Applied Energy, Volume 360, 2024, 122807, ISSN 0306-2619, DOI:

State of Charge Estimation for a Parallel Battery Pack Jointly by Fuzzy-PI model Regulator and Adaptive Unscented Kalman Filter

Simin Penga b, Yifan Miaoa, Rui Xiongb, Jiawei Baib, Mengzeng Chengc, and Michael Pechtd
aSchool of Electrical Engineering, Yancheng Institute of Technology, Yancheng 224051, China
bJoint Laboratory for Advanced Energy Storage and Application, School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, China
cEconomic Research Institute of State Grid Liaoning Electric Power Supply Co., Ltd., Shenyang 110015, China
dCenter for Advanced Life Cycle Engineering (CALCE), University of Maryland, College Park 20742, USA

For more information about this article and related research, please contact Michael G. Pecht.


Parallel battery pack (PBP) is an important unit for its application in electric vehicles and energy storage, and precise state of charge (SOC) is the basic parameter for battery efficient operation. However, the SOC is an internal hidden immeasurable variable, and the measurable battery parameters of the PBP are limited, which makes it difficult to precisely estimate SOC for the PBP. The main efforts are as follows: An improved equivalent circuit model of the PBP is first established on the basis of the fuzzy-proportional integral model regulator, which can accurately describe the influence of battery cell inconsistency on the PBP discharging characteristics. Under constant current and UDDS operating conditions, the battery model voltage can accurately capture the measured voltage during the discharging process, especially at the final stage of discharge with the maximum voltage absolute error below 0.12 V (about 3.2%). A model-based SOC prediction algorithm using an adaptive unscented Kalman filter (AUKF) with a sliding window noise estimator is developed for the PBP. It can adaptively achieve accurate process and measurement noise statistics of the PBP for the AUKF. The SOC of the PBP can be precisely estimated using the developed method with the absolute errors below 2% even if the noise statistics are randomly given respectively. Moreover, compared to the unimproved AUKF and the Sage-Husa method, the presented algorithm illustrates the highest SOC prediction precision with the lowest root mean square error of 1.12% and the minimum mean absolute error of 1.08%.

This article is available online here for free until March 28, 2024.

[Home Page] [Articles Page]
Copyright © 2024 by CALCE and the University of Maryland, All Rights Reserved