An Integrated Method for Day-ahead Photovoltaic Power Prediction Considering Error Correction Strategy

PV power is difficult to predict accurately during fluctuation, and considering the correction of prediction error in PV power prediction is an effective countermeasure to increase PV power prediction accuracy. For this reason, this essay suggests a day-ahead PV power combination prediction method considering prediction error correction. Firstly, Numerical Weather Prediction (NWP) data and measured PV power are utilized to obtain the PV predicted power of every individual model through the single forecasting model. Then, the weight coefficients of each single prediction model are determined according to the gray wolf optimization algorithm, and the PV predicted power of the combined model is obtained by weighting. Finally, based on the PV predicted power acquired from the combined model and the measured PV power, an error prediction model is established using the AdaBoost algorithm to obtain the corrected intraday PV power prediction results. Validation is performed using a two-year dataset that was gathered from a PV field station located in a region of China. The prediction accuracies using different prediction models and with or without prediction error correction are compared to comprehensively assess the suggested method’s efficacy. The prediction results under diverse weather types demonstrate that the suggested error correction technique can significantly increase the accuracy of the PV power forecast and that the suggested prediction method’s performance is stable.