On-line Identification of Liquid Metal Battery Model Using Bias Compensation Recursive Least Squares Method with Forgetting Factor

• Published in: 2018 IEEE 2nd International Electrical and Energy Conference (CIEEC) pp. 114 - 119
• Main Authors: Wang, Xian, Song, Zhengxiang, Geng, Yingsan, Wang, Jianhua
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.11.2018
• Subjects: Batteries; Bias compensation recursive least squares method; Forgetting factorcomponent; Least mean squares methods; Li-Bi liquid metal battery; Liquids; Metals; On-line parameter identification; Parameter estimation
• DOI: 10.1109/CIEEC.2018.8745921

Liquid metal battery is a new battery with high current charging and discharging capability, low cost and long service life. It has a large capacity and is suitable to be used in power grid. An accurate online identification of battery model parameters is the basis of the state of charge and state of health estimation. However, there is presently no published literature for the on-line estimation of the parameters in the liquid metal battery model. To improve the suitability of liquid metal battery model under various scenarios, such as fluctuating and SoC variation, dynamic model with parameters updated on-time is developed, based on second order RC model, using bias compensation recursive least squares method with forgetting factor (FF-BCRLS). Open circuit voltage (OCV) of this device is also estimated as a parameter of the model. Three designed working scenarios are adopted to examine the performance of the algorithm and general recursive least squares method is used as a comparison. The root mean square error and the mean relative error of the estimation using this algorithm is less than 0.01 V and 0.16%, both less than that using general RLS algorithm. The parameters of the battery, internal resistance, polarization capacitances and resistances, and OCV, are proved to be obtained easily and accurately and time-varying by this algorithm, and the maximum estimation error of the OCV is about 0.07 V. The algorithm has of high accuracy and good adaptability to different battery conditions.