Hybrid State of Charge Estimation for Lithium-Ion Batteries Using Data-Driven and Model-Based Fusion

Le résumé fourni par la source

Accurately estimating the state of charge (SOC) is essential for effective lithium-ion battery management. This work proposes an innovative hybrid SOC estimation method that integrates model-driven and data-driven techniques through a Bayesian fusion strategy to enhance performance under varying operational conditions. Specifically, the Extended Kalman Filter (EKF) and a Long Short-Term Memory (LSTM) neural network are used independently to estimate SOC, leveraging their respective strengths in system modeling and pattern recognition. To improve estimation reliability, a probabilistic fusion mechanism is introduced, dynamically weighting each model's output based on uncertainty, computed using a rolling standard deviation. This enables the framework to emphasize the most trustworthy estimate at each timestep, effectively reducing noise and drift. Experimental validations reveal superior accuracy and robustness compared to standalone EKF or LSTM approaches. The proposed fusion framework provides a reliable SOC estimation solution tailored for real-world battery management applications.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.