Transfer Function Method - Solar Thermal | IEA-SHC
Transcription
Transfer Function Method - Solar Thermal | IEA-SHC
1 MONTREAL SUMMER SCHOOL 25-29 June 2011 Transfer Function Method Dipartimento dell’Energia Prof. Maurizio Cellura mcellura@dream . unipa .it Ph +39-091-23861931; Fax +39-091-484425; PhD Summer School Montreal 25-29 June 2011 GENERAL INFORMATION Analysis of the thermal performances of building during the project allows the developer: the choice of the best materials for the local climatic characteristic; energy saving and a correct bio-climatic approach. 2 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Introduction Simulation and analysis of the thermal fluxes in a building help the developer to choose the best materials for the local climatic characteristics Many software packages for the thermal dynamic simulation of buildings have been developed to better understand the thermal behavior of buildings 3 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Introduction A caravan has walls with a great thermal insulation, but the thermal comfort conditions are often poor in summer conditions On the contrary, in sunny hot days, inside a massive building, with heavy walls and without insulations, thermal comfort conditions are excellent – Radiant mean temperature of surfaces and comfort? How the HWAC control temperature of surfaces? Are the north architectural styles successfully transferrable to the south of the world? 4 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Introduction Many thermal process are relevant in the assessment of building thermal behavior including: heat conduction through exterior walls, roofs, ceilings, floors and interior partitions; solar radiation through transparent surfaces; latent or sensible heat generated in the space by occupants, lights, and appliances; heat transfer through ventilation and infiltration of outdoor air and other miscellaneous heat gains. 5 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Introduction The Thermal Inertia – Developer could Use the delay 6 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Introduction The Thermal Inertia The thermal inertia, linked to conductive heat transfer, can: • Control the inner temperature variations • Realize the optimal comfort conditions • Limit the HVAC costs By planning the right value of thermal inertia of walls, the maximum requested power for HVAC during summer season can be reduced 7 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Introduction Considering an homogeneous and isotropic element.. 8 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Fourier’s Law: Recognized that the flux is a vectory quantity, we can write a more general statement of the conduction rate equation as: ∂T ∂T ∂T q = −k∇T = −k i +j +k ∂y ∂z ∂x Where: ∇ is the three dimensional del operator T (x, y, z ) is the scalar temperature field. 9 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 The non–steady–state heat problems in multilayered walls • For a homogeneous isotropic element, the change of temperature and of flux with distance are: ∂θ ( x, t ) ∂q ( x, t ) ∂θ ( x, t ) 1 = − q ( x, t ) ; = − ρC ∂x λ ∂x ∂t Combination of these equation gives the Fourier equation: ∂ 2θ ( x, t ) 1 ∂θ ( x, t ) = 2 ∂x α ∂t 10 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Methods: To solve the dynamic heat transmission it is possible to use: Laplace’s Method; Z-Transform Method; Finite element Method; Finite difference Method. 11 University of Palermo, Italy Analytics' Methods Numerical Methods Discussion on Analitical’s Methods Maurizio Cellura 12 MONTREAL SUMMER SCHOOL 25-29 June 2011 LAPLACE’S METHOD Prof. Maurizio Cellura mcellura@dream . unipa .it cell-. +393204328228 PhD Summer School Montreal 25-29 June 2011 Laplace’s Method: (continuous signals) The Laplace transform is an algorithm commonly used in the solution of differential equations Given a function of time f(t) for 0 ≤ t < ∞, the Laplace transform F(s) is: = F ( s ) L= [ f (t )] 13 University of Palermo, Italy ∫ ∞ 0 e −st f (t ) d t Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Laplace’s Method: (continuous signals) Some fundamentals L-transformations Step function: 1 = f (t ) u= (t ) F (s) s * 1 f (t ) Au (t ) F (s) A = = s * 14 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Laplace’s Method: (continuous signals) Some fundamentals L-transformations Ramp function: 1 f (t ) t = F (s) = s2 1 f (t ) A= t F ( s) A 2 = s 15 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Laplace’s Method: (discrete signals) Other more complicated function are known only in fixed instants of time. These functions (signals) are called discrete 16 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Laplace’s Method: (continuous signals) To sample a continuous signal, the first step is to interpolate data between two values. This steps is very important because the area under the interpolating line is linked to the energy. Without interpolating lines, there will be no energy linked to the thermal flux (that is continuous). In the figure, a continuous signal sampled with step interpolating function 17 University of Palermo, Italy Maurizio Cellura 18 Laplace’s Method: (continuous signals) In place of linear or step interpolating function, a train of linear ramps can be used. Combining different elementary functions, linear, triangular, parabolic, it is possible to create more interpolating functions. 18 PhD Summer School Montreal 25-29 June 2011 Laplace’s Method: (continuous signals) For a function such as temperature T(x, t), which has transformed θ (x, s), the following relations: ∂T ( x, t ) L = s θ ( x, s ) − T ( x, 0 ) ∂t ∂T ( x, t ) L ∂ x 19 ∂ 2T ( x, t ) ∂ 2θ ( x, s ) ∂θ ( x, s ) = L 2 2 ∂x ∂ ∂ x x University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Laplace’s Method: (continuous signals) Using Laplace Transform, the Fourier’s Law is: ∂T ∂ 2T = −α 2 0 ∂t ∂x ∂ 2θ ( x, s ) s −= θ ( x, s ) 0 2 ∂x α Assumption: the temperature at the initial instant is zero. The general integral is: βx θ= ( x, s ) M e + N e where 20 β= −β x s α University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Laplace’s Method: (continuous signals) For a homogeneous isotropic element, changes with the distance of the temperature and flux are: ∂θ ( x, t ) ∂q ( x, t ) ∂θ ( x, t ) 1 = − q ( x, t ) ; = − ρC ∂x λ ∂x ∂t The solutions of this system was obtained using Laplace Transform in 1964 21 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Laplace’s Method: (continuous signals) The non–steady–state heat problems in multilayered walls • Solution of these equations into an imaginary space was obtained by Carlslaw and Jaeger 1 2 s x q ( 0, s ) sinh 1 α 2 s θ ( x, s ) cosh x θ ( 0, s ) − 1 α s 2 λ α 1 1 1 2 2 s s s 2 −λ sinh x θ ( 0, s ) + cosh x q ( 0, s ) q ( x, s ) = α α α 22 University of Palermo, Italy Maurizio Cellura 23 MONTREAL SUMMER SCHOOL 25-29 June 2011 Z-TRANSFORM’S METHOD Prof. Maurizio Cellura mcellura@dream . unipa .it cell-. +393204328228 PhD Summer School Montreal 25-29 June 2011 Z-transform Method: (discrete signals) Considering that, usually the available signals in thermophysics of buildings, such as inner temperatures, insolation, outdoor temperature, etc…, are discrete signals, it is better to use a more suitable mathematical transformation (Z transform method). Furthermore, the transformation must be suitable even to be implemented in software codes. To solve the Carlslaw and Jager equations by the Ztransform method 24 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Z-transform Method: (discrete signals) Consider a function f(t), for t ≥ 0, where are known f(nΔ) for a time data set Δ , its Z-transform function is: Z [ f (= t ) ] F (= z) ∞ ∑ f (n∆) z= −n f (0) + f (∆ ) z −1 + f (2∆ ) z −2 + f (3∆ ) z −3 + ⋅ ⋅ ⋅ n =0 where z is a complex variable Z-transform is used for the following functions: 1 1 ( ) = F s 1 − z −1 s 1 1 ( ) = f (t ) t = F ( z) = F s z (1 − z −1 ) 2 s2 1 1 − at ( ) = f (t ) e= F ( z) = F s 1 − e − a∆ z −1 s+a * (t ) = f (t ) u= F ( z) 25 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Z-transform Method: (discrete signals) The non–steady–state heat problems in multilayered walls • The solution for a multi-layered wall can be expressed in matrix form: A B Ti x = Qe C D Qi Te 26 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Z-transform Method: (discrete signals) THE WALL TRANSMISSION MATRIX A, B, C and D are the coefficients of the wall transmission matrix reached through the product of transmission matrixes, of each n layers forming the wall: n a b A B m m =∏ C D m =1 cm d m 27 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Z-transform Method: (discrete signals) THE WALL TRANSMISSION MATRIX where: s s s ; cm = λ ; am = cosh L senh L α α α s senh L α bm = s λ α λ d m = am ; α = ρC p 28 University of Palermo, Italy Maurizio Cellura 29 Z-TRASFORM METHOD After many mathematical manipulations Transfer function in a polynomial form Very fast response – Coefficients decrease their values soon and this fact implies a limited employment of them; usually less than 10 Phd summer school on solar NZEBs 30 METODO DELLA Z-TRASFORMATA With the following substitution And the following The heat flux Qi could be calculated like the ASHRAE manual Know temperatures till the current nΔ and thermal flux of the inner surface till (n-1)Δ Phd summer school on solar NZEBs PhD Summer School Montreal 25-29 June 2011 Z-transform Method: (discrete signals) THE WALL TRANSMISSION MATRIX The system can be described using the ZT at the LT place. For a sampled, time-continuous function f(t), LT is,: f (0 ) + f (∆ )e − s∆ + f (2∆ )e −2 s∆ + .... Z-T of function f(t), is: f (0 )z + f (∆ )z + f (2∆ )z 0 −1 −2 z + .... Using Z-T implies modifications of previous wall transmission matrix formula 29 University of Palermo, Italy Maurizio Cellura 32 MONTREAL SUMMER SCHOOL 25-29 June 2011 An Application: Heat Balance Method Prof. Maurizio Cellura mcellura@dream . unipa .it cell-. +393204328228 PhD Summer School Montreal 25-29 June 2011 Heat Balance Method (ASHRAE) Conduction heat flux through a multilayered wall is of paramount importance in the ASHRAE algorithm. 31 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Evaluation of conduction thermal flux. The central element of the HB method is the dynamic evaluation of conduction thermal fluxes through boundary walls. The methods to calculate the conduction heat flux through a multilayered wall are manifold and one of the most employed is the Transfer Function Method ( TFM). (TRNSYS, ENERGY PLUS) 33 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Transfer Function Method The TFM allows the calculation of the thermal flux or of the wall surface temperature employing Conduction Transfer Function – CTFs that depends on thermo-physical characteristics of the layers. exact determination of these coefficients would require a theoretical infinite numbers of calculus; to evaluate CTFs a numerical approach is used, The numerical approach can be a source of error. The choice of time step is really very important, 34 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Transfer Function Method The thermal simulations with TFM applied to very massive walls, characterized by high thermal mass, strongly depends on time step! DOE-2, TRNSYS e ENERGY PLUS Massive wall with 80% insulation Error > 44% Massive walls composed by concrete Error > 27% Wood 35 University of Palermo, Italy Error < 2 % Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Transfer Function Method The determination of the thermal flux through multilayered walls was solved by Carslaw and Jager in 1959. θe θi ϕe ϕi 0 36 L University of Palermo, Italy x Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Transfer Function Method The application of Mitalas resolution method to the matrix system of Carslaw and Jager permits to obtain the inner thermal flux or the inner surface temperature by means of some coefficients called Transfer Function Coefficients, Transfer Function Coefficients are numerically calculated 37 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Quality Assessment of CTFs Time step Thermophysical characteristics Resolution of the system Choice of poles number 38 University of Palermo, Italy Quality of system resolution depends on many parameters!!! Choice of interpolating function Choice of coefficients number Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 TFM and Fourier Transforms To evaluate the quality of resolution by TFM we made a comparison with the results obtained in Periodic Steady State (Fourier Transform). Percentage Mean Error where: PME 1 24/ ∆ QZ (τ ) − QF (τ ) ⋅100 ∑ QF (τ ) 24∆ τ =1 QZ: thermal flux emerging from the inner surface of the wall calculated with TFM ; QF: thermal flux emerging from the inner surface of the wall calculated with Fourier Transforms (Periodic Steady State Solution) 39 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Accuracy of TFM We applied the TFM to typical Mediterranean buildings. Some considerations: 1. 2. 3. 4. 40 Simulations with the greatest number of poles not always are the best; Simulations with the greatest number of coefficients not always are the best; A larger time step (2 or 3 hours) generally decreases the PME but nevertheless decreases the quality of information; Choice of a larger time step or of a different number of poles and coefficients is function of high inertia mass of the wall. University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Who is really an expert in the field of Transfer Function Method? 1. How to choice the best time step? 2. How to determine the optimum number of poles? 3. How to choice the best number of coefficients? The optimum choice of these parameters is crucial to obtain a good and reliable simulation using TFM. Often in the commercial Building Simulation Software is impossible the “tuning“ of these parameters. Even if possible, normal user would not be able to choice the optimal values. 41 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Software CATI Client-server structure 42 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Software CATI s S Density Asymmetry ∑ρ Ls DA = i i ∑ρ L 3 s i i =1 S i i =1 2 [ m] i s 1 S Specific heat Asymmetry HA = ∑c L s i i =1 S i i ∑c L i i =1 Thermal resistance Asymmetry S RA = Thermal Inertia 43 University of Palermo, Italy −1 i i i Ls ∑λ i =1 o i ∑λ i =1 S [ m] −1 i i [ m] L S TI = ∑c ρ L i i =1 i i 1 1 3600 ⋅ + ∑ λi−1 Li + hout hin i =1 S −1 [h] Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Il Software CATI The software CATI records: the coefficients of the CTFs numerator; the coefficients of the CTFs denominator; the Transfer Function in the Laplace domain; the temperature time series (or thermal flux) obtained as output of the simulation with TFM; the temperature time series (or thermal flux) obtained as output of the simulation with Fourier Other data linked to the used algorithms 44 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Results: The accuracy of a simulation performed with TFM strongly depends on the number of coefficients!! Wall Sampled period Number of poles Number of coefficients PME A0:A1:B6:C11:E0 1 10 6 0,067 A0:A1:B6:C11:E0 A0:A1:B6:C11:E0 A0:A1:B6:C11:E0 A0:A1:B6:C11:E0 A0:A1:B6:C11:E0 A0:A1:B6:C11:E0 A0:A1:B6:C11:E0 A0:A1:B6:C11:E0 1 1 2 2 2 3 3 3 10 10 10 10 10 10 10 10 5 4 5 4 3 4 3 2 5,395 767,828 1,435 0,645 175,077 2,227 6,868 490,801 Reliability of performed simulations A decrease of only two or three coefficients implies a significant increase in the PME. A number of coefficients lower than the threshold value makes the simulation completely unreliable. 45 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Results: Elimination of only two or three coefficients provokes a drastic increase of PME. Composizione parete: A0:A1:B6:C11:E0 Numero di poli: np= 10 nc= 4 767.828 800 700 600 nc= 2 490.801 500 400 175.077 nc= 5 5 nc= 3 nc= 3 6.868 200 5.395 PME (%) 300 4 3 0 1 1 1 nc= 4 2 2 2.227 1 nc= 5 0.645 nc= 6 1.435 2 0.067 nc= 4 2 3 3 3 Periodo di campionamento (h) Wall composition: Table 1 in ASHRAE 1994 . 46 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Results: Furthermore, a sort of asymptotic trend in the PME can be also observed. The increase of the number of poles or the coefficients does not increase the quality of the simulation: Wall Sampled period Number of poles Number of coefficients PME A0:A1:B6:C11:E0 1 16 12 0,065 A0:A1:B6:C11:E0 1 16 11 0,065 A0:A1:B6:C11:E0 1 16 10 0,065 A0:A1:B6:C11:E0 2 16 12 1,434 A0:A1:B6:C11:E0 2 16 11 1,434 A0:A1:B6:C11:E0 2 16 10 1,434 A0:A1:B6:C11:E0 3 16 9 2,229 A0:A1:B6:C11:E0 3 16 8 2,229 A0:A1:B6:C11:E0 3 16 7 2,229 47 University of Palermo, Italy Excerpt Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Aim: provide an expert tool to help the user. 1. An expert tool able to automatically select the best time step 2. An expert tool able to suggest the optimum number of poles 3. An expert tool able to suggest the optimum number of coefficients 48 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Data-driven approach data-driven: an approach characterized by a “self-learning” process based upon the results and the simulations performed by CATI Reliability of a CTF set (Artificial Neural Network- ANN) 49 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Neural network and Classification “Pattern Recognition: A neural network is able to recognize similar features of input data and to classify them in categories neuroni f(.) funzione non lineare j Input l wijl wnil+1 i n w pesi sinaptici Input 2 Input 3 Strato l-1 Strato l Strato l+1 A classifier is an expert tool that creates a relation between the variables of a domain “A” and a vector of labels “B”, In other words it gives a label “B” to every sample of domain “A”, 50 University of Palermo, Italy Σ f(.) Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Neural Networks Classifier Using the software CATI, were performed more than 54000 simulations. It was possible to build a large databases to train a neural network classifier. Time step Thermal inertia of the wall; Have been defined two classes (labels): Number of poles of CTF; Number coefficients of CTF; 1) Theofreliable simulations (PMEAsymmetry; <5%); Density Specific Heat Asymmetry 1) The unreliable simulations Thermal Resistance Asymmetry (PME >5%). Input vector Other data 51 University of Palermo, Italy Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Neural Network Classifier Output Correlation analysis CLASSIFICATOR JUDGEMENT Confusion Matrix 52 Reliable simulation TRUTH University of Palermo, Italy Reliable Simulation Unreliable simulation Unreliable simulation 98,2% 1,8% 2,8% 97,2% Maurizio Cellura PhD Summer School Montreal 25-29 June 2011 Software development: Evaluation of a CTF with 1 hour time step, 11 poles and 5 coefficients. The CTF will be UNRELIABLE! Evaluation of a CTF with 1 hour time step, 14 poles and 13 coefficients. The CTF will be reliable! 53 University of Palermo, Italy Maurizio Cellura 55 MONTREAL SUMMER SCHOOL 25-29 June 2011 Thanks for your attention! Dipartimento dell’Energia Prof. Maurizio Cellura Ph +39-091-23861931; e-mail: mcellura@dream .unipa .it