A method of grey forecasting with wavelet transform for energy source consumption in china

The GM (1,1) grey forecasting model with wavelet transform was presented for predicting energy source consumption in China. In order to reduce the random of series, the time series data was decomposed onto different scales by wavelet analysis, and the decomposed time series were forecasted by grey forecasting model to obtain predictive results of original time series. In the process of wavelet transform, A trous wavelet algorithm previous of Mallat algorithm in the parallel movement of time series was adopted for wavelet decomposition and reconstruction. The data of energy source consumption in China was forecasted by this reliable model. The forecasting precision of the data from 2000 to 2002 was 89.86%, 91.06% and 97.92%, which was higher than that in GM (1,1) grey model. It shows that grey forecasting model with wavelet transform is a good method for forecasting non-stationary time series with serious random fluctuation.