Among the direct measurement methods used to estimate SOC, the most common are the ampere-hour integration method and the open-circuit voltage (OCV) method. The ampere-hour integration method is a simple open-loop estimation method that integrates current to obtain the current SOC of the battery.
However, accurate estimation of battery SOC remains challenging due to the highly dynamic and nonlinear characteristics of lithium batteries . Currently, methods for estimating SOC mainly include direct measurement, model-based methods, data-driven approaches, and hybrid methods [7, 8].
To enhance the accuracy of battery SOC estimation and its resistance to outlier interference, this study treats the estimated SOC by the deep learning model TGMA as noisy predicted observations, and introduces robust and adaptive factors into the Kalman filter. Consequently, the closed-loop SOC estimation method TGMA-RAKF is proposed.