Residual Bias Compensation Filter for Physics-Based SOC Estimation in Lithium Iron Phosphate Batteries

TL;DR

This paper introduces a dual EKF approach with residual bias compensation for more accurate SOC estimation in LFP batteries, especially in flat OCV regions, validated across various temperatures and operating conditions.

Contribution

The paper proposes a novel dual-filter structure that decouples residual bias estimation from electrochemical state estimation, improving SOC accuracy in flat OCV regions.

Findings

Significantly reduces SOC RMSE from 3.75% to 0.20%.

Decreases voltage RMSE from 32.8 mV to 0.8 mV.

Enhances SOC estimation accuracy across temperature ranges.

Abstract

This paper addresses state of charge (SOC) estimation for lithium iron phosphate (LFP) batteries, where the relatively flat open-circuit voltage (OCV-SOC) characteristic reduces observability. A residual bias compensation dual extended Kalman filter (RBC-DEKF) is developed. Unlike conventional bias compensation methods that treat the bias as an augmented state within a single filter, the proposed dual-filter structure decouples residual bias estimation from electrochemical state estimation. One EKF estimates the system states of a control-oriented parameter-grouped single particle model with thermal effects, while the other EKF estimates a residual bias that continuously corrects the voltage observation equation, thereby refining the model-predicted voltage in real time. Unlike bias-augmented single-filter schemes that enlarge the covariance coupling, the decoupled bias estimator refines the voltage observation without perturbing electrochemical state dynamics. Validation is conducted on an LFP cell from a public dataset under three representative operating conditions: US06 at 0 degC, DST at 25 degC, and FUDS at 50 degC. Compared with a conventional EKF using the same model and identical state filter settings, the proposed method reduces the average SOC RMSE from 3.75% to 0.20% and the voltage RMSE between the filtered model voltage and the measured voltage from 32.8 mV to 0.8 mV. The improvement is most evident in the mid-SOC range where the OCV-SOC curve is flat, confirming that residual bias compensation significantly enhances accuracy for model-based SOC estimation of LFP batteries across a wide temperature range.