Hardware-in-the-Loop simulation for online identification of lithium-ion battery model parameters and state of charge estimation

Abstract

• Propose a modified AUKF (mAUKF) with exponential weighting to dynamically adjust noise covariance for better adaptability to varying conditions. • Establish an HIL platform integrating real-world battery data with simulation to validate SOC estimation under diverse scenarios. • Integrate an adaptive mechanism for real-time identification of battery parameters to ensure accurate model alignment with the battery's state and aging effects. • Combine Kalman filters with recursive least squares for better real-time parameter tracking and noise suppression. Accurately estimating the state of charge (SOC) of Lithium batteries remains a major challenge due to their nonlinear behavior, especially when the initial SOC value significantly deviates from the actual one. To address this issue, this study introduces a new algorithm to estimate the battery SOC based on adaptive unscented Kalman filter (AUKF). The modified AUKF (mAUKF) begins by constructing a second-order Thevenin model of the batteries through the forgetting factor recursive least squares (FFRLS) method, and then implements the novel approach for SOC estimation. The proposed method retains the strengths of the AUKF while dynamically adjusting the process and measurement noise covariance to improve SOC estimation robustness and accuracy. The effectiveness of the method is validated using both simulation and experimental results. Additionally, to avoid the limitations of pure simulation or experimentation in validating when there are novel SOC estimation algorithms, a hardware-in-the-loop (HIL) test bench is also introduced. The final results demonstrate that the mAUKF algorithm achieves faster convergence rates of about 45 % and 49 % when handling large initial errors and transient responses compared to the original methods, in simulation and experiment cases, respectively.

Topics & Concepts