국내 하절기 일별 최대전력 예측에 관한 연구

초록

In thistudy, we consider forecasting of daily peak load of Korea in summer season. In recent years, the highestpeak load of Korea has ben kept renewed in summer times, especialy in July and August. For this reason,forecasting the electricity demand with acuracy in summer season has become more important. In this study,we propose a peak load forecasting method based on regresion models. To do this, we first analyze summerelectricity demand in several respects, and identify features, such as, day factor, autocorelation, seasonal factor,temperature, cumulative temperature, vapor presure, summer vacation efect and so on. Then, we propose aregresion model with those factors as wel as interactions betwen them. To evaluate the performance of theproposed model, we conducted forecasting of daily peak load in July and August for 10 years (208~2017). Testresults demonstrate the MAPE (Mean Absolute Percentage Eror) of les than 2% in al years, which proves thesuperiority of the proposed method to other benchmarks including a dep learning aproach in the literature.

키워드