<![CDATA[Reliable control-oriented models of PEM fuel cells remain challenging because PEMFC dynamics are nonlinear, coupled, and hard to excite under practical constraints. This paper presents a hybrid identification workflow in a controlled MATLAB/Simulink simulation environment. After discretization, bounded multisine excitation is applied, and correlation-based response analysis (CRA) is used to obtain non-parametric dynamics; low-order parametric structures (ARX, ARMAX, Box–Jenkins, OE, and FIR) and a grey-box state-space model are then estimated and validated using Fit%, information criteria (AIC/BIC), and residual diagnostics. In SISO, ARMAX provides the best accuracy–parsimony compromise (Fit = 96.84% with the lowest AIC/BIC and residuals mostly within confidence bounds), while Box–Jenkins achieves the highest fit (i.e., 98.75%) at higher complexity. In MIMO, most channels achieve an accuracy over 92% fit, with the most coupled pathway remaining the limiting case (best fit = 86.38% with BJ), and ARMAX/BJ emerging as the dominant structures across channels. The grey-box model attains 97.35% fit for voltage and 86.47% for power. This paper establishes a unified, control-oriented hybrid workflow that links CRA non-parametric estimation with low-order parametric and grey-box models, providing compact, physically interpretable PEMFC dynamics and practical model-selection guidance for control and energy-management applications.]]>
Copyrights © 2026
