Hydrogen internal combustion engine H2-SCR tail gas treatment control method based on model predictive control and neural network optimization

Abstract

The invention discloses a hydrogen internal combustion engine H2-SCR tail gas treatment control method based on model predictive control and neural network optimization. The method comprises the steps that a coupling kinetic model of a hydrogen internal combustion engine combustion system and an H2-SCR catalytic reduction system is built; based on the simulation data of the coupling dynamics model, training an adaptive neural network model for correcting a prediction output parameter of the coupling dynamics model; on the basis of the corrected prediction output parameters, constructing a state-space equation of model prediction control, and designing a target function and constraint conditions; engine working condition parameters and SCR system state parameters are collected in real time, the H2-SCR hydrogen injection amount and the catalyst heating power are dynamically optimized through the state-space equation and the target function, and a control instruction is output. By means of the method, NOx emission can be prevented from exceeding the standard, the temperature working interval of the catalyst is ensured, then control accuracy and robustness are ensured, and the overall performance and economical efficiency of the system are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of tail gas treatment, and particularly relates to a control method for H2-SCR tail gas treatment of a hydrogen internal combustion engine based on model predictive control and neural network optimization. Background Technique

[0002] As a key technology of a clean energy power system, the hydrogen internal combustion engine is becoming the core breakthrough point for the low-carbon transformation in the transportation field. Traditional diesel internal combustion engines rely on fossil fuels and emit a large amount of carbon dioxide (CO2) in the tail gas. However, the hydrogen internal combustion engine uses hydrogen as fuel and can achieve nearly zero carbon emissions throughout the life cycle. The hydrogen internal combustion engine is not only an important carrier for hydrogen energy application, can follow the existing internal combustion engine industrial chain to reduce the transformation cost, but also can form a complement with fuel cell technology. In addition, the combustion product of the hydrogen internal combustion engine is mainly water vapor, and the generation amount of nitrogen oxides (NOx, including NO, NO2, etc.) in its tail gas is much lower than that of traditional diesel internal combustion engines. In terms of NOx post-treatment, the hydrogen internal combustion engine utilizes the strong reducibility of hydrogen and treats NOx in the tail gas through hydrogen selective catalytic reduction (H2-SCR) technology, realizing the integrated supply of "fuel - reducing agent" without an additional urea supply system, further simplifying the tail gas treatment system.

[0003] However, there are still some deficiencies in the existing technology. Although the combustion product of the hydrogen internal combustion engine is mainly water vapor, under high-temperature combustion and stoichiometric air-fuel ratio conditions, nitrogen (N2) in the air will still react with oxygen (O2) to generate NOx (mainly in the form of NO). Therefore, NOx post-treatment is still a necessary link for the hydrogen internal combustion engine to meet environmental protection regulations. Due to the flammable characteristics of hydrogen, the hydrogen internal combustion engine has a wider air-fuel ratio operation range compared with the diesel internal combustion engine and can more flexibly control the combustion strategy to adapt to changing working conditions. Under certain working conditions, the tail gas of the hydrogen internal combustion engine may contain unburned hydrogen, and these unburned hydrogen can also be used as a reducing agent for SCR, thus reducing the hydrogen injection amount in the SCR system. However, under flexible and variable working conditions, it is challenging to precisely control the hydrogen injection amount in the SCR system to balance NOx treatment and unburned hydrogen emissions. Existing control strategies often have difficulty in accurately predicting and optimally controlling NOx emissions and catalyst temperature under multi-transient working conditions, resulting in insufficient system robustness and unable to fully exert the environmental protection advantages of the hydrogen internal combustion engine. Summary of the Invention

[0004] To solve the above technical problems, the present invention proposes a control method for H2-SCR tail gas treatment of a hydrogen internal combustion engine based on model predictive control and neural network optimization to solve the problems existing in the above-mentioned existing technology.

[0005] To achieve the above object, the present invention provides a control method for H2-SCR tail gas treatment of a hydrogen internal combustion engine based on model predictive control and neural network optimization, including:

[0006] Build a coupled kinetic model of a hydrogen internal combustion engine combustion system and an H2-SCR catalytic reduction system, where the coupled kinetic model includes an engine combustion model, a NOx generation model, an exhaust gas temperature model, and an SCR reaction kinetic model;

[0007] Based on the simulation data of the coupled kinetic model, train an adaptive neural network model to correct the predicted output parameters of the coupled kinetic model;

[0008] Based on the corrected predicted output parameters, construct a state space equation for model predictive control, and design an objective function and constraint conditions;

[0009] Collect the engine operating condition parameters and SCR system state parameters in real time, and dynamically optimize the hydrogen injection amount and catalyst heating power of H2-SCR through the state space equation and the objective function, and output control instructions.

[0010] Preferably, building the coupled kinetic model includes:

[0011] Establish a one-dimensional simulation model of a hydrogen internal combustion engine, calibrate the hydrogen injection strategy parameters, Wiebe combustion model parameters, and Zeldovich NOx generation model parameters, and predict the NOx raw emissions and unburned hydrogen concentration under different engine operating conditions;

[0012] Construct an H2-SCR reaction kinetic model, calibrate the catalytic reaction rate equation and the catalyst temperature equation, and couple it with the one-dimensional simulation model of the hydrogen internal combustion engine to predict the NOx final emissions of the SCR system.

[0013] Preferably, the hydrogen injection strategy parameters include hydrogen injection timing, injection duration, injection pressure, and air-fuel ratio.

[0014] Preferably, the input parameters of the adaptive neural network model include the hydrogen injection timing, injection duration, ignition timing, intake pressure, and engine speed at the current moment, and the output parameters include the NOx raw emissions, unburned hydrogen concentration, SCR catalyst temperature, and SCR conversion rate at the future moment.

[0015] Preferably, the input variables of the state space equation include the hydrogen injection amount and the catalyst heating power, and the output variables include the SCR catalyst temperature and the NOx final emission amount.

[0016] Preferably, the objective function is to minimize the weighted sum of the NOx final emission amount, hydrogen injection amount consumption, and the deviation of the catalyst temperature from the optimal working range.

[0017] Preferably, the constraint conditions of the objective function include the upper limit of the hydrogen injection amount, the catalyst temperature range, and the NOx instantaneous emission limit.

[0018] Preferably, the parameters collected in real time include the NOx concentration, unburned hydrogen concentration, and exhaust gas temperature at the SCR inlet, as well as the final NOx emission at the SCR outlet.

[0019] Preferably, the generation of the dynamic optimization control instruction includes: based on the state - space equation and objective function of model predictive control, solving for the optimal hydrogen injection amount and heating power through a rolling - horizon optimization algorithm.

[0020] Preferably, the hydrogen injection device of the H2 - SCR catalytic reduction system is installed in front of the SCR catalyst inlet, and the power of the catalyst heater is dynamically adjusted according to the corrected SCR temperature model.

[0021] Compared with the prior art, the present invention has the following advantages and technical effects:

[0022] The present invention provides a control method for H2 - SCR tail - gas treatment of a hydrogen internal combustion engine based on model predictive control and neural network optimization. First, a coupled kinetic model of the hydrogen internal combustion engine combustion system and the H2 - SCR catalytic reduction system is constructed. The coupled kinetic model includes an engine combustion model, a NOx generation model, an exhaust gas temperature model, and an SCR reaction kinetic model. Secondly, based on the simulation data of the coupled kinetic model, an adaptive neural network model is trained to correct the predicted output parameters of the coupled kinetic model. Thirdly, based on the corrected predicted output parameters, a state - space equation of model predictive control is constructed, and an objective function and constraint conditions are designed. Finally, the engine operating parameters and SCR system state parameters are collected in real time, and the hydrogen injection amount and catalyst heating power of H2 - SCR are dynamically optimized through the state - space equation and the objective function, and control instructions are output.

[0023] The present invention optimizes the hydrogen injection amount and the start - stop decision of the heater in the SCR system through model predictive control (MPC) and an adaptive neural network model (ANN) to avoid exceeding the NOx emission standard and ensure the catalyst temperature working range, thereby ensuring the accuracy and robustness of control and improving the overall performance and economy of the system. Description of the Drawings

[0024] The drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments and descriptions thereof of this application are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0025] Figure 1 It is a flowchart of the control method for H2 - SCR tail - gas treatment of a hydrogen internal combustion engine according to an embodiment of the present invention;

[0026] Figure 2Schematic diagram of the control principle for the tail gas treatment of a hydrogen internal combustion engine H2-SCR according to an embodiment of the present invention. Specific embodiments

[0027] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will describe this application in detail with reference to the accompanying drawings and in combination with the embodiments.

[0028] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0029] Embodiment 1

[0030] Considering that unburned hydrogen will be generated in the cylinder under certain combustion conditions of the engine, calibrating the hydrogen injection amount of the H2-SCR system only based on the tail gas temperature will lead to excessive hydrogen emissions. Since hydrogen is combustible within a relatively wide air-fuel ratio range, it is somewhat dangerous that the high-temperature tail gas contains excessive hydrogen.

[0031] This embodiment precisely controls the H2 injection amount of H2-SCR based on the combustion strategy to improve the stability and safety of the system. Specifically, this embodiment optimizes the hydrogen injection amount and the start-stop decision of the heater in the SCR system through model predictive control (MPC) and an adaptive neural network model (ANN) to avoid exceeding the NOx emission standard and ensure the working range of the catalyst temperature, thereby ensuring the accuracy and robustness of the control and improving the overall performance and economy of the system.

[0032] To achieve more precise model predictive control (Model Predictive Control, MPC), first, a coupled model capable of real-time predicting the combustion of the engine and the dynamic behavior of the SCR system needs to be built. This model needs to accurately describe the hydrogen combustion of the engine, NO x generation, the change of the tail gas temperature, and the SCR reduction reaction, and use the data of this model to train the ANN model to correct the model prediction value and prevent the model predictive control from failing under extreme conditions.

[0033] As Figure 1-2 shown, this embodiment provides a control method for the tail gas treatment of a hydrogen internal combustion engine H2-SCR based on model predictive control and neural network optimization, including:

[0034] S1. Construct a coupled kinetic model of the combustion system of the hydrogen internal combustion engine and the H2-SCR catalytic reduction system, where the coupled kinetic model includes an engine combustion model, a NOx generation model, a tail gas temperature model, and an SCR reaction kinetic model;

[0035] Furthermore, constructing the coupled dynamics model includes:

[0036] S101. Establish a one-dimensional simulation model of a hydrogen internal combustion engine, calibrate the parameters of the hydrogen injection strategy, Wiebe combustion model, and Zeldovich NOx generation model, and predict the original NOx emissions and unburned hydrogen concentration of the engine under different operating conditions;

[0037] Specifically, in this embodiment, taking GT-SUITE as an example, determine the target engine configuration (such as setting the number of engine cylinders, cylinder diameter, stroke, and compression ratio, etc.), set the hydrogen injection strategy (such as direct injection or intake port injection, injection pressure, air-fuel ratio, and hydrogen injection timing), use bench test data to calibrate the Wiebe combustion model (such as combustion duration and ignition delay), and finally use the Zeldovich NOx generation model to calculate the in-cylinder NOx production. This model needs to be calibrated using the emission data of the target engine, and its function is to reduce the number of experimental groups and predict the original tail gas emissions of the engine under different operating conditions.

[0038] S102. Construct an H2-SCR reaction kinetics model, calibrate the catalytic reaction rate equation and the catalyst temperature equation, and couple it with the one-dimensional simulation model of the hydrogen internal combustion engine to predict the final NOx emissions of the SCR system.

[0039] Specifically, calibrate k1 and k2 of the NOx and H2 reaction kinetics equation according to the performance data of the H2-SCR catalyst sample at different temperatures, as shown in formula (1), and calculate the catalyst temperature using formula (2). Couple the model of H2-SCR with the one-dimensional simulation model of the engine to obtain the final tail gas emissions of the engine under different operating conditions.

[0040]

[0041] Among them, T SCR is the SCR catalyst carrier temperature, Q rxn is the NO x reduction heat release, Q exhaust is the exhaust heat input, Q loss is the heat dissipation loss.

[0042] S2. Based on the simulation data of the coupled dynamics model, train an adaptive neural network model to correct the predicted output parameters of the coupled dynamics model;

[0043] Furthermore, the input parameters of the adaptive neural network model include the hydrogen injection timing, injection duration, ignition timing, intake pressure, and engine speed at the current moment, and the output parameters include the original NOx emissions, unburned hydrogen concentration, SCR catalyst temperature, and SCR conversion rate at the future moment.

[0044] Specifically, traditional physical models (such as the Zeldovich NO x formation equation and the Arrhenius catalytic reaction equation) are prone to prediction errors under complex working conditions, resulting in the failure of MPC optimization. ANN can correct the predicted value of NOx final emissions by learning the complex non-linear relationship of NO x formation under different engine working conditions through training.

[0045] In this embodiment, an appropriate number of ANN layers is set, and the test set output by the calibrated one-dimensional H2-SCR and internal combustion engine coupling model is used for training. The test set needs to cover a wide range of transient working condition changes, and the bench emission test data of a hydrogen internal combustion engine equipped with H2-SCR is used as the validation set for testing. Among them, the input parameters of the neural network are the hydrogen injection timing T inj_start at time k, the hydrogen injection duration T inj_dur at time k, the ignition timing T ignition at time k, the intake pressure P intake at time k, and the engine speed N. The output parameters of the neural network are the raw NOx emissions NO x_origin at time k + n, the unburned hydrogen concentration H 2_unburn at time k + n, the SCR temperature T SCR at time k + n, and the SCR conversion rate η SCR at time k + n. Here, n is the time domain required for prediction. Therefore, through the above input parameters at time k, the output parameters at time k + n can be predicted using the neural network.

[0046] S3. Based on the corrected predicted output parameters, construct the state space equation of model predictive control, and design the objective function and constraint conditions;

[0047] Furthermore, the input variables of the state space equation include the hydrogen injection amount and the catalyst heating power, and the output variables include the SCR catalyst temperature and the NOx final emission amount.

[0048] Furthermore, the objective function is to minimize the weighted sum of the NOx final emission amount, the hydrogen injection amount consumption, and the deviation of the catalyst temperature from the optimal working range.

[0049] Furthermore, the constraint conditions of the objective function include the upper limit of the hydrogen injection amount, the catalyst temperature range, and the NOx instantaneous emission limit.

[0050] In essence, this embodiment still describes the dynamic behavior of the system through the MPC state - space model, and predicts and controls the hydrogen injection amount of H2 - SCR according to the state - space equation and the objective function. The ANN model trained with the one - dimensional internal combustion engine and SCR coupling simulation dataset is used to correct the parameters predicted by MPC to improve the control accuracy. An example of the state - space model of MPC is as follows:

[0051] x k+n = Ax k + Bu k (3)

[0052] y k = Cx k + Du k (4)

[0053] Where x k is the state variable, u k is the control variable, y k is the output, and A, B, C, D are all system matrices, which can be calibrated through experiments or simulations.

[0054] In this embodiment, the output is the final NOx emission NO x_final after the SCR catalyst, the SCR device temperature T SCR , the control variables are the hydrogen injection amount H 2_SCR_inject of H2 - SCR, and the power P _SCR_heater of the H2 - SCR heater. The state variables are various variables related to the engine performance, such as the hydrogen injection timing T inj_start , the hydrogen injection duration T inj_dur , the ignition timing T ignition , the intake pressure P intake , and the engine speed N, etc. In addition, emission data needs to be obtained in the physical engine. A NOx sensor needs to be installed in front of the H2 - SCR device to measure the original NOx emission NO x_origin , a hydrogen concentration sensor is installed to measure the unburned hydrogen concentration H 2_unburn , and a temperature sensor is installed to measure the exhaust gas temperature T exhuast_in at the inlet of the H2 - SCR. A NOx sensor is installed behind the H2 - SCR device to measure the final NOx emission NO x_final , and a temperature sensor is installed to measure the exhaust gas temperature T exhuast_out at the outlet of the H2 - SCR. The hydrogen nozzle is installed in front of the H2 - SCR device to provide the hydrogen H 2_SCR_inject required to reduce NOx. The H2 - SCR device also needs to set the heater power P _SCR_heater to maintain a suitable catalyst working temperature during cold start.

[0055] The MPC design and workflow calibrated by ANN are used in this embodiment, specifically including:

[0056] S301. Construct a state-space equation, establish a combustion system equation for the in-cylinder internal combustion engine to predict the future raw NOx emissions NO x_origin and the concentration H of unburned hydrogen 2_unburn , and the equations are shown in Formulas (5) and (6). The functions f1(...) and f2(...) therein are mechanism models corrected by fitting functions trained in combination with ANN.

[0057]

[0058] To predict the future final NOx emissions NO x_final , it is necessary to construct an SCR system equation. Before that, it is necessary to first construct an SCR temperature model as shown in Formula (7), and the function h(…) in this formula is corrected by combining the SCR temperature trained using ANN.

[0059]

[0060] Finally, use Equation (8) to predict the final NOx emissions NO x_final , where the function f3(...) also combines the SCR conversion rate η SCR trained using ANN for fitting.

[0061]

[0062] S302. Construct an MPC objective function, optimize the variables within the prediction range to minimize the objective function J as shown in Equation (9).

[0063]

[0064] Among them, it is necessary to set the weight w1 to represent the minimum final NOx emissions, the weight w2 to represent the minimum hydrogen injection consumption of H2-SCR, and the weight w3 to represent the optimal temperature range suitable for the catalyst to work, where T opt represents the optimal temperature for the SCR catalyst to work.

[0065] S303. Add constraints to the variables, such as setting the maximum constraint of the hydrogen injection amount, the minimum and maximum temperature constraints of the SCR device, and the maximum constraint of the instantaneous final NOx emissions.

[0066] S4. Real-time collect the engine operating condition parameters and the SCR system state parameters, dynamically optimize the hydrogen injection amount and the catalyst heating power of H2-SCR through the state-space equation and the objective function, and output control instructions.

[0067] Furthermore, the parameters collected in real time include the NOx concentration, unburned hydrogen concentration, and exhaust gas temperature at the SCR inlet, as well as the final NOx emissions at the SCR outlet.

[0068] Furthermore, the generation of the dynamic optimization control instruction includes: based on the state space equation and objective function of model predictive control, solving for the optimal hydrogen injection amount and heating power through a rolling horizon optimization algorithm.

[0069] Furthermore, the hydrogen injection device of the H2-SCR catalytic reduction system is installed in front of the SCR catalyst inlet, and the power of the catalyst heater is dynamically adjusted according to the corrected SCR temperature model.

[0070] Specifically, to implement MPC predictive control, by optimizing the objective function (9), the controller determines the optimal H2-SCR hydrogen injection amount H 2_SCR_inject and the power P of the catalyst heater _SCR_heater , and the user can also set weights according to requirements to meet the needs of different specifications of hydrogen nozzles, heaters, and SCR materials.

[0071] Advantages of this embodiment:

[0072] This embodiment proposes a control method for the tail gas NOx treatment system of a hydrogen internal combustion engine H2-SCR system based on the collaborative optimization of model predictive control (MPC) and neural network (ANN). Aiming at problems such as large fluctuations in NOx generation and low utilization rate of unburned hydrogen under wide air-fuel ratio conditions of hydrogen internal combustion engines, the traditional NH3-SCR and urea supply strategies are abandoned, and H2-SCR is adopted. By constructing a coupled kinetic model of engine combustion and the H2-SCR system, combined with the non-linear correction ability of ANN for traditional mechanism models, accurate prediction of NOx emissions and catalyst temperature is achieved. The MPC controller is based on the state space equation, with the hydrogen injection amount and heating power as control variables, dynamically optimizing the balance between NOx conversion efficiency and hydrogen consumption, while ensuring the catalyst active temperature range. This method innovatively combines physical models and data-driven models, trains ANN through bench data to compensate for prediction deviations under extreme conditions, and solves the problem of insufficient robustness of traditional control strategies under multi-transient conditions, providing an intelligent control solution for the efficient and clean emissions of hydrogen internal combustion engines.

[0073] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A control method for the H2-SCR tail gas treatment of a hydrogen internal combustion engine based on model predictive control and neural network optimization, characterized in that, It includes the following steps: Construct a coupled kinetic model of a hydrogen internal combustion engine combustion system and an H2-SCR catalytic reduction system, where the coupled kinetic model includes an engine combustion model, a NOx generation model, an exhaust gas temperature model, and an SCR reaction kinetic model; Train an adaptive neural network model based on the simulation data of the coupled kinetic model to correct the predicted output parameters of the coupled kinetic model; Construct a state space equation for model predictive control based on the corrected predicted output parameters, and design an objective function and constraint conditions; Real-time collect the engine operating condition parameters and SCR system state parameters, and dynamically optimize the hydrogen injection amount and catalyst heating power of H2-SCR through the state space equation and the objective function, and output control instructions.

2. The method according to claim 1, wherein: Constructing the coupled kinetic model includes: Establish a one-dimensional simulation model of a hydrogen internal combustion engine, calibrate the hydrogen injection strategy parameters, Wiebe combustion model parameters, and Zeldovich NOx generation model parameters, and predict the NOx raw emissions and unburned hydrogen concentration under different engine operating conditions; Construct an H2-SCR reaction kinetic model, calibrate the catalytic reaction rate equation and the catalyst temperature equation, and couple it with the one-dimensional simulation model of the hydrogen internal combustion engine to predict the NOx final emissions of the SCR system.

3. The method according to claim 2, wherein: The hydrogen injection strategy parameters include hydrogen injection timing, injection duration, injection pressure, and air-fuel ratio.

4. The method according to claim 1, wherein: The input parameters of the adaptive neural network model include the hydrogen injection timing, injection duration, ignition timing, intake pressure, and engine speed at the current moment, and the output parameters include the NOx raw emissions, unburned hydrogen concentration, SCR catalyst temperature, and SCR conversion rate at a future moment.

5. The method according to claim 1, wherein: The input variables of the state space equation include the hydrogen injection amount and the catalyst heating power, and the output variables include the SCR catalyst temperature and the NOx final emission amount.

6. The method according to claim 1, wherein: The objective function is to minimize the weighted sum of the NOx final emission amount, hydrogen injection amount consumption, and the deviation of the catalyst temperature from the optimal operating range.

7. The method according to claim 6, wherein: The constraint conditions of the objective function include the upper limit of the hydrogen injection amount, the catalyst temperature range, and the NOx instantaneous emission limit.

8. The method according to claim 1, wherein: The real-time collected parameters include the NOx concentration, unburned hydrogen concentration, exhaust gas temperature at the SCR inlet, and the NOx final emission amount at the SCR outlet.

9. The method according to claim 1, wherein: The generation of the dynamically optimized control instruction includes: based on the state space equation and the objective function of model predictive control, solve for the optimal hydrogen injection amount and heating power through a rolling horizon optimization algorithm.

10. The method according to claim 1, wherein: The hydrogen injection device of the H2-SCR catalytic reduction system is installed in front of the SCR catalyst inlet, and the power of the catalyst heater is dynamically adjusted according to the corrected SCR temperature model.

Images