Explicit Model Predictive Control Method for Turbocharged Diesel Engine Gas Path

Abstract

This invention relates to an explicit model predictive control method for the airflow path of a turbocharged diesel engine, belonging to the field of turbocharged diesel engine airflow path control technology. The purpose of this invention is to address the tracking control problem of intake manifold pressure and exhaust manifold pressure in the airflow path system of a turbocharged diesel engine. A data-driven control-oriented model is established, and a linear controller based on the explicit model predictive control method is designed for this turbocharged diesel engine airflow path. The steps of this invention are: system identification and data acquisition, system identification, setting the matrix to be identified, setting the measurement matrix and input / output data matrix, and dividing the control process into offline and online parts. This invention offers higher accuracy and better reflects the transient and steady-state characteristics of the system. Compared to traditional MPC, it has a faster response speed, occupies less ECU memory, significantly reduces computational requirements, and lowers hardware costs.

Description

Technical Field

[0001] This invention belongs to the field of turbocharged diesel engine air circuit control technology. Background Technology

[0002] Due to the enormous and ever-increasing number of vehicles in my country, excessive consumption of fuel resources has occurred. Incomplete combustion of fuel produces polluting exhaust gases, which are not only harmful to human health but also cause serious environmental pollution. Currently, vehicle exhaust emissions have become a major source of smog. Since the 1980s and 90s, my country has formulated specific laws and regulations to effectively control pollution from vehicle exhaust emissions. Under the dual pressures of increasingly severe environmental pollution and a growing energy shortage, improving the fuel efficiency of internal combustion engines and reducing pollutant emissions is an imperative choice for the engine industry. Energy conservation, emission reduction, and vehicle pollution control are urgent priorities, making the research and development of advanced engine technologies an inevitable trend.

[0003] Diesel engines offer significant advantages in power, economy, and environmental friendliness, leading to their widespread application in the automotive industry and their gradual emergence as a mainstream vehicle power source. The development of technologies such as Variable Geometry Turbochargers (VGTs) and Exhaust Gas Recirculation (EGR) valves has further propelled the adoption of diesel engines. However, turbocharged diesel engines are typical multiple-input multiple-out (MIMO) systems, exhibiting strong coupling, multiple constraints on actuators, and time-varying dynamic characteristics. Since both the EGR valve and VGT blades are within the exhaust gas flow, a strong coupling exists between the EGR and VGT flows. Reduced fresh air pumped into the intake manifold leads to increased PM emissions, while a lower EGR gas flow fraction results in higher NOx emissions. This dilemma is known as the NOx-PM trade-off. This strong coupling exists not only between the VGT and EGR actuators but also along the fuel path. Its highly nonlinear characteristics mean that traditional control methods are unable to cope with increasingly stringent pollution and environmental protection requirements.

[0004] Model Predictive Control (MPC) is one of the most promising control strategies in industrial applications, capable of handling constraints on manipulated variables in MIMO systems. MPC optimizes at each time step and offers greater flexibility in handling input, output, and state constraints. However, the real-time implementation of MPC incurs a high computational burden due to solving a finite-range optimal control problem in each sampling period. This high computational burden places greater demands on the processing power of the ECU, increasing hardware costs. Therefore, computationally expensive control methods are unsuitable for diesel engine production. Current control methods for diesel engine air circuits can control strongly coupled and highly nonlinear diesel engine air circuit systems to a certain extent, but the following problems still exist:

[0005] 1. For diesel engine gas circuit systems with strong coupling and strong nonlinearity, existing mechanism modeling and controller design are very complex, and there is still a large gap between theoretical control methods and engineering implementation.

[0006] 2. The storage and computing capabilities of vehicle ECUs are limited. For existing methods, if one only pursues engineering implementation, such as using PID to simplify the diesel engine air circuit system into a SISO system for control, although the feasibility is enhanced, the dynamic characteristics of the diesel engine air circuit are greatly abandoned. If one pursues control precision, such as using nonlinear MPC for control, it greatly exceeds the storage and computing capabilities of the vehicle ECU and cannot be used in practice.

[0007] 3. The control of the diesel engine's air circuit must be carried out under strong constraints to ensure its safety and reliability. Existing methods, to some extent, avoid or weaken the influence of constraints, which is not conducive to practical applications. Summary of the Invention

[0008] The purpose of this invention is to address the problem of tracking and controlling the intake manifold pressure and exhaust manifold pressure in the air circuit system of a turbocharged diesel engine. A data-driven control-oriented model is established, and an explicit model predictive control method for the air circuit of a turbocharged diesel engine with a linear controller is designed based on the explicit model predictive control method.

[0009] The steps of this invention are:

[0010] S1. System Identification Data Acquisition

[0011] We chose to apply the stimulus using APRBS and control variables in order to measure the system response under more operating conditions.

[0012] S2. System Identification

[0013] The model for the turbocharged diesel engine's air passage system is set as follows:

[0014]

[0015] Where m is the system order; k is the current sampling time of the system, k = 1, 2, ..., M; M > 2m + 1; M is the total sampling duration; a i (i = 1, 2, ..., m), b j (j=0,1,…,m) is a 2×2 parameter matrix; Let ξ(k) be the intake manifold pressure and exhaust manifold pressure output by the system at time k; ξ(k) is the prediction noise at time k.

[0016] S3. Let θ = [a1, a2, ..., a m ,b0,b1,…,b m ] T The matrix to be identified is a 2×(2m+1) row, 2 column matrix;

[0017] The observation matrix consists of 2 rows and 2×(2m+1) columns, where and These are the actual output and input values ​​of the system at time k-1 obtained from data acquisition.

[0018] The least squares identification form of the gas path system is:

[0019]

[0020] S4. Order For measurement matrix; H M =[h(1)h(2)…h(M)] T Given the input and output data matrix, the least squares error J of the gas path system is:

[0021]

[0022] The extreme points obtained by taking the derivative of the error with the matrix to be identified as the independent variable are the optimal parameters to be identified.

[0023] S5. The control process is divided into offline and online parts.

[0024] a. During the offline process, the state-space equation of the turbocharged diesel engine system is first obtained. This equation is discrete, as shown in the following equation:

[0025]

[0026] Where, let n x n u n y These represent the spatial dimensions of state variables, control variables, and output variables, respectively. These are the state variables of the controlled system; The control inputs to the controlled system are the openings of VGT and EGR. The output of the controlled system is the pressure in the intake and exhaust manifolds.

[0027] Define a quadratic cost function to describe the performance metrics:

[0028]

[0029]

[0030] Where N is the prediction time domain length, U N ={u0,...,u N-1 Let Y be the control sequence, Q and R be the weight matrices of the control output matrix and control input matrix, respectively. min Y max For output vector constraints; U min U max To control input constraints, based on the physical characteristics of the control input and output, the VGT opening constraint is set to 20%-100%; the EGR opening constraint is set to 0%-100%; b. In the online process, the controller first obtains the input, executes the lookup table command based on the input, searches each partition and determines whether the partition is feasible. If feasible, the control law for this partition is executed; if not feasible, the search and determination of the next partition continues until all partitions are traversed. If no partition meets the requirements, the control law is not executed.

[0031] The beneficial effects of this invention are:

[0032] 1. Using the variable geometry turbocharger and the exhaust gas recirculation valve opening as control variables, and the intake manifold pressure and exhaust manifold pressure as output variables, a model predictive control method is adopted to better solve the control problems of multi-input multi-output, multi-constraint, strong coupling and strong nonlinearity of the turbocharged diesel engine air circuit system.

[0033] 2. To address the system's data identification requirements, a method combining APRBS and control variable step inputs is used to apply excitation to the controlled model. Compared to control-oriented models identified using traditional data acquisition methods, this method offers higher accuracy and better reflects the system's transient and steady-state characteristics.

[0034] 3. The explicit model predictive control method used divides traditional MPC into offline and online parts. The offline part solves the QP problem to form a control law table, while the online part looks up the table to obtain the optimal control law for control. Compared with traditional MPC, it has a faster response speed, occupies less ECU memory, greatly reduces computational requirements, and lowers hardware costs. Attached Figure Description

[0035] Figure 1This is a schematic diagram of the air passage structure of a turbocharged diesel engine;

[0036] Figure 2a This is a schematic diagram of APRBS excitation signal data acquisition;

[0037] Figure 2b This is a schematic diagram of step excitation signal data acquisition;

[0038] Figure 3 It is a comparison of model outputs;

[0039] Figure 4 This is a block diagram of the air circuit system control for a turbocharged diesel engine;

[0040] Figure 5 This is a schematic diagram of the working principle of EMPC;

[0041] Figure 6a This is a verification diagram of the tracking effect of the identification system model controller under the operating conditions of 1300 rpm and 50 mg / cycle fuel injection.

[0042] Figure 6b This is a verification diagram of the tracking effect of the identification system model controller under the operating conditions of 1850 rpm and 95 mg / cycle fuel injection. Detailed Implementation

[0043] This invention effectively solves the problems of high time and manpower costs in mechanism modeling, while also avoiding, to some extent, the memory consumption and computational burden of traditional MPC control methods. The novel control method proposed in this invention is easier to implement in engineering, and simulation results demonstrate the effectiveness of the designed control system. Furthermore, this control framework can effectively address the three problems mentioned above.

[0044] The present invention may include the following components in terms of structure: a turbocharged diesel engine air circuit system oriented towards a control model, a controlled object, and an EMPC controller.

[0045] A 12.7-liter six-cylinder high-precision turbocharged diesel engine model was selected as the controlled object. By collecting the input and output data of the controlled object model and performing system identification, the control-oriented model obtained can effectively replace the complex model for controller design. Throughout the control process, the opening degree of VGT and EGR is used as the control variable, and the intake manifold pressure and exhaust manifold pressure are used as the output variable. The EMPC controller uses predictive control to provide feedback control of the opening degree of VGT and EGR at the next moment under the condition that the input and output meet the constraints, forming a control closed loop to ensure its safety and reliability.

[0046] The implementation method of the present invention includes the following parts:

[0047] The pressure tracking of the intake and exhaust manifolds in a turbocharged diesel engine, controlled by the explicit model predictive control method described in this invention, is achieved through a software system. This software system consists of the advanced MATLAB / Simulink simulation software. MATLAB / Simulink is used for building the simulation model of the controller and identifying the turbocharged diesel engine's airflow path model, and provides a simulation experimental environment.

[0048] First, the control-oriented model of the turbocharged diesel engine's airflow system is identified using MATLAB / Simulink software. The data basis for system identification can be obtained from the high-fidelity, high-precision Lars model. The opening degrees of VGT and EGR are used as control variables, and the pressures of the intake and exhaust manifolds are used as output variables. The control objective is to track the pressures of the diesel engine's intake and exhaust manifolds.

[0049] Diesel engines operate and respond differently under various conditions, so system identification should cover as many operating conditions as possible. Theoretically, a turbocharger consists of a VGT (Vehicle Gas Turbine) and a compressor. The VGT extracts energy from the exhaust gases to power the compressor, mounted on the same shaft, which in turn compresses more fresh air, resulting in higher pressure in the intake manifold. The EGR (Exhaust Gas Recirculation) circuit feeds some of the combustion exhaust gases back into the intake manifold to dilute the fresh air, thereby reducing peak combustion temperature and NOx concentration. Air and combusted gases mix and are pumped into the cylinders from the intake manifold. When the piston reaches the top of its compression stroke, fuel is injected into the cylinders and combusted in the compressed air, generating torque on the crankshaft. The hot exhaust gases are pumped from the cylinders into the exhaust manifold, with some flowing out of the engine through the VGT and the rest recirculated back into the intake manifold through the EGR valve. Therefore, in order to better meet its dynamic characteristics, the method of combining amplitude-modulated pseudo-random binary sequence (APRBS) and control variable step is selected to apply excitation to the system and collect data. The output data obtained is then used for system identification to obtain the system's state-space equation.

[0050] The EMPC-based controller is designed based on the identified system-oriented control model. The offline part solves a quadratic programming (QP) problem to obtain offline partitioned control laws. The online part selects an appropriate control law based on engine operating conditions through a lookup table, thereby achieving the goal of controlling the output to track the desired value and realizing the control loop of the entire process. Based on the above principles, the entire control system is built in the MATLAB / Simulink environment.

[0051] The present invention will now be described in detail with reference to the accompanying drawings:

[0052] The working principle of a turbocharged diesel engine is as follows: Figure 1 As shown, fresh air is pumped into the intake manifold by the compressor. c This refers to the compressor's air mass flow rate. After the combustion gas enters the cylinder, the exhaust gas produced by combustion flows into the exhaust manifold, where W... e It is the total mass flow rate of the engine gas, p in p ex These are the pressures of the intake and exhaust manifolds, respectively. Part of the exhaust gas flows to the VGT valve to provide some power to the turbocharger, while the other part flows to the EGR valve for secondary utilization. egr It is the mass flow rate of gas flowing into the intake manifold via the EGR valve, W f W is the engine fuel injection quantity. t It is the turbine gas mass flow rate, N e T is the engine speed. in T ex The temperatures of the intake manifold and exhaust manifold are λ, respectively. a F1 is the air-fuel ratio in the cylinder, and F1 is the fraction of gases participating in combustion in the intake manifold. The control block diagram of the turbocharged diesel engine control system based on explicit model predictive control in this invention is as follows: Figure 4 As shown, the opening degree of VGT and EGR is used as the control variable, and the intake manifold pressure and exhaust manifold pressure are used as the output variable. The system identification result is used as the control-oriented model of this invention, and the entire control framework is built in Simulink.

[0053] The control objective of this invention is that the controller controls the opening of the VGT valve and EGR valve according to the operating state of the turbocharged diesel engine, thereby tracking the intake manifold pressure and exhaust manifold pressure.

[0054] This invention provides a device based on the above operating principles and processes. The setup and operation process are as follows:

[0055] 1. Software Selection

[0056] The simulation models of the controlled object and controller of this control system were built using Matlab / Simulink software, specifically Matlab R2022a. The simulation step size was fixed at 0.01 s.

[0057] 2. System Identification

[0058] 2.1 System Identification Data Acquisition

[0059] To ensure the accuracy and reliability of the simulation results, the data acquisition step should follow these four principles: 1. The applied input / excitation during data acquisition should fully reflect the system's dynamics; 2. The measurement time should be long enough; 3. The acquired system should have a good signal-to-noise ratio; 4. An appropriate sampling time interval should be selected. As mentioned above, to cover more engine operating states, this invention does not use traditional data acquisition methods, but instead chooses to apply excitation through APRBS and control variables to measure the system response under more operating conditions, as shown in Figure 2. APRBS is a pseudo-random binary sequence with adjustable amplitude. In reality, all data combinations appear randomly. The PRBS code stream largely possesses this "random data" characteristic. Adjusting the excitation amplitude based on PRBS allows the obtained excitation signal to be suitable for exciting the transient characteristics of turbocharged diesel engines. The control variable step excitation includes all the execution action combinations of VGT and EGR under specific operating conditions, obtaining more steady-state information of the system. The resulting control-oriented model has higher accuracy and can better reflect the transient and steady-state dynamic characteristics of the controlled system, such as... Figure 3 As shown, the accuracy of the obtained control-oriented model is above 95%.

[0060] 2.2 System Identification

[0061] For the intake system of a turbocharged diesel engine, its model can be set as follows:

[0062]

[0063] Where m is the system order; k is the current sampling time of the system, k = 1, 2, ..., M; M > 2m + 1; M is the total sampling duration; a i (i = 1, 2, ..., m), b j (j=0,1,...,m) is a 2×2 parameter matrix; Let ξ(k) be the intake manifold pressure and exhaust manifold pressure output by the system at time k; ξ(k) is the prediction noise at time k.

[0064] Let θ = [a1, a2, ..., a m ,b0,b1,...,b m ] T The matrix to be identified is a 2×(2m+1) row, 2 column matrix;

[0065] This is an observation matrix with 2 rows and 2×(2m+1) columns. and These are the actual output and input values ​​of the system at time k-1 obtained from data acquisition.

[0066] Therefore, the least squares identification form of the gas path system is:

[0067]

[0068] make For measurement matrix; H M =[h(1) h(2)…h(M)] T Given the input and output data matrix, the least squares error J of the gas path system is:

[0069]

[0070] The extreme point obtained by taking the derivative of the error with the matrix to be identified as the independent variable is the optimal parameter to be identified, and the gas path explicit model predictive controller is designed.

[0071] As mentioned above, this invention, taking into account the modeling characteristics of diesel engine air circuit systems, employs an explicit model predictive control method to design a controller that controls the intake manifold pressure and exhaust manifold pressure of the controlled object. The working principle of explicit model predictive control is as follows: Figure 5 As shown, its control process is divided into offline and online parts.

[0072] The offline EMPC controller pre-solves the mp-QP problem. For each input within a given range, the EMPC pre-calculates the optimal solution. The solution consists of a linear function, is piecewise affine and continuous. The constraints divide the solution into different regions, and each region maps to a unique optimal solution.

[0073] During the offline process, the state-space equation of the turbocharged diesel engine system is first obtained. This equation is discrete, as shown in the following equation:

[0074]

[0075] Where, let n x n u n y These represent the spatial dimensions of state variables, control variables, and output variables, respectively. These are the state variables of the controlled system; The control inputs to the controlled system are the openings of VGT and EGR. Let A represent the output of the controlled system, i.e., the pressure in the intake manifold and exhaust manifold. Let A, B, and C be the system state space matrices identified for the control model.

[0076] Define a quadratic cost function to describe the performance metrics:

[0077]

[0078]

[0079] Where N is the prediction time domain length, U N ={u0,…,u N-1 Let} represent the control sequence, and Q and R be the weight matrices of the control output matrix and control input matrix, respectively. min Y max For output vector constraints; U min U max To control input constraints, this invention sets the VGT opening constraint to 20%-100% and the EGR opening constraint to 0%-100% based on the physical characteristics of the control input and output.

[0080] Based on the state-space equations of the control-oriented model, the predicted output at the current moment can be inferred as:

[0081]

[0082] The above formula represents the predicted output Y(k+n|k) obtained by performing n-step prediction on the output Y at time k, where

[0083] n = 1, 2, ..., N, τ = 1, 2, ..., n-1, X(k) is the state quantity of the system at time k. In this invention, the state quantity is obtained by the system and is not given a specific physical meaning.

[0084] Substituting equation (1.1) into equation (5) and rearranging, we get:

[0085]

[0086] stGU N ≤W+LX(k) (1.2)

[0087] Where D, E, F, G, W, and L are constant matrices, which can be derived and simplified using equations (5) and (1.1) as well as Q and R. Since in equation (1.2) The term is determined by the state variable at the current moment, and the optimal control law U N The solution has no impact, so it is omitted in the calculation.

[0088] Define z = U N +E -1 F T Substituting X(k) into equation (1.2), and simplifying, we obtain the optimal form of the standard quadratic form:

[0089]

[0090] stGz≤W+SX(k) (1.3)

[0091] Where S = L + GE -1 F T,

[0092] EMPC introduces multi-parameter quadratic programming theory to perform convex partitioning of the system's state region, obtaining piecewise affines within the multi-cell partition, and then solves the above-mentioned optimal problem offline to obtain the corresponding partition control law.

[0093] During the online process, the EMPC locates the current state region online to select the appropriate control operation. First, the controller receives input and executes a lookup command based on it, searching each partition and determining its feasibility. If feasible, the control law for that partition is executed; otherwise, the search continues to the next partition until all partitions have been traversed. If no partition meets the requirements, the control law is not executed.

[0094] Experimental verification and analysis

[0095] To verify the tracking performance of the controller designed based on the explicit model predictive control method for the intake manifold pressure and exhaust manifold pressure of the turbocharged diesel engine's air circuit system, this invention sets up two sets of tracking experiments to verify the tracking accuracy. The first set identifies the system model under the condition of 1300 rpm and a fuel injection quantity of 50 mg / cycle. The parameters of the identified fourth-order state-space equation are as follows:

[0096]

[0097]

[0098]

[0099] Steady-state conditions were selected for step tracking, as shown in Figure 6(a). The second group identified the model under conditions of 1850 rpm and 95 mg / cycle fuel injection. The parameters of the fourth-order state-space equation obtained were:

[0100]

[0101]

[0102]

[0103] Similarly, steady-state conditions are selected for step tracking, as shown in Figure 6(b).

[0104] As shown in Figure 6, after applying a step change at 27s, the system can quickly identify and respond to the change in the tracked value within 1s. Figure (a) shows 27–30s, and Figure (b) shows 27–32s, indicating that the system exhibits minimal jitter and almost no overshoot during the response process. Both experimental conditions achieve rapid tracking within 5s, while maintaining system stability during changes. Figures (a) show 0–27s and 30–60s, and Figure (b) show 0–27s and 32–60s, indicating that the tracking error is controlled within 1% during the tracking process, demonstrating good tracking performance. Therefore, the controller designed in this invention for turbocharged diesel engines based on the explicit model predictive control method can meet the tracking requirements of the intake manifold pressure and exhaust manifold pressure of turbocharged diesel engines.

[0105] A table of definitions for the proper nouns involved in this invention:

[0106] Abbreviations of proper nouns Specific meaning VGT Variable geometry turbocharger EGR Exhaust gas recirculation valve MIMO Multiple Input Multiple Output SISO Single Input Single Output MPC Model predictive control EMPC Explicit Model Predictive Control APRBS Pseudo-random binary sequence mp-QP Multi-parameter quadratic programming problem

[0107] A table showing the meanings of the variables involved in this invention:

[0108]

[0109]

Claims

1. A predictive control method for the explicit model of the air path of a turbocharged diesel engine, characterized in that: The steps are as follows: S1. System Identification Data Acquisition We chose to apply the stimulus using APRBS and control variables in order to measure the system response under more operating conditions. S2. System Identification The model for the turbocharged diesel engine's air passage system is set as follows: Where m is the system order; k is the current sampling time of the system, k = 1, 2, ..., M; M > 2m + 1; M is the total sampling duration; a i (i = 1, 2, ..., m), b j (j=0,1,...,m) is a 2×2 parameter matrix; Let ξ(k) be the intake manifold pressure and exhaust manifold pressure output by the system at time k; ξ(k) is the prediction noise at time k. S3. Let θ = [a1, a2, ..., a m ,b0,b1,...,b m ] T The matrix to be identified is a 2×(2m+1) row, 2 column matrix; The observation matrix consists of 2 rows and 2×(2m+1) columns, where and These are the actual output and input values ​​of the system at time k-1 obtained from data acquisition. The least squares identification form of the gas path system is: S4. Order For measurement matrix; H M =[h(1) h(2) … h(M)] T Given the input and output data matrix, the least squares error J of the gas path system is: The extreme points obtained by taking the derivative of the error with the matrix to be identified as the independent variable are the optimal parameters to be identified. S5. The control process is divided into offline and online parts. a. During the offline process, the state-space equation of the turbocharged diesel engine system is first obtained. This equation is discrete, as shown in the following equation: Where, let n x n u n y These represent the spatial dimensions of state variables, control variables, and output variables, respectively. These are the state variables of the controlled system; The control inputs to the controlled system are the openings of VGT and EGR. The output of the controlled system is the pressure in the intake and exhaust manifolds. Define a quadratic cost function to describe the performance metrics: Where N is the prediction time domain length, U N ={u0,...,u N-1 Let Y be the control sequence, Q and R be the weight matrices of the control output matrix and control input matrix, respectively. min Y max For output vector constraints; U min U max To control input constraints, based on the physical characteristics of the control inputs and outputs, the VGT opening constraint is set to 20%-100%; the EGR opening constraint is set to 0%-100%. b. In the online process, the controller first obtains input and executes a lookup command based on the input to search each partition and determine whether the partition is feasible. If it is feasible, the control law for this partition is executed. If it is not feasible, the search and judgment of the next partition continues until all partitions are traversed. If no partition meets the requirements, the control law is not executed.

Images