A Study Of Feature And Forecasting For China Electricity Consumption Based On Seasonal Structural Time Series Models

In China,electricity consumption is the main way of energy consumption.On the one hand,the electricity consumption usually has two kinds of features of periodicity and trend.How to reasonably and accurately describe the characteristics of electricity consumption,has been one of the important research topics of economic accounting departments and research institutions.On the other hand,during the "13th Five-Year" electricity planning,electricity consumption in China faces many challenges.The urgent task is to forecast the electricity demand in the future.It is also an important research topic that how to forecast the demand of electricity consumption and realize the steady economic growth.In this context,this paper discusses the features of China's electricity consumption from a new perspective on the measurement and prediction of two aspects.This paper from the component decomposition and state space model,introduces the theory and method of basic structure of four kinds of static STM,and applied to the electricity consumption demand of China,feature measure from 2005 to 2016 China's electricity quarter consumer demand,the selection of the optimal static state STM.Then,on the basis of the model,the STM of fixed parameter and time-varying parameters are constructed,which is applied to the forecast of electricity consumption demand in China,and to forecast the electricity consumption in 2016.Finally,compared with STM+X and static STM,based on the optimal model selection index,the effects of these two STM models are compared from the fitting effect and prediction accuracy.This paper's main conclusion are as follows:First,through the ?(1)and ?(2)processes of trend components,seasonal dummy variables and seasonal trigonometric function of two seasonal components,this paper constructed four kinds of static STM model based on state space form.The hyper parameters were estimated by the maximum likelihood method in order to draw four electric consumption demand features in China.These four kinds of static STMs describe the features of China's electric consumption,indicating that China'selectricity consumption has the trend and seasonal characteristics.According to four kinds of evaluation indies,it found that ?(2)combined with seasonal trigonometric function in STM static model,can be very good to describe China's electric consumption.Second,based on optimal static STM model in the measurement of features,the promotion of the model,get the fixed parameter form and time-varying seasonal STM+X model under the framework of state space form.Considering economic growth and climate factors,two STM+X models were used to depict the electric consumption,which the hyper parameters are estimated by maximum likelihood.These two STM+X models also can well depict the demand for electricity consumption in China.According to the evaluation index of four kinds,time-varying parameter model had significant advantages in two aspects of fitting and prediction,namely the electric consumption in the prediction of the optimal model for ?(2)process and seasonal trigonometric function of the time-varying parameter model.Third,the empirical results showed that,China electric consumption demands had stable increasing trend and influence in external shocks easily.In addition,there were seasonal effects in China electric consumption demands obviously.With effected by global warming,this seasonal effect also existed Seasonal expansion.Forth,the empirical results showed that,China macro-economic inner development and China climate condition together decided electric consumption demands.Meanwhile,economic growth and climate condition had a shock effect on electric consumption demands,and this shock effect was time-varying.