摘要
为了减少实际行驶排放试验(real driving emission,RDE)受到驾驶行为、车型等干扰的情况,降低试验耗时和测试成本,基于BP神经网络建立了重型车的排放预测模型,引入遗传算法(genetic algorithm,GA)和蚁群算法(ant colony optimization,ACO)进行优化。使用便携式排放测试系统(portable emissions measurement system,PEMS)对某重型车进行RDE试验,并将试验数据划分为训练集、测试集、验证集,通过B型关联度算法提取数据主要成分用于训练与预测。结果表明:瞬时排放水平上,NO_(x)预测结果与样本数据的皮尔逊相关系数为0.9686,线性高度相关;整体误差水平上,NO_(x)排放因子的最大相对误差为2.36%。该模型对重型车的瞬时排放和整体排放特性预测准确性较好,对辅助RDE试验具有参考意义。
This paper aims to reduce the interference of driving behaviors,vehicle type,etc.in real driving emission(RDE)and to reduce test time and cost.The emission prediction model of heavy-duty vehicles is established based on BP neural network and optimized by a combination of genetic algorithm(GA)with Ant Colony Optimization(ACO).A portable emission measurement system(PEMS)is used to conduct RDE tests on a heavy-duty vehicle.The test data are divided into training,test,and validation sets.Then,the main components of the data are extracted by a B-type correlation algorithm for training and prediction.The results show that the Pearson correlation coefficient of NO_(x) prediction results and sample data is 0.9686 at the instantaneous emission level,which is highly linearly correlated;the maximum relative error of the NO_(x) emission factor is 2.36%at the overall error level.The model accurately predicts heavy-duty vehicles’instantaneous emission and overall emission and has some theoretical significance for assisting RDE tests.
作者
闻增佳
谭建伟
王怀宇
余浩
常虹
孙文强
WEN Zengjia;TAN Jianwei;WANG Huaiyu;YU Hao;CHANG Hong;SUN Wenqiang(School of Mechanical Engineering,Beijing Institute of Technology,Beijing 100081,China;China Automotive Engineering Research Institute Co.,Ltd.,Chongqing 401122,China;OBD Calibration Room,Weichai Power Co.,Ltd.,Weifang 261001,China)
出处
《重庆理工大学学报（自然科学）》
CAS
北大核心
2022年第12期202-209,共8页
Journal of Chongqing University of Technology：Natural Science
基金
国家自然科学基金面上项目(52172337)
中国汽车工程研究院股份有限公司创新课题项目(JCCXKT-2021-002)。
