Research On Influence Factors And Forecast Of Electric Power Demand In Beijing

The development of modern society could not be sustainable without the electricity.In addition to supporting the development of industry and business,electric power resources are closely related to people's daily life.In recent years,with the continuous improvement of the level of industrialization and urbanization in China,the demand for electricity has also been increasing.And the power demand has gradually fluctuate in some metropolis.Studying the inherent cyclical fluctuations of electric power demand,analyzing the impact of some relevant factors,and forecasting future electricity demand,which would help us grasp the law of electricity demand changes,allocate electricity resources scientifically,and ensure a stable economic growth.This paper identified the different inherent cycles of electricity demand in Beijing based on the complete ensemble empirical mode decomposition with adaptive noise method(CEEMDAN).The results showed that in addition to a long-term trend item,the decomposition of the Beijing electricity demand sequence consisted of four inherent cycles,about one quarter,six months,one year and two years.The results of the nonlinear Granger test showed that the inherent cycle of a quarter is mainly caused by meteorological factors such as sunshine and precipitation;the inherent cycle components of the six months are closely related to the average temperature and pressure,corresponding to the summer and winter electricity demand peak;while the annual cycle reflected temperature change.There existed a long-term cointegration relationship between economic growth and the reconstructed component sequences stem from inherent cycle of two years and trend item,and linear Granger test results indicated that economic growth will promote electricity consumption,and electricity demand will also stimulate economic growth.This article firstly built an OLS regression model,the DWH test result showed that there existed an endogenous problem between electricity consumption and economic growth.Then we relaxed some assumptions and established a non-parametric generalized additive model(GAM),the results implied that the average temperature generated significant non-linear effect on electricity consumption;the decline in electricity consumption caused by the decrease of temperature is flat,while the average temperature rise will lead to an rapidly rise in electricity consumption.However,when the average temperature between 14 to 22 degrees Celsius,the effect is not obvious.And we also found that economic growth may have a nonlinear effect on electricity consumption even the relevant statistics are not significant.Finally,we controlled the endogenous problem based on the two-stage generalized additive model(2SGAM),the results was consistent with the GAM model when investigating the exogenous meteorological factors.However,the nonlinear effect of economic growth on electricity consumption was more significant,showing a certain smooth transition characteristics,and the time period was also in line with the policy of energy conservation in Beijing.In this paper,a multi-scale forecasting model was built to predict electricity demand in Beijing based on decomposition-reconstruction-forecast-integration method.Firstly,electricity demand series was decomposed by complete ensemble empirical mode decomposition with adaptive noise method(CEEMDAN),which performs better than EMD and EEMD methods;then we reconstructed the component sequences decomposed from the original series according to MIC and economic implications of the component sequences;finally the reconstructed component sequences was forecasted by Elman neural network and integrated by support vector regression,in which process a genetic algorithm was used to optimize the parameters.The results showed that the model presented in this article performed better than some single model prediction method such as ARIMA and other multi-scale forecast model based on EMD,EEMD and CEEMDAN methods.