Approach for Short-Term Power Load Prediction Utilizing the ICEEMDAN–LSTM–TCN–Bagging Model

Aiming at the problems that the current approaches to short-term load prediction cannot fully extract the temporal features of the historical data of electric load, and have lower forecasting accuracy and poorer generalization ability. In this paper, we propose a prediction method for the ICEEMDAN–LSTM–TCN–Bagging model, which is a blend of improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN), ensemble Learning with Bagging, long short term memory network (LSTM), and temporal convolutional network (TCN). The method first applies the ICEEMDAN method to decompose the initial load sequence in order to reduce the intricacies of the load sequence. Then, the LSTM–TCN hybrid model is utilized to fully extract the temporal features of the input data and establish the connection between the features and the output. Finally, Bagging ensemble learning algorithm is used to re-fuse multiple predictions results to improve the prediction precision and generalization ability of the model. Simulation validation using different datasets, and the results of the experiment show that the proposed method has higher prediction precision compared to the commonly used prediction methods, such as ICEEMDAN–LSTM, TCN and convolutional Neural Network–LSTM, with a root mean square error of 31.47 kW, an mean absolute error of 17.32 kW.