공간 통계 기법을 적용한 기상 데이터 기반 태양광 발전량 예측 모델 연구

초록

Accurate photovoltaic (PV) power forecasting has become increasingly critical with the rapid expansion of renewable energy. In Korea, the Korea Power Exchange has reinforced error thresholds and settlement rules to improve forecasting accuracy, reducing the average error rate from 8.3% to 5.7%. However, stricter regulations may impose disproportionate burdens on small-scale producers and exacerbate market polarization, highlighting the need for more precise forecasting approaches. Previous studies have largely relied on short-term deep learning models using meteorological data from nearby stations, which fail to capture actual generation characteristics when geographical factors are excluded. This study proposes a deep learning-based forecasting model integrating geographical attributes of PV plants and meteorological variables such as temperature, precipitation, wind direction, wind speed, humidity, solar irradiance, sunshine hours and cloud cover. Geostatistical interpolation was applied to estimate these meteorological variables at unsampled locations, and the results were incorporated into the forecasting dataset. Experimental results demonstrate that the proposed model consistently outperforms conventional approaches, confirming that the joint consideration of geographical and meteorological factors substantially improves the accuracy of PV power forecasting.

키워드