SARIMAX 모형을 이용한 발전용 가스 수요 분석 및 예측 연구

In this study, we consider the problem of forecasting daily gas demand for power generation. Natural gas, which is one of the major energy sources of South Korea, is classified into two categories by its usage: city gas and gas for power generation. Natural gas for power generation is used for producing electricity and the amount of its usage keeps increasing. It is very important to accurately forecast the natural gas demand for stable operations of the gas supply system as well as prompt response to request for power generation. In this study, time series analysis models are used for forecasting daily gas demand of power generation, that is, we use SARIMA model (Seasonal ARIMA) model and the SARIMAX model, in which an exogenous variable is additionally considered. In this study, the models were constructed through the process involving several sophisticated phases, such as, stationarity and seasonality tests, order selection, estimating parameters, and diagnostic checking of the models. To evaluate the performance of the proposed model, computational comparison experiments with various existing forecasting methods were conducted. Results of the empirical analysis show that the proposed SARIMAX outperformed benchmark methods in terms of the several accuracy measures.