基于双隐含层GA-BP神经网络的重型柴油车排放预测
|
摘要
为了建立一种能够预测柴油车道路排放特性的模型,文章采用便携式车载(汽车尾气)排放测量系统(portable emission measurement system, PEMS),对某重型柴油车进行道路污染物排放特性测试;利用测得的试验数据,在双隐含层反向传播(back propagation,BP)神经网络的基础上,引入Levenberg-Marquardt(LM)优化算法,用遗传算法(genetic algorithm,GA)优化网络的权值与阈值;以车辆比功率(vehicle specific power,VSP)为输入,搭建CO、NO_x排放预测模型,并用试验数据对模型进行训练、验证。结果表明,CO、NO_x的预测结果与样本数据之间的皮尔逊相关系数分别为0.855 3、0.851 2,线性高度相关;在整体误差水平上,CO、NO_x排放因子的相对误差分别为2.61%、6.71%。该方法对车辆CO、NO_x的瞬时排放和整体排放特性的预测准确性较好,具有一定的理论意义和工程应用价值。
In order to establish a model to predict the road emission characteristics of diesel vehicles, a portable emission measurement system(PEMS) is used to test the pollutant emission characteristics of a heavy duty diesel vehicle. Based on the experimental data of the double hidden layer back propagation(BP) neural network, the Levenberg-Marquardt(LM) optimization algorithm is introduced to build the prediction model of CO and NO_x emission with the input of the vehicle specific power(VSP). Genetic algorithm(GA) is used to optimize the network parameters of initial weights and thresholds and the model is trained and validated by the test data. The simulation results demonstrate good agreements between the raw data and the model predictions with Pearson correlation coefficients of 0.855 3 and 0.851 2 for CO and NO_x, respectively. And the relative error of the emission factor of CO and NO_x is 2.61% and 6.71%, respectively. The prediction accuracy of the instantaneous emission and the overall emission characteristics of the vehicle CO and NO_x is good, which shows that the method has certain theoretical significance and engineering application value.
In order to establish a model to predict the road emission characteristics of diesel vehicles, a portable emission measurement system(PEMS) is used to test the pollutant emission characteristics of a heavy duty diesel vehicle. Based on the experimental data of the double hidden layer back propagation(BP) neural network, the Levenberg-Marquardt(LM) optimization algorithm is introduced to build the prediction model of CO and NO_x emission with the input of the vehicle specific power(VSP). Genetic algorithm(GA) is used to optimize the network parameters of initial weights and thresholds and the model is trained and validated by the test data. The simulation results demonstrate good agreements between the raw data and the model predictions with Pearson correlation coefficients of 0.855 3 and 0.851 2 for CO and NO_x, respectively. And the relative error of the emission factor of CO and NO_x is 2.61% and 6.71%, respectively. The prediction accuracy of the instantaneous emission and the overall emission characteristics of the vehicle CO and NO_x is good, which shows that the method has certain theoretical significance and engineering application value.
引文
[1] JOHN C,FRIEDRICH R,STAEHELIN J,et al.Comparison ofemission factors for road traffic from a tunnel study (Gubrist tunnel,Switzerland) and from emission modeling [J].Atmospheric Environment,1999,33:3367-3376.
[2] 葛蕴珊,王爱娟,王猛,等.PEMS用于城市车辆实际道路气体排放测试[J].汽车安全与节能学报,2010,1(3):141-145.
[3] 俞明,林冬燕,孙国斌,等.基于广义回归神经网络的发动机排放预测[J].华南理工大学学报(自然科学版),2003,31(1):51-56.
[4] 胡杰,林峰,王天田.基于神经网络偏最小二乘法的柴油机NOx排放预测模型[J].内燃机学报,2015,33(6):510-515.
[5] LI H Y,BUTTS K,ZASECK K.Emissions modeling of a light-duty diesel engine for model-based control design using multi-layer perceptron neural networks:SAE Paper 2017-01-0601[R].[S.l.]:SAE,2017.
[6] 周品.Matlab神经网络设计与应用[M].北京:清华大学出版社,2013.
[7] 李楠,汤东,李昌远,等.基于BP神经网络的柴油机掺烧生物柴油排放特性预测研究[J].车用发动机,2013(10):48-53.
[8] 简晓春,王利伟,闵峰.基于LM算法的BP神经网络对汽车排放污染物的预测[J].重庆理工大学学报(自然科学),2012,26(7):11-16.
[9] CHRIS M A.Virtual sensing:a neural network-based intelligent performance and emissions prediction system for on-board diagnostics and engine control:SAE Paper 980516[R].[s.l.]:SAE,1998.
[10] JIMENEZ J L.Understanding and quantifying motor vehicle emissions with vehicle specific power and TILDAS remote sensing[D].Boston:Massachusetts Institute of Technology,1999.
[11] JIMENEZ J L,MCCLINTOCK P,MCRAE G L,et al.Vehicle specific power:a useful parameter for remote sensing and emission studies[R].San Diego US:9th CRC On-Road Vehicle Emissions Workshop Council,1999.
[12] 闻新,李新,张兴旺,等.应用MATLAB实现神经网络[M].北京:国防工业出版社,2015.
[13] 丁红,董文永,吴德敏.基于LM算法的双隐含层BP神经网络的水位预测[J].统计与决策,2014(15):16-19.
[14] 卢建中,程浩.改进GA优化BP神经网络的短时交通流预测[J].合肥工业大学学报(自然科学版),2015,38(1):127-131.
[15] HAO L J,CHEN W,LI L,et al.Modeling and predicting low-speed vehicle emissions as a function of driving kinematics[J].Journal of Environmental Sciences,2017,55(5):109-117.
[2] 葛蕴珊,王爱娟,王猛,等.PEMS用于城市车辆实际道路气体排放测试[J].汽车安全与节能学报,2010,1(3):141-145.
[3] 俞明,林冬燕,孙国斌,等.基于广义回归神经网络的发动机排放预测[J].华南理工大学学报(自然科学版),2003,31(1):51-56.
[4] 胡杰,林峰,王天田.基于神经网络偏最小二乘法的柴油机NOx排放预测模型[J].内燃机学报,2015,33(6):510-515.
[5] LI H Y,BUTTS K,ZASECK K.Emissions modeling of a light-duty diesel engine for model-based control design using multi-layer perceptron neural networks:SAE Paper 2017-01-0601[R].[S.l.]:SAE,2017.
[6] 周品.Matlab神经网络设计与应用[M].北京:清华大学出版社,2013.
[7] 李楠,汤东,李昌远,等.基于BP神经网络的柴油机掺烧生物柴油排放特性预测研究[J].车用发动机,2013(10):48-53.
[8] 简晓春,王利伟,闵峰.基于LM算法的BP神经网络对汽车排放污染物的预测[J].重庆理工大学学报(自然科学),2012,26(7):11-16.
[9] CHRIS M A.Virtual sensing:a neural network-based intelligent performance and emissions prediction system for on-board diagnostics and engine control:SAE Paper 980516[R].[s.l.]:SAE,1998.
[10] JIMENEZ J L.Understanding and quantifying motor vehicle emissions with vehicle specific power and TILDAS remote sensing[D].Boston:Massachusetts Institute of Technology,1999.
[11] JIMENEZ J L,MCCLINTOCK P,MCRAE G L,et al.Vehicle specific power:a useful parameter for remote sensing and emission studies[R].San Diego US:9th CRC On-Road Vehicle Emissions Workshop Council,1999.
[12] 闻新,李新,张兴旺,等.应用MATLAB实现神经网络[M].北京:国防工业出版社,2015.
[13] 丁红,董文永,吴德敏.基于LM算法的双隐含层BP神经网络的水位预测[J].统计与决策,2014(15):16-19.
[14] 卢建中,程浩.改进GA优化BP神经网络的短时交通流预测[J].合肥工业大学学报(自然科学版),2015,38(1):127-131.
[15] HAO L J,CHEN W,LI L,et al.Modeling and predicting low-speed vehicle emissions as a function of driving kinematics[J].Journal of Environmental Sciences,2017,55(5):109-117.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |