State of Charge Estimation of Lithium-Ion Batteries Based on Temporal Convolution Network and Unscented Kalman Filter

The state of charge (SOC) stands as a pivotal parameter within the lithium-ion battery management system (BMS). Precise estimation of SOC is imperative for safeguarding the safety and ensuring the reliability of lithium-ion batteries. Nonetheless, the SOC in lithium-ion batteries cannot be directly measured, necessitating its indirect estimation through observable parameters. Considering the nonlinear relationship between the measured values of lithium-ion batteries and SOC, this study introduces a neural network approach integrating Temporal Convolution Network (TCN), Genetic Algorithm (GA), and Unscented Kalman Filter (UKF) to estimate the SOC in lithium-ion batteries. This method precisely correlates the measured values of current, voltage, and temperature in the operation of lithium-ion batteries with their SOC. The experimental results indicate that, under various operating conditions and environmental temperatures, the proposed model exhibits an average MAE estimation of less than 0.5% across all test data. In comparison to traditional approaches, the proposed method demonstrates superior accuracy and robustness.