PV Power Prediction Method Based on EMD-KPCA-SSA-LSTM

In order to meet the precision requirements of PV power forecasting, this paper proposes a combined PV power forecasting method based on LSTM combined with the characteristics of EMD, KPCA and SSA. Firstly, EMD is used to decompose the time series of environmental parameters to obtain feature series with different time scales. Secondly, KPCA is used to screen the key features that affect photovoltaic power, which eliminates the data redundancy and correlation caused by EMD decomposition and reduces the input dimension of the model. Finally, LSTM optimized by SSA is used to model the multi-dimensional time series. The simulation experiment results of a photovoltaic power station in Shanxi Province show that the method can effectively improve the prediction accuracy of photovoltaic power.