An Electric Bicycle Charging Load Forecasting Method Based on CNN-Seq2Seq-Attention

To address the fire risks caused by improper electric bicycle charging, accurate load forecasting is crucial. As existing methods are difficult to apply directly, this paper proposes a load forecasting method based on CNN-Seq2Seq-Attention. The core of the method is a sequence-to-sequence (Seq2Seq) architecture that integrates a convolutional neural network (CNN), a bidirectional long short-term memory (BiLSTM), and an additive attention mechanism to improve forecasting accuracy. Experiments on a real-world dataset show that ablation studies confirm the effectiveness of the CNN and the additive attention mechanism. Comparative experiments demonstrate that the model achieves optimal MAE and MAPE metrics across different prediction horizons, confirming its practical utility under complex conditions.