Multi-fault synergistic diagnosis of battery systems based on the modified multi-scale entropy

Faults of lithium batteries in their early stage in electric vehicles (EVs) are usually undetectable, and their characteristics are difficult to be extracted by conventional methods. This paper presents a novel synergistic diagnosis scheme for multiple battery faults using the modified multi-scale entropy (MMSE). The proposed MMSE can effectively extract the multi-scale features of complex battery signals in the early stages of battery faults as well as overcome the shortage of the coarse-grained mode in the standard multi-scale entropy. The simulation results on experimental data and the real-world operational vehicles show that the proposed method can effectively detect and locate multiple battery faults/abnormities before they trigger the alarm thresholds. The defined sensitivity factor can implement real-time evaluation on abnormities with high efficiency and stability, and the developed variable-calculation-window diagnosis scheme can synchronously detect and locate different fault types in real time. Furthermore, feasibility, stability, reliability, versatility, robustness, and practicality of the proposed method are separately verified using multiple sets of real-world operation data. More importantly, the proposed method also provides feasibility to effectively prevent battery thermal runaway caused by multiple battery abnormities/faults. The applications of multi-scale entropy theory is the first of its kind to battery fault diagnosis on the real-world operational vehicles.