Power Consumption Behavior Analysis And Load Forecasting Of Comprehensive Building Based On Data Mining Technology

With the development of China’s modernization,building energy consumption is rising rapidly,and the energy consumption of large-scale public buildings is much higher than that of ordinary buildings,which has a large space for energy optimization.Analyzing the energy consumption of large-scale buildings is conducive to achieving the goal of energy conservation and emission reduction.In order to establish a load model that can reflect the time and space distribution law,and then deduce the effect of energy system operation of large-scale public buildings,and lay the analysis foundation for the energy conservation of large-scale public buildings.The comprehensive building is a multi-functional commercial building with a large scale,this paper takes the comprehensive buildings as the research object and mainly does the following work:(1)This paper introduces the theoretical basis of data mining in the field of building load research.According to the research ideas,it introduces the traditional clustering algorithm,correlation coefficient and Apriori algorithm,BP neural network prediction model and depth learning model.On this basis,the method adopted in this paper is pointed out.(2)Analysis of load characteristics and power consumption behavior of integrated buildings.The comprehensive building load is analyzed statistically.The improved k-means algorithm considering the similarity of curves is constructed: the selection of K value is optimized by DBI index,and the composite measurement takes into account the characteristics of load curve in value and shape contour,which has better clustering accuracy.On this basis,using the idea of fusion clustering,the improved k-means clustering and the traditional clustering method are combined to get more reasonable clustering results,and finally complete the power consumption behavior analysis of the total load and sub load of the comprehensive building.(3)To explore the influence of external meteorological factors and air quality factors on the load of comprehensive buildings.The correlation coefficient method is used to analyze the correlation of different factors,and the apriori multi-dimensional association rule algorithm is used to analyze the correlation of different factors.The important rules are listed,and the role and significance of the rules are analyzed,so as to better understand the correlation degree between the influencing factors and the building load and the logical relationship between them.(4)Short term load forecasting of integrated buildings.Based on the previous power consumption mode division and correlation analysis results,we can get the most relevant factors that affect the building load,and then further explore the influence degree of external meteorological factors on the load change of comprehensive buildings in each power consumption mode,and on this basis,select the daily load curve similar to the predicted daily meteorological conditions as the input training sample of the prediction model,and complete For the shortage of shallow neural network,the improved deep learning model(based on Elman’s improved deep belief network and attention mechanism’s long-term and short-term memory network)is used to predict building load.Finally,it is found that the improved depth belief network model and long-term memory network model maintain better prediction accuracy for the comprehensive building load.