Short-Term PV Prediction Based on a Hybrid Algorithm Based on EMD-KPCA-LSTM Network
摘要
Photovoltaic power generation is greatly affected by multi-dimensional environmental factors, and the accuracy of PV power prediction brings about great challenges on the scheduling, safe operation and stability. Therefore, this paper uses empirical mode decomposition (EMD) integrating with KPCA to treat the original data with various environmental factors. Afterward, LSTM is used for prediction. The test results have shown that the EMD-KPCA-LSTM model has the best prediction performance by comparing with other fellow algorithms. The prediction errors, including the root mean square error (RMSE), average absolute error (MAE) and goodness of fit (R2) have been improved from 1.85, 1.42, 87.33% to 0.52, 0.42, 92.54%, respectively.