two-step optimal design using genetic algorithms and cfd-coupled simulation for indoor thermal environments

1、Applied Thermal Engineering 27 (2007) 311/locate/apthermengTwo-step optimal design method using genetic algorithms and CFD-coupled simulation for indoor thermal environmentsTaeyeon Kim a, Doosam Song b,*, Shinsuke Kato c, Shuzo Murakamia Yonsei University, 134 Shinchon-dong, Seodaemu

2、n-gu, Seoul 120-749, Republic of Koreab Sungkyunkwan University, 300 Chunchun-dong, Jangan-gu, Suwon 440-746, Republic of Korea c Institute of Industrial Science, University of Tokyo, 4-6-1 Komaba Meguro-ku, Tokyo 153-8505, Japan d Keio University, 4-1-1 Hiyoshi, Kohoku-ku, Yokohama, Kanagawa 223-85

3、21, Yokohama, JapanReceived 28 February 2006; accepted 18 May 2006Available online 1 August 2006dAbstractThe optimal design method of the indoor thermal environment using CFD coupled simulation and genetic algorithms (GA) is developed in this study. CFD could analyze the thermal environment consider

4、ing the distribution of temperature and air ow in a room. Therefore, it would be appropriate to use CFD for the optimal design method considering their distribution. In this study, the optimal design means the most appropriate boundary conditions of the room among the conditions where the design tar

5、get of the indoor thermal environment is achieved.The authors have been examined many types of the optimal design methods using CFD, and nd that the high calculating load and the dierent results according to the initial conditions, among other factors, are the diculties of the optimal method. Consid

6、ering these dif- culties, a two step optimal design method for the indoor thermal environment is proposed. It includes the GA for determining the optimal design for the indoor thermal environment. To examine the performance of this method, the optimal design of a hybrid ventilation system, which use

7、s the natural cross ventilation and the radiant cooling panel, is completed. The optimal design which satises the design target (thermal comfort, minimum cooling load, and minimum vertical temperature dierence) is found using a two step optimal design method.2006 Elsevier Ltd. All rights reserved.Ke

8、ywords: Optimal design; Genetic algorithms (GA); CFD; Coupled simulation1. Introductionconditioning system, location of the supply inlet or exhaust outlet, wind velocity and outlet temperature, with the ulti- mate goal of eciently controlling the indoor environment and saving energy. In the process

9、of designing the optimal method for the indoor thermal environment, therefore, it is benecial to use CFD coupled simulation method.In general, the process of the optimal design method for the indoor thermal environment can be (1) an expert sys- tem which incorporates the designers decision making pr

10、o- cess. In this case the applicable design conditions are examined one by one according to the expert system 2, and (2) an optimizing method which could nd the most rational design among the combinations of all the design conditions (for example, generic algorithms) 3.The optimal design method usin

11、g designers decision making process divides the design process into each step,This paper is a basic study for developing an optimal design method using CFD (computational uid dynamics) coupled simulation method for the indoor thermal environ- ment. Because it allows a detailed analysis considering t

12、he distribution of the indoor physical environment, the CFD coupled simulation method is widely used to predict indoor thermal environments and to analyze the performance of air conditioning systems 1. The design of the optimal thermal indoor environment comprises the process of nding the most appro

13、priate boundary conditions considering the air* Corresponding author. Tel.: +81 31 290 7551; fax: +82 31 290 7570.E-mail addresses: tkimyonsei.ac.kr (T. Kim), (D. Song).1359-4311/$ - see front matter 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.applthermaleng.2006.05.0224T. K

14、im et al. / Applied Thermal Engineering 27 (2007) 311Table 1Organism and genetic algorithmsand repeats the process of designing, analyzing, evaluating and modifying of the indoor thermal environment. When the thermal environment satises certain condition, the method judges it as the optimal design a

15、nd ends the evalu- ation. During the process, the indoor thermal environment is evaluated by a CFD coupled simulation using the distri- bution of the indoor physical environments. The result can be considered to be the optimal design method for the ther- mal environment. However, unlike the dynamic

16、thermal load calculation tools which are widely used for the analy- sis of energy, CFD coupled simulation requires large a computational load. Therefore, it is not suitable to com- plex analysis which requires an extremely large computa- tional load to investigate the optimal conditions with feedbac

17、k. Also, the result of the optimal design often depends on the initial conditions for the analysis. There- fore, dierent initial conditions would create dierent design methods.After considering these problems, this study proposes a two-step optimal design using genetic algorithms (GA) and CFD couple

18、d simulation analysis. The approach is then examined. GA is often used in optimal design of non-archi- tecture areas as the optimal search method. In architecture, GA has been used in optimal green building design, opti- mal work process management, and optimal building material assessment 35.Organi

19、smGenetic algorithmsGenotypeTT, Tt, tt, . (T: dominance, t: recessive)Characteristics according to genes Human race,type of owers, .Ability of adaptations, .000, 001, 010, . .PhenotypeConditions according to genesType of HVAC system, . .FitnessEciency of HVAC systems, . .0, 1, . .GeneT, t, . . cross

20、 over(a) Cross overmutation(b) MutationFig. 1. Examples of GA operators.2. Indoor thermal environment design and genetic algorithmsincludecrossover and mutation. Manipulation frequencyand method are generally determined by constant probabil- ity decided in advance.2.1. Introduction of genetic algori

21、thmsBased on the generic principles of organisms, genetic algorithms (GA) is used to solve optimization problems. GA is the engineering application of the evolutionary pro- cess in which organisms adapt to the environment through crossover and mutation.2.4. Sequence of GAThe sequence of GA is shown

22、in Fig. 2. If the group of gene type, M(n), is considered to be the population of a generation, n, the suitability of an individual to a given problem is determined by each characteristic of gene. Manipulation of GA, is generally applied to gene types of high suitability. As a result, newly created

23、gene types are replacements of gene types of low suitability. Through the manipulation, the new generation, n + 1, forms the group, M(n + 1), and this process repeats as necessary.2.2. Genotype and phenotype of genetic algorithmsCharacteristics of an organism are determinedby itstypes of genes. Simi

24、lar to the living organisms, GA uses two of genetic types (genotype, organisms gene) and expressive type (phenotype, gene created characteristics of an organism). Genotype is the group of the most funda- mental principles in optimizing variables, and is modied by a genetic operator. Phenotype repres

25、ents the character- istics of genotype. The suitability to a given problem, degree of the optimization, is dependant on the character- istics. The comparison between an organism and the GA is shown in Table 1.calculate fitness value2.3. Gene manipulation by genetic algorithmsselect genesSimilar to t

26、he evolutionary process of an organism, in the GA process, gene manipulation on genotype creates a new genotype of the next generation, shown in Fig. 1. The most represented methods of gene manipulationGenerate new group M(n+1) by genetic operator using selected genesgenerate genesFig. 2. Analysis o

27、w of GA.Choose the genes which have high fitness value using GA selection methodCalculate the fitness of each gene in the current group M(n)Generate initial group M(1) randomlyaBcdefabcdefab345612cdefabcdef123456T. Kim et al. / Applied Thermal Engineering 27 (2007) 31152.5. Selection of gene typemet

28、hods of investigation and to reduce the calculating load. The following methods can be considered: (1) Using the designers decision-making models. During the process of calculation, the CFD boundary conditions are modied by the feedback system to nd the optimal design. (2) Using the optimal method t

29、o nd the boundary conditions for the optimal design.The authors conducted a basic study of the optimal design method using a feedback system. The indoor thermal environment was analyzed by the CFD-coupled simulation. The design conditions were modied by the feedback system in the method according to

30、 the simple decision-making model of the designer during the process of CFD-coupled simulation. It investigated the design con- ditions that achieved the design goal in this process. This method was eective for relatively simple design condi- tions. However, it was dicult to apply to designs with co

31、mplicated boundary conditions, because of the large vol- umes of computations. Also, depending on the decision- making model used, it did not provide a sucient range of investigation, and this created a problem of the initial conditions inuencing the investigated optimal design.The two-step optimal

32、design method was introduced to overcome these problems. In the method, the optimal design was examined by the genetic algorithms (GA) which was an optimal investigation method.Selecting a method for gene type to be manipulated is closely related to searching eciency. Widely used methods include rou

33、lette wheel selection, rank-based selection and tournament selection methods.2.6. Application of GA for design of the optimal thermal comfortCases of genotype and phenotype of GA for the optimal thermal comfort design are shown in Fig. 3. Genotype in this design is organized in arrangements of six g

34、enes. Each gene has discrete values, and has information including the shape of room and location of radiant panel. The arrange- ment of the genes reects the condition of thermal comfort design. The design conditions are modied based on gene manipulation including crossover and mutation. The change

35、of design conditions through crossover of gene type is shown in Fig. 4.2.7. Optimal design method using CFD-coupled analysis for indoor thermal environments and its problemsThere are generally many parameters to be examined in designing an indoor environment, such as the air condi- tioning system, r

36、oom shape, and cooling/heating loads of the room. It is not easy to examine all possible combina- tions of these design parameters (referred to here as the design conditions) to nd the optimal design and almost impossible to do this in a practical application, especially when using CFD, which involv

37、es a large volume of compu- tation in analyzing the indoor thermal environment. There- fore, in the optimal design method using CFD for the indoor thermal environments it is essential to use ecient3. Two-step optimal design method using genetic algorithms3.1. Introduction of GA and a simplied analys

38、is method for indoor thermal environments the rst step of investigationThe rst investigation of the optimal design was con- ducted using genetic algorithms (GA). A simplied indoor thermal environment analysis method was used for this investigation, which required a relatively small volume of computa

39、tion. Many kinds of methods were considered for this simplied analysis, however, this research used radiation/conduction coupled simulation with one point model for temperature and humidity. Even though this method presumed the perfect mixing of air temperature and humidity, it conducts a detailed a

40、nalysis of the unevenoffice configuration1Phenotype:3 location of supply inlet6 width of supply inlet (0.5m)Genotype:1 2 3 42 4 4 25 61 1location of cooling panel panel type (single type)245 surface temp.(24C)radiationeldand the thermal transmission due to theFig. 3. Phenotype and genotype for optim

41、al design of HVAC system.0.5mDouble type 20C0.5mSingle type 21C2 311222 3 1 11 10.1mSingle type21C3 40.1mDouble type 20C3 4 4Mutation412Fig. 4. Changing design parameters by GA operator.6T. Kim et al. / Applied Thermal Engineering 27 (2007) 3113.4. Procedures of two-step optimal design methodtempera

42、ture distribution on the wall surface. In the case of the radiation analysis, Monte Carlo method 6 for the view factor and Gebharts absorption factor method 7 for the radiation heat transfer were used. In this step, the optimal design candidates exceeding the predened level were selected by the GA.P

43、rocedures for the two-step optimal design method are shown in Fig. 5. Detailed analysis procedures are as follows:(1)Determine elements to evaluate the indoor thermal environment for the optimal design and its numerical evaluation method. Also, establish the design param- eters and reference value t

44、o select the optimal design candidates for numerical evaluation in the rst step (Fig. 5(a)(d). This reference value inuences the number of optimal design candidates (or volume of computation) in the second step, however, it may not be possible to predene the value in some cases. In such cases, it is

45、 sucient to dene the number of optimal design candidates (e.g., top 10 designs, top 5% designs, etc.) instead of the reference value.Investigate the optimal designs by GA. In the rst step of the investigation, the optimal design candi- dates, whose evaluation values of the indoor thermal environment

46、 were above the reference value deter- mined before the simulation, were selected. The sim-3.2. Optimal design investigation by CFD-coupled analysis the second step of investigationOnly the optimal design candidates selected in the rst step were analyzed by CFD-coupled analysis in the second step. T

47、he calculating load could be changed depending on the number of optimal design candidates. However, it was very small compared to the optimal design method using the feedback system. The candidate which had the highest evaluation value in the second step was selected as the opti- mal design.(2)3.3.

48、Evaluation of indoor thermal environment based on optimal designThere would be many kinds of the numerical methods to evaluate the indoor thermal environment quantitatively in the optimal design process. The method was selected con- sidering the objectives of the design, the performance of the simul

49、ation, and other factors. In the paper, the indoor thermal environment was evaluated based on the following functions shown in Eq. (1).pliedcoupledsimulationofradiationandconduction, which uses the one point model for tem- perature and humidity, and the Gebharts absorptioncoecient method for radiati

50、on, was used (Fig. 5(e) (g).In the second step of the investigation, the thermal environment of the optimal design candidates selected in the rst step were analyzed by the CFD- coupled simulation method. The CFD-coupled simu- lation was used to analyze the environment from the design condition selec

51、ted by the GA system as the optimal candidates in the second step. The same method as in the rst step was used for evaluation function of the thermal environment. Numerical eval- uation values dierent from those in the rst step may be obtained because the indoor temperature and humidity distribution

52、s were analyzed in detail.(3)XiDi NiOGi1totalwhere,Ototalevaluation value to determine the level of optimi- zationweighting function for the design target i evaluating function for the design target i function for dimensionlessG(i)D(i)N(i)Second step investigationFirst step investigationa. Thermal i

53、ndoor environment evaluation elements foroptimal design- Thermal comfort of human ( e.g. PMV)- Cooling/heating loads- etc.g. HVAC optimal design candidate satisfying reference valueh. Detailed indoor th ermal environmentanalysis using CFDi. Decision on final optimal HVAC design and analysis of indoo

54、r thermal environmente. Simplified indoor thermal environment analysis method- CFD with coarse mesh- Coupled simulation based onpredetermined temperature and humidity distributionCase 1b. Thermal indoor environment evaluation methods foroptimal designFrom the optimal HVAC design candidateselected us

55、ing GA, choose the one that has the highestevaluation value using CFD as the final optimal designEvaluation valuefor HVAC optimal designCase 2Optimal value 2HVACdesignc. Design elements- Room configuration- Air conditioning system- etc.Case 3.f. GA(Genetic Algorithms) Searching HVAC optimal designd.

56、 Reference value for HVAC optimal designFig. 5. Optimal indoor thermal environment design system.Optimal value 3Optimal value 1T. Kim et al. / Applied Thermal Engineering 27 (2007) 3117(4) The optimal design (Fig. 5(i) was determined with the design condition of the highest numerical evalua-Opening

57、for natural ventilationSymmetric planetion value from among the design conditions lyzed by CFD method.ana-WindowHuman model4. Examples of optimal design investigation: the optimal design of a hybrid air conditioning system with natural ventilation and radiant panels1.0In order to examine the eectiveness of the optimal design method developed, the optimal design was