An intelligent fault diagnosis method for lithium-ion battery pack based on empirical mode decomposition and convolutional neural network

The rapid detection and accurate identification of the safety state of lithium-ion battery systems have become the main bottleneck of the large-scale deployment of electric vehicles. To solve this problem, an intelligent fault diagnosis method based on deep learning is proposed. In order to avoid the influence of noise signals on fault identification, firstly, the high-frequency noise signal is filtered by the empirical mode decomposition algorithm and Pearson correlation coefficient. Secondly, an improved voltage data processing method is proposed for the first time, which can expand the relative voltage difference between the monomer voltages in the system, facilitate CNN to quickly extract the characteristic parameters of voltage data. Thirdly, in order to meet the requirements that the training model of CNN needs a large number of samples, the method of expanding the number of samples by using a sliding window is proposed. Finally, samples are input into the trained CNN model for fault type identification, and the results show that the method has high accuracy and timeliness. In summary, the proposed method is feasible, which provides the theoretical basis for the battery system's future fault hi-erarchical management strategy.

Automated summary

This research presents an intelligent fault diagnosis method based on deep learning. The method removes high-frequency noise signals and proposes an improved voltage data processing method to achieve rapid detection and accurate identification of the safety state of lithium-ion battery systems.