Prediction of Day-ahead Photovoltaic Output Based on FCM-WS-CNN

In order to deal with the challenges brought by large-scale photovoltaic grid-connection to the power grid dispatching, an FCM-WS-CNN model is proposed to predict the day-ahead and minute-level photovoltaic outputs. First, the distance correlation coefficient and the principal component analysis are used to extract the comprehensive meteorological factors from the original meteorological data. Then, taking the five statistical indicators of the comprehensive meteorological factors and the historical power data as the clustering features, the fuzzy C-means clustering is used to classify the historical data according to the different weather types, and the training samples are weighted based on the membership matrix. Finally, the FCM-WS-CNN model is constructed using the training data. In the experimental analysis, the above method is compared with the CNN model and the FCM-CNN model. The results show that the effectiveness of the proposed method is verified with its higher accuracy and robustness.