<p>An accurate evaluation of the state of charge (SOC) for lithium-ion batteries is crucial for electric vehicles. However, dynamic temperature variations significantly impact both battery model parameters and available capacity, leading to cumulative errors in existing methods. To address this, a novel dual time-scale framework is proposed. This temperature-adaptive SOC estimation technique utilizes a second-order RC equivalent circuit model (ECM). First, a mapping relationship among SOC, open-circuit voltage (OCV), and temperature was established. Subsequently, a hybrid particle swarm optimization and Levenberg-Marquardt (PSO-LM) algorithm was employed to identify ECM parameters at different temperatures. Moreover, capacity tests under varying temperatures and discharge rates were conducted to quantify variations in available capacity. The dual time-scale framework implements microscale updates for ECM parameters and SOC via adaptive filtering, and macroscale updates for battery available capacity. Experimental results demonstrate that under varying temperatures, the proposed method achieves a mean absolute error below 1% and a root mean square error below 1.6%, significantly outperforming the traditional fixed-capacity cubature Kalman filter (CKF). This study provides a comprehensive solution for practical SOC estimation, thereby enhancing the dependability and performance of battery management systems.</p>