The TA-PS Model For Energy Demand Forecasting Of China

Energy is an important material resource of existence, economy development, social progress and modern civilization. The scienctific forecast of the energy demand of our country has important meaning for establishing energy development program and promoting the fast development of national economy. Based on the analysis of influencing factors of energy demand, this text presents the TA-PS model with the combination of time series forecasting and regression forecasting.The first, the structures of energy supply and comsumption are analyzed to give an overview of the energy system. Qualitative and quantitative analyses are applied to analyze the relationship between the energy demand and the influencing factors. The correlation matrix of these facotors reveals that there is redundant information among them.The second, the TA-PS model are construsted to forecast the total energy demand of China. On one hand, by making use of the time series theory, we establish a trend extrapolation model alone, and then analyse the non-trend component based on the combination of trend exoplation and ARMA, which is called the TA model. On the other hand, principal component analysis method is used to pre-process the input variables. Considering that the artificial neural network is unstable in the condition of small samples, the support vector machine is used to establish a regression model between the total energy demand and the influencing factors. The model presented in this chapter, called the PS model, has avoiding the overfitting phenomenon which is often in the artificial neural network. The TA model and PS model are combined in two types, the paralleled one and the in-series one.At last, the forecast is made according the energy data from 1978 to 2006. Result reveals that the TA-PS model in series is of more meaning than the paralleled one. We can use this model to forecast that the total energy demand of our contoury in 2010 and 2020 are 2.656 and 4.545 billion ton standard coal respectively, which is consistent which the result of CErCmA system development by Wei Yiming.