Observer design for an electrochemical model of lithium ion batteries based on a polytopic approach

Joint Estimation of Battery Parameters and State of Charge Using an Extended Kalman Filter: A Single-Parameter Tuning Approach

“…As mentioned previously, robust-observer approaches to battery SoC estimation have been presented in the literature, see, e.g., [3,4,16]. In [3,4], a robust observer is used in combination with electrochemistry-based battery models to estimate the SoC using estimates of the Lithium-ion concentrations. In [16], a robust observer is presented for estimating the SoC using an ECM.…”

“…In case of joint estimation of states and parameters, a typical approach in the literature is to use an EKF-type of observer, such as the joint EKF, see, e.g., [1,8]. However, based on the advantages of the robust-observer approaches as presented in [3,4,16] (i.e., straightforward tuning and an explicit interpretation of convergence and disturbance), a robust observer is preferred for joint estimation. To achieve this, the nonlinear system can be described as an uncertain linear system, which can be done using a polytopic system description, where the polytope is defined by a number of vertices based on the lower bounds and upper bounds of the linearised system.…”

“…Therefore, the SoC is typically estimated using the measured quantities in combination with model-based techniques, see, e.g., [1,2] and references therein. Although SoC estimation can be done with observers designed for accurate electrochemistry-based battery models, see, e.g., [3,4], an Equivalent-Circuit Model…”

“…In summary, the aforementioned EKF-based methods require extensive tuning of the covariance matrices, which is a tedious task without explicit tuning guidelines. To address the aforementioned drawbacks of the EKF in case of offline estimation of the model parameters, a robust-observer approach based on Linear Matrix Inequalities (LMIs) can be used to perform SoC estimation, see, e.g., [3,4,16,17]. This approach results in observers with a structured way of capturing disturbances and model uncertainty and, consequently, tuning of the observer is likely more straightforward.…”

“…Note that the term straightforward is used in the sense that (1) the number of tuning parameters is limited and (2) there are clear and explicit guidelines for tuning these parameters so that the user knows what to expect when changing a tuning parameter. However, as we will show in Section II-C, extending the robust-observer approach presented in [16], or adapting other robust-observer approaches, see, e.g., [3,4,17], towards joint estimation of ECM parameters and SoC is not feasible. Consequently, an observer for nonlinear systems such as the joint EKF should be considered, which suffers from non-obvious tuning as explained previously.…”