Heating Season Natural Gas Consumption Forecasting Technology Based on Prophet+XGBoost

Natural gas, as an important clean energy source, has become increasingly prevalent in heating applications in recent years. Accurate forecasting of natural gas consumption during the heating season is crucial for ensuring heating safety and energy supply. This paper develops a model based on Prophet+XGBoost to forecast natural gas consumption for the heating season. The Prophet model captures the time series characteristics of gas consumption, including trend, seasonality, holiday effects, and error terms. The XGBoost model then corrects the error terms predicted by Prophet, enhancing overall forecast accuracy. Using a city in central-southern China as a case study, the model achieved an average absolute percentage error (MAPE) of only $2.39 \%$ and a coefficient of determination $\mathrm{(R}^{{2}}$) of ${9 4 . 2 4 \%}$ for the 2024 heating season. Comparisons with other single and combined models show that the Prophet+XGBoost combination excels across all evaluation metrics, significantly outperforming other models. This study provides strong support for natural gas companies in optimizing supply chain management and inventory control, reducing the risk of supply shortages during the heating season, and offers a basis for developing reasonable pricing strategies and marketing plans.