For the realization of carbon neutrality, cities are switching from fossil fuels to renewable energy such as solar energy. The introduction of photovoltaic (PV) power generation systems into electric power system increases every year. However, PV power generation is difficult to forecast precisely as it’s estimated by many factors. One main factor is solar radiation which is also estimated by multiple factors, and thus its forecast errors vary.
The input data included forecast radiation, forecast clearness index, 1 hour’s change of forecast radiation, 1 hour’s change of forecast clearness index, current total cloud cover (TCDC), maximum of differences of observation and surrounding lattice points’ TCDC, the standard deviation of TCDC, and change of sky conditions.
Radiation forecast errors were divided into several categories based on underestimation or overestimation and the degree of estimation error. A model was built using the random forest to forecast divided possible range of incoming radiation forecast errors. Our proposed method for estimation of the radiation forecast error range is expected to serve as a basis for future work on PV power generation forecasting.
