A <scp>SOC</scp> and <scp>SOH</scp> Estimation Method for Energy Storage Batteries Based on the Fusion of a Reference Difference Model and Kolmogorov‐Arnold Networks

ABSTRACT Addressing the issue of excessive computational cost in Kalman filter algorithms for state of charge (SOC) and state of health (SOH) estimation in battery energy storage systems based on equivalent circuit models, this paper introduces a novel approach. The proposed method integrates a reference difference model with Kolmogorov‐Arnold Networks (KAN) to achieve rapid and cost‐effective SOC and SOH estimation for energy storage batteries. By employing a dual adaptive extended Kalman filter (DAEKF) algorithm and constructing a reference difference model, the computational burden of the Kalman filter algorithm decreases. Simultaneously, this methodology estimates battery voltage and SOC, while also determining battery parameters and capacity, thereby enabling joint estimation of SOC and SOH. Furthermore, the approach incorporates KAN to establish a voltage difference compensation mechanism, effectively correcting voltage errors caused by the difference model's neglect of polarization voltage differences and enhancing the accuracy of SOH estimation. The efficacy of this method is validated through testing on three datasets(University of Aachen, NASA random walk, and University of Wisconsin‐Madison). The results demonstrate that the proposed method significantly reduces computational burden compared to the first‐order RC circuit model and achieves superior SOH estimation performance after KAN compensation, thus providing a feasible technical approach for real‐time state monitoring of large‐scale energy storage power stations.