A Study On The Medium And Long-term Prediction Of Residential Demand For Electricity In Beijing

Having the advantage of convenience of use and clean sanitation, electric energy, as one of important energy in our country, acts as a very important role in the modern life. With the continuous development of economy, The proportion of residential electricity consumption accounted for the social electricity consumption in Beijing increases year by year, and has a tendency to continue to increase. So, analyzing of the historical characteristics of Beijing residential electricity consumption and forecasting can either provide a reference for the power sector to formulate reasonable energy-saving and emission reduction policies at macro level, or to determine reasonable step tariff when implementing ladder electricity price policy at micro level.This paper aimed at the mid-long term forecast of residential electricity consumption in Beijing by a combined model based on grey correlation analysis, grey forecasting and support vector machine (SVM) optimized by particle swarm optimization. Firstly, it analyzed the influence factors of Beijing residential electricity demand and its related properties from seven aspects. And on the basis of computing the elasticity of income, electricity price and alternative energy price of residential electricity demand, this paper analyzed the historical characteristics of Beijing residential electricity demand in different stages from 1990 to 2012. Secondly, it filtrated the key influence factors of residential electricity demand of urban and rural areas in Beijing by grey correlation analysis and generated the original data series into data of regularity by the first-order accumulative method of grey prediction to make preparation for model training. The purpose of this data processing was to weaken the influence of the random disturb factors of the original data series, and gained the new series with monotone increasing rule. Afterwards, it optimized the parameter of the support vector machine (SVM) by the particle swarm algorithm. And, it combined the key factor screening and the data processing results and the optimized support vector machine to forecast the residential electricity demand of Beijing from 2013 to 2022.To validate the feasibility and scientificity of the combination forecast model, it compared the predicted values and the real value with artificial neural network model, conventional support vector machine, and multiple linear regression model. Finally, based on the research conclusion, it also put forward the corresponding countermeasures and suggestions in terms of electricity standards defining of step tariff, the combination of step tariff policy and The peak valley price policy, and Electrical energy alternative in residential areas.