Electric Vehicle Load Prediction Based on CNN-BILSTM and XGBoost Combined Model

To address the issue of low prediction accuracy in single models due to the high volatility and redundancy of electric vehicle (EV) load data, this paper proposes a combined prediction method using CNN-BILSTM and XGBoost. The original electric vehicle (EV) load sequence is initially split into high-frequency and low-frequency components using adaptive noise complete ensemble empirical mode decomposition. The high-frequency components are processed by the CNN-BILSTM model, whereas the low-frequency components are handled by the XGBoost model. Hyperparameter optimization is refined using the improved Dung Beetle Optimization (DBO) algorithm. Experimental results indicate that this approach surpasses traditional methods in performance, including single CNN-BILSTM, CNN-GRU, and XGBoost models, in both prediction accuracy and speed.