국내 동절기 일별 최대전력 예측에 관한 연구 :개방형 공공 데이터 활용 사례

In recent years, many organizations, especially those in the public sector, have opened their data, which are used in various management science methods for data-based decision-making. In this study, we use open public data to forecast the daily peak load of South Korea during the winter season. An accurate forecast of power demand is crucial for a stable power supply and demand, especially during the summer and winter seasons when power demand reaches its peak. We analyze the characteristics of power demand in winter and identify the autocorrelation among temperature, day, and special day factors, as well as the effects of their interaction. Based on the analysis results, we propose a regression model, which has various independent variables, including not only first-order terms but also interaction terms. To evaluate the performance of the proposed model, we gather the forecasts during the winter seasons from December 2009 to February 2019. The proposed model shows a very low forecast error in terms of the mean absolute percentage error. Comparisons with several existing forecasts, including those obtained using deep learning methods, also confirm that the proposed model shows superior forecast performance in all cases.