Short Term Forecast of Photovoltaic Power Generation Based on WOA-LSTM

Short-term prediction of photovoltaic (PV) power is vital for ensuring the safe and Uninterrupted functioning of the power system and optimizing resource allocation. In this study, we propose a short-term PV power prediction method based on WOA-LSTM. To accurately capture the dynamic performance of PV power generation systems, We carefully selected the sample data set, conducted data analysis on it, and utilized the Spearman correlation coefficient to identify valuable features for training the prediction model. The proposed WOA-LSTM model utilizes the whale optimization algorithm to determine optimal hyperparameters for LSTM, including neuron count, learning rate of the model, and batch size. A comparative analysis with other prevalent models, including LSTM, GRU, BP, RNN, WOA-GRU, WOA-RNN, and PSO-LSTM, demonstrates the superiority of our proposed model in terms of power prediction accuracy. Furthermore, our integrated model surpasses individual models in prediction precision: The whale optimization algorithm effectively reduces model errors during LSTM hyperparameter search. Therefore, our WOA-LSTM model holds significant promise for short-term PV power prediction research.