Abstract: Accurate estimation of the state of charge (SOC) of lithium-ion batteries is essential for ensuring the safe, reliable, and efficient operation of electric vehicles. SOC estimation provides critical information for energy management within the battery management system (BMS), directly influencing driving range prediction and preventing overcharge or over-discharge. Conventional estimation methods based on the RC equivalent circuit model require the identification of all RC branch parameters, which increases computational complexity and propagates intermediate errors, ultimately limiting real-time performance. To overcome these challenges, this paper proposes a simplified modeling and adaptive algorithm framework that achieves high-precision, robust, and computationally efficient SOC estimation. The first contribution is the design of a linear simplified first-order equivalent circuit model. By considering the RC network as an integral polarization element, the method directly identifies the polarization voltage rather than individual RC parameters. This reduces the number of estimated variables, eliminates the repeated recalculation of state transition matrices, and significantly lowers computational burden. In addition, the nonlinear relationship between open-circuit voltage (OCV) and SOC is approximated using a piecewise linear regression, enabling the application of a linear state-space model and facilitating the efficient use of Kalman filtering. Second, to address the time-varying nature of battery parameters, an adaptive forgetting factor recursive least squares (AFFRLS) algorithm is developed. Unlike fixed forgetting factor methods, the adaptive design dynamically adjusts the factor based on estimation error, allowing precise tracking of parameter variations while preserving stability. This ensures reliable online parameter identification under varying operating conditions. Third, a noise-adaptive Kalman filtering (AKF) method is integrated for SOC estimation. By adaptively updating process and measurement noise covariances according to residual statistics, the AKF compensates for model uncertainties and external disturbances. This strategy enhances robustness and maintains high estimation accuracy even under highly dynamic operating conditions and extreme temperature scenarios, where conventional methods often suffer from accuracy degradation due to intensified nonlinearities and noise effects. To validate the approach, extensive experiments were conducted using datasets from the CALCE Center, including Federal Urban Driving Schedule (FUDS), Dynamic Stress Test (DST), and temperature-varying profiles at 0 ℃ and 45 ℃. Comparative results with three benchmark approaches—linear model with FFRLS-KF, first-order RC model with FFRLS-EKF, and first-order RC model with UKF—confirm the superiority of the proposed method. Specifically, the method reduces mean absolute error (MAE) and root mean square error (RMSE) by about 30% and 25% respectively, while also shortening runtime by more than 60% compared to the UKF algorithm. Even under extreme temperature conditions, it maintains stable performance, demonstrating robustness and adaptability. In summary, this paper contributes: (1) a simplified polarization voltage-based equivalent model that reduces the parameter identification burden, (2) an adaptive forgetting factor mechanism for accurate tracking of time-varying parameters, and (3) a noise-adaptive Kalman filtering approach to handle model uncertainty and external disturbances. Together, these innovations improve both estimation accuracy and computational efficiency, offering a practical solution for real-time SOC estimation in electric vehicle applications.