An improved grey model WD-TBGM (1,1) for predicting energy consumption in short-term

The traditional grey model has been widely used for predicting energy consumption (EC) in short-term with a small sample-size, but its accuracy is greatly affected by data fluctuation. In order to further improve the prediction performance while considering the data fluctuation, in this study, the wavelet de-noising is introduced to pre-processing the EC data as the input of a modified grey model, leading to an improved novel grey model WD-TBGM (1, 1). It is found that using a wavelet decomposition algorithm can denoise the data and then the data fluctuation is effectively reduced. After illustrating effectiveness by numerical simulation and case study, the prediction performance of this newly proposed hybrid model can be enhanced with approximately 5% compared with the classical grey models. Furthermore, this newly proposed hybrid model is used to address the issues of EC prediction in China which is one of the worldwide top ten energy consumers and in Shanghai city which is one of the top energy consumers in China. The forecasting results show that the total EC of China and Shanghai will slow down in the next few years, which is in line with their actual development situation. This research also explains the effectiveness of the energy conservation and emission reduction policies that China and Shanghai are taking.