Correlation Analysis And Forecast On Industrial Electricity Demands Considering Seasonal Factors

Analysis and forecasting of different electricity demand characteristic is the basis of power grid planning and construction, and also the effective guide to the power demand side management. With the high-speed development of economy and the complex changes in industrial structure in recent years, change has occurred in certain degree about the internal structure of electricity demand. Meanwhile, the seasonal fluctuation of electricity consumption also caused a seasonal lack of electricity demand in some areas. Therefore, it is necessary to forecast and analyze the electricity consumption of different types considering seasonal factors, so as to realize detailed power consumption management and develop more economical strategy for power purchasing and selling. The following research works has been done in terms of analyzing and forecasting on industrial and domestic electricity demands.Firstly, correlation analysis is carried out on the electricity consumption of the first industrial, the second industrial, the third industrial, the urban and rural residents from three aspects of auto correlation, cross-correlation and external temperature influences. To be specific, this thesis analyzes the correlative relationship about industrial and domestic consumption in historical data, the correlative relationships among those five electricity demands during different seasons and discusses the external temperature influences to it. Correlation relationships of electricity consumptions in above three situations can be obtained after preliminary judging the power consumption trend using the correlation analysis trend chart, then calculating the Pearson coefficients of them and quantifying the closely related degree.In addition, on the basis of cross-correlative relationship between electricity consumptions and external temperature influences, vector error correction (VEC) models in seasonal divisions for electricity demand forecasting are proposed. Model building process includes the following three parts:the first part is that making a stationary test to electricity consumptions data and temperature data during different seasons; the second part is that making a co-integration test to electricity consumptions which has passed the stationary test and establishing a co-integration equation based on co-integration test result; the third part is that building VEC models in seasonal divisions based on co-integration analysis. Comparing with the traditional methods, outcomes of the case study by these models and results show that VEC models have higher accuracy.Lastly, the thesis makes a forecasting blend of two aspects of auto correlation relationship and cross-correlation relationship about industrial and domestic electricity consumptions. On the basis of auto correlation relationship, an auto-regressive integrated moving average (ARIMA) model is built according to stationary test result and correlation figures about different electricity consumptions. Monthly combined forecasting model can be obtained after calculating the monthly optimal weight distribution results according to forecasted relative errors of ARIMA model and VEC model. Finally, using these models to forecast and analyze the monthly electricity demands of one province. Comparing with single forecasting model, the results provided that the proposed models have higher accuracy and reliability.This thesis makes a correlation analysis on industrial and domestic electricity consumptions according to internal relations and external impacts, and establishes electricity demand forecasting models based on the correlation. The example results prove to be accurate and effective of the built models. In general, this thesis is of significance for electricity enterprises to control and improve economical power grid.