CN112149209B - An Optimization Method for Multi-performance Oriented Design of Buildings - Google Patents

An Optimization Method for Multi-performance Oriented Design of Buildings Download PDF

Info

Publication number
CN112149209B
CN112149209B CN202010921204.4A CN202010921204A CN112149209B CN 112149209 B CN112149209 B CN 112149209B CN 202010921204 A CN202010921204 A CN 202010921204A CN 112149209 B CN112149209 B CN 112149209B
Authority
CN
China
Prior art keywords
optimization
building
performance
model
sample
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
CN202010921204.4A
Other languages
Chinese (zh)
Other versions
CN112149209A (en
Inventor
吕石磊
王冉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University
Original Assignee
Tianjin University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tianjin University filed Critical Tianjin University
Priority to CN202010921204.4A priority Critical patent/CN112149209B/en
Publication of CN112149209A publication Critical patent/CN112149209A/en
Application granted granted Critical
Publication of CN112149209B publication Critical patent/CN112149209B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Hardware Design (AREA)
  • Biophysics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Software Systems (AREA)
  • Evolutionary Biology (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Analysis (AREA)
  • Physiology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Computational Mathematics (AREA)
  • Structural Engineering (AREA)
  • Civil Engineering (AREA)
  • Architecture (AREA)
  • Medical Informatics (AREA)
  • Genetics & Genomics (AREA)
  • Biomedical Technology (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Air Conditioning Control Device (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明公开了一种建筑多性能导向设计的优化方法，步骤1、确定建筑设计决策目标及其潜在的影响因素分别作为优化目标和优化变量，并明确约束条件，从而构建优化模型；步骤2、获取具有统计代表性的建筑性能样本库，该步骤包括优化变量样本矩阵的构建、建筑物理模型的设定、建筑性能输出矩阵的构建以及建筑多维输入‑多维输出样本矩阵的构建；步骤3，采用机器学习算法构建优化变量和优化目标之间的简化关系模型，作为代理模型；步骤4，将代理模型作为优化算法的适应度函数，执行建筑多目标优化，获取帕累托优化结果。本发明能够使得建筑性能的多目标优化过程更加科学和标准，通过耦合代理模型和多目标遗传算法减少了优化过程中对EnergyPlus软件的调用，缩短了优化时间。

Figure 202010921204

The invention discloses an optimization method for architectural multi-performance-oriented design. Step 1: Determine the architectural design decision-making target and its potential influencing factors as the optimization target and the optimization variable, and clarify the constraints, so as to construct an optimization model; Step 2, Obtaining a statistically representative building performance sample library, this step includes the construction of the optimization variable sample matrix, the setting of the building physical model, the construction of the building performance output matrix, and the construction of the building multi-dimensional input-multi-dimensional output sample matrix; Step 3, using The machine learning algorithm builds a simplified relationship model between the optimization variables and the optimization objective as a proxy model; in step 4, the proxy model is used as the fitness function of the optimization algorithm to perform multi-objective optimization of buildings and obtain Pareto optimization results. The invention can make the multi-objective optimization process of building performance more scientific and standard, reduce the call to the EnergyPlus software in the optimization process and shorten the optimization time through the coupling proxy model and the multi-objective genetic algorithm.

Figure 202010921204

Description

Optimization method for multi-performance oriented design of building
Technical Field
The invention relates to the technical field of building optimization design, in particular to an optimization method of building multi-performance guide design.
Background
Decisions at the building design stage may involve a variety of building performance indicators that affect, and even determine, the future performance of the building. Different building performance indexes may have a competitive relationship, so that different design guides should be considered during building design, and the overall performance of the building is improved. Factors influencing the building performance relate to the heat insulation performance, the air tightness, the window-wall ratio, the building orientation and the like of the outer wall, and have the characteristics of high dimensionality, high uncertainty and interactivity. The interaction between these influencing factors constitutes a very complex design space, and it is therefore time-consuming and laborious to identify the optimal design among the many alternatives.
At present, the green building design in China still mainly adopts manual decision, namely the traditional trial and error method, and designers need to continuously adjust the design scheme to meet the requirements of design standards. This conventional approach relies heavily on the designer's own experience, making manual decision-making difficult. Moreover, the method can not solve the multi-objective optimization problem and lags behind the requirements of the current market on the comprehensive performance of the building.
With the rapid development of the computer field and the introduction of high-tech building technologies and concepts such as building information models and building industrialization, the construction of an integrated and automatic green building comprehensive performance optimization system by coupling building performance simulation software and a computer algorithm gradually becomes a research hotspot. Currently, the field of academic research integrates optimization algorithms into simulation-based design processes to achieve automatic optimization of building performance. Although this method can achieve the purpose of solving the comprehensive performance of the building, it has not been widely adopted in the engineering industry because of the need to sacrifice a lot of time to perform building simulation.
Disclosure of Invention
The invention aims to provide an optimization method for building multi-performance oriented design, which constructs the optimization method for building multi-objective oriented design by coupling a building information modeling technology, an agent model technology, a machine learning algorithm and a multi-objective optimization algorithm and can enhance the decision efficiency and decision reliability of a building department.
The invention is realized by the following technical scheme:
a method for optimizing a multi-performance oriented design of a building comprises the following steps:
step 1, determining a building design decision target and potential influence factors thereof as an optimization target and an optimization variable respectively, and defining constraint conditions so as to construct an optimization model;
the optimization model comprises three parts of an optimization target, an optimization variable and a constraint condition:
the optimization target at least comprises the annual heating energy consumption demand of the building, the annual cooling energy consumption demand of the building and the annual thermal comfort level of the building, and the annual comfort hour proportion, namely the heating energy consumption density HEUI, the cooling energy consumption density CEUI and the annual comfort hour proportion are respectively used as measurement indexes; the calculation formulas of HEUI and CEUI are as follows:
Figure BDA0002666789630000021
Figure BDA0002666789630000022
wherein, EUhiAnd EUciRespectively representing the time-by-time heating energy consumption demand and the cooling energy consumption demand, NhRepresents the heating hours of the whole year, NcRepresents annual cooling hours, and M represents air-conditioning area;
the annual cumulative comfort CTR is adopted as a thermal comfort measure, and the CTR calculation formula is as follows:
Figure BDA0002666789630000023
Figure BDA0002666789630000024
wherein T represents a comfort index, NPDenotes the total number of hours, TupperAnd TlowerThe upper and lower boundaries of the thermal comfort zone;
Tupperand TlowerThe calculation formula of (a) is as follows:
Figure BDA0002666789630000031
Figure BDA0002666789630000032
wherein, TcIndicating the temperature of the indoor thermal center, TdRepresents the thermal comfort bandwidth, which is 7 ℃ when considering 80% acceptability;
the optimization variables at least comprise design parameters of heat preservation, density and specific heat of an outer wall and a roof, heat preservation of an outer window, solar heat gain coefficient, window-wall ratio and air tightness related to building form and building layout;
the constraint condition is the value range of the optimization variable and is mainly limited by the building design standard required to be met;
step 2, obtaining a building performance sample library with statistical representativeness, wherein the step comprises the construction of an optimized variable sample matrix, the setting of a building physical model, the construction of a building performance output matrix and the construction of a building multidimensional input-multidimensional output sample matrix:
the method comprises the following steps of obtaining an optimized variable sample, wherein the steps of specifying the distribution form of the optimized variable, determining a sampling method, determining the size of the sample and executing a sampling process are as follows: sampling is performed on the optimized variables, and an optimized variable sample matrix X is constructed, wherein the optimized variable sample matrix X is represented by the following formula:
Figure BDA0002666789630000033
wherein m and n are respectively an optimized variable number and a sample number;
setting a physical model of the building: setting boundary conditions in EnergyPlus software, wherein the boundary conditions at least comprise weather files, building sizes, layout and enclosure information as non-optimized parameter information;
the construction process of the building performance output matrix comprises the following steps: sequentially reading each group of optimized variables in the X and writing the optimized variables into an IDF file of EnergyPlus software, and sequentially executing annual simulation of building performance according to time step length to obtain building performance output;
constructing an optimized performance index matrix Y as shown in the following formula:
Figure BDA0002666789630000041
wherein, p is the number of the performance index;
constructing a sample matrix (total sample) of multidimensional optimization variable input-multidimensional performance output:
Figure BDA0002666789630000042
step 3, constructing a simplified relation model between the optimization variables and the optimization targets by adopting a machine learning algorithm to serve as an agent model;
and 4, taking the agent model as a fitness function of the optimization algorithm, executing building multi-objective optimization, and obtaining a pareto optimization result.
The agent model building process of step 3 comprises the following steps:
establishing a model performance measurement index, adopting an error index specified in ASHRAE criterion 14-2002 as the performance measurement index, and when the standard mean deviation NMBE and the root mean square error variation coefficient CV-RMSE are respectively less than +/-5% and +/-15%, the model is accurate and reliable;
the formula for NMBE and CV-RMSE is as follows:
Figure BDA0002666789630000043
Figure BDA0002666789630000044
wherein, yi，
Figure BDA0002666789630000051
Respectively representing a software simulation actual value, a predicted value and a mean value of the actual value, wherein n is a sample size;
dividing the total sample into two parts according to the proportion of 3:1, and respectively taking the two parts as a training set and a test set;
respectively carrying out agent model training on each building performance by taking the training set as a basis;
testing the trained agent model by taking the training set as a data base to obtain NMBE and CV-RMSE indexes of the training set;
finally, it is judged whether NMBE and CV-RMSE are less than + -5% and + -15%, respectively, to evaluate the accuracy of the model.
The optimization solution identification process of the step 4 comprises the following steps:
setting hyper-parameters including competition scale, population size, cross probability, mutation probability and maximum genetic algebra;
setting optimization variables and ranges;
setting a fitness function of the NSGA-II algorithm;
randomly generating a first generation population, wherein the population is a design scheme; calculating a fitness function value according to the fitness function;
judging whether the evolution reaches the maximum algebra; if not, selecting, crossing and mutating to generate a new population, and entering the next cycle; if yes, an optimization solution is generated.
The method has positive guiding significance for early design of the building, the provided optimization method enables the multi-objective optimization process of the building performance to be more scientific and standard, the invocation of the Energyplus software in the optimization process is reduced through the coupling agent model and the multi-objective genetic algorithm, and the optimization time is shortened.
Drawings
FIG. 1 is an overall flow chart of the optimization method of the multi-objective guidance design of the building of the present invention;
FIG. 2 is a schematic diagram of a building optimization model;
FIG. 3 is a sample database acquisition flow chart;
FIG. 4 is a flow chart of proxy model construction;
FIG. 5 is a simplified flow diagram of pareto solution set acquisition;
FIG. 6 is a flow chart of a computational process of a multi-objective optimization method;
FIG. 7 is a schematic diagram of a process in which a proxy model participates in optimization as a fitness function;
fig. 8 is a diagram illustrating the pareto optimization result of the embodiment.
Detailed Description
The following detailed description of embodiments of the invention will be made with reference to the accompanying drawings.
As shown in FIG. 1, the present invention is an overall flow chart of a method for optimizing a multi-performance oriented design of a building. The process specifically comprises the following steps:
step 1, determining a building design decision target and potential influence factors thereof as an optimization target and an optimization variable respectively, and defining constraint conditions;
and 2, acquiring a building performance sample library with statistical representativeness. The method mainly comprises the steps of constructing an optimized variable sample matrix, setting a building physical model, constructing a building performance output matrix and constructing a building multidimensional input-multidimensional output sample matrix.
Obtaining the optimized variable sample includes specifying a distribution form of the optimized variable, determining a sampling method, determining a sample size, and performing a sampling process. For the building performance optimization problem, the probability of each design scheme appearing can be considered as equal possibility, so the distribution form of the design variables is a continuous uniform variable. The sampling method used is Latin Hypercube Sampling (LHS), which is a method for approximating random sampling from multivariate parameter distribution and belongs to a layered sampling technology. Under the same precision, the sample size required by LHS is smaller than that of the simple random sampling, and is generally not less than 10 times of the number of sampling variables.
Adopting LHS to perform sampling on the optimized variables, and constructing an optimized variable sample matrix X as follows:
Figure BDA0002666789630000061
wherein m and n are respectively an optimized variable number and a sample number.
And setting a physical model of the building. Setting a physical model of the building, namely setting boundary conditions in EnergyPlus software. The method mainly comprises weather files, building size, layout, building envelope information and other non-optimized parameter information.
And (5) building a building performance output matrix. Sequentially reading each group of optimized variables in the X by adopting a Python language and writing the optimized variables into an IDF file of EnergyPlus software; and sequentially executing annual simulation of the building performance by the EnergyPlus software according to the time step length to obtain the building performance output.
The optimized performance index matrix Y is constructed as follows:
Figure BDA0002666789630000071
wherein p is the number of the performance index.
Constructing a sample matrix (total sample) of multidimensional optimization variable input-multidimensional performance output:
Figure BDA0002666789630000072
step 3, constructing a simplified relation model between the optimization variables and the optimization targets by adopting a machine learning algorithm to serve as an agent model;
and 4, taking the agent model as a fitness function of the optimization algorithm, executing building multi-objective optimization, and obtaining a pareto optimization result.
As shown in FIG. 2, the optimization model includes three parts, an optimization goal, an optimization variable and a constraint condition.
(1) Optimizing the target: the building performance greatly influenced by outdoor environment is mainly considered, the heating energy consumption requirement of the building all the year round, the cooling energy consumption requirement of the building all the year round and the thermal comfort level of the building all the year round are respectively used for heating energy consumption density HEUI (unit: kWh/m)2a) Cooling energy consumption density CEUI (unit: kWh/m2a) And annual comfort hour ratio (%) as a measure.
The calculation formulas of HEUI and CEUI are as follows:
Figure BDA0002666789630000081
Figure BDA0002666789630000082
wherein, EUhiAnd EUciRespectively representing the heating energy consumption demand and the cooling energy consumption demand (unit: kWh) by time; n is a radical ofhRepresents the heating hours of the whole year, NcNumber of annual cooling hours, M tableIndicating the air conditioning area (unit: m)2)。
On the basis of the adaptive model, the annual cumulative Comfort (CTR) is taken as a thermal comfort measure, and the calculation formula is as follows:
Figure BDA0002666789630000083
Figure BDA0002666789630000084
wherein T represents a comfort index, here room temperature, NPDenotes the total number of hours, TupperAnd TlowerThe upper and lower boundaries of the thermal comfort zone.
TupperAnd TlowerThe calculation formula of (a) is as follows:
Figure BDA0002666789630000085
Figure BDA0002666789630000086
wherein, TcIndicating the temperature of the indoor thermal center, TdRepresenting a thermal comfort bandwidth of 7 c when considering 80% acceptability.
It should be noted that the default is that the indoor hot environment meets the thermal comfort requirement during building heating and cooling. The building is in a self-running state in the transition season, and can be naturally ventilated to adjust the indoor thermal environment. The invention adopts ASHRAE-55 self-adaptive model to establish the indoor heat center temperature TcAnd the outdoor monthly average temperature TrThe relationship between the indoor temperature and the ambient temperature, to evaluate the indoor thermal environment in the transition season. Indoor neutral temperature TcThe calculation formula of (a) is as follows:
Tc＝0.31×Tr+17.8
(2) optimizing variables: passive parameters are taken as main parameters and are related to performance parameters of building components, such as design parameters of heat preservation, density, specific heat and the like of an outer wall and a roof; design parameters such as heat preservation, solar heat gain coefficient and the like of the external window, design parameters such as window-wall ratio, air tightness, building layout (orientation) and the like related to the building form.
(3) Constraint conditions are as follows: the boundaries of the optimized variables should preferably meet the requirements in the current relevant design standards, and for variables that are not well defined in the standards, it should be ensured that the prior art in the market can achieve this.
The objective function (fitness function) is as follows:
Figure BDA0002666789630000091
fig. 3 is a flowchart illustrating the sample database acquisition process according to the present invention. The method specifically comprises the following steps:
obtaining an optimized variable sample with statistical significance comprises the steps of specifying the distribution form of optimized variables, determining a sampling method, determining the size of the sample and executing a sampling process; this step is done in Python programming language, all alternatives are equally possible for early building design, thus setting the optimization variables to be continuously and uniformly distributed to fully cover the building feature space. The sampling method used is Latin Hypercube Sampling (LHS), which is a method for approximating random sampling from multivariate parameter distribution and belongs to a layered sampling technology. Under the same precision, the sample size required by LHS is smaller than that of the simple random sampling, and the sample size is generally not less than 10 times of the number of sampling variables.
Adopting LHS to perform sampling on the optimized variables, and constructing an optimized variable sample matrix X as follows:
Figure BDA0002666789630000092
wherein m and n are respectively an optimized variable number and a sample number. Each sample is a set of design solutions.
And (5) building a physical model of the building. And setting a physical model of the building, namely simulating the set boundary conditions of the building in EnergyPlus software. The method mainly comprises weather files, building size, layout, building envelope information and other non-optimized parameter information.
And (5) building a building performance output matrix. Sequentially reading each group of optimized variables in the X by adopting a Python language and writing the optimized variables into an IDF file of EnergyPlus software; and sequentially executing annual simulation of the building performance by the EnergyPlus software according to the time step length to obtain the building performance output. The step consists of two cycles, which are completed by coupling Python programming language and EnergyPlus;
the building performance output matrix Y is shown as follows:
Figure BDA0002666789630000101
wherein p is the number of the performance index.
Constructing a sample matrix of multi-dimensional optimization variable input-multi-dimensional performance output:
Figure BDA0002666789630000102
as shown in fig. 4, the agent model building process mainly includes the following five steps:
establishing a model performance measurement index: the error index specified in ASHRAE criterion 14-2002 is used as a performance measurement index, and when NMBE and CV-RMSE are respectively less than +/-5% and +/-15%, the model is accurate and reliable;
dividing samples: total samples were as follows 3:1, dividing the ratio into two parts which are respectively used as a training set and a test set;
training a model: respectively carrying out agent model training on each building performance by taking the training set as a basis;
test model and evaluation model: testing the trained agent model by taking the training set as a data base to obtain NMBE and CV-RMSE indexes of the training set;
finally, it is judged whether NMBE and CV-RMSE are less than + -5% and + -15%, respectively, to evaluate the accuracy of the model.
As shown in fig. 5, a simplified flowchart of the pareto solution set acquisition includes the following specific operation steps:
and taking the constructed proxy model of the building performance as a fitness function of the NSGA-II, setting a super parameter, executing an optimization process, and finally obtaining a group of pareto optimization solutions. In the NSGA-II algorithm, a roulette selection method and a two-point crossover are selected, and main super parameters comprise competition scale, population size, crossover probability, variation probability and maximum genetic algebra.
FIG. 6 is a flow chart of a computational process of the multi-objective optimization method. First, the optimization variables and ranges are set, and a fitness function is specified. A first generation population is then randomly generated, which is the design solution in this study. And calculating a fitness function value according to the fitness function. And judging whether the evolution reaches the maximum algebra. And if not, selecting, crossing and mutating to generate a new population, and entering the next cycle. If yes, generating an optimized solution.
FIG. 7 is a schematic diagram of a process in which a proxy model participates in optimization as a fitness function. And inputting the design scheme into the fitness function to obtain a corresponding fitness function value. The fitness function value is the optimized target value. Screening the individuals with high fitness function values is to select some design schemes with better performance according to the optimization target. Cross-mutation refers to changing component variable parameters according to cross-probability exchange and according to mutation probability, respectively. A new generation of population is a new set of design solutions.
The invention mainly provides guidance for early design of buildings and serves building designers; the method provided by the invention is not only suitable for residential buildings, but also suitable for other building types such as large public buildings and the like. The following description will be given only by taking a certain residential building as an example.
The optimization method is shown by taking a certain residential building in Tianjin city as an example. The building has three layers, namely, the north is sitting and the south is facing, and each layer is provided with three units. Total building area 1029.6m2The floor area is 31.2m × 11.0m (length × width), the layer height is 2.9m, and the figure coefficient is 0.299. The aim is to optimize the heating energy consumption, the cooling energy consumption and the indoor comfort of the building on the premise of meeting the requirements of the passive house.
(1) Firstly, determining an optimization variable and a value range. The optimization variable range mainly refers to relevant regulations of 'residential building energy-saving design standards in severe cold and cold regions' JGJ26-2010 and 'passive low-energy-consumption residential building energy-saving design standards'. The optimized variables and value ranges are shown in Table 1
TABLE 1
Figure BDA0002666789630000111
Figure BDA0002666789630000121
(2) And sampling the optimized variable by adopting an LHS method, wherein the sample size is 1100. And sequentially executing the annual dynamic performance simulation of the building by coupling Python and EnergyPlus software to obtain an optimized performance output matrix. For the convenience of optimization, heating and cooling energy consumption is integrated into an optimization target, namely the annual HVAC energy consumption (EUI) of the building.
And combining the optimized variable matrix and the optimized performance output matrix to form a total sample. In the sample, the EUI distribution ranged from about 13 to 27kWh/m2a, CTR of about 0.82-0.87%.
(3) And constructing a proxy model of the building performance by adopting a Gradient Boosting Decision Tree (GBDT) algorithm. The agent model has better fitting performance between the predicted value and the simulated value, and the fitting coefficient (R) of EUI and CTR2) 0.996 and 0.935, respectively. The performance of the proxy model meets the requirements of ASHRAE 14-2002: the EUI and CTR had 2.45% and 1.78% NMBE and 3.36% CVRMSE and 2.36% CVRMSE, respectively.
(4) And (3) performing building performance optimization by taking the building energy consumption and comfort agent model constructed by adopting the GBDT algorithm as a fitness function of the NSGA-II algorithm. Since the NSGA-II algorithm is typically used to optimize both minimization objectives simultaneously, the thermal comfort level maximization is replaced with a thermal Discomfort (DCTR) minimization, DCTR 100% -CTR. As shown in Table 2, the hyper-parameter setting for NSGA-II was determined.
TABLE 2
Figure BDA0002666789630000131

Claims (3)

1.一种建筑多性能导向设计的优化方法，其特征在于，该方法包括以下步骤：1. an optimization method of building multi-performance-oriented design, is characterized in that, this method comprises the following steps: 步骤1、确定建筑设计决策目标及其潜在的影响因素分别作为优化目标和优化变量，并明确约束条件，从而构建优化模型；Step 1. Determine the architectural design decision-making objective and its potential influencing factors as the optimization objective and optimization variable respectively, and clarify the constraints, so as to construct the optimization model; 所述优化模型包括优化目标、优化变量和约束条件三部分：The optimization model includes three parts: optimization objective, optimization variables and constraints: 优化目标至少包括建筑全年供暖能耗需求、建筑全年供冷能耗需求和建筑全年热舒适性水平，分别用供暖能耗密度HEUI、供冷能耗密度CEUI和全年舒适性小时比例作为度量指标；HEUI和CEUI的计算公式如下：The optimization objectives include at least the annual heating energy demand of the building, the annual cooling energy demand of the building, and the annual thermal comfort level of the building. As metrics; HEUI and CEUI are calculated as follows:
Figure FDA0003490791780000011
Figure FDA0003490791780000011
Figure FDA0003490791780000012
Figure FDA0003490791780000012
 其中，EUhi和EUci分别表示逐时的供暖能耗需求和供冷能耗需求，Nh表示全年采暖时数，Nc表示年度冷却小时数，M表示空调面积；Among them, EU hi and EU ci represent the hourly heating energy demand and cooling energy demand, respectively, N h represents the annual heating hours, N c represents the annual cooling hours, and M represents the air-conditioning area; 采用年度累积舒适度CTR作为热舒适措施，CTR计算公式如下：The annual cumulative comfort level CTR is used as the thermal comfort measure, and the formula for calculating CTR is as follows:
Figure FDA0003490791780000013
Figure FDA0003490791780000013
Figure FDA0003490791780000014
Figure FDA0003490791780000014
 其中，T表示舒适度指标，NP表示总小时数，Tupper和Tlower为热舒适区的上下边界；Among them, T represents the comfort index, NP represents the total number of hours, and T upper and T lower are the upper and lower boundaries of the thermal comfort zone; Tupper和Tlower的计算公式如下：The formulas for calculating T upper and T lower are as follows:
Figure FDA0003490791780000015
Figure FDA0003490791780000015
Figure FDA0003490791780000016
Figure FDA0003490791780000016
 其中，Tc表示室内热中心温度，Td表示热舒适带宽，当考虑到80％的可接受性，热舒适带宽为7℃；Among them, T c represents the indoor thermal center temperature, and T d represents the thermal comfort bandwidth. When 80% acceptability is considered, the thermal comfort bandwidth is 7°C; 优化变量至少包括外墙和屋顶的保温、密度、比热，外窗的保温、太阳得热系数，建筑形态相关的窗墙比、气密性、建筑布局这些设计参数；The optimization variables include at least the thermal insulation, density, and specific heat of external walls and roofs, thermal insulation of external windows, solar heat gain coefficient, and design parameters such as window-to-wall ratio, air tightness, and building layout related to building form; 约束条件为优化变量的取值范围，其主要受限于需要满足的建筑设计标准；The constraint condition is the value range of the optimization variable, which is mainly limited by the architectural design standards that need to be met; 步骤2、获取具有统计代表性的建筑性能样本库，该步骤包括优化变量样本矩阵的构建、建筑物理模型的设定、建筑性能输出矩阵的构建以及建筑多维输入-多维输出样本矩阵的构建：Step 2: Obtaining a statistically representative building performance sample library, this step includes the construction of the optimization variable sample matrix, the setting of the building physical model, the construction of the building performance output matrix, and the construction of the building multi-dimensional input-multi-dimensional output sample matrix: 其中，获取优化变量样本包括指定优化变量的分布形式、确定抽样方法、确定样本大小和执行抽样过程为：对优化变量执行抽样，构建优化变量样本矩阵X，如下式所示：Among them, obtaining the optimization variable samples includes specifying the distribution form of the optimization variables, determining the sampling method, determining the sample size, and executing the sampling process: sampling the optimization variables, and constructing the optimization variable sample matrix X, as shown in the following formula:
Figure FDA0003490791780000021
Figure FDA0003490791780000021
 其中，m、n分别为优化变量编号和样本个数；Among them, m and n are the optimization variable number and the number of samples, respectively; 建筑物理模型的设定：在EnergyPlus软件中设定边界条件，至少包括天气文件、建筑尺寸，布局，围护结构信息作为非优化参数信息；Setting of building physical model: Set boundary conditions in EnergyPlus software, including at least weather file, building size, layout, and envelope structure information as non-optimized parameter information; 建筑性能输出矩阵的构建过程为：依次读取X中的每组优化变量并将其写入EnergyPlus软件的IDF文件，根据时间步长依次执行建筑性能的全年模拟，得到建筑性能输出；The construction process of the building performance output matrix is as follows: read each group of optimization variables in X in turn and write them into the IDF file of the EnergyPlus software, and execute the annual simulation of building performance according to the time step to obtain the building performance output; 构建优化性能指标矩阵Y，如下式所示:Construct the optimized performance index matrix Y, as shown in the following formula:
Figure FDA0003490791780000022
Figure FDA0003490791780000022
 其中，p为性能指标的编号；Among them, p is the number of the performance index; 构建多维优化变量输入-多维性能输出的样本矩阵：Build a sample matrix of multidimensional optimization variable input-multidimensional performance output:
Figure FDA0003490791780000031
Figure FDA0003490791780000031
 步骤3，采用机器学习算法构建优化变量和优化目标之间的简化关系模型，作为代理模型；Step 3, using a machine learning algorithm to build a simplified relationship model between the optimization variables and the optimization target, as a proxy model; 步骤4，将代理模型作为优化算法的适应度函数，执行建筑多目标优化，获取帕累托优化结果。Step 4, take the surrogate model as the fitness function of the optimization algorithm, perform multi-objective optimization of the building, and obtain the Pareto optimization result. 2.如权利要求1所述的一种建筑多性能导向设计的优化方法，其特征在于，所述步骤3的代理模型构建流程包括以下步骤：2. the optimization method of a kind of building multi-performance-oriented design as claimed in claim 1, is characterized in that, the proxy model construction process of described step 3 comprises the following steps: 建立模型性能度量指标，采用ASHRAE准则14-2002中规定的误差指标作为性能度量指标，当标准平均偏差NMBE和均方根误差变异系数CV-RMSE分别小于±5％和±15％时，模型是准确可靠的；Model performance metrics are established, and the error metrics specified in ASHRAE guidelines 14-2002 are used as performance metrics. When the standard mean deviation NMBE and the root mean square error coefficient of variation CV-RMSE are less than ±5% and ±15%, respectively, the model is accurate and reliable; NMBE和CV-RMSE的计算公式如下：The calculation formulas of NMBE and CV-RMSE are as follows:
Figure FDA0003490791780000032
Figure FDA0003490791780000032
Figure FDA0003490791780000033
Figure FDA0003490791780000033
 其中，yi，
Figure FDA0003490791780000034
分别代表软件模拟实际值，预测值，实际值的均值，n为样本量；Among them, y i ,
Figure FDA0003490791780000034
Represents the actual value, predicted value, and the mean of the actual value simulated by the software, and n is the sample size; 将总样本按照3:1的比例划分为两部分，分别作为训练集和测试集；Divide the total sample into two parts according to the ratio of 3:1, which are used as training set and test set respectively; 以训练集作为基础分别对每个建筑性能进行代理模型训练；Based on the training set, surrogate model training is performed for each building performance separately; 以训练集作为数据基础对训练好的代理模型进行测试，得到训练集的NMBE和CV-RMSE指标；Test the trained surrogate model with the training set as the data base, and obtain the NMBE and CV-RMSE indicators of the training set; 最后，判断NMBE和CV-RMSE是否分别小于±5％和±15％，以评估模型的精度。Finally, judge whether the NMBE and CV-RMSE are less than ±5% and ±15%, respectively, to evaluate the accuracy of the model. 3.如权利要求1所述的一种建筑多性能导向设计的优化方法，其特征在于，所述步骤4的优化解识别流程包括以下步骤：3. the optimization method of a kind of building multi-performance oriented design as claimed in claim 1 is characterized in that, the optimization solution identification process of described step 4 comprises the following steps: 设定超参数，包括竞赛规模、种群大小、交叉概率、变异概率和最大遗传代数；Set hyperparameters, including competition size, population size, crossover probability, mutation probability and maximum genetic generation; 设定优化变量及范围；Set optimization variables and ranges; 设定NSGA-II算法的适应度函数；Set the fitness function of the NSGA-II algorithm; 随机产生第一代种群，种群即为设计方案；根据适应度函数计算适应度函数值；The first generation population is randomly generated, and the population is the design plan; the fitness function value is calculated according to the fitness function; 判断是否进化到最大的代数；若否，则进行选择、交叉和变异进而生成新的种群，并进入下一个循环；若是，则生成优化解。Judge whether it has evolved to the maximum algebra; if not, select, crossover and mutate to generate a new population and enter the next cycle; if so, generate an optimal solution.
CN202010921204.4A 2020-09-04 2020-09-04 An Optimization Method for Multi-performance Oriented Design of Buildings Expired - Fee Related CN112149209B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010921204.4A CN112149209B (en) 2020-09-04 2020-09-04 An Optimization Method for Multi-performance Oriented Design of Buildings

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010921204.4A CN112149209B (en) 2020-09-04 2020-09-04 An Optimization Method for Multi-performance Oriented Design of Buildings

Publications (2)

Publication Number Publication Date
CN112149209A CN112149209A (en) 2020-12-29
CN112149209B true CN112149209B (en) 2022-03-22

Family

ID=73890716

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010921204.4A Expired - Fee Related CN112149209B (en) 2020-09-04 2020-09-04 An Optimization Method for Multi-performance Oriented Design of Buildings

Country Status (1)

Country Link
CN (1) CN112149209B (en)

Families Citing this family (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112818458B (en) * 2021-02-26 2023-11-21 西安建筑科技大学 A building green performance design optimization method and system
CN113297660A (en) * 2021-06-05 2021-08-24 西北工业大学 Multi-objective-based construction scheme stage energy-saving optimization design mode construction method
CN114117608B (en) * 2021-11-26 2022-06-03 中国矿业大学 A fast green building energy-saving evolutionary design method based on multi-agent assistance
CN114662201B (en) * 2022-03-31 2024-01-19 中国石油大学(华东) Optimizing method for intelligent regulation and control of natural ventilation
CN114741766B (en) * 2022-04-25 2023-03-21 清华大学 Building structure design method and device based on joint structure optimization neural network
CN114896664B (en) * 2022-05-12 2023-07-11 浙江大学 Park building photovoltaic integrated enclosure structure optimization method and system
CN114611201B (en) * 2022-05-12 2022-07-22 北京中建协认证中心有限公司 Multi-objective energy-saving optimization method and system for future climate buildings
CN115935466A (en) * 2022-11-14 2023-04-07 博迈科海洋工程股份有限公司 A Method for Optimizing the Modular Enclosure Structure in the Arctic Region
CN116720247B (en) * 2023-06-12 2024-01-16 浙江大学 Multi-objective optimization design method of office building envelope based on random operation
CN116976199B (en) * 2023-07-07 2024-03-29 同济大学 Optimization method of PCM-TCG photothermal performance based on non-dominated multi-objective genetic algorithm
CN116976155B (en) * 2023-09-25 2023-12-26 青岛理工大学 BIM (building information modeling) -based building energy consumption simulation method and system
CN117574785A (en) * 2024-01-16 2024-02-20 天津大学 Zero-carbon building multi-objective optimization method based on machine learning hybrid modeling
CN118734383A (en) * 2024-05-10 2024-10-01 天津大学 Comprehensive performance multi-objective optimization method for design decision of community elderly care facilities
CN118410721B (en) * 2024-06-28 2024-09-10 安徽省交通控股集团有限公司 Energy consumption diagonal bracing design parameter optimization method for highway pile plate structure

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2002005106A1 (en) * 2000-07-07 2002-01-17 Consilient, Inc. Method and apparatus for providing process-container platforms
CN106295029A (en) * 2016-08-16 2017-01-04 重庆大学 A kind of building multi-objective optimization design of power method guided based on performance
CN109492878A (en) * 2018-10-17 2019-03-19 天津大学 A kind of evaluation method of super low energy consumption public building energy technical solution

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10937211B2 (en) * 2018-11-09 2021-03-02 Autodesk, Inc. Automated parametrization of floor-plan sketches for multi-objective building optimization tasks

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2002005106A1 (en) * 2000-07-07 2002-01-17 Consilient, Inc. Method and apparatus for providing process-container platforms
CN106295029A (en) * 2016-08-16 2017-01-04 重庆大学 A kind of building multi-objective optimization design of power method guided based on performance
CN109492878A (en) * 2018-10-17 2019-03-19 天津大学 A kind of evaluation method of super low energy consumption public building energy technical solution

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
"Evaluation of energy supply and demand in solar neighborhood";Caroline Hachem.etc;《Energy and Buildings》;20120228;全文 *
"多目标优化在建筑表皮设计中的应用";袁栋等;《城市建筑》;20180630;全文 *

Also Published As

Publication number Publication date
CN112149209A (en) 2020-12-29

Similar Documents

Publication Publication Date Title
CN112149209B (en) An Optimization Method for Multi-performance Oriented Design of Buildings
Liu et al. Enhancing building energy efficiency using a random forest model: A hybrid prediction approach
Attia et al. Assessing gaps and needs for integrating building performance optimization tools in net zero energy buildings design
CN111639823B (en) Building cold and heat load prediction method constructed based on feature set
CN112926795B (en) High-rise residential building group heat load prediction method and system based on SBO optimization CNN
CN106295029A (en) A kind of building multi-objective optimization design of power method guided based on performance
Song et al. Framework on low-carbon retrofit of rural residential buildings in arid areas of northwest China: A case study of Turpan residential buildings
CN113868954B (en) Air conditioner load prediction method and system
CN117574785A (en) Zero-carbon building multi-objective optimization method based on machine learning hybrid modeling
Lawrence et al. Application of surrogate modelling to improve the thermal performance of single-family homes through archetype development
Chen et al. Carbon emission and thermal comfort prediction model for an office building considering the contribution rate of design parameters
CN118551669A (en) Multi-objective energy-saving optimization method and device for building envelope structure based on climate prediction
Wang et al. Low-carbon-oriented commercial district urban form optimization and impact assessment analysis
CN112712213A (en) Method and system for predicting energy consumption of deep migration learning of centralized air-conditioning house
CN118916965A (en) Green building multi-objective optimization design method and device, electronic equipment and readable storage medium
CN118395576B (en) Energy-saving control method and system based on building group space morphology
CN118921726B (en) A concentrator adaptive electric energy meter data transmission method
Zhang et al. Multi-objective optimization prediction model for building parameters of photovoltaic windows based on NSGA II-BP
CN119442403A (en) A zero-carbon building design method based on AI optimization
Zhang et al. A study on source domain selection for transfer learning-based cross-building cooling load prediction
Yan et al. Developing an integrated prediction model for daylighting, thermal comfort, and energy consumption in residential buildings based on the stacking ensemble learning algorithm
Wagner et al. Towards a classification of the building stock in terms of potential energy savings through advanced control
Pan et al. Multi-objective optimization for building performance design considering thermal comfort and energy consumption
CN117349390A (en) Climate zoning refinement method based on building performance index under refinement design
CN119129081B (en) Existing building envelope thermal parameter optimization method based on genetic algorithm

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20220322