Towards robust state estimation for LFP batteries: Model-in-the-loop analysis with hysteresis modelling and perspectives for other chemistries

• Published in: Journal of energy storage 2024-07, Vol.92, p.112042, Article 112042
• Main Authors: Jöst, Dominik, Palaniswamy, Lakshimi Narayanan, Quade, Katharina Lilith, Sauer, Dirk Uwe
• Format: Article
• Language: English
• Subjects: Battery management system; Hysteresis modelling; Kalman filter; Lithium iron phosphate; Particle filter; Stationary storage system
• ISSN: 2352-152X; 2352-1538
• DOI: 10.1016/j.est.2024.112042

The accurate estimation of a battery’s state of charge (SOC) is critical in battery management systems for various applications. Lithium Iron Phosphate (LFP) batteries, preferred for their long cycle life, cost efficiency, and enhanced safety, have emerged as favourable choices for stationary storage. Yet, they still face challenges in precise SOC estimation due to the flatness and hysteresis of their open circuit voltage. Addressing this, our study integrates a hysteresis model into a third-order battery model for BMS controlling a stationary storage system in frequency containment reserve (FCR) application. We analysed three advanced SOC estimation techniques — extended Kalman filter (EKF), dual unscented Kalman filter (DUKF), and particle filter (PF) — with the hysteresis model using a model-in-the-loop (MiL) toolchain. Performance testing under a 48-hour FCR load profile showed EKF with a 4% error, DUKF achieving the best result with a 1.1% error, and PF’s performance varying between 2.9% and 4% depending on particle count. Robustness tests against initialization and current sensor errors under an 8hr profile revealed DUKF maintained a 2% error boundary irrespective of the error introduced, highlighting the hysteresis model’s effectiveness. Broadening the scope, the study also explores extending the method to other lithium-ion chemistries. •Incorporation of a hysteresis model into a battery model for BMS integration.•Integration of hysteresis model into different model-based SOC estimation approaches.•Validation of algorithms under multiple conditions by a Model-in-the-Loop toolchain.•Perspective on application of improved estimation methods on other cell chemistries.