Real Electric Vehicle Lithium-Ion Battery Capacity Estimation Method Based on the Fusion Model of Discrete Arrhenius and Double Extended Kalman Filter

Accurate capacity estimation of lithium-ion batteries (LIBs) in real-world electric vehicles (EVs) remains a critical challenge due to complex degradation mechanisms, sparse cloud-platform data, and the limitations of existing single-model approaches. To address these issues, this paper proposes a novel fusion estimation method that integrates a discretized Arrhenius empirical lifetime model with a first-order equivalent circuit model (ECM) identified via particle swarm optimization (PSO) and employs a dual extended Kalman filter (DEKF) to synergistically update model parameters and fuse capacity estimates. Then, the proposed method is validated using real-world charging data from 36 EVs across three vehicle brands. Experimental results demonstrate that the fusion model achieves capacity estimation errors within 2.974% maximum and delivers state-of-health (SOH) estimation with R² values of 0.966, 0.983, and 0.981 for the three vehicle types, respectively. The method effectively overcomes data sparsity and noise issues inherent in cloud-platform data, and provides accurate, adaptive, and robust capacity tracking under diverse real-world operating conditions. This work offers a practical and scalable solution for cloud-based battery health management in next-generation EV applications.