Towards State-of-Charge Estimation for Battery Packs: Reducing Computational Complexity by Optimising Model Sampling Time and Update Frequency of the Extended Kalman Filter

Accurate State-of-Charge (SoC) estimation remains a challenge for large battery packs. This paper aims to reduce the computational complexity of single-cell estimation, which already achieves satisfactory performance, such that it can be more easily scaled to large arrays of cells inside battery packs. This is done by experimenting with a range of sampling times for the models used in an Extended Kalman Filter (EKF) and by adjusting the update frequency of this estimator. The EKF is tested with linear time-invariant and linear parameter-varying models and also in joint state-parameter estimation form. Results show that adjusting the sampling time and update frequency can result in a significant reduction of computational complexity, around a factor of 148, while only suffering a minor increase in SoC estimation error. This means that a relatively small micro-controller can be employed to estimate the SoC of an entire battery pack.