Online multi-fault detection and diagnosis for battery packs in electric vehicles

Abstract

Rapid detection and accurate diagnosis of faults are essential to safe operation of battery packs in electric vehicles. However, the misdiagnosis happens occasionally because of similar signatures of cell faults, sensor faults and connection faults. In this paper, an online multi-fault diagnostic method is proposed based on a non-redundant crossed-style measurement circuit and improved correlation coefficient method. In the measurement circuit, each sensor measures the voltage sum of two neighboring cells and one connection part without increasing the hardware cost. The correlation coefficient method is used to catch fault signatures and assess the fault degree. By applying these two methods, the cell faults can be distinguished from other faults by identifying the correlation coefficient of neighboring voltages with fault flags. Furthermore, connection faults and voltage sensor faults are isolated by the correlation coefficient of the neighboring voltages difference and current. The multi-fault diagnostic method can avoid false fault detection among different faults, and ensure high robustness to normal measurement errors and battery inconsistencies of ambient temperature, state of charge, and state of health. The feasibility and advantage are validated by theoretical analysis and comparative study of experimental results.