Identification And Repair Of Abnormal Data Of Multi-split Air Conditioner Energy Consumption Monitoring

With the advancement of building energy conservation work,most cities across the country have successively established energy consumption monitoring platforms for large office buildings.After more than ten years of development,various platforms have accumulated a large amount of energy consumption monitoring data.Energy consumption data is the basis and guarantee for building energy efficiency.Whether it is in the analysis of energy-saving potential of the building,the implementation and decomposition of energy-saving tasks or the evaluation of energy-saving amounts,etc.,energy consumption data needs to be consulted.However,the current energy consumption monitoring platforms all have the problem of "heavy quantity and light quality".The existence of a large number of problem data makes the deep-level value of energy consumption data not fully utilized.In order to solve the above-mentioned problems,this paper conducts abnormal data identification and repair research based on the energy consumption data of multi-spit air conditioners in the energy consumption monitoring platform.First of all,in view of the complex of influential factors of air conditioning energy consumption,this paper explores the energy consumption characteristics and the degree of correlation between the energy consumption of multi-line air conditioning and its actual easily accessible influencing factors.According to the problem that the lighting socket energy consumption and the indoor unit energy consumption share a monitoring circuit in the actual project,a separation method of indoor unit energy consumption and lighting socket energy consumption is proposed.The calculated indoor unit energy consumption value can fill the gap in the current repair model lacking to measure indoor influencing factors.Secondly,on the basis of clear data characteristics of abnormal energy consumption data,combined with analysis of air conditioning energy consumption characteristics,this paper based on threshold method and k-means clustering method to identify non-working time energy consumption data and working time energy consumption data.Finally,in this paper,multiple regression equations,knn clustering and BP neural network are applied to the repair of abnormal energy consumption data of multi-line air conditioning,and the repair results of the three methods are compared.BP neural network proved to be more suitable for non-linear problems of air conditioning energy consumption data repair.It is verified in the form of a case analysis that calculating the indoor unit energy consumption value helps improve the repair accuracy.And this article discusses the repair results when the proportion of abnormal data is different and the size of historical data sets is different.Suggestions are made on the capacity of historical data sets and the update cycle selection in actual projects.The research work in this paper provides a solution and theoretical support for the quality of multi-line air conditioning energy consumption data in the energy consumption monitoring platform,and is of great significance to the development of monitoring technology for energy consumption monitoring platforms in China.