Day-Ahead Correction of Numerical Weather Prediction Solar Irradiance Forecasts Based on Similar Day Analysis

Solar forecast based on Numerical weather prediction (NWP) is widely recognized and applied for a safer and more sufficient usage of solar energy sources at present. However, there are many negative cases of forecasts due to the unescapable inherent errors of numerical techniques. Thus, aiming to correct the error of NWP, this paper proposes a day-ahead correction model, which makes use of actual observation, for correcting NWP global horizontal irradiance (GHI) forecast. A dynamic time-window is adopted to select the similar day, whose size can be adjusted adaptively. The similar days of the corrected day constitute the training set of a generalized regression neural network (GRNN) model. This work is well demand-oriented, which can improve the day-ahead correction results of GHI forecast with 15-minute interval. The data under different weather conditions are selected to test the model performance, evaluating with root-mean-square error (RMSE) and mean absolute error (MAE). The results clearly demonstrate that the accuracy has been effectively improved by combining the actual observation and NWP data. Moreover, the proposed correction model allows more accurate forecast of photovoltaic (PV) power generation.