Search

Based on the characteristics of natural gas demand trend, this paper proposed ARIMA model which can predict China's natural gas demand as an effective tool. Compared with the RBF neural network model and combined model, empirical results show that the accuracy and stability of the ARIMA model is best.

Energy demand system is a complex system, which is affected and controlled by many factors and external environmental in its development and evolution process. This paper selected the prediction method of correlation, in the way of literature review at first, preliminary qualitatively choose factors which influence the energy demand. Then the direct, indirect and total effect degree of each factor on energy demand were measured by the method of path analysis. On the basis of path analysis, used ridge regression to eliminate multicollinearity to forecast China's energy demand, the prediction accuracy is high, and its practicability in this model is good.

Choose factors which influence the energy demand by the method of path analysis, build radial basis function (RBF) neural network model to predict energy demand in China. The RBF neural network is trained with the actual data of the main factors affecting energy demand during 1989-2003 and energy demand during 1993-2007 as learning sample with a good fitting effect. After testing network with the actual data of the main factors affecting energy demand during 2004-2007 and energy demand during 2008-2011, higher prediction accuracy can be obtained. By comparison with the BP network, RBF network prediction model outperforms BP network prediction model, finally RBF network is applied to make prediction of energy consumption for the year 2013-2015.