Forecasting China's electricity consumption using a new grey prediction model

Abstract

A modified grey prediction model is employed to accurately forecast China's overall and industrial electricity consumption. To this end, a novel optimized grey prediction model, combining a new initial condition and rolling mechanism, is designed on the principle of “new information priority”. The previous initial conditions possess the inherent deficiencies of having a fixed structure and poor adaptability to changing raw data. To overcome these deficiencies, the new initial condition, possessing alterable weighted coefficients, is proposed. Its generating parameters can be optimally determined by employing a particle swarm optimization algorithm according to various characteristics of the input data. In addition, to demonstrate its efficacy and applicability, the novel model is utilized to predict China's total and industrial electricity consumption from 2012 to 2014 and then compared to forecasts obtained from a range of benchmark models. The two empirical results illustrate that the novel initial condition with dynamic weighted coefficients can better adjust to the features of electricity consumption data than the previous initial conditions. They also show the superiority of the newly proposed model over the benchmark models. Within this paper, the new model is used for predicting the future values of China's total and industrial electricity consumption from 2015 to 2020.