Abstract: Accurate and reliable state-of-charge (SOC) estimation is essential for lithium-ion battery management systems (BMS), especially in electric vehicle applications, where both estimation accuracy and real-time performance are critical. In conventional SOC estimation approaches based on resistance-capacitance (RC) equivalent circuit models, the internal RC network parameters typically must be identified online. However, the identification errors of these parameters are closely linked to SOC estimation errors, and increasing the number of parameters to be estimated significantly increases the computational burden. These issues limit the robustness and real-time applicability of traditional methods under dynamic operating conditions. To address these challenges, this study proposes an online SOC estimation method based on a simplified linear model integrated with a dual adaptive estimation mechanism. First, motivated by the observation that polarization-related parameters exhibited relatively stable characteristics over short timescales under practical operating conditions, a simplified first-order linear equivalent model was constructed, with the polarization voltage as the core state variable. In this modeling framework, the parallel RC network is represented as an integrated polarization voltage component, rather than being explicitly parameterized. This structural simplification reduces the number of parameters required for online identification and weakens the coupling between parameter identification and SOC estimation at the model level, thereby improving numerical stability and computational efficiency. Based on the simplified model, an adaptive forgetting factor recursive least squares (AFFRLS) algorithm was developed to perform online identification of the polarization dynamic parameters. Contrary to conventional recursive least squares (RLS)-based approaches with fixed forgetting factors, the proposed AFFRLS algorithm dynamically adjusts the forgetting factor according to the estimation error characteristics, allowing the algorithm to balance tracking capability and noise suppression under time-varying operating conditions. This adaptive mechanism allows the proposed method to respond to load fluctuations and moderate parameter variations without introducing excessive estimation oscillations. Furthermore, a noise-adaptive kalman filter (AKF) was employed to estimate the SOC within a simplified modeling framework. Contrary to the traditional Kalman filtering approaches with fixed noise covariance matrices, the proposed AKF dynamically updates the process and measurement noise covariances based on the statistical properties of the terminal voltage residuals. This noise-adaptive strategy enhances the robustness of the SOC estimation against measurement noise, modeling uncertainties, and external disturbances, particularly under highly dynamic current profiles. The proposed method was experimentally validated using lithium-ion battery data collected under multiple operating conditions, including different temperature environments and dynamic load profiles. Comparative experiments with three representative benchmark methods demonstrate that the proposed approach achieves a reduction of approximately 30% in the SOC estimation mean absolute error (MAE) while also reducing the overall online computational load. These results indicate that the proposed method can achieve a favorable tradeoff between estimation accuracy and real-time performance. Overall, by integrating a simplified polarization-voltage-based linear model with adaptive parameter identification and noise-adaptive state estimation, the proposed method provides an efficient and robust solution for online SOC estimation. The proposed framework is particularly suitable for real-time BMS applications, where computational resources are limited and operating conditions are highly dynamic, offering clear practical value for embedded implementations in automotive battery management systems.