An adaptive estimation approach based on fisher information to overcome the flat voltage plateau challenges of SOC estimation in LFP batteries

Accurate and robust state-of-charge (SOC) estimation remains a critical challenge for lithium iron phosphate (LFP) batteries due to their flat SOC–open-circuit-voltage (OCV) characteristics, pronounced hysteresis, and non-ideal operating conditions such as current sensor bias, voltage quantization, temperature variation, and insufficient excitation. This paper proposes an adaptive SOC estimation framework that addresses these challenges through an information-aware fusion strategy. The method adaptively fuses Coulomb counting and voltage-based SOC estimation using Fisher information, allowing the estimator to automatically adjust its reliance on each source according to excitation conditions and SOC–OCV observability. To explicitly capture hysteresis effects, a three-dimensional OCV–hysteresis–SOC (OCV–H–SOC) mapping is introduced for reliable SOC inversion within extended flat voltage plateaus. The proposed method is experimentally validated on an LFP cell using four realistic driving cycles under a wide range of challenging scenarios, including large initial SOC errors, prolonged operation in SOC–OCV flat zones, current bias, voltage quantization noise, low-temperature operation (10 °C), and insufficient current excitation. Compared with state-of-the-art benchmarks, including the unscented Kalman filter (UKF), LSTM, and Transformer-based estimators, the proposed approach consistently achieves superior accuracy and robustness across all tested conditions. In particular, the proposed method reduces SOC estimation error by up to approximately 80% relative to the UKF, while maintaining real-time computational efficiency.