Adaptive estimation of battery pack state of charge with optical fibre strain measurements

Battery packs are critical components in electric vehicles and energy storage systems, yet reliable pack state-of-charge (SOC) estimation remains challenging due to cell-to-cell heterogeneity, dense sensor wiring, and computation that scales with pack size when monitoring all individual cells. This study introduces an optical fibre sensing approach that replaces conventional multi-sensor networks with a more compact set of sensors capable of monitoring entire cell modules while maintaining only pack-level voltage measurements. This allows the development of an innovative strain–charge sensitivity (SCS) analysis methodology that identifies representative cells by capturing subtle mechanical behaviour changes during charging cycles. More specifically, the SCS analysis reveals distinctive peak patterns that correlate with battery ageing states, providing an accurate diagnostic indicator for cell degradation assessment. When the SCS analysis is assisted with a Gaussian Process Regression-based adaptive Unscented Kalman Filter, more accurate and robust battery pack SOC estimation can be achieved under variable operating conditions. Experimental validation using a battery pack composed of two NCR18650 cylindrical cells demonstrates exceptional performance, achieving SOC estimation with 1.28 % Mean Absolute Percentage Error under dynamic conditions, significantly outperforming conventional methods. Under static discharging conditions, the adaptive model maintains a Root Mean Square Error of 0.46, representing improvements exceeding 67 % compared to existing approaches. Additional validation using a LiFePO4 pouch cell pack confirms the effectiveness of the proposed method for different cell technologies, achieving 77.27 % improvement compared to existing methods. This integrated sensing–modelling framework represents a significant advancement in battery-pack state estimation for large-scale applications.