An Integrated Energy Generation Prediction Method Based on Federated Learning

An integrated energy generation prediction method based on federated learning is proposed to address the current focus on single energy sources and low prediction accuracy in power generation prediction. Firstly, a prediction model based on Stacked Recurrent Neural Network (SRNN) is established based on the strong correlation between new energy generation data. Then, an integrated energy generation prediction framework based on federated learning is constructed using the federated average algorithm. SRNN is used as the local model, and the SRNN models trained on multiple different client power generation data are fused to obtain a global prediction model. Finally, input the integrated energy generation data into the prediction model to obtain long-term and short-term prediction results. Based on the Pytorch platform, experimental verification is conducted on the proposed method, and the results shows that the predicted values of the method are closest to the true values, with an accuracy of 96.52% and a MAPE of 2.09%, which is superior to other comparative methods.