Multi-Objective Optimization of Roots Hydrogen Circulation Pumps Using Gaussian Process Regression

Abstract

Abstract. With the increasing demand for hydrogen fuel cell vehicles, Roots hydrogen circulation pumps play a crucial role in ensuring the efficiency and reliability of hydrogen fuel cell systems. Traditional optimization methods for Roots pumps struggle to balance multiple objectives effectively during parameter optimization. This research integrates Gaussian process regression with particle swarm optimization (PSO), applying this optimization framework to the multi-objective parameter optimization of a Roots hydrogen recirculation pump. Flowrate and shaft power were selected as optimization objectives, while optimization variables—including diameter–distance ratio, three different clearances, and rotating speed—were investigated. A training database was constructed using greedy algorithm-optimized Latin hypercube sampling. Gaussian process regression was validated as a surrogate model compared with radial basis function. Multi-objective PSO was then applied to obtain Pareto solutions, from which the technique for order preference by similarity to ideal solution method selected the optimal solution. The optimized rotor achieved a 24% increase in flowrate, a 21% reduction in flow pulsation, and stable shaft power. Flow field analysis confirmed reduced leakage and vortex formation, improving volumetric efficiency and flow stability. The optimization method used in this research demonstrates high accuracy and reliability, providing an effective approach for enhancing hydrogen circulation pump design and offering valuable guidance for future fuel cell system applications.