Impact Factor:5.6

DOI number:10.1109/TIM.2023.3346509

Teaching and Research Group:Hou, W., Shi, Q., Liu, Y., Guo, L., Zhang, X., & W

Journal:IEEE Transactions on Instrumentation and Measurement

Key Words:State of charge, Estimation, Batteries, Machine learning algorithms, Mathematical models, Kalman filters, Temperature measurement

Abstract:State of charge (SOC) plays a crucial role in battery management systems (BMSs) as it significantly impacts battery lifespan and energy efficiency. However, accurately estimating SOC is challenging due to the highly nonlinear electrochemical characteristics of batteries. Traditional machine learning algorithms struggle with substantial SOC estimation errors and limited convergence capability, particularly when faced with significant current fluctuations. Recursive algorithms heavily depend on battery models, which can introduce uncertainties during computation, potentially leading to system instability or divergence. Therefore, achieving accurate SOC monitoring has become a significant technical obstacle. To address these challenges, this study presents a novel SOC estimation algorithm named XGBoost-ACKF, which combines the strengths of extreme gradient boosting (XGBoost) and adaptive cubature Kalman filter (ACKF). XGBoost establishes a nonlinear mapping model between input and output characteristics, while ACKF filters and estimates the SOC approximation obtained from XGBoost. This integration yields highly accurate SOC estimation. Experimental results demonstrate that the proposed algorithm surpasses existing approaches in terms of estimation accuracy, generalization ability, and error convergence. Across various temperatures and testing conditions, the algorithm achieves a mean absolute error (MAE) and root-mean-square error (RMSE) of less than 1.06% and 1.25%, respectively. Furthermore, both RMSE and MAE are reduced by over 20% compared to the extended Kalman filter (EKF) and gradient-boosted decision tree (GBDT) methods. These findings establish the proposed algorithm as a promising solution for accurate SOC estimation in BMS applications.

Indexed by:Journal paper

Discipline:Engineering

Document Type:J

Volume:73

Page Number:2504611

Translation or Not:no

Date of Publication:2024-01-06

Included Journals:SCI

Links to published journals:https://ieeexplore.ieee.org/document/10373171