Short-Term Power Load Forecasting based on KPCA-LSTM

In the area of energy management, the short term forecast of power load is very important, which is of great significance to the safe and efficient operation of motor system. Aiming at the defects of current load forecasting methods in the accuracy of feature extraction and sequence prediction, this paper proposes a short-term load forecasting model based on KPCA and LSTM. KPCA is used to reduce the dimensionality of the original PV power data, process the nonlinear separable data, and reduce the characteristic dimension. LSTM, with its advantages of processing long sequences and avoiding gradient problems, is suitable for prediction scenarios considering time dimension and sequence information. Through the analysis of the data of a photovoltaic power station in southern China, the results show that KPCA-LSTM model has higher accuracy and stability in short-term load prediction. Compared with RF, SVM and single LSTM model, the evaluation indexes such as mean absolute error, error of mean square and root-mean-square error are reduced, and R2 and PRD are significantly improved. The efficiency and superiority of the model are verified.