An energy entropy-based diagnostic method for voltage abnormality in real-world battery systems

Lithium-ion batteries, as crucial energy storage solutions in the context of renewable energy and electric vehicle development, present significant risks of fault and malfunction, highlighting the urgent need for in-depth analysis to ensure their performance, safety, and longevity. To address the challenges faced by traditional fault diagnosis methods in detecting early battery abnormalities, this paper proposes an innovative diagnostic method based on energy entropy to detect voltage abnormalities in battery systems. By combining energy entropy with Z-scores, the method effectively enhances the sensitivity and effectiveness of voltage abnormality detection. The energy entropy significantly reduces system response time, improving the timeliness in fault detection and preventing potential hazards. The Z-score reduces the discrepancies impact between cells, enabling precise abnormality detection. By incorporating sliding and calculation windows, the method achieves long-term voltage abnormality diagnosis. The effectiveness of the proposed method is verified through processing two complete charge-discharge cycles consisting of 6000 data points. Subsequently, its superiority is demonstrated by comparing it with the sample entropy, information entropy, Shannon entropy, and local outlier factor methods. Furthermore, application to other real-world vehicle datasets further confirms the adaptability in real-world battery troubleshooting scenarios. Finally, this study proposes a voltage abnormality diagnosis strategy based on energy entropy using real-world vehicle battery system data.