Forecasting of natural gas based on a novel discrete grey seasonal prediction model with a time power term

To effectively manage natural gas dispatchability, the energy industry needs to precisely analyze the seasonal trends of natural gas production and consumption. However, most predictive models demand data pretreatment and are labor-intensive. To address this, a novel seasonal discrete grey prediction model with a time power term was developed based on grey modeling theory. This model can identify seasonal characteristics of seasonal time series data without preprocessing thanks to its unique structural parameters. It is unbiased and efficient. Through ablation and the grey wolf optimization technique, its versatility was demonstrated. An empirical analysis of US seasonal natural gas data showed it outperformed 10 other models under five evaluation criteria. Based on the findings, policy suggestions are made. The government should build a better monitoring system, increase investment in prediction research, and formulate smart policies. The energy industry and enterprises need to boost technological innovation investment, conduct R&D (Research and Development) cooperation, and improve the market mechanism. In international cooperation, the government should actively engage in projects to ensure energy supply diversity and promote the internationalization of technologies and standards. • A novel discrete seasonal grey prediction model with time power terms is constructed in this study for long-term time series data with seasonality and trend. • The grey wolf optimizer algorithm is used to optimize the hyperparameters of the proposed DGSTM(1,1, k α ) model. • The proposed DGSTM(1,1, k α ) model achieve the SOTA (State of the Art) effect in prediction performance.