Petroleum products consumption forecasting based on a new structural auto-adaptive intelligent grey prediction model

Abstract

The demand for petroleum products has risen rapidly over the last two decades, owing mostly to the residential, transports and industrial sectors. Precise mid and long-term petroleum products consumption prediction is therefore of crucial importance for energy planning and management of strategic reserves. Thus, a new structural auto-adaptive intelligent grey model (SAIGM) is proposed as a solution to this problem. Initially, a novel time response function for predictions is theoretically determined, which addresses the standard grey model's main weaknesses. Subsequently, SAIGM is employed to determine the optimum parameter values to improve the adaptability and flexibility to confront diverse forecasting issues. Altogether, the proposed model offers a triple contribution. Firstly, SAIGM improves the predictive capabilities of intelligent grey models by making it capable of fully extracting the laws of evolution of a system, regardless of time series characteristics. Secondly, structural inflexibility and parametrization issues are addressed, so that SAIGM does not only apply to pure exponential series. Thirdly, the proposed model does not need to preprocess data nor determine the characteristics of input data. Consequently, SAIGM decreases the reliance on modeling knowledge from the standpoint of expert systems. A case study is used to check the feasibility and superiority of SAIGM. Simulation results are compared with recent intelligent grey-based models. The new model appears to benefit from its structural flexibility, since it can generate forecasts with MAPE as low as 1.54% and RMSE of 3.10. Hence, SAIGM outperforms the majority of grey models and intelligent systems known so far