Cell-level state-estimation in parallel connected lithium-ion battery packs

State estimation is essential when deploying lithium-ion (Li-ion) battery packs in the field as it enables accurate predictions of key properties, such as the remaining range of electric vehicles. Most existing studies on state estimators for battery packs have used simple, lumped models for the pack, with each cell considered equivalent. These low-resolution lumped models are not able to capture the inherent cell-to-cell variability in packs, a feature which has limited the effectiveness of state estimators. To address this issue, a Hermite polynomial-based Extended Kalman filter (HP-EKF) is proposed to estimate the states of each cell in a parallel connected battery pack described by descriptor system dynamics. The performance of the proposed cell-level state-estimator is validated in experiments with two LiNiMnCoO${}_2$ Li-ion batteries connected in parallel. The model demonstrated high accuracy in predicting the response of the two parallel-connected Li-ion batteries, with root mean squared error of 0.00345V between experimental and modeled voltages. The proposed HP-EKF significantly reduces the estimation error compared to the conventional EKF while achieving accuracy comparable to the Cubature Kalman filter (CKF). Moreover, the HP-EKF exhibits computational complexity similar to the CKF while offering enhanced numerical stability by preserving the desirable properties of the error covariance matrices during implementation. This advantage, which typically requires the square-root variant of the CKF (SR-CKF), is inherently retained in the HP-EKF without the additional computational burden of the SR-CKF. These results highlight the potential of implementing cell-level estimation in parallel connected battery packs to provide information-rich estimates of its states.