Sensor fault diagnosis for lithium-ion battery packs based on thermal and electrical models. (October 2020)

• Record Type: Journal Article
• Authors: Tian, Jiaqiang; Wang, Yujie; Chen, Zonghai
• Is Part Of: International journal of electrical power & energy systems. Volume 121(2020)
• Publication Date: 2020-10
• Subjects: Battery pack -- Sensor fault diagnosis -- Particle filter -- Studentized residual -- Monte Carlo Simulation -- Residual log-likelihood ratio; Electrical engineering -- Periodicals; Electric power systems -- Periodicals; Électrotechnique -- Périodiques; Réseaux électriques (Énergie) -- Périodiques; Electric power systems; Electrical engineering; Periodicals; 621.3
• DOI: 10.1016/j.ijepes.2020.106087
• Languages: English
• ISSNs: 0142-0615

Abstract

Highlights: The thermal and electrical models are developed for the series battery pack. The multi-feature fault reasoning mechanism is established. The proposed fault diagnosis scheme can isolate sensor and battery faults. The residual evaluation method improves the accuracy of sensor fault diagnosis. The Monte Carlo simulation and studentized residual methods improve robustness. Abstract: Sensor fault diagnosis is a crucial technology for the battery management system. In this work, a sensor fault diagnosis scheme is proposed for the battery pack using equivalent models and particle filters. The thermal and electrical models of the battery pack are developed based on the Thevenin equivalent circuit model and the radial equivalent thermal model. The model parameters are identified by the recursive least square algorithm. The particle filter estimates the temperature and voltage of the battery pack, which overcomes the problems of system noise and nonlinearity. The studentized residual method based on a sliding window is developed to eliminate the residual outliers caused by uncertainty. Monte Carlo simulation is used to calculate the specific values of distribution function, mean and standard deviation for different sequences. The critical value table of the absolute value method of studentized residuals is obtained by interpolating. The cumulative sum of residual log-likelihood ratio method based on sliding windows is developed to calculate residuals. By multiple residual assessments, faults of voltage, current, temperature sensors, and batteries are detected and isolated. Different fault cases are simulated to verify the proposed scheme. The results of the experiment and simulation results verify the effectiveness of the proposed sensor fault diagnosis scheme.