Dynamic Multi-Scale Dual Kalman Filtering for State of Charge and Capacity Estimation on Lithium Titanium Oxide Cells for Traction Applications

This paper addresses the estimation of battery state variables (state of charge, overpotentials and remaining capacity) of a battery cell based on lithium titanium oxide anode chemistry using measurements of cell current and voltage. The algorithms developed use an equivalent circuit model that mimics the electrochemical processes in the cell while being simple and computationally efficient. The observer for estimating the state of charge, namely an adaptive extended Kalman Filter, is combined with a dual estimator that runs in parallel on a slower and asynchronous time scale and estimates the cell capacity. The novelty of this approach consists in a formulation that computationally simplifies the algorithm while maintaining good estimation accuracy as the cell ages. The results obtained are robust and reliable, and are validated using operative data from batteries deployed in rolling stock fleets.