Research On Multi-objective Optimization Of Modern Residential Building Performance In Extremely Hot And Cold Climate Region

The progress of human society is also reflected in the improvement of the adaptability of living space to the environment.The various forms of residential buildings are not only the embodiment of architectural aesthetics,but also the integration of a large number of passive design methods and concepts.In the course of the development of modern architecture,highly standardized forms,materials and techniques still have defects in coping with the changes of different climates and environments.In order to maintain the quality of indoor living environment,building space mostly relies on active energy supply,which leads to high energy consumption.However,if the building is regarded as a machine under the comprehensive action of multiple parameters,its performance evaluation and optimization in the design process can effectively improve the efficiency in the operation process.Based on this,this paper,from the perspective of building technology,conducts field investigation on modern residential buildings in the climate zone of "extreme hot and cold alternating",and establishes the original building performance model.The multi-objective optimization model of building envelope performance was reconstructed by fitting and verifying the simulation data and operation data.LHS technology was used to sample the optimization parameters,and the parametric performance optimization tool(LBT)was used to simulate the solution,and the simulated optimization data set was obtained.Stepwise regression(SR)method was used to analyze the correlation of the characteristic parameters,and the multi-collinear relationship among the parameters was excavated,and the correlation coefficients and modeling accuracy of each parameter were obtained.Multiple machine learning algorithms(ML)are used to establish a fast prediction model of building performance,in which the artificial neural network(ANN)has the best goodness of fit(R~2)and the smallest error(RMSE,MAE).Parametric sensitivity analysis and feature importance analysis were carried out to enhance the interpretability of the model.On this basis,Geatpy in Python is used to solve the multi-objective optimization.Through testing a variety of algorithms,it is concluded that NSGA-II has the best comprehensive performance(GD,IGD,Spacing is low).Combining with the DE2 analysis platform,the non-dominated solution set obtained from multi-objective optimization is classified and discussed,and the different parameter combination forms conforming to the architectural design process are obtained,and the performance of the original building is compared and analyzed.The results show that the rapid prediction model of building performance constructed by ML can significantly improve the speed of multi-objective optimization and ensure the accuracy of optimization results.The results show that the combination forms satisfying the optimal performance have good diversity,and the appropriate design parameter combination can be selected according to different design dominant factors.The parameter combination recommended in this paper improves the spatial comfort level by 14.68% and reduces the energy consumption by 17.46% while improving the comprehensive lighting performance and ensuring the visual field requirements.In order to verify the accuracy of the results,the optimization of the three combined use of N apartment type upstairs and get better results.From the perspective of energy consumption,the decrease is about 11.37-18.22%,the DWS of comprehensive lighting performance is about 20.37-24.74%,and the PMV of comfort level is about 16.71-22.67%.