基于不同人工神经网络的CNG发动机排气温度预测模型研究  

Research on Prediction Model of Exhaust Temperature of CNG Engine Based on Different Artificial Neural Networks

作　　者：张鹏 张铖 毛功平[2] ZHANG Peng;ZHANG Cheng;MAO Gongping(Shanghai GL-automotive Electronics Co.,Ltd.,Shanghai 201101;School of Automotive and Transportation Engineering,Jiangsu University,Zhenjiang 212013)

机构地区：[1]上海格令汽车电子有限公司,上海201101 [2]江苏大学汽车与交通工程学院,江苏镇江212013

出　　处：《内燃机》2020年第2期19-24,共6页Internal Combustion Engines

摘　　要：为了提高CNG发动机排气温度预测精度,基于BP、RBF和GRNN神经网络建立了3种排气温度的预测模型。开展了CNG发动机台架实验,测量了不同工况条件下发动机的排气温度,利用实验值对模型进行训练,并预测了不同发动机转速、空气进气量、点火提前角等条件下的排气温度,将预测值与实验值进行了对比分析,评估了不同预测模型的准确性。结果表明:BP、RBF和GRNN 3种神经网络的误差分别为3.5%、2.8%和3.1%。RBF神经网络的预测误差比BP和GRNN神经网络的误差小,稳定性强,更适合CNG发动机的排气温度预测。In order to improve the prediction accuracy of exhaust gas temperature for CNG engines, three types of exhaust gas temperature prediction models are established based on BP, RBF and GRNN artificial neural networks. This paper carried out CNG engine bench experiment, measured the exhaust temperature of the engine under different operating conditions, trained the model using the experimental values, and predicted the exhaust gas temperature under different conditions such as engine speed, air intake, and ignition advance angle. The exhaust gas temperature was compared with the experimental value to evaluate the accuracy of different prediction models. The results show that the errors of the three neural networks of BP, RBF and GRNN are 3.5%, 2.8% and 3.1%, respectively. The prediction error of the RBF neural network is smaller than that of the BP and GRNN neural networks, and the stability is strong, which is more suitable for the exhaust gas temperature prediction of CNG engines.

关 键 词：CNG发动机 排气温度模型 人工神经网络 

分 类 号：TK413[动力工程及工程热物理—动力机械及工程]