Data-driven Based Diesel Engine Combustion Process Modeling And Control

With the emissions regulations becoming more and more strict,National V emission standard have already been implemented and the National VI emission standard will be implemented in the next stage.For the National VI emission standard,only using combustion control or exhaust gas after-treatment can hardly meet the emission requirements.Currently a feasible technical is that reduce the the generation of nitrogen oxides(NO_x)to 2g/kwh by combustion control and then use selective catalytic reduction(SCR)and Diesel Particulate Filter(DPF)to further process of exhaust gas,to meet the National VI emission standards.Traditional control method of diesel combustion is the open loop calibration map or simple PID control.But as the emission regulations become more and more strict,the above control method can hardly meet the requirements of the national five standards.Therefore,it is necessary to adopt advanced control theory and method to precisely control the combustion process of diesel engine to meet the requirements of emission regulations.But diesel engine combustion is a complex process,involving complicated chemical process and complicated physics knowledge,such as decomposition,combustion and heat transfer,thermodynamics,and in order to improve the economy,power performance and emission performance,diesel engines are usually equipped with exhaust gas recirculation(EGR)and turbo(VGT).This causes the mechanism model of diesel engine combustion process is very complex,and presents the challenge to control the combustion process effectively.For above question,this paper design a predictive control algorithm based on data driven for diesel combustion process to realize the coordinated control of diesel torque and emissions.Firstly,for the question of the choice of the inspiration input,in this paper,the combustion process of diesel engine is analyzed in detail from the angle of mechanism,and the oil injection quantity,oil injection angle,the opening of EGR valve and the opening of VGT valve are selected as the inspiration input.This paper select fuel consumption rate,NO_xemissions and diesel engine crankshaft output torque as output.This paper utilizes a 2.0 L diesel engine GT model with EGR and VGT as plant,and in the condition of open loop utilizes gaussian white noise to incentive it,and obtain the experimental data.Secondly,subspace identification method can hardly describe the nonlinear characteristic of diesel combustion process.For above question,this article describes nonlinear system as linear varying system,and under the framework of linear parameter variable systems use multiple local linear model with parameter dependence to deduces the output prediction equation of the form of linear parameter variation.This paper utilized particle swarm optimization to optimize the center point and bandwidth of radial basis function and obtain the dependency parameter.Multiple local model with parameter dependency will cause the problem of matrix dimension explosion,in order to solve above problem,the dual problem of the original problem is constructed and the kernel function is introduced,and the system output prediction equation of the kernel method is derived.At the same time,a set of working conditions is designed to validation the accuracy of output prediction equation in the form of kernel function,and compared it with the linear subspace identification method.The experimental results show that the accuracy of the output prediction equation in the form of kernel function is better than that of linear model,and its prediction accuracy can be improved by 10% to 30%.Finally,in view of the coordinated control problem of diesel power performance and diesel emissions,this paper utilized system prediction equation,took the requirement of torque and the emissions of NO_xless than 2g/kwh as optimization,took the control input and its change rate as constrains,and chose appropriate control time domain to realize of coordinated control of power performance and emission.ETC conditions include cities,villages and highways,in this paper,the simulation experiment is carried out in three working conditions.The experimental results show that the relative torque tracking error of the controller based on the linear parameter variable output prediction equation is reduced by 57.14%,61.78%,64.53% respectively.At the same time,the crankshaft output torque can track the expected torque,and the mean value of NO_xis less than 2g/ kwh meeting the control requirement of diesel combustion process.