Parameter optimization calibration of low-pressure fuel cell energy system based on adaptive response surface algorithm

Abstract

This study addresses the high time, material, and labor costs associated with the non-automated online parameter optimization calibration of fuel cell systems. An adaptive response surface method (ARSM) is proposed to automate the calibration process and significantly reduce these costs. The low-pressure fuel cell system is modeled using the lumped parameter approach, including segmented models for stack channels, with optimization parameters and constraints analyzed comprehensively. The ARSM algorithm dynamically shrinks the parameter search space during each iteration and employs a sample point inheritance strategy following space reduction, further minimizing time costs. Quantitative analysis demonstrates that the ARSM reduces calibration time by 31% compared to traditional RSM. Case studies reveal that as current load increases, the optimal oxygen excess ratio rises, while the optimal stack temperature declines. Latin hypercube sampling is identified as more efficient than full factorial sampling, and initializing optimization with the previous condition’s optimal values is shown to outperform using the parameter space center point. The proposed methodology not only enhances the efficiency of the calibration process but also offers a scalable and practical solution for real-time applications. These findings provide a robust framework for reducing costs and improving efficiency in fuel cell system parameter optimization.