Monthly Henry Hub natural gas spot prices forecasting using variational mode decomposition and deep belief network. (15th July 2021)

• Record Type: Journal Article
• Authors: Li, Jinchao; Wu, Qianqian; Tian, Yu; Fan, Liguo
• Is Part Of: Energy. Volume 227(2021)
• Publication Date: 2021-07-15
• Subjects: Natural gas prices forecasting -- Hybrid model -- Variational mode decomposition -- Particle swarm optimization algorithm -- Deep belief network; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2021.120478
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: The global trade scale of natural gas is expanding, and its price forecasting has become one of the most critical issues in the planning and operation of public utilities. In this paper, a hybrid forecasting model of monthly Henry Hub natural gas prices based on variational mode decomposition (VMD), particle swarm optimization (PSO) and deep belief network (DBN) is proposed. In addition, influencing factors of the long-term natural gas price variation are investigated and considered on the natural gas price forecasting. Empirical forecasting results validate that the newly proposed hybrid forecasting model has better forecasting performance than the traditional models. The results also show that natural gas consumption, natural gas gross withdrawals, monthly West Texas Intermediate (WTI) crude oil spot prices, the proportion of extreme high temperature weather, and the proportion of extreme low temperature weather all contribute to long-term Henry Hub natural gas spot prices forecasting to varying degrees. By comparing the accuracy of forecasting models with different combinations of influencing factors, it is found that the hybrid model with natural gas consumption and WTI crude oil spot prices has the best forecasting performance. Highlights: A hybrid model is put forward for Henry Hub natural gas price forecasting. Influencing factors are considered for irregular subcomponents forecasting. The performance of models with different influencing factors is compared. Empirical study proves the superior accuracy and stability of this hybrid model.