An adaptive wolf pack algorithm based on a diversity-driven mechanism and its application in PEMFC parameter identification

Accurate parameter identification is crucial for building reliable proton exchange membrane fuel cell (PEMFC) models, which underpin efficiency improvement and system optimization. However, the strong nonlinearity and high coupling of parameters pose significant challenges for conventional methods. To overcome this, a diversity-driven adaptive wolf pack algorithm (ADWPA) is proposed. Its key innovations include a region-divided initialization, a reward-driven wandering behavior, and an entropy-based population update strategy, which collectively enhance global search capability and convergence precision. Evaluations on CEC2017 benchmarks and multiple PEMFC stacks demonstrate that ADWPA achieves superior accuracy, faster convergence, and greater robustness compared to state-of-the-art algorithms. This work provides an effective and efficient tool for high-precision PEMFC modeling and optimization.