Research On Building Energy Consumption Evaluation Method Based On Data-driven

Building Energy Benchmarking is an important measure to evaluate the energy use of the building.Though many building energy benchmarking programs have been developed during the past decades,they hold certain limitations.Most of the existing evaluation methods for Building Energy Benchmarking are based on expert experience.These methods classify buildings according to only one feature of building,i.e.the use type,and then use statistical methods to analyze the energy consumption of buildings with different use type.However,in practical applications,due to the complexity of building structures and the diversity of building uses,there are many factors that may affect building energy consumption.Ignoring these factors may result in a misleading benchmarking.Therefore,how to find the key factors affecting energy consumption and design a reasonable evaluation model are key issues of establishing an effective building energy benchmarking.Based on the analysis and summarization of existing methods,this paper proposes a novel data-driven framework for building energy use benchmarking called DdBB.DdBB incorporates various data mining approaches and mainly includes four parts:data cleansing and statistical analysis,feature selection,building classification,and building energy benchmarking and model evaluation.Unlike traditional expert system,statistical models or simulation models,in this work,the core problems of building energy benchmarking are handled by data mining techniques combining building data characteristics,we take sensitivity analysis as a feature selection problem,and building grouping as a clustering problem.The proposed feature selection method is applied to find the most informative features impacting building energy performance,and clustering method is used to classify buildings with similar energy consumption characteristics into one class.In addition,CBECS2012 is adopted as experimental and evaluation data.Compared with widely used benchmarks provided by Energy Star,DdBB shows better performance on energy prediction accuracy and model robustness.Moreover,this paper also studied other important issues related to energy consumption,i.e.short-term electrical demand forecasting for the community.Due to a large number of households in a community and the different electricity consumption behavior of households,it would be a tedious and time-consuming task to directly forecast the electrical demand of a community.To solve this problem,this paper proposes a cluster-based aggregation forecasting strategy.The strategy first grouping the households,then forecasting the clusters’energy consumption separately,and finally aggregating the forecast results.Experimental results show that the strategy can improve the efficiency of Community electricity demand forecasting.