Attention-Based CNN-BiGRU Hybrid Model for Enhanced Parameter Prediction and Early Fault Warning in Electric Vehicle Charging

An innovative Attention-based CNN-BiGRU hybrid model (Atten-CNN-BiGRU) that enhances the accuracy of parameter prediction and the efficiency of early fault warning during the electric vehicle charging process is proposed in this paper. The model seamlessly integrates the robust feature extraction capabilities of Convolutional Neural Networks (CNNs), the bidirectional information processing capabilities of Bidirectional Gated Recurrent Units (BiGRUs), and the precision of attention mechanisms in focusing on key information. A notable innovation lies in the dynamic allocation of computational resources based on the importance of input data, which not only strengthens the ability of the model to capture important information in input sequence data but also reduces unnecessary computations, thereby improving the efficiency of model training. Additionally, a sliding window analysis method is utilized to dynamically generate early warning thresholds, further enhancing the adaptability and accuracy of the warning system. Cross-validation experiments with various models have demonstrated the superior performance of the Atten-CNN-BiGRU hybrid model in the early warning of electric vehicle charging fault, attesting to its advantages in predictive accuracy and early warning capabilities.