Research on fault diagnosis and control strategies for battery packs

With the rapid electrification of the automotive industry, the traction battery pack has become the most safety-critical subsystem of an electric vehicle, and any insulation or communication fault must be identified and managed reliably. This work proposes a fuzzy-theory-based diagnosis and control strategy for battery packs that detects and confirms unpredictable faults under diverse operating conditions. The method couples a fuzzy inference system (FIS) with a calibrated expert database: an adaptive fuzzy matching mechanism continuously adjusts the process and observation noise in the state-of-charge and terminal-voltage correction, while triangular membership functions, an IF–THEN rule base, Mamdani inference and centroid defuzzification convert the diagnostic signals into graded fault levels, with rule weights being refined after each confirmed event. Fault confirmation follows the debounce intervals and consecutive-cycle counts defined in the unified threshold tables, and the resulting fault levels activate power-limitation responses. Validation through simulation and vehicle experiments reveals that the judgment time for insulation faults lengthens by 4.6 s (from 3.0 s to 7.6 s), whereas the corresponding accuracy climbs from 81.1% to 92.3%; for communication faults, the accuracy rises from 91.4% to 97.7%, and the interval between fault confirmation and completed power limitation is 16.1 s in simulation, 8.2 s on the test bench, and 24 s under real-world driving. This modest rise in diagnostic time remains within design tolerances, while the achieved accuracy outperforms that of comparable studies and the protective actions taken after confirmation safeguard vehicle occupants and the overall system.