Comparison of ensemble forecasting of solar irradiance with different numbers of ensemble members

Abstract As photovoltaic (PV) power generation systems become more widespread, the instability of the electric power grids with PV connection is becoming an issue. For appropriate management of the grids, probability prediction of solar irradiance is proposed. The lagged average forecast (LAF) method is used for ensemble forecasting. 72 h ahead forecasting of solar irradiance with a numerical weather prediction (NWP) model is carried out in Thailand. Two different ensemble forecastings with NWP are performed. One has three ensemble members, and the interval of the forecasting cycle is 24 h, while the other has 12 members, and the interval is 6 h. Both forecastings work properly, and the actual frequencies of forecasting inside of the confidence interval are almost the same as the specified confidence levels. The difference in the accuracy of the two ensemble forecastings is very small, and increasing the number of ensemble members did not contribute to improved forecasting accuracy in this study. A large number of members with different characteristics are effective for efficient ensemble forecasting. However, the LAF method used in this study does not improve as the number of members is icreased. This indicates that the members generated by the LAF method have similar characteristics.