An intelligent diagnosis method for battery pack connection faults based on multiple correlation analysis and adaptive fusion decision-making

Multiple lithium-ion battery cells and multi-contact connection methods increase the chances of connection failures in power battery packs, posing a significant threat to the operational safety of electric vehicles. To this end, the study proposes an intelligent diagnosis method for battery pack connection faults based on multiple correlation analysis and adaptive fusion decision-making. The method uses Pearson correlation coefficients (PCC), Spearman correlation coefficients (SCC), and Kendall correlation coefficients (KCC) to simultaneously quantify the voltage synchronization between different cells in the battery pack, which is used to shield against the voltage measurement noise. The images of the correlation series transformed by the Gramian angular field are used as the input values of the convolutional neural network to classify the fault states by combining the adaptive weights of the three correlation series images on the fault levels. The experimental data shows that the accuracy of the method is 97.75 %, which is an improvement of 4.56 %, 5.98 %, and 2.39 % over the fault diagnosis accuracy using only PCC, SCC, and KCC, respectively, and effectively avoids misdiagnosis and omission due to the limitation of a single measurement. The proposed battery pack connection fault diagnosis method is robust and reliable, and has great practical application value. •Establishment of a fault diagnosis framework based on multiple correlation analysis and adaptive fusion decision-making.•Conversion of one-dimensional electrical signals into synchronized two-dimensional images.•Feature self-learning using convolutional neural networks.•A new adaptive fusion decision-making mechanism is proposed.