A hybrid paradigm combining model-based and data-driven methods for fuel cell stack cooling control.

The cooling of the open-cathode proton exchange membrane fuel cell (PEMFC) is critical for the operational safety and overall efficiency. However, its control is challenging because of the model uncertainties and frequent disturbances caused by the power adjustment. To this end, this paper proposes a hybrid cooling control strategy by combining the merits of the model-based and data-driven methods. Firstly, a simplified nonlinear mechanistic model is used to exhibit the dynamic perturbations in terms of the different fan speeds and power conditions. Secondly, a modified active disturbance rejection control (ADRC) is developed by incorporating an identified nominal linear model into extended state observer. The external disturbances and the internal uncertainties beyond the nominal model are lumped as a total term, which will be estimated and mitigated in a real-time data-driven manner. The simulation results show that the proposed hybrid method is able to give a faster response with stronger robustness and less noise sensitivity against the uncertainties than the conventional PI and ADRC methods. The experimental test on a 500W open-cathode PEMFC verifies the simulation merits in both set-point tracking and disturbance rejection, depicting a promising prospect of the proposed hybrid method in the open-cathode PEMFC cooling control practice. • The nonlinearity of the fuel cell stack temperature is exhibited. • The uncertainties are accommodated by the data-driven estimation. • The effect of measurement noise is reduced by incorporating model information. • Experiments validate the efficacy of the proposed method. [ABSTRACT FROM AUTHOR]