Estimating state of charge and state of health of rechargable batteries on a per-cell basis

Abstract

Much of current research on State-of-Charge (SOC) and State-of-Health (SOH) tracking for rechargeable batteries such as Li-ion focuses primarily on analyzing single cells, or otherwise treat a set of series-connected cells as a homogeneous unit. Since no two cells have precisely the same properties, for applications involving large batteries this can severely limit the accuracy and utility of the approach. In this paper we develop an model-driven approach using a Dual Unscented Kalman Filter to allow a Battery Monitoring System (BMS) to monitor in real time both SOC and SOH of each cell in a battery. A BMS is an example of a Cyber-Physical System (CPS) which requires deep understanding of the behavior of the physical system (i.e., the battery) in order to obtain reliability in demanding applications. In particular, since the SOH of a cell changes extremely slowly compared to SOC, our dual filter operates on two timescales to improve SOH tracking. We show that the use of the Unscented Kalman Filter instead of the more common Extended Kalman Filter simplifies the development of the system model equations in the multiscale case. We also show how a single "average" cell model can be used to accurately estimate SOH for different cells and cells of different ages.

Toward Composing Efficient FPGA-Based Hardware Accelerators for Physics-Based Model Predictive Control Smart Sensor for HEV Battery Cell Management

“…From our investigation on related work (presented in Section V.C), we could not find any FPGA-based embedded hardware architectures, specifically for PB-MPC. Our investigation further revealed that most of the existing works [10], [11], [12], [13] on EKF for batteries, were based on the ECM approach for estimating the state of charge (SOC) and state of health (SOH) of the batteries, and not much research has been done on physics-based (PB) EKF. Existing works in the published literature for PB-MPC designs, such as [7], [9], [15] were focused mainly on using Matlab software models to verify and improve the physics-based (PB) approaches.…”

“…The effectiveness of the MPC algorithm depends on the accuracy of the mathematical model of the battery cell. Two main mathematical models of the battery cell are the equivalent circuit model (ECM) [10], [11], [12], [13] and the electro-chemical physics-based model (PBM) [4], [7]. The ECM model comprises simplified algorithms and provides solid performance.…”

Design and implementation of an FPGA-based piecewise affine Kalman Filter for Cyber-Physical Systems

“…A similar example appears in [95,81] which is concerned with assessing the state of health (e.g. degree of degradation from nominal capacity) of rechargeable cells.…”

“…The following section details an extended case study on piecewise-affine Kalman Filtering for tracking cell voltages in large batteries. It is based on results appearing in [80] and [81].…”

“…The model used for the case study is a that of a rechargeable cell (Fig. 5.7) and is based on prior work on the subject [81]. It is only non-linear in one dimension, and that non-linearity must be identified empirically; thus it is well-suited for piecewise-affine approximation, using a simple point-to-point partitioning procedure similar to that described in [61].…”