Photovoltaic Power Forecasting Model based on Random Forest and BiLSTM Neural Network

Due to the strong randomness and volatility of the photovoltaic power generation process, it is difficult to accurately and effectively predict power. In response to this issue, This paper presents a photovoltaic power prediction model based on random forest (RF) and bidirectional long short term memory (BiLSTM) neural network. Firstly, fuzzy C-means (FCM) clustering is used to obtain similar day data under different meteorological conditions. Then, the random forest model is trained using the photovoltaic power feature information data of similar days to be predicted. By comparing the importance evaluation indicators of feature information, the optimal set of input feature information is obtained, and the historical data of this set is input into the BiLSTM neural network model to obtain the photovoltaic power of the predicted period. Finally, the model proposed in this paper was used for photovoltaic power prediction in different meteorological environments, and compared with several commonly used algorithms to verify the effectiveness and accuracy of the method.