Air compressor is an important sub-component in fuel cell systems. Previous experimental and numerical studies still cannot provide an effective method for the quick and deep modification for fuel cell centrifugal compressor structure. To improve the compressor pressure ratio and efficiency, this paper develops a novel artificial intelligence (AI) framework for fuel cell centrifugal compressors by integrating three-dimensional (3D) computational fluid dynamics (CFD) models, machine learning and genetic algorithm. Based on the simulated results, the surrogate model is developed, which is subsequently coupled with NSGA-III algorithm to find the Pareto-optimal front. A set of optimized parameters are obtained. Based on the CFD model, the performance of the optimized compressor is comprehensively compared with the original compressor design. It is found that the flow uniformity inside compressor is greatly enhanced, leading to a better performance.
Keywords centrifugal compressor, AI, data-driven surrogate model, SVM, NSGA-â…¢