The aim of this study is to analyze the exergy efficiency of a high-temperature proton exchange membrane fuel cell. To do this purpose, meta-heuristic technique has been used. First, the model of a membrane fuel cell is simulated and the polarization diagram shows a potent agreement with empirical data. Then, a new improved version of collective animal behavior algorithm is utilized for evaluating and optimizing the thermodynamic irreversibility, exergy efficiency, and work of the fuel cell. The algorithm uses opposition-based learning and Levy flight for improving the algorithm's premature convergence shortcoming. The result of this study shows that by comparison with standard collective animal behavior algorithm, genetic algorithm, and empirical results, the proposed algorithm has better achievements for both terms of optimal value finding and convergence strength.