Fault detection and diagnosis of electric vehicle battery pack combining neural network and CRDAN algorithm
TL;DR
This study proposes a two-stage approach that combines a convolutional neural network and bidirectional long short-term memory (BiLSTM) for fault detection, followed by a fault diagnosis model integrating a domain-adaptive neural network with channel attention, temporal attention, and category enhancement mechanisms.
Abstract
With the widespread use of electric vehicles, higher standards for battery pack safety and reliability have emerged, making fault detection and diagnosis essential for stable operation. This study proposes a two-stage approach that combines a convolutional neural network (CNN) and bidirectional long short-term memory (BiLSTM) for fault detection, followed by a fault diagnosis model integrating a domain-adaptive neural network with channel attention, temporal attention, and category enhancement mechanisms. The detection model achieved maximum accuracy of 97.53%, precision of 98.03%, F1 score of 0.998, and recall of 99.31%, with minimum RMSE of 0.004 and time consumption of 49 ms, significantly outperforming comparison models. For diagnosis, the model achieved an AUC of 0.987, diagnostic accuracy of 98.33%, and time consumption of 66 ms, while demonstrating higher precision in identifying short-circuit, over-charging, over-discharging, and capacity fading faults. The proposed detection and diagnostic framework operates with high efficiency and robustness, offering reliable technical support for the safe operation and maintenance of electric vehicle battery packs.