A Bayesian Inferred Health Prognosis and State of Charge Estimation for Power Batteries

• Published in: IEEE transactions on instrumentation and measurement Vol. 74; pp. 1 - 12
• Main Authors: Dong, Guangzhong, Shen, Fukang, Sun, Li, Zhang, Mingming, Wei, Jingwen
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.01.2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Adaptation models; Algorithms; Batteries; Bayes methods; Bayesian analysis; Bayesian optimization algorithm (BOA); Complexity; Computational complexity; Computational modeling; Electric charge; Estimation; fractional-order modeling; Gaussian process; Lithium-ion batteries; Management systems; Monitoring; Normal distribution; Parameter estimation; Parameter identification; Power management; Power sources; State of charge; state-of-charge (SOC); state-of-health (SOH); Statistical inference
• ISSN: 0018-9456; 1557-9662
• DOI: 10.1109/TIM.2024.3497053

Lithium-ion batteries serve as vital power sources across various industrial sectors, necessitating accurate modeling and state monitoring via battery management systems to ensure reliable and efficient operations. This article proposes a co-estimation scheme for state-of-charge (SOC) and state-of-health (SOH) using Bayesian inference. First, a fractional-order model (FOM) is introduced to capture battery dynamics due to its capability to describe both time-domain and frequency-domain characteristics. To address the complex parameter identification challenges inherent in FOMs, the article proposes a Bayesian optimization algorithm (BOA) which efficiently reduces computational complexity and time associated with evaluating fractional-order functions. Next, a combination of Gaussian-sum particle filter (GSPF) and recursive total least squares (RTLSs) is proposed to simultaneously estimate battery SOC and SOH. The principle of GSPF is to approximate posterior distribution by weighted Gaussian mixtures, which can avoid the time-consuming resample process of sequential-importance-resample PF while retaining its advantages. The RLTS can fully consider biased noises of SOC estimation and accumulated ampere hour measurements. Additionally, the co-estimation algorithm provides accurate estimates of crucial battery aging parameters such as capacity and internal resistance, facilitating enhanced model adaptation and estimation accuracy over the battery's entire lifespan. Finally, the proposed method is compared with several available technologies to highlight its superiorities in terms of accuracy, complexity, and robustness.