Study On The Correlation Between The Residential Buildings' Electricity Consumption And The Intrinsic Attributes Of Buildings And The Personal Characteristics Of Households

The annual consumption of various energy sources continues growing around the world,which causes severe problems such as significant increase in sea level,melting glaciers,deteriorating ecological environment and obvious reduction of agricultural production.These problems do pose great challenge on the sustainable development of our national economy.Given the urgent need of energy saving and considerable energy saving potential of residential dwellings,it is necessary to have an in-depth understanding of the relevant impact factors(IFs)of residential electricity consumption(REC),including dwelling intrinsic property and household characteristics.The correlation analysis between REC and its IFs is conducive to the government to formulate effective energy-saving policies,power companies to provide efficient personalized power services,and residential users to implement powerful energy-saving measures.However,at present,relevant literature work at home and abroad still focuses on the establishment of regional residen t load forecasting models,as well as the resident load elasticity models under time-sharing electricity prices.Little attention has been paid to the correlation analysis between REC and its IFs.In this paper,we firstly master the common IFs of REC,correlation measures,typical regression interpretation models and feature selecting models.Then based on the REC data and questionnaire data(p.s.the questionnaire data contains the information of residential characteristics.)that are downloaded from the official website of Ireland,we carried out correlation analysis between REC and its IFs.Firstly,based on the literature review work,we determine the IFs of REC that need to be studied in our research work,and then find the corresponding questionnaire data.The matching process of REC data and questionnaire data is done according to the ID number of Ireland residential users.And the data missing problem is also addressed.Secondly,in order to ensure the reliability of the nonlinear regression interpretation model,a nonlinear wrapper feature selecting(WFS)model based on genetic algorithm(GA)and support vector machine(SVM)is proposed to select important IFs that are linearly or nonlinearly related to REC,simultaneously removing the redundancy.In order to further verify the pros and cons of the feature selecting results of the nonlinear WFS model,this model is then compared to two common linear filtering methods and stepwise linear regression.The comparison results primarily depend on maximum information coefficient(MIC).Finally,the dependent and independent variables are preprocessed,including classification of residential customers and introduction of dummy variables.Based on the multivariate logistic algorithm(MLR),a nonlinear regression interpretation model is established to further analyze the specific impact mechanism between REC and the selected IFs.The IFs studied in our paper are mainly into four categories,including dwelling characteristics,appliance and cooking-Heating Methods,social demographics and energy-saving attitudes.The results of Ireland dataset based case study show that different IFs do have different effects on REC.Compared with the traditional linear feature selecting methods,the nonlinear WFS model proposed in this paper can effectively identify the nonlinear complex relationship between REC and its IFs,simultaneously removing the redundancy among the remained IFs.According to the MLR based explanatory model,it is found that the residents' energy-saving attitude almost has no impact on the electricity consumption.Some of the rest IFs that belong to the other three categories have influence on REC to different extent.