In order to improve the performance of hydrogen fuel cell ejector，the hydrogen fuel cell ejector under rated operating conditions was taken as the research object,an optimization method of ejector structure parameters based on EBF neural network model and NLPQL algorithm was proposed.Based on orthogonal test,the EBF neural network model was established,the nonlinear relationship between the structural parameters of the ejector and the ejection coefficient was described. The precision of the proxy model was verified by comparing the simulated value of the ejection coefficient with the predicted value of the proxy model and the complex correlation coefficient. The NLPQL algorithm was applied to make global optimization, and the structure parameter combination with the maximum ejection coefficient was obtained and verified by simulation.The results indicate that,based on EBF neural network and NLPQL algorithm, the ejection coefficient of fuel cell ejector is improved,compared with the maximum value of the orthogonal test scheme, the ejection coefficient is increased by 3.9%.The method based on orthogonal test and EBF neural network can expand the research scope and level of ejector structural parameters,save the time of CFD simulation calculation.