The invention discloses a lithium ion battery fault diagnosis method based on voltage reconstruction, and the method comprises the steps: carrying out the processing of voltage values collected by a sensor in batches, and guaranteeing that the format of the voltage values can be input into a subsequent network for training and reconstruction; a voltage reconstruction TransAD network is constructed, reasoning is quickly executed according to wider time trend information in data by using an attention-based sequence encoder, and multi-modal feature extraction and adversarial training are realized by using a self-regulation mechanism based on a target score; a TransAD network trained by normal voltage data is used for reconstructing voltage, then the reconstructed voltage data and actual voltage data are subtracted, and the difference is subjected to fault division through a dynamic threshold judgment module.
本发明公开了一种基于电压重构的锂离子电池故障诊断方法,对传感器采集到的电压值进行分批次处理,保证其格式可以输入到后续的网络中进行训练和重构;构建了一个电压重构TranAD网络,通过使用基于注意力的序列编码器,根据数据中更广泛的时间趋势信息快速执行推理,使用基于目标分数的自调节机制来实现多模态特征提取和对抗训练;将用正常电压数据训练好的TranAD网络用于重构电压,然后将重构的电压数据与实际电压数据进行作差,并将差值通过动态阈值判断模块进行故障划分。
Lithium ion battery fault diagnosis method based on voltage reconstruction
一种基于电压重构的锂离子电池故障诊断方法
2024-11-29
Patent
Electronic Resource
Chinese
Model-Based Multi-Fault Diagnosis for Lithium-Ion Battery Systems
SAE Technical Papers | 2022
|Model-Based Multi-Fault Diagnosis for Lithium-Ion Battery Systems
British Library Conference Proceedings | 2022
|Lithium ion secondary battery diagnosis device and lithium ion secondary battery diagnosis method
European Patent Office | 2019
|