关键词: 锂离子电池, Thevenin模型, 分数阶扩展卡尔曼, HPPC实验

Abstract:

The state-of-charge (SOC) estimation of lithium batteries is an important part of battery management, and the accuracy of SOC estimation results can directly affect the performance of battery management systems. To address the difficult problem of improving the accuracy of the SOC estimations, a time-varying equivalent circuit model and an improved fractional extended Kalman algorithm are proposed. Varying the time in model parameters was used to accurately describe the SOC of lithium batteries over the entire cycle, and the memory characteristics of fractional derivatives were used to improve the state prediction equation. Given the estimation error usually found with the traditional fractional extended Kalman reference to historical data, the adaptive noise factor was added to improve the accuracy of the algorithm. To overcome the data redundancy caused by a fractional order algorithm and the short-term memory characteristics of lithium batteries, a fixed window with a size of 20 (M=20) was designed using the sliding window idea, and the data in the window was updated in real time with the charging and discharging of the battery. Fractional order operation was carried out with 20 data in the window, which reduces the error caused by data redundancy and improves the estimation accuracy. The feasibility and accuracy of the proposed algorithm were verified using two different working conditions. The experimental results showed that the maximum error of the fractional extended Kalman estimation was 0.02, while the maximum error of the traditional extended Kalman algorithm was 0.05. The error fluctuation of the proposed algorithm was also smaller than the traditional algorithm. Overall, the proposed estimation method has a high accuracy and anti-interference ability, which will be helpful to promote the development of lithium battery management systems and the application of more accurate state-of-charge estimation methods.

Key words: lithium-ion battery, Thevenin model, fractional-order extend Kalman filtering algorithm, HPPC experiment