An improved variational mode decomposition neural network intelligent diagnosis method for battery connection faults based on real vehicle data

Continuous vibrations/impacts in vehicle Li-ion packs loosen connectors, which may lead to connection failures and subsequent thermal runaway hazards. The characteristics of battery pack connection failures are often hidden within signals of different frequencies, making them difficult to detect promptly. Therefore, this paper proposes an intelligent fault diagnosis method for lithium-ion batteries based on an improved variational mode decomposition neural network, which can identify fault information promptly and accurately. Firstly, the Archimedes optimization algorithm is used to optimize the parameters of variational mode decomposition in order to obtain optimal parameters, and the impact of extracting different levels of intrinsic mode functions on feature extraction is analyzed. The dimensionality of extracted multi-high frequency fault features is reduced using an autoencoder, and a sliding window is introduced to recombine input signals in order to expand samples. Finally, the processed sample is input into a one-dimensional convolutional neural network model for classification, and a confusion matrix is introduced to explain reasons for diagnostic errors while real vehicle verification is conducted. The results show that this method has high accuracy and real-time performance, providing a theoretical basis for future battery management system intelligence and safety.