With the rapid proliferation of electric vehicles, accurate charging load forecasting has become critical for grid stability and infrastructure planning. In light of the limitations of traditional forecasting methods in addressing the complex nonlinear coupling relationships between charging loads and multi-source influencing factors, as well as their inadequacies in handling distribution heterogeneity across different spatial regions, this paper proposes a hybrid model based on the IVY-optimized Temporal Convolutional Network (TCN) and Bidirectional Long Short-Term Memory (BiLSTM). The IVY optimization algorithm is employed to adaptively determine the hyperparameters of Variational Mode Decomposition (VMD), mitigating over-decomposition or under-decomposition caused by empirical settings. This process decomposes the original load series into multiple stationary modal components. These components are then fed into the TCN-BiLSTM model, where TCN extracts local temporal features and BiLSTM captures bidirectional long-term dependencies, enabling collaborative modeling of multi-scale temporal characteristics. Experimental results based on a real electric vehicle charging load dataset collected from Shenzhen, a major metropolitan city in China, demonstrate that the proposed model significantly outperforms the comparative models in terms of prediction accuracy and stability. The achieved Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Coefficient of Determination (R2) are 0.98308, 1.3208, and 0.99039, respectively, confirming the model's high-precision forecasting capability and strong generalization performance across different spatial regions with heterogeneous charging characteristics.
Zhang et al. (Thu,) studied this question.