Abstract: This study proposes an enhancement for the performance of the neural voltage-tracking controller based on different types of on-line optimization algorithms for nonlinear Proton Exchange Membrane Fuel Cell (PEMFC) system. The goal of this research is to employ the NARMA-L2 neural model in order to identify and control the nonlinear system. The task of the proposed nonlinear adaptive neural inverse voltage-tracking controller is to find precisely and quickly the optimal hydrogen partial pressure action which is used to control the (PEMFC) stack terminal voltage. Three intelligent optimization algorithms are used to learn and tune the weights of the neural model, the first one is the FireFly Algorithm (FFA), the second one is the Chaotic Particle Swarm Optimization (CPSO) algorithm and the third one is the Hybrid Firefly-Chaotic Particle Swarm Optimization (HFF-CPSO) algorithm. The numerical simulation results show that the NARMA-L2 controller with (HFF-CPSO) algorithm is more accurate than CPSO and FFA in terms of quickly obtaining the neural controller’s parameters with high reduction for the number of function evolutions and moreover in its capability of generating smooth partial pressure control response for the nonlinear (PEMFC) system without voltage oscillation in the output through investigating under random load-current variations.