Electricity Price Prediction Model Based on EEMD-KPCA-LSTM

The traditional electricity price prediction methods in the electricity market lack a certain degree of stability. In order to improve the accuracy of electricity price prediction during peak and valley periods, this paper proposes a electricity price prediction method based on EEMD-KPCA-LSTM. Firstly, quantifying the correlation between electricity price data and its influencing factors by calculating correlation, selecting main features as inputs, and using EEMD to decompose the sequence of main environmental factors. Then, KPCA is used to reduce the dimensionality of the model input and preserve the main features of the data. Finally, an EEMD-KPCA-LSTM prediction model was constructed using the LSTM network. A case study was conducted on the measured data of a certain regional electricity market, and compared with LSTM, the proposed method was verified to have higher prediction accuracy and computational efficiency. The RMSE and MAE of the EEMD-KPCA-LSTM prediction model proposed in this paper were reduced by 5.7682 and 2.4438, respectively.