Research On Multi-objective Optimization Model Of Energy-saving Residential Buildings Based On GA-BPNN

Type along with our country economy and industrial structure adjustment,the development of the urbanization and human pursuit of high content of residential building,the current social terminal energy consumption accounts for 20% of the proportion of building energy consumption will rise to 30% ~ 40%.Residential building as an important part of building,which lead to the residential building energy consumption is not only with the individual characteristics of the building itself,more with the occupant behavior consciousness,habits,etc are closely related.In the current market economy condition,multi-objective optimization of energy saving residential building has become a research hotspot.People living standard unceasing enhancement,the targets in a single,double optimization design of residential building,it is difficult to meet the people of the housing economy,comfort,solar utilization aspects of requirements.For energy saving residential building is the problem of multi-objective optimization,this paper made the following research:Firstly,to set up the typical residential model,simulation experiment was carried out,the training data.First of all,through a hierarchical sampling method,statistics of residential building is the original data,using the theory of similar inductive residential building features related to factors,multi-objective optimization analysis of similar value validation,factors can accurately reflect the features of the prototype of the simplified model to achieve purpose.Then,typical building model is set up,using Designbuilder simulation,collect data,as GA-BPNN multiobjective optimization model of training data set and test data set.Secondly,combined with the basic principle of BP neural network,genetic algorithm,using the Python language loading Tensorflow framework to build the GA-BPNN multi-objective optimization model,and then input training set training network model and optimize the network parameters,to improve model accuracy,again through the test data set to test the model,so as to determine the performance of the trained neural network model is good or bad.Can be seen from the test results,after a number of 37504 round of training,the accuracy of the model for more than98%,satisfies the requirement of experiment,so the GA-BPNN the construction of a multiobjective optimization model is feasible.Thirdly,to further verify the efficiency and practicability of the model,by contrast test of GA-BPNN multi-objective optimization model for training convergence test,the test accuracy validation,simulation verification.Finally,combined with the concrete case,carries on the multiple objective optimization analysis,the final out of the integrated optimal solution set,the designer can choose according to actual demand,proved that the model is effective and practical.