Scholarlink

<p>Accurate joint estimation of State of Charge (SOC) and State of Health (SOH) is essential for safe and efficient Lithium-ion (Li-ion) battery operation. However, strong system nonlinearities and non-Gaussian disturbances challenge traditional filters, causing particle degeneracy and sample impoverishment in Particle Filters (PF). We propose the Unscented Kalman–Genetic Algorithm Particle Filter (UKF-GA-PF) as a hybrid Bayesian estimator. An Unscented Kalman Filter (UKF) generates measurement-informed proposal distributions for individual particles, improving sampling efficiency. A Genetic Algorithm (GA) replaces classical resampling to preserve particle diversity and actively mitigate degeneracy. The architecture retains a coherent Bayesian framework while combining guided proposals and evolutionary resampling. We validate the UKF-GA-PF on NASA datasets B0005, B0006, and B0007 and compare it with PF variants. Results show RMSE as low as 0.57% for SOC and 0.18% for SOH. Compared to a classical PF baseline, the proposed method reduces SOC and SOH errors by up to 78% and 73%, respectively. This work establishes a new benchmark for PF-based joint state estimation and offers practical guidance for high-fidelity Battery management system (BMS) implementation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel hybrid particle filter optimized by an unscented Kalman filter and genetic algorithm for joint SOC/SOH estimation of li-ion batteries

  • 期刊论文

  • 发表日期：

    2025年8月29日

查看全文
  • Ismain Guedaouria,
  • Noureddine Doghmane,
  • Mohamed-Faouzi Harkat

摘要

Accurate joint estimation of State of Charge (SOC) and State of Health (SOH) is essential for safe and efficient Lithium-ion (Li-ion) battery operation. However, strong system nonlinearities and non-Gaussian disturbances challenge traditional filters, causing particle degeneracy and sample impoverishment in Particle Filters (PF). We propose the Unscented Kalman–Genetic Algorithm Particle Filter (UKF-GA-PF) as a hybrid Bayesian estimator. An Unscented Kalman Filter (UKF) generates measurement-informed proposal distributions for individual particles, improving sampling efficiency. A Genetic Algorithm (GA) replaces classical resampling to preserve particle diversity and actively mitigate degeneracy. The architecture retains a coherent Bayesian framework while combining guided proposals and evolutionary resampling. We validate the UKF-GA-PF on NASA datasets B0005, B0006, and B0007 and compare it with PF variants. Results show RMSE as low as 0.57% for SOC and 0.18% for SOH. Compared to a classical PF baseline, the proposed method reduces SOC and SOH errors by up to 78% and 73%, respectively. This work establishes a new benchmark for PF-based joint state estimation and offers practical guidance for high-fidelity Battery management system (BMS) implementation.