Distributed State Estimation of PEM Fuel Cells using Multi-Agent Unscented Kalman Filtering with Hydrogen-State Consensus

Abstract

The hydrogen state in proton-exchange membrane (PEM) fuel cells is essential for their efficient operation. Con-ventional estimators face challenges due to model nonlinearities, limited sensor measurements, and cell-to-cell variability in the fuel cell stack. This paper proposes a distributed estimation framework based on multi-agent unscented Kalman filter (MAS-UKF) for PEM fuel cells with state-dependent dynamics. Each cell in the stack is modeled as an agent to estimate all of its internal states, including hydrogen availability. A consensus protocol is implemented among the cells to capture the shared hy-drogen supply dynamics, thereby enhancing estimation accuracy through distributed information fusion. Simulation results show that the proposed MAS-UKF achieves robust state estimation while ensuring convergence of the hydrogen state across all agents.