The Short-Term Photovoltaic Forecasting Method Based on the F-ISSA-LSTM Model

Historical power and past meteorological data can effectively reflect the fluctuation of photovoltaic power within a certain period. By combining the FCM (Fuzzy C-Means) clustering algorithm to cluster the original data, a long short-term memory neural network (LSTM) is then utilized to predict photovoltaic power generation. Concurrently, an improved sparrow search algorithm (ISSA) is applied for optimizing the hyperparameters of the neural network, aiming to achieve optimal hyperparameters for different power characteristic scenarios. A novel F-ISSA-LSTM combined model for short-term photovoltaic forecasting is established. Finally, measured data from a photovoltaic power station in East China are adopted for verification. The results indicate that compared with traditional prediction models such as BP, LSTM, and ISSA-LSTM, the proposed prediction model achieves the lowest root mean square error (RMSE), mean absolute percentage error (MAPE), and the best fit. This suggests that the prediction model exhibits higher prediction accuracy for this set of data compared to traditional prediction methods, making it more suitable for application in short-term photovoltaic forecasting.