Study On Energy Consumption Benchmark Of Campus Buildings Based On The Importance Of Electricity Consumption Behavior And Characteristics

With the gradual expansion of Chinese schools,the proportion of school building energy consumption in total social energy consumption is increasing,so it is particularly important to study energy conservation of school buildings.The energy consumption benchmark of buildings is the basis for energy conservation calculation of buildings and the standard for measuring the reasonable level of building energy consumption.However,the existing school building energy consumption benchmark has low accuracy and poor practicality,so this paper researches the energy consumption benchmark of school buildings based on the electricity consumption behavior and the importance of building feature.First,based on the analysis of building time-series energy consumption data,a method of reclassifying buildings according to electricity behavior considering seasonal factors is proposed.Through the clustering algorithm,the buildings are reclassified according to the Euclidean distance between the electricity behaviors of different buildings at different moments,and the actual working mode of buildings in the new category is closer to the original classification.On this basis,the energy consumption benchmark for building time-series energy consumption data is determined by quartile method.Second,based on the analysis of building feature data,a regressionpeer comparison method for determining energy consumption benchmark is proposed by combining the advantages of regression analysis and peer comparison.A regression model with building energy intensity and building original classification as target values is constructed by random forest algorithm to calculate the importance of each feature.After feature selection based on the importance of features,the building features are reduced by factor analysis,and then the buildings are reclassified by clustering algorithm.Compared with the original classification,the buildings in the new classification are closer to the actual feature.On this basis,the building energy consumption benchmark for building feature data is determined.Finally,In order to focus on some special architectural feature data sets,a multi-algorithm fusion model based on linear weighted fusion is proposed based on the regression-peer comparison method and the advantage intervals of different machine learning algorithms are considered.The results of feature selection based on the importance of features of three machine learning algorithms are taken as input.The buildings are reclassified by clustering algorithm according to the reclassification effect,and the weight values of the results of each algorithm are allocated.The artificial neural network,random forest and support vector machine are weighted and fused,combining the advantages of various algorithms on different data types,and finally determining the energy consumption benchmark of the target building.This study of the energy consumption benchmark of school buildings based on electricity behavior and importance of features provides a more accurate and practical energy consumption benchmark determination method,and also provides a case support for determining the energy consumption benchmark of campuses.