Integrating surrogate modeling and dynamic crayfish optimization for enhanced performance of proton exchange membrane fuel cells

Abstract

As a renewable energy solution, proton exchange membrane fuel cells (PEMFCs) deliver high operational effectiveness and reduced environmental discharges, supporting deployment in transportation and fixed power infrastructures. Effective industrial deployment requires optimizing performance by carefully selecting structural parameters and operational conditions. This study proposes a Dynamic Crayfish Optimization (DCO) technique to evaluate and enhance PEMFC performance. Unlike conventional evolutionary algorithms, which are complex and require extensive parameter tuning, the DCO algorithm is simpler, converges faster, and improves optimization efficiency. Key performance-influencing parameters are identified using variance-based feature selection, reducing dimensional complexity, while surrogate learning models accelerate fitness evaluation. The DCO-based optimization simultaneously enhances critical PEMFC metrics: voltage efficiency improves from 60% to 70%, fuel utilization from 85% to 90%, system efficiency from 40 % to 50 %, [w1.1][w1.2]with notable increases in power density and oxygen consistency. The DCO model achieved ξ₂ = 2.2158×10⁻³, ξ₃ = 3.4205×10⁻⁵, ξ₄ = −1.5679×10⁻⁴, Rc= 0.95×10⁻⁴, λ = 25, b = 0.0321. These improvements establish that the DCO outperforms existing techniques across all evaluated criteria. Overall, the proposed method provides a systematic, effective, and robust framework for PEMFC performance optimization, offering significant potential for engineering applications in renewable energy systems. The results highlight the practical value of integrating advanced optimization algorithms with data-driven surrogate models in fuel cell research.