Natural gas demand forecasting based on a subdivided forecasting model and rule-based calibration

In South Korea, with growing volatility in natural gas demand owing to the implementation of eco-friendly energy policies, accurate demand forecasting is becoming more essential. Natural gas demand in South Korea is divided into city and power generation gas. To forecast the volatile energy demand considering daily and regional characteristics, detailed mathematical models and rules to calibrate subtle variations are needed. Power generation gas is more difficult to predict because of exceptional conditions changing the demand pattern owing to sudden weather changes. We propose a subdivided mathematical model that reflects use and daily and regional characteristics. Additionally, adopting rule-based calibration improved forecasting accuracy compared with using only the mathematical model. We performed a forecasting test for one year and confirmed that the average error rate was approximately 2.9%, a substantial reduction in mean absolute percentage error (MAPE) compared to the previously employed moving average method, which validates our proposed method. [Received: October 26, 2021; Accepted: August 13, 2022]