Study On Neural Network Modeling And Receding Horizon Control Of Diesel Engine

With the progress of The Times and the development of industry,the problem of energy shortage and environmental pollution is becoming more and more serious.China’s automobile industry is faced with strict national six emission standards,especially the nitrogen oxide emission of diesel engines.In order to solve this problem,the most commonly used solution in the project is to install the selective catalytic reduction system(SCR)of urea,and then through the precise control of the SCR system to achieve the purpose of cleaning outside the machine to achieve a significant reduction of NO_xemissions.Under the premise of meeting the national emission standard,it is necessary to make a unified planning for the economy,emission and power performance of automobiles.Traditional control methods are difficult to control the increasingly complex diesel engine system accurately.In addition,the complexity of diesel engine not only leads to the difficulty of its mechanism modeling,but also poses a challenge to the design of its controller.To solve the above problems,a diesel engine with a post-processing system is taken as the controlled object,and a method of neural network modeling and rolling time domain control for the diesel engine is proposed in this paper.Specific plans are as follows:(1)Based on the engine dynamics simulation software GT-Power,a 2L displacement diesel engine simulation model is established,which includes two parts:diesel engine module and SCR module.On this basis,the effectiveness of the established model was verified by analyzing the effects of EGR valve opening,injection advance angle,VGT valve opening,injection volume on the engine output performance,as well as the effects of exhaust temperature and ammonia nitrogen ratio on the NO_xconversion and ammonia escape of the post-treatment system.(2)Aiming at the complex modeling of diesel engine combustion process and chemical reaction process in post-processing system,a control-oriented prediction model of diesel engine based on neural network was established.First,fuel injection volume,fuel injection time,EGR and VGT valve opening were taken as input signals of diesel module,NO_xemissions of original engine,crankshaft output torque and exhaust temperature of diesel engine were taken as output signals of diesel module,and relevant experimental data were collected by GT-Power simulation model.Then,the prediction models of diesel engine and SCR system based on BP neural network were established respectively through normalization processing of the data.Finally,the validity of the prediction model is verified by simulation.(3)For diesel engine combustion and emission control target and constraints,multivariable coupling problem,using the neural network prediction model,the rolling horizon optimization control framework,the diesel engine combustion and emissions layered coordination control method,including the upper controller to expect torque tracking as the optimization goal,the lower controller with the optimization target of NO_xconversion rate and amount of ammonia escape.Then by solving the optimization problem of the target,the control input that meets the requirements is obtained.The simulation results show that the proposed layered control method can meet the driver’s torque demand while effectively reducing the NOx emissions below the national standard VI,which verifies the effectiveness of the controller.For diesel engine combustion and emission control target and constraints,multivariable coupling problem,using the neural network prediction model,the rolling horizon optimization control framework,the diesel engine combustion and emissions layered coordination control method,including the upper controller to expect torque tracking as the optimization goal,the lower controller with the optimization target of NO_xconversion rate and amount of ammonia escape.Then by solving the optimization problem of the target,the control input that meets the requirements is obtained.The simulation results show that the proposed layered control method can meet the driver’s torque demand while effectively reducing the NO_xemissions below the national standard VI,which verifies the effectiveness of the controller.