Researchers from Taiyuan University of Technology suggested an ensemble model based on rolling grey model (RGM) and support vector regression (SVR) to forecast primary energy demand
Economic development of most countries around the world relies on adequate energy supply, especially for emerging economies that are under a transition period. Novel approaches that can estimate energy demand can aid in ensuring economic development and predicting carbon emission. Therefore, several studies have focused on research and development of models that accurately forecast energy demand. These approaches can aid policy makers regarding energy development planning and policy formulation. Now, a team of researchers from Taiyuan University of Technology suggested an adaptive inertia weight particle swarm optimization algorithm (APSO).
Moreover, the team also developed an adaptive variable weight ensemble forecasting model (APSO-RGM-SVR) to predict primary energy demand and energy structure. This model was enhanced to more accurately predict China’s primary energy demand and primary energy structure. A ‘momentum’ factor was added to particle swarm optimization (PSO) to adjust the adaptive inertia weight of the particles, which in turn can enhance the global search ability of PSO. The team compared several single-prediction models and fixed-weight ensemble models and found that the results of the APSO-RGM-SVR model demonstrated the smallest error. The ensemble model effectively integrated the characteristics of a time-series model and artificial intelligence. Moreover, the method also revealed higher accuracy and better generalization.
According to the researchers, APSO-RGM-SVR can be used for future short-term and medium-term primary energy demand forecasting and primary energy structure prediction in China. In the current test in which the team used APSO-RGM-SVR to forecast China’s primary energy consumption during 2017-2025, the model predicted significant increase in China’s energy demand. China’s primary energy demand is expected to reach 4656.41 mtce, by 2020. The method can also be used in other parts of the world for similar purposes. In further research, the team plans to test the ability of the method to predict large samples. The research was published in the journal MDPI Energies on April 8, 2019.