Aiming at the low accuracy of State of charge (SOC) estimation for auxiliary batteries of hydrogen fuel cell heavy-duty trucks under complex dynamic conditions, this paper proposes an SOC estimation method based on a fractional-order equivalent circuit model, fusing an H∞ filter and an adaptive unscented Kalman filter to enhance estimation performance and robustness. A capacity composite correction coefficient is introduced to couple aging, temperature and current ratio effects, with a genetic algorithm and recursive least squares with forgetting factor used to determine model order and estimate coefficients online, respectively. Pulse excitation and dynamic condition experiments confirm that the method optimizes the fractional-order model, and the correction coefficient exhibits high accuracy and good working condition adaptability, providing theoretical guidance for accurate SOC estimation and supporting improved Battery Management System reliability and lifespan