Constrained extended Kalman filter design and application for on-line state estimation of high-order polymer electrolyte membrane fuel cell systems. (20th May 2021)

• Record Type: Journal Article
• Authors: Böhler, Lukas; Ritzberger, Daniel; Hametner, Christoph; Jakubek, Stefan
• Is Part Of: International journal of hydrogen energy. Volume 46:Number 35(2021)
• Page Start: 18604
• Page End: 18614
• Publication Date: 2021-05-20
• Subjects: PEMFC -- Kalman filter -- Successive linearization -- Observer -- Constrained estimation; Hydrogen as fuel -- Periodicals; Hydrogène (Combustible) -- Périodiques; Hydrogen as fuel; Periodicals; 665.81
• DOI: 10.1016/j.ijhydene.2021.03.014
• Languages: English
• ISSNs: 0360-3199

Abstract

Abstract: In this paper an alternative approach to extended Kalman filtering (EKF) for polymer electrolyte membrane fuel cell (FC) systems is proposed. The goal is to obtain robust real-time capable state estimations of a high-order FC model for observer applications mixed with control or fault detection. The introduced formulation resolves dependencies on operating conditions by successive linearization and constraints, allowing to run the nonlinear FC model at significantly lower sampling rates than with standard approaches. The proposed method provides state estimates for challenging operating conditions such as shut-down and start-up of the fuel cell for which the unconstrained EKF fails. A detailed comparison with the unscented Kalman filter shows that the proposed EKF reconstructs the outputs equally accurate but nine times faster. An application to measured data from an FC powered passenger car is presented, yielding state estimates of a real FC system, which are validated based on the applied model. Graphical abstract: Image 1 Highlights: An alternative EKF formulation utilizing successive linearization is proposed. Robustness of the state estimation is improved by introducing inequality constraints. The constrained EKF is nine times faster than the comparable constrained UKF. The proposed concept is validated using drive-cycle measurements of a real FC stack. Estimation results during shut-down and start-up of the stack are presented.