Online Multi-Fault Detection and Isolation for Battery Systems Using Improved Model-Based and Signal Processing Methods

Fast and accurate battery system fault diagnosis is essential to ensure electric vehicles' safe and reliable operation. This paper proposes an online multi-fault detection and isolation method for battery systems by combining improved model-based and signal-processing methods, which eliminates the limitation of interleaved voltage measurement topologies on traditional multiple-fault diagnostic algorithms. Residuals are generated by model-based online state estimation. An improved cumulative sum method for residual evaluation is proposed to achieve accurate sensor fault detection and isolation. According to the different residual variation characteristics of shortcircuit and connection faults, a signal processing method based on a two-sided cumulative sum is proposed to capture the deviation trend of the mean residual and further isolate short-circuit and connection faults. The effectiveness of the proposed diagnostic method is verified by the multi-fault experiment tests.