Real-time application of Koopman-based optimal control strategies for fuel cell stack thermal management

Abstract

In this study, a Koopman operator was applied in conjunction with linear optimal control algorithms, specially linear quadratic regulator (LQR), to enable real-time control of a 5 kW open cathode proton exchange membrane fuel cell (PEMFC) stack. A proportional–integral (PI) controller and two Koopman-based linear quadratic regulator (KLQR) controllers were developed. These controllers were designed to focus on minimizing in-stack temperature and temperature variations across the PEMFC stack, while also reducing auxiliary power consumption. To evaluate the real-time performance of these controllers, they were integrated with associated hardware using a National Instruments data acquisition system (DAQ) through LabVIEW. Various real-world scenario test conditions were conducted to assess the controllers’ performance and disturbance rejection, including evaluations based on the net power output of the PEMFC stack, as well as temperature and temperature variations in transient and steady states. Additionally, a comparative analysis of the controllers’ hardware-in-loop performance was conducted and compared to simulation results. The results demonstrate that the proposed Koopman-based LQR approach offers several advantages in thermal management and control of temperature variations across the PEMFC stack, leading to improved net power outputs.