단기 태양광 발전량 예측을 위한 국내 기상자료별 일사량의 유사성 비교 분석

초록

Photovoltaic power generation varies depending on the weather conditions. The selection of meteoro logical data is important for securing reliable short-term forecasts. This study analyzed errors in forecasting global horizontal irradiance (GHI) and PVWatts model-based power generation for 11 sites in Korea. Automated synoptic observing system (ASOS) 2025 data were used as reference against which past ASOS data (2021–2024), a typical meteorological year, and the local data assimilation and prediction system (LDAPS) 2025 were compared. Errors were evaluated for all periods as well as sunshine periods using the mean absolute percentage error (MAPE) and R². The comparisons revealed that the predictive performance of the LDAPS 2025 is the best. In the case of sunshine periods, for GHI and power generation, the R² values were 0.832 and 0.829, and the MAPE values were 15.9% and 14.8%, respectively. These results show that the LDAPS 2025 accurately estimates weather variability at the time of forecasting . Overall, the results can be used as a basis for selecting meteorological input data for short-term photovoltaic power generation forecasting.

키워드