摘要
为提高燃料电池氢气循环系统引射器设计效率,提出一种结合机器学习和数据驱动的技术。在大幅度降低计算需求的前提下,精确地预测引射器的一次流和二次流的质量流量并计算引射比。采用10个选定特征作为输入层,一次、二次流体质量流量作为输出层,建立基于麻雀搜索算法优化的BP神经网络模型进行预测,得出训练和测试集的平均绝对误差(MAPE)都低于10%。分别在40、70、100、150 kW的工况下进行验证,平均相对误差均不超过11%。所提方法能够准确预测引射器一、二次流质量流量及引射比,为引射器设计优化提供了一种高效工具,以快速评估不同结构参数及工作参数对引射比的影响。
To improve the design efficiency of ejectors in fuel cell hydrogen circulation systems,a novel approach integrating machine learning and data-driven techniques was proposed.This method significantly reduced computational demands while accurately predicting the mass flow rates of the primary and secondary flows,as well as calculating the entrainment ratio.In this study,a SSA-BP neural network model was developed using ten selected features as the input layer and the mass flow rates of the primary and secondary fluids as the output layer.The results demonstrated that the mean absolute percentage error(MAPE)for both the training and test sets was below 10%.The model was validated under operating conditions of 40,70,100,and 150 kW,with the average relative error remaining within 11%in all cases.This approach provides an efficient tool for ejector design and optimization,enabling rapid assessment of the impact of various structural and operational parameters on the entrainment ratio.
作者
邹润翔
李汝宁
冯兴
ZOU Runxiang;LI Runing;FENG Xing(School of Automobile and Transportation,Tianjin University of Technology and Education,Tianjin 300222,China;College of Transportation Science and Engineering,Civil Aviation University of China,Tianjin 300300,China)
出处
《农业装备与车辆工程》
2025年第5期95-102,共8页
Agricultural Equipment & Vehicle Engineering
关键词
质子交换膜燃料电池
引射器
机器学习
性能预测
计算流体动力学
proton exchange membrane fuel cell
ejector
machine learning
performance prediction
computational fluid dynamics
