Research On Energy Consumption Prediction Method Of College Dormitory Building In Severe Cold Area

With the proposal of"carbon peaking and carbon neutrality"goals,the construction field,which accounts for a large proportion of total carbon emission in China,will be the key field to achieve the"double carbon"goal and energy transformation.Building energy consumption prediction is an indispensable part for both new building energy-saving design and existing building energy-saving transformation.It is particularly important to establish a building energy consumption prediction model with high prediction accuracy and wide application range.As a special kind of residential buildings,college dormitory buildings lack systematic research on their energy consumption prediction methods and energy consumption prediction models.Therefore,taking a college dormitory building in Jilin City as the research object,this paper aims to establish a prediction model of college dormitory building energy consumption in severe cold areas,summarize the influence law of building energy consumption influencing factors,further improve the prediction method of building energy consumption,and provide a basis for college dormitory building energy-saving planning and building energy-saving transformation.Firstly,referring to the architectural drawings,the reference model of college dormitory building in severe cold area is established by using De ST-h software.According to the relevant standards and specifications and the field investigation of the reference building,the parameters of the building model are determined.Through De ST-h energy consumption simulation,the average heat load index of the reference building in the heating season is 16.19 W/m~2,and the total building energy consumption index in the heating season is 38.08 k W·h/（m~2·a）.The reference building model is verified from three aspects:the standard value of building energy consumption,the standard value of heat load and the actual value of heat load.The results show that the heat load error between the model building and the actual building is 3.96%,and the heat load index error is 0.53%.The established reference model of College dormitory building is representative and can be used for the establishment of building energy consumption prediction model.Secondly,according to the model parameters of the reference building model and the corresponding standards and specifications,13 influencing factors of building energy consumption are selected,and the orthogonal test is designed and carried out.The correlation analysis and significance analysis of the test results eliminate three insignificant factors affecting energy consumption.Then the regression analysis method is used to establish the regression prediction model of college dormitory building energy consumption in severe cold area,and 54 simulation tests are carried out with De ST-h software to verify the prediction accuracy of the model.The results show that the MAPE of the regression prediction model is 1.34%,and the relative error range is 0.04%～3.13%.The model can be applied to the prediction of energy consumption of college dormitory buildings in severe cold areas.Thirdly,a variety of building energy consumption neural network prediction models are established for different energy consumption types.For static energy consumption,BP neural network is used to establish the prediction model of total energy consumption of college dormitory buildings in severe cold area.The MAPE of the model is 0.94%,which improves the prediction accuracy by 29.8%compared with the regression prediction model.For dynamic energy consumption,NARX neural network is selected to establish the hourly energy consumption prediction model of college dormitory buildings in severe cold areas,and the MAPE of model is9.63%.After network optimization,the improved NARX neural network prediction model is finally obtained.The MAPE of model is 4.4%and the prediction accuracy is improved by 54.3%.For the situation that there is no external input and the prediction is only based on the historical energy consumption data,the deep neural network LSTM is selected to predict the hourly energy consumption of buildings.The MAPE of LSTM neural network model is 3.57%,and the prediction accuracy is improved by 18.86%compared with the improved NARX neural network.Finally,this paper discusses the application of energy consumption prediction model in building energy conservation.According to the model standardization coefficient and F value,the influence degree is divided into different influence levels,and the corresponding energy-saving scheme is proposed for each influence factor according to influence level,so as to provide a basis for energy-saving planning and energy-saving transformation of college dormitory buildings.This paper summarizes the building energy consumption prediction models established in this paper,analyzes the applicability of different models from two aspects of model characteristics and prediction accuracy,and further improves the building energy consumption prediction method.