Multi-dimension Association Analysis And Application For Load Of Buildings Considering Characteristics Of Crowd Behavior

With the rapid development of the economy,various international conferences and large-scale events are held in China,and various large-scale venues and commercial buildings are established.The amount of electricity of buildings is huge and growing rapidly,which have a bad impact on the safety,stability and economic operation of local grid.Therefore,multi-dimension correlation analysis for load of buildings,further understanding of behavior of energy consumers,improving the accuracy of short-term forecasting for load of buildings,and improving the reliability and economy of power system dispatching are important issues for the current power grid development.At the same time,because of the development of sensor technology and big data technology,multi-source heterogeneous user-side power big data is gradually formed.However,the current research on building load correlation analysis still stays at the level of meteorological factors and historical data.There is no characteristic factor from the big data to characterize the impact of crowd behavior on building load.Moreover,the existing method of correlation analysis is less robust and cannot adapt well to the multi-source heterogeneity of power big data.In view of the above situation,the method of correlation analysis for the data of mobile base station,which is related to the crowd behavior,and power load are studied in this thesis,based on the measured data of four buildings in Shanghai.The main contents include:First,preprocessing of the data is performed on the measured data of four buildings in Shanghai by the techniques of data cleaning,screening and integration.With the help of statistical analysis and cluster analysis method,the characteristics of the crowd behavior,which are characterized by the data of mobile base station,are analyzed.Then,considering the multi-heterogeneity of power big data,the initial data is divided into numerical data set and non-numeric data set.And an improved method of multi-dimension correlation analysis for power load is proposed.By this method,the correlation analysis for the pre-processed measured data is conducted,and the relationship between characteristics of the crowd behavior and the load of buildings is obtained.Finally,with the above relationship,the input variables of short-term forecasting for load of buildings are selected.For the above measured data,short-term forecasting for load is performed by BP neural network.The results show that considering the behavior characteristics of the crowd,carrying out the multi-dimension correlation analysis for load of buildings,and considering the data of inobile base station to be the input variables of the short-term forecasting for load of the buildings according to the analysis results,improve the accuracy of the short-term forecasting for load of buildings and benefit the safe operation and economical efficiency of the power system.