Short-term Natural Gas Sales Forecasting under Cold Wave Weather Based on Error Correction Model

With the frequent occurrence of extreme weather events such as cold waves, natural gas consumption has experienced significant fluctuations, posing challenges to the forecasting model performance. To address this issue, this paper proposes a framework based on the Prophet-LSTM baseline prediction model and the Transformer-LSTNet error correction model, specifically designed to handle conditions of large number of outliers during cold waves. First, the Prophet model is used for trend decomposition, providing more effective features for the LSTM model, which generates the baseline prediction values. Then, the Transformer-LSTNet model is employed for error correction. By adding the prediction errors to the baseline forecast, the model adjusts the natural gas demand during cold wave periods, thus enhancing the stability and accuracy of natural gas load forecasting. Experimental results show that, after error correction, the MAPE for cold wave periods decreased from 7.07% to 5.54%. In conclusion, this method significantly improves the accuracy of natural gas load forecasting during cold wave conditions.