Design of PSO-tuned FOPI & Smith predictor controller for nonlinear polymer electrolyte membrane fuel cell

Abstract

ABSTRACT The polymer electrolyte membrane (PEM) fuel cell or proton exchange membrane fuel cell (PEMFC) has sparked a lot of interest in renewable electricity generation due to high efficiency, quick start up time, and low operating temperature. However, they exhibit highly nonlinear behavior with degraded power quality under uncertain input conditions. The hydrogen, oxygen, and air feed determine the PEMFC system’s performance. Controlling the manifold pressure at the cathode side is essential in preventing the problem of oxygen starvation. In this paper, a nonlinear dynamic model of the PEMFC is considered to control the supply manifold pressure. The performance of the nonlinear PEMFC with various classical controllers (PI, PD, PID, and PID with filter derivative), Smithpredictor, and Fractional Order PI (FOPI) controllers are compared. The optimization of classical controllers is performed by PID tuner. On the other hand, Smith predictor and FOPI controller are tuned by artificial intelligence technique, namely, Particle Swarm Optimization (PSO). The FOPI controller is also tuned with Genetic Algorithm (GA). The PEMFC output is compared in terms of performance evaluation parameters, i.e. settling time, overshoot, and steady state error. The computer simulations suggest that in comparison with GA-FOPI [settling time (ts) = 1.374 s, overshoot (Mp) = 18.452%, steady state error (ess) = 0.1%] and PSO-Smith predictor (ts = 1.262 s, Mp = 5.581%, ess = 0%), the PSO-FOPI controller results in much better settling time (ts = 1.008 s), overshoot (Mp = 2.577%), and steady state error (ess = 0%). Thus, PSO-optimized FOPI controller with nonlinear PEMFC outperforms the classical, PSO-Smith predictor, and GA-FOPI control schemes.