Fast and accurate state-of-charge estimation for lithium iron phosphate batteries for battery energy storage systems

Lithium iron phosphate (LFP) batteries have rapidly become a cornerstone technology in both automotive and grid energy storage due to their safety, longevity, affordability, and supply-chain stability. Inaccurate State of Charge (SOC) estimates, which in real-world LFP deployments can reach up to 30%, can impair system reliability, reduce usable capacity, accelerate battery aging, and undermine operator revenue. Conventional methods fail to maintain reliable long-term accuracy due to weak voltage observability and hysteresis effects intrinsic to LFP chemistry.

This project employs a hybrid approach combining machine learning, electrochemical impedance spectroscopy, and physics-based electrochemical and mechanistic models to enhance SOC estimation, State of Health (SOH) assessment, and Remaining Useful Life (RUL) prediction for LFP batteries. The approach integrates real-world operational data with characteristic operational signatures linked to battery performance and degradation in different applications (e.g., EV vs. stationary storage). These algorithms are validated in a hardware-in-the-loop environment to ensure readiness for real-world deployment. Accurate estimation of SOC, SOH, and RUL unlock multiple operational and economic benefits, including increased usable capacity, enhanced market participation, and streamlined battery repurposing and resale. The team plans to collaborate closely with battery manufacturers, electricity trading companies, and dispatch operators to inform the approach and facilitate integration of our solution into real-world battery operations.