Modeling a hybrid methodology for evaluating and forecasting regional energy efficiency in China

This study proposes a new hybrid methodology for short-term prediction of energy efficiency. This new method consists of the stochastic frontier analysis-generalised autoregressive conditional heteroskedasticity (SFA-GARCH) model and the radial basis function neural (RBFN) model. The study finds that 30 regions (provinces and municipalities) in China have cluster-hetergeneity, and the different levels of industry structure, technology content and energy resources in the different regions lead to dissimilar energy saving quotas. In addition, through fair comparison between the traditional GARCH model and the new hybrid model, it is proved that the new hybrid model shows good performance and the results are reasonable. The energy efficiency indicators predicted by the hybrid model appear to be more reliable than the summation of the individual forecasts because it avoids the superposition of errors.