent can be observed, where there is also a clear difference in delays mainly because all
messages are sent at the same moment, and thus some of them are kept up in queues
before they can be processed. In Fig. 5.35 and Fig. 5.36 the same results for UDP
communication are presented. In this case the delays are smaller than the corresponding of
the TCP communication.
31.08.2015
66
Fig. 5.33: End-to-end delays via TCP measured at 2 regular agents located at the closest
and furthest from the aggregator agent
Fig. 5.34: End-to-end delays via TCP measured at the aggregator agent
31.08.2015
67
Fig. 5.35: End-to-end delays via UDP measured at 2 regular agents located at the closes and
furthest from the aggregator agent
Fig. 5.36: End-to-end delays via UDP measured at the aggregator agent
31.08.2015
68
31.08.2015
69
6.
Conclusions
This report presents a detailed description of the Overlaying Control and of the
communication network simulator that is incorporated in the INCREASE simulation platform.
The full version of the Overlaying Control consists of the OLTC scheme, the congestion
control technique, and the FPS algorithm, which are developed to cope with overvoltages
and congestion problems, when integrating multiple DRES units in the distribution grid. The
combination of Local and Overlaying Control is also discussed, presenting the consecutive
implementation of both control schemes and the update of the various control parameters
in each simulation time interval. Furthermore, the actual incorporation of MAS in JADE
environment is analyzed, highlighting the bi-directional communication between the Core
and JADE components of the INCREASE simulation platform.
The performance of the full Overlaying Control is demonstrated on a selected pilot
installation and is compared to the cases of no DRES power curtailment as well as to the
Local Control scheme. Results show the efficiency of the proposed algorithm in mitigating
overvoltages by changing the OLTC state and enforcing a fair contribution of the curtailed
power among the installed PV inverters, while maintaining low levels of power losses in the
distribution network. Moreover, the efficiency of the congestion control technique is also
shown, assuming a high DRES penetration level.
Next, the LAN simulator component is presented aiming to enhance the features of the
INCREASE simulation platform, by focusing on the evaluation of the communication
infrastructure and of the requirements posed by the MAS control system and the
implemented control algorithms. The general framework and the different simulation
programs are described, while the simulation functionalities and features of the open-source
OMNeT++ software are further analyzed. The necessary operational functions and their
implementation in the INCREASE simulation platform are thoroughly discussed, while the
actual integration is presented, focusing on the interface between GUI, Core and LAN
components. The performance evaluation of the communication layer is conducted in the
examined pilot installation using the LAN simulator component, while assuming different
wired and wireless communication technologies.
31.08.2015
70
References
[1] X.-P. Zhang, “Fast three phase load flow methods,” IEEE Trans. Power Syst., vol. 11, no.
3, pp. 1547-1553, Aug. 1996.
[2] G. C. Kryonidis, E. O. Kontis, A. I. Chrysochos, C. S. Demoulias, D. Bozalakov, B.
Meersman, T. L. Vandoorn, and L. Vandevelde, “A Simulation tool for extended
distribution grids with controlled distributed generation,” in IEEE Eindhoven PowerTech
2015, Eindhoven, Netherlands, June 29 - July 2, 2015, accepted for presentation.
[3] B. Meersman, B. Renders, L. Degroote, T. Vandoorn, and L. Vandevelde, “Three-phase
inverter-connected DG-units and voltage unbalance,” Elec. Power Syst. Res., vol. 81,
pp. 899-906, 2011.
[4] B. Renders, W. R. Ryckaert, K. De Gussemé, K. Stockman, and L. Vandevelde,
“Improving the voltage dip immunity of converter-connected distributed generation
units,” Renewable Energy, vol. 33, pp. 1011-1018, 2008.
[5] B. Renders, K. De Gussemé, W. R. Ryckaert, and L. Vandevelde, “Converter-connected
distributed generation units with integrated harmonic voltage damping and harmonic
current compensation function,” Elec. Power Syst. Res., vol. 79, pp. 65-70, 2009.
[6] T. L. Vandoorn, B. Meersman, L. Degroote, B. Renders, and L. Vandevelde, “A control
strategy for islanded microgrids with DC-link voltage control,” IEEE Trans. Power Del.,
vol. 26, no. 2, pp. 703-713, 2011.
[7] D. Bozalokov, T.L. Vandoorn, B. Meersman, C. Demoulias, L. Vandevelde, “Voltage dip
mitigation capabilities of three-phase damping control strategy”, Elec. Power Syst. Res.,
vol. 121, pp. 192-199, 2015
[8] V. M. A. M. Mansoor, P. H. Nguyen, and W. L. Kling, “An integrated control for
overvoltage mitigation in the distribution network,” in IEEE PES Innov. Smart Grid
Tech., 2014.
[9] M. M. M. A. Viyathukattuva, P. H. Nguyen, W. L. Kling, A. I. Chrysochos, T. A.
Papadopoulos, and G. K. Papagiannis, “Fair power curtailment of distributed
renewable energy sources to mitigate overvoltages in low-voltage networks,” in IEEE
Eindhoven PowerTech 2015, Eindhoven, Netherlands, June 29 - July 2, 2015, accepted
for presentation.
[10] A. F. Gubina, A. Tuerk, and G. Taljan, “Policy framework conditions for provision of
ancillary services in a distribution network by distributed RES generation,” in IEEE
Eindhoven PowerTech 2015, Eindhoven, Netherlands, June 29 - July 2, 2015, accepted
for presentation.
[11] http://sourceforge.net/projects/electricdss/
[12] http://jade.tilab.com/
[13] http://inet.omnetpp.org/
31.08.2015
71
[14] http://www.mathworks.com/
[15] http://smartgrid.epri.com/SimulationTool.aspx
[16] P. Kadurek, J. F. G. Cobben, and W. L. Kling, “Smart MV/LV transformer for future
grids,” in SPEEDAM 2010, Pisa, Italy, June 14-16, 2010.
[17] P. Kadurek, J. F. G. Cobben, and W. L. Kling, “Future LV distribution network design and
current practices in the Netherlands,” in 2nd IEEE PES Innovative Smart Grid
Technologies, Manchester, UK, December 5-7, 2011.
[18] P. Bach Andersen, J. Hu, and K. Heussen, “Coordination strategies for distribution grid
congestion management in a multi-actor, multi-objective setting,” in 3rd IEEE PES
Innovative Smart Grid Technologies, Berlin, Germany, October 14-17, 2012.
[19] B. Asare-Bediako, W. L. Kling, and P. F. Ribeiro, “Future residential load profiles:
Scenario-based analysis of high penetration of heavy loads and distributed
generation,” Energy Build., vol. 75, pp. 228–238, 2014.
[20] K. Mets, J. Aparicio and C. Develder, “Combining power and communication network
simulation for cost-effective smart grid analysis,” IEEE Commun. Surveys Tutorials, vol.
16, no. 3, pp. 1771-1796, 2014.
[21] T. R. Henderson, M. Lacage, and G. F. Riley, “Network simulations with the ns-3
simulator,” in Proc. ACM Conf. of the Special Interest Group on Data Communication
2008 (SIGCOMM ’08), August 17-22, 2008.
[22] J. Bergmann, C. Glomb, J. Gö, J. Heuer, R. Kuntschke, and M. Winter, “Scalability of
smart grid protocols: Protocols and their simulative evaluation for massively
distributed DERs,” in Proc. 1st IEEE Int. Conf. on Smart Grid Communications 2010
(SmartGridComm ’10), Oct. 2010.
[23] H. Lin, S. Sambamoorthy, S. Shukla, J. Thorp, and L. Mili, “Power system and
communication network co-simulation for smart grid applications,” in Proc. IEEE/PES
Innovative Smart Grid Technologies 2011 (ISGT ’11), 2011.
[24] A. Varga, “An overview of the OMNeT++ simulation environment,” in Proc. 1st Int. Conf.
Simulation Tools and Techniques for Communication Networks and Systems
(SIMUTools ’08), Marseille, France, March 3-7, 2008.
[25] “INET Framework: An open-source OMNeT++ model suite for wired, wireless and
mobile networks”, https://inet.omnetpp.org/.
[26] J. Juarez, C. Rodrıguez-Morcillo, and J. A. Rodrıguez-Mondejar, “Simulation of IEC
61850-based substations under OMNeT++,” in Proc. of the 5th Int. ICST Conference on
Simulation Tools and Techniques. ICST (Institute for Computer Sciences, SocialInformatics and Telecommunications Engineering), pp. 319-326, 2012.
[27] K. Mets, T. Verschueren, C. Develder, T. L. Vandoor, and L. Vandevelde, “Integrated
simulation of power and communication networks for smart grid applications,” in Proc.
31.08.2015
72
IEEE 16th Int. Workshop on Computer Aided Modeling and Design of Communication
Links and Networks 2011 (CAMAD ’11), Kyoto, Japan, June 10-11, 2011.
[28] C. Müller, H. Georg, and C. Wietfeld, “A modularized and distributed simulation
environment for scalability analysis of smart grid ICT infrastructures,” in Proc. 5th Int.
ICST Conf. on Simulation Tools and Techniques (SIMUTools ’12), Brussels, Belgium, pp.
327–330, 2012.
[29] T. Konnerth, J. Chinnow, D. Grunewald, and S. Kaiser, “Integration of simulations and
MAS for smart grid management systems,” in Proc. 3rd Int. Workshop on Agent
Technologies for Energy Systems (ATES 2012) at 11th Int. Conf. on Autonomos Agents
and Multi-agent Systems (AAMAS ’12), Valencia, Spain, June 5 2012.
31.08.2015
73

Report on simulation results and evaluation of the

Transcription

Similar documents

iSIM iSIM

Power devices - Crosslight Software

VISUAL SIMULATION As viewed from Orono Fairgrounds South

When Colette Foisy-Doll arrived in Qatar last year

ASE2000 Version 2 - Applied Systems Engineering Inc.

sendai mediatheque - toyo ito

6SN7 SE TJ FullMusic

To a brochure about our facility, resources

How to Tie a Woggle

Simulation Experience