Simultaneous Cell State Estimation via Dense Adaptive Extended Kalman Filter
Luke Nuculaj, Jun shuo Chen · IEEE Transactions on Control Systems Technology · 2025
This work addresses the computational intractability apropos of extended Kalman filters (EKFs) in the context of battery cell state estimation under limited voltage measurement. A novel, compact variation of the Kalman filter, namely the “dense EKF” (DEKF) is proposed, which leverages unique information about each of the cell’s inherent physical properties and net currents at each time step to compress sparsely populated covariance matrices and state vectors into a dense form whose size does not vary with the number of cells in the pack. The computational savings in terms of floating-point operations (FLOPs) reduction are analytically compared and illustrated through simulation. More specifically, the DEKF offers significant resource savings while maintaining estimation accuracy, reducing the estimation algorithm’s time complexity from$\mathcal {O}(N^{3})$to$\mathcal {O}(N)$, whereNis the number of cells in a serial-connected string. Furthermore, a special case where all serial-connected cells share the same discharge current, that is, no balancing or leakage, is also studied and demonstrated.