<p>Proton exchange membrane fuel cells (PEMFCs) are an important technology of clean energy because of their efficiency, low emissions and the ability to start up quickly. To achieve accurate performance predictions, it is crucial to have precise parameter estimation to come up with accurate PEMFC models. This paper proposes a new hybrid optimization algorithm, differential evolution-based artificial rabbits optimization (DEARO), with the specific purpose of extracting the seven main unknown parameters in PEMFC models. The DEARO algorithm suggested is a strategic combination of the advantages of differential evolution (DE) and artificial rabbits optimization (ARO) to address the shortcomings of the current approaches. The DEARO algorithm proposes three major novelties an adaptive mutation mechanism which combines several DE strategies (DE/rand/1, DE/best/1 and DE/rand/2) to provide global search capabilities and prevent premature convergence, a dynamic crossover operation which intelligently combines the characteristics of the solutions to improve local refinement and a failure rate–based adaptation scheme that automatically adapts the control parameters to maintain the best balance between exploration and exploitation during the optimization process. The features allow DEARO to effectively traverse the multimodal, multimodal search space that is typical of PEMFC parameter estimation problems. The validation was carried out comprehensively with 12 different PEMFC stacks of the major manufacturers, such as BCS 500W, NedStack PS6, and Ballard Mark V systems. The experimental findings prove that DEARO is more accurate with SSE being up to 92.5% lower than traditional approaches (e.g. 0.0255 vs 0.0412 in the case of BCS 500W). The statistical analysis of 40 independent runs showed an outstanding consistency with standard deviations as low as 5.9110 − 5. The computational efficiency of the algorithm is also impressive with most test cases being solved to optimal solutions within 0.4&#xa0;s and with a robust performance in a wide range of operating conditions. The results of comparative studies with nine state-of-the-art algorithms (RUN, HHO, RIME and PSO) proved the advantage of DEARO in terms of accuracy and reliability. Convergence analysis demonstrated that DEARO usually converges to optimal solutions in less than 50 iterations, which is a lot faster compared to other methods. Further confirmation of the effectiveness of the algorithm is made by I-V/P–V characteristic matching, boxplot analyses and non-parametric statistical tests. The developments make DEARO an effective means of designing, optimizing and real-time controlling PEMFC systems and possibly other complex energy system modelling problems.</p>