Forecasting and Evaluation on Energy Efficiency of China by a Hybrid Forecast Method

This study proposes a new hybrid forecasting methodology for short-term energy efficiency prediction, this new method composes stochastic frontier analysis-generalized autoregressive conditional heteroskedasticity (SFA-GARCH) model and radial basis function neural (RBFN) model. A three-step procedure is implemented. First, the selected independent variables are analysed via SFA-GARCH model, to present their casual relations. Second, regional energy efficiency level is evaluated based upon the time series data obtained from past ten years. Finally, the proposed hybrid model considers a 6-years ahead prediction of regional energy efficiency level. The result demonstrates good performance according to tail loss test when compared with normal SFA method, it proves that the hybrid methodology should be an appropriate measure.