A Joint Online Strategy of Measurement Outliers Diagnosis and State of Charge Estimation for Lithium-Ion Batteries

This article develops a joint diagnosis and estimation algorithm for state of charge of lithium-ion batteries subject to sensor measurement outlier. By means of the chi-square test mechanism, an online-outlier-detection method is put forward to detect and further diagnose the type of outliers. Compared with the traditional data-driven-based fault detection approach that relies on a great amount of historical data for training in which each iteration only requires the information from the previous moment such that the computational complexity relieves fairly. Different from the existing filtering methods, which are vulnerable to the corrupted measurements from the current and voltage sensor caused by unexpected outliers, this research involves measurement outliers in the design of the estimator. Then, combined with the extended Kalman filtering algorithm and the Holt's two-parameter linear exponential smoothing method (Holt method), an outlier-resistant Kalman filtering algorithm is proposed to prevent the outlier-induced effect from degrading the estimation performance. Finally, extensive experiments are conducted to validate the serviceability and practicability of the proposed strategy.

Automated summary

This article proposes a joint diagnosis and estimation algorithm for lithium-ion batteries' state of charge with sensor measurement outlier. An online outlier detection method is introduced using the chi-square test mechanism to diagnose outlier types. The computational complexity is reduced by requiring only the previous moment's information for each iteration. An outlier-resistant Kalman filtering algorithm is developed by combining the extended Kalman filtering algorithm and the Holt method to prevent the degradation of estimation performance caused by outlier-induced effects. Extensive experiments validate the effectiveness and practicability of the proposed strategy.