An intelligent control method and system based on diesel engine after-treatment system emissions

Abstract

The present invention belongs to the technical field of intelligent control of diesel engine emissions and discloses a method and system for intelligent control of diesel engine aftertreatment system emissions. The method comprises constructing a multi-head perceptron neural network model, wherein the model has multiple hidden layers, and starting from the second hidden layer, the neural network is divided into multiple channels, each of which predicts different parameters; inputting training data into the model to obtain a prediction result of the diesel engine aftertreatment system emissions; defining a loss function, optimizing the model parameters, and using the trained model to perform predictions; and adjusting relevant inputs of the aftertreatment system based on the prediction results to achieve intelligent control of the diesel engine aftertreatment system emissions. The present invention uses a method combining a neural network and feedback control to predict diesel engine aftertreatment system emissions, and then adjusts relevant inputs of the aftertreatment system based on the predicted data information to achieve intelligent control of diesel engine aftertreatment system emissions.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control of diesel engine emissions, and in particular to an intelligent control method and system based on diesel engine after-treatment system emissions. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Diesel engines have the advantages of high power density and low fuel consumption, and are widely used in medium and heavy-duty road vehicles, construction machinery, ships and other fields. x It is the main pollutant in diesel engine exhaust, which not only causes acid rain and smog, but also causes serious harm to the human body (Reference: Deepak Singh, Amit Kumar, Krishan Kumar et al. Statistical modeling of O3, NOx, CO, PM2.5, VOCs and noise levels in commercial complexand associated health risk assessment in an academic institution [J]. Science of The Total Environment, 2016, 572:586-594.). At present, the standards for exhaust emissions are becoming increasingly stringent: the subsequent emission standards in Europe include NO xThe emission limit of NO was lowered, and the emission limit of N2O was increased. N2O, commonly known as "laughing gas", is a recognized strong greenhouse gas. The greenhouse effect it produces is 298 times that of the same amount of CO2 (Reference: Ashok Kumar, Krishna Kamasamudram, Neal Currier et al. SCR Architectures for Low N2OEmissions[R], SAE International, 2015.). At the same time, N2O can remain in the atmosphere for about 150 years and can diffuse into the stratosphere to destroy the atmospheric ozone layer. It is considered to be an important atmospheric stratospheric ozone depleting substance in the 21st century (Reference: AR Ravishankara, John S. Daniel, and Robert W. Portmann. Nitrous Oxide(N2O): The Dominant Ozone-Depleting Substance Emitted in the 21st Century[J]. science, 2009, 326 (5949):123-125.). Therefore, it is necessary to control NO x and N2O emissions are controlled.

[0004] Currently, the general NO x The emission control method is to use closed loop control to monitor the NO x The concentration signal and NH3 concentration signal are used to feedback and adjust the injection amount of urea. The feedback signal has a time delay, which is not conducive to real-time and accurate adjustment. This single control method cannot meet the increasingly stringent emission standards. It is difficult to meet the requirements of the standards for the emission control of the newly emerging N2O gas and lacks an adaptive mechanism. At present, the NO x The accuracy of gas sensors is insufficient, and N2O gas sensors are still in the initial development stage. Currently, only gas analyzers and other equipment under laboratory conditions can ensure the accuracy of NO in the post-treatment system. x High accuracy for gases and N2O gases.

[0005] How to realize NOx in diesel engine after-treatment system x More precise control of N2O gas to meet increasingly stringent emission standards is an urgent problem that needs to be solved by relevant technical personnel. Summary of the Invention

[0006] In order to solve the above problems, the present invention proposes an intelligent control method and system based on diesel engine after-treatment system emissions, which adopts a method combining neural network and feedback control, builds a training neural network model through laboratory related data, and controls NOx The conversion rate, N2O concentration at the end of the post-treatment system and NH3 concentration are predicted, and then the relevant inputs of the post-treatment system are adjusted according to the predicted data information to achieve NO x , N2O and NH3 precise control.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] In a first aspect, the present invention provides an intelligent control method for emissions from a diesel engine aftertreatment system, comprising the following steps:

[0009] Constructing a multi-head perceptron neural network model, the model comprising an input layer, a hidden layer, and an output layer, wherein the hidden layer has multiple layers, and starting from the second hidden layer, the neural network is divided into multiple channels, each of which predicts different parameters;

[0010] Input the training data into the model to obtain the prediction results of diesel engine aftertreatment system emissions;

[0011] Define the loss function, optimize the model parameters, train the model, and use the trained model to predict diesel engine aftertreatment system emissions;

[0012] The relevant inputs of the after-treatment system are adjusted according to the predicted results to realize the intelligent control of the emissions of the diesel engine after-treatment system.

[0013] As an optional embodiment, the training data includes exhaust gas temperature, space velocity, ammonia nitrogen ratio and NO2 ratio before the selective catalytic reduction converter.

[0014] As an optional embodiment, the prediction results of diesel engine after-treatment system emissions include NO x Conversion rate, N2O concentration and NH3 concentration at the end of the after-treatment system.

[0015] As an optional implementation, the root mean square error and the coefficient of determination are used to evaluate the accuracy of the prediction results.

[0016] As an optional implementation, the relevant inputs of the post-processing system are adjusted according to the prediction results, specifically:

[0017] By controlling the throttle opening, changing the diesel engine rail pressure and controlling the post-injection, the air speed, exhaust temperature and the NO2 ratio before the selective catalytic reduction converter are adjusted, and the ammonia nitrogen ratio is controlled by controlling the urea injection amount.

[0018] As an optional implementation method, the multi-head perceptron neural network model adopts an error back propagation algorithm to adjust weights and biases to reduce the error, and optimizes the algorithm using a gradient descent method.

[0019] In a second aspect, the present invention provides an intelligent control system based on diesel engine after-treatment system emissions, comprising:

[0020] The model construction module is configured to: construct a multi-head perceptron neural network model, the model including an input layer, a hidden layer, and an output layer, the hidden layer having multiple layers, and starting from the second hidden layer, the neural network is divided into multiple channels, and the multiple channels respectively predict different parameters;

[0021] The prediction module is configured to: input the training data into the model to obtain a prediction result of the diesel engine aftertreatment system emissions;

[0022] The model training module is configured to: define a loss function, optimize model parameters, train the model, and use the trained model to predict emissions from a diesel engine aftertreatment system;

[0023] The intelligent control module is configured to adjust relevant inputs of the after-treatment system according to the predicted results to achieve intelligent control of emissions from the diesel engine after-treatment system.

[0024] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.

[0025] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the method described in the first aspect is performed.

[0026] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which implements the method described in the first aspect when executed by a processor.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The present invention proposes an intelligent control method and system based on diesel engine after-treatment system emissions, which uses MHP neural network to perform NOx detection on the end of after-treatment system based on historical data and real-time sensor data. x, N2O and NH3 concentrations, thereby achieving early adjustment and optimization of temperature, air velocity, ammonia nitrogen ratio and NO2 ratio, effectively improving the timeliness and accuracy of control. The MHP neural network adopted in the present invention shows faster training speed and higher prediction accuracy compared with the traditional MLP neural network. Even when the sample size is small, the MHP neural network can still maintain excellent prediction performance, thereby reducing the pressure on the laboratory in data collection and greatly reducing the experimental cost of collecting gas parameters in the post-processing system. At the same time, the MHP neural network adopted in the present invention also has decoupling capabilities. When processing multivariable prediction tasks, it can separate the relationship and influence between different variables, reduce mutual interference between variables, and maintain high stability when facing noisy data. The MHP neural network can better capture the potential patterns of the data during the training process, thereby showing better generalization ability when facing new data and reducing the risk of overfitting.

[0029] This invention proposes an intelligent control method and system for diesel engine aftertreatment system emissions. This system utilizes a combined MHP neural network and ECU control to achieve precise and optimized control. The neural network's powerful learning and adaptability allows it to handle nonlinear relationships and complex dynamics, improving the adaptability and robustness of the aftertreatment system. This invention not only overcomes the limitations of traditional NOx gas control strategies due to their single-source nature, but also effectively avoids accuracy loss by reducing reliance on sensor quality. Furthermore, by reducing the number of sensors, this invention further reduces installation and maintenance costs.

[0030] This paper proposes an intelligent control method and system based on diesel engine after-treatment system emissions. Based on the in-depth analysis of a large amount of experimental data, the key factors affecting the after-treatment system are revealed. By finely adjusting the diesel engine parameters and the amount of urea injection, the temperature, air velocity, NO2 / NO x ratio and precise control of the ammonia nitrogen ratio, thereby achieving NO x This innovative approach provides a new approach and practical path for the diesel engine industry to achieve ultra-low emission targets.

[0031] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0033] Figure 1A schematic diagram of the test bench instrument layout that is the source of experimental data required for the neural network of the present invention;

[0034] Figure 2 A schematic diagram of the MHP neural network structure provided in Example 1 of the present invention;

[0035] Figure 3 This is a flow chart of the intelligent control method for diesel engine after-treatment system emissions provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0036] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0037] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0038] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0039] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0040] Example 1

[0041] like Figure 3 As shown, this embodiment provides an intelligent control method based on diesel engine after-treatment system emissions, comprising the following steps:

[0042] S1. Constructing a multi-head perceptron neural network model, the model comprising an input layer, a hidden layer, and an output layer, wherein the hidden layer has multiple layers, and starting from the second hidden layer, the neural network is divided into multiple channels, wherein the multiple channels respectively predict different parameters;

[0043] S2. Input the training data into the model to obtain the prediction results of diesel engine after-treatment system emissions;

[0044] S3. Define the loss function, optimize the model parameters, train the model, and use the trained model to predict the emissions of the diesel engine aftertreatment system;

[0045] S4. Adjust relevant inputs of the after-treatment system according to the predicted results to achieve intelligent control of emissions from the diesel engine after-treatment system.

[0046] The present invention adopts a method combining neural network and feedback control, relying on the classic layout of DOC+DPF+SCR+ASC post-processing system, and builds a training neural network model through laboratory related data to control NO x The conversion rate, N2O concentration at the end of the post-treatment system and NH3 concentration are predicted, and then the relevant inputs of the post-treatment system are adjusted according to the predicted data information to achieve NO x , N2O and NH3 control.

[0047] (1) Construction of neural network

[0048] The training samples are all scattered data (not pictures and do not have sequence characteristics), which are more suitable for multi-layer perceptron (MLP) models with simple structure, flexibility and strong nonlinear mapping capabilities. The present invention adopts a multi-head perceptron (MHP) neural network model developed on the basis of the MLP model. This model draws on the multi-head attention mechanism. Compared with the traditional MLP model, it can have faster training speed and higher prediction accuracy even with a small sample size. It can also maintain high stability when facing noisy data during training. The MHP model used in the present invention consists of 5 layers of neurons, including an input layer, three hidden layers and an output layer, where the number of neurons in the hidden layers is 6, 9 and 9 respectively. Starting from the second hidden layer, the neural network is divided into three channels to achieve accurate prediction of each output parameter, such as Figure 2 shown.

[0049] Through the collation and analysis of a large amount of experimental data and extensive reference to literature in related fields, it is finally determined that exhaust temperature, air velocity, ammonia nitrogen ratio and NO2 ratio before the selective catalytic reduction converter (SCR) are the factors affecting NO x The conversion rate, N2O and NH3 concentration at the end of the post-treatment system are important factors affecting the conversion rate. Therefore, the exhaust temperature, air velocity, ammonia nitrogen ratio and NO2 ratio before SCR are selected as the input of the MHP neural network, which are denoted as x1, x2, x3 and x4 respectively; NO x The conversion rate, N2O concentration at the end of the post-treatment system, and NH3 concentration are output from the neural network and are denoted as y1, y2, and y3, respectively. Figure 2 The training samples used in this invention are all obtained through experiments. The layout diagram of the test bench instrument from which the experimental data is derived is shown in Figure 1As shown, the test bench instruments include a dynamometer and a fuel consumption meter. The dynamometer is used to measure the output power, speed and torque of the diesel engine, and the fuel consumption meter is used to measure the fuel consumption of the diesel engine. These two devices can be used to determine the working condition of the diesel engine. The data description is as follows:

[0050] Filter out the required data. The ammonia nitrogen ratio requires the NH3 sensor and NOx under laboratory conditions. x NH3 and NO obtained by the sensor x The calculation formula of concentration and ammonia nitrogen ratio is:

[0051]

[0052] Among them, ANR is the ammonia nitrogen ratio, and The unit of concentration is ppm.

[0053] NO x The data required for conversion rate is NO before SCR x concentration and NO after SCR x The concentration is calculated as follows:

[0054]

[0055] in, As a percentage, it means NO x conversion efficiency.

[0056] The data required for the NO2 ratio is NO before SCR x concentration and NO x The concentration is calculated as follows:

[0057]

[0058] in, Indicates NO before SCR x The concentration of Indicates the concentration of NO before SCR, in ppm.

[0059] (2) Neural network training

[0060] Before training begins, all data sets must be divided, and 75% of the total sample size is selected as the training data for the neural network. The remaining samples are used in the neural network prediction stage, and it is ensured that the training data set and the test data set have the same distribution characteristics.

[0061] set up represents the weight of the connection from the jth neuron in the kth layer to the ith neuron in the k+1th layer, represents the bias from the kth layer to the k+1th layer, represents the number of neurons in the k-th layer, represents the weighted input of the k-th layer of neurons, represents the activation value (output value) of the kth layer, where , the learning rate is The transfer function selects the sigmoid excitation function with nonlinear amplification gain. The sigmoid function is defined by the following formula:

[0062]

[0063] This function has the following characteristics:

[0064]

[0065] The weighted input of any neuron in the second layer of the neural network and activation value for:

[0066]

[0067]

[0068] in, is the sample data input to the input layer. Since the number of neurons in the second layer of the neural network is 6, i=1, 2, ..., 6. Based on the data output by the neurons in the second layer, the weighted input and activation value of any neuron in the third layer are calculated as follows:

[0069]

[0070]

[0071] in, is the output value of the neurons in the second layer of the neural network. Since the number of neurons in the third layer of the neural network is 9, i=1, 2,…, 9.

[0072] Since the third and fourth layers of the neural network are connected separately, the weighted input and activation value of each neuron in each segment of the fourth layer must be given separately.

[0073] The weighted input and activation value of any neuron in the fourth layer connected to the first neuron in the output layer is:

[0074]

[0075]

[0076] in, is the output value of the third layer of neurons in the neural network, and i=1, 2, 3 at this time.

[0077] The weighted input and activation value of any neuron in the fourth layer connected to the second neuron in the output layer is:

[0078]

[0079]

[0080] in, is the output value of the third layer of neurons in the neural network, and i=4, 5, 6 at this time.

[0081] The weighted input and activation value of any neuron in the fourth layer connected to the third neuron in the output layer is:

[0082]

[0083]

[0084] in, is the output value of the third layer of neurons in the neural network, and i=7, 8, 9 at this time.

[0085] The weighted input and activation values ​​of each neuron in the output layer of the neural network are:

[0086]

[0087]

[0088]

[0089]

[0090]

[0091]

[0092] The MHP neural network uses the error back propagation algorithm to adjust the weights and biases to reduce the error. The algorithm is optimized by the gradient descent method.

[0093]

[0094] Bundle Recorded as , the loss function is , we can get the error of each neuron and the error of the output layer neurons:

[0095]

[0096] The hidden layer neuron error is:

[0097]

[0098] From the above error formula, the parameter update formula can be obtained as follows:

[0099]

[0100]

[0101] (3) Neural network prediction

[0102] After completing the training of the MHP neural network, the neural network prediction is performed on the test data set, using the root mean square error RMSE and the coefficient of determination R 2 To evaluate the accuracy of the prediction results, the calculation formulas of the two parameters are:

[0103]

[0104]

[0105] Where p is the number of samples, is the experimental value of the i-th sample, is the predicted value of the i-th sample, is the average value of the neural network output, SSE is the residual sum of squares, which represents the error between the experimental value and the predicted value, and SST is the total sum of squares, which represents the sum of the deviations between the experimental value and its average value.

[0106] Based on the above prediction model, N2O, NO x The comprehensive control method of CO2 and NH3 is carried out through the following process:

[0107] Invented an optimization control method based on Cu-based molecular sieve SCR catalyst to improve NO x Through in-depth analysis of experimental data, we determined the optimal ranges of several key operating parameters, including space velocity, ammonia nitrogen ratio, NO2 ratio, and temperature, and proposed relevant control methods.

[0108] Optionally, as the airspeed increases, the SCR system's NO x The conversion efficiency will decrease. At the same time, high air velocity under low temperature conditions will lead to increased NH3 leakage, while low air velocity under high temperature conditions will help reduce NH3 leakage. It is particularly noteworthy that under low temperature and low air velocity conditions, N2O emissions increase significantly, especially in the temperature range of 200℃ to 250℃. Therefore, the present invention preferably sets the air velocity to 6000h -1 , to achieve the best balance between emissions control and efficiency.

[0109] Optionally, at low temperatures, the increase in the ammonia-nitrogen ratio has an effect on NOx The impact on conversion efficiency is small, and the activity of the catalyst plays a dominant role at this time; in high temperature environments, appropriately increasing the ammonia nitrogen ratio can significantly increase NO x conversion rate. However, an excessively high ammonia-nitrogen ratio will lead to increased NH3 leakage at low temperatures, while at high temperatures the leakage amount will be relatively stable due to the oxidation of NH3. N2O emissions are mainly concentrated in the range of 200°C to 250°C. Within this range, the standard SCR increases the amount of N2O generated as the ammonia-nitrogen ratio increases, while the fast SCR decreases the amount of N2O generated as the ammonia-nitrogen ratio increases. Based on the above findings, the present invention proposes an adaptive ammonia-nitrogen ratio control strategy: maintain the ammonia-nitrogen ratio at 1 at low temperatures; when the temperature exceeds 350°C, gradually increase the ammonia-nitrogen ratio until it reaches above 1.2 at 550°C.

[0110] Optional, after analyzing the experimental data, it was found that NO2 / NO x When it is around 0.2, NO x The temperature range of 95% and 90% conversion efficiency is wider; at low temperature, with the NO2 / NO x The amount of N2O generated increases first and then decreases. At high temperatures, the NO2 / NO x Therefore, the present invention proposes to increase the NO2 / NO x The ratio is maintained at about 0.2 to ensure efficient conversion of NO x while reducing N2O emissions.

[0111] Optionally, based on analysis of experimental data, high N2O emissions are mainly concentrated in the temperature range of 200°C to 250°C, so the operating temperature of the SCR system is kept away from this temperature range as much as possible and operated above 250°C.

[0112] In order to achieve real-time monitoring and precise control of the above key parameters, the present invention constructs a prediction model based on the MHP neural network and transmits the prediction results of the model to the ECU (engine control unit). The ECU timely adjusts and controls the throttle opening, changes the rail pressure of the diesel engine and the post-injection control airspeed to maintain at 6000h -1 The exhaust temperature is controlled above 250℃, and the NO2 / NO x Maintain at around 0.2; variable ammonia nitrogen ratio control is achieved by controlling the urea injection amount, and the urea injection rate is adjusted in real time according to temperature changes to achieve the target ammonia nitrogen ratio.

[0113] Example 2

[0114] This embodiment provides an intelligent control system based on diesel engine aftertreatment system emissions, including:

[0115] The model construction module is configured to: construct a multi-head perceptron neural network model, the model including an input layer, a hidden layer, and an output layer, the hidden layer having multiple layers, and starting from the second hidden layer, the neural network is divided into multiple channels, and the multiple channels respectively predict different parameters;

[0116] The prediction module is configured to: input the training data into the model to obtain a prediction result of the diesel engine aftertreatment system emissions;

[0117] The model training module is configured to: define a loss function, optimize model parameters, train the model, and use the trained model to predict emissions from a diesel engine aftertreatment system;

[0118] The intelligent control module is configured to adjust relevant inputs of the after-treatment system according to the predicted results to achieve intelligent control of emissions from the diesel engine after-treatment system.

[0119] It should be noted that the above modules correspond to the steps described in Example 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above Example 1. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0120] In further embodiments, there is also provided:

[0121] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed by the processor, wherein when the computer instructions are executed by the processor, the method described in Example 1 is performed. For the sake of brevity, no further details are given here.

[0122] It should be understood that in this embodiment, the processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), off-the-shelf field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0123] The memory may include a read-only memory and a random access memory, and provides instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0124] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the method described in Example 1 is performed.

[0125] The method in Example 1 can be directly implemented as a hardware processor, or can be implemented using a combination of hardware and software modules in the processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, it will not be described in detail here.

[0126] A computer program product includes a computer program, which implements the method described in embodiment 1 when executed by a processor.

[0127] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions contained in program modules, which are executed in a device on a real or virtual processor of a target to perform the process / method described above. Generally, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or divided between program modules as needed. The machine-executable instructions for the program modules can be executed in local or distributed devices. In distributed devices, program modules can be located in local and remote storage media.

[0128] The computer program code for implementing the method of the present invention can be written in one or more programming languages. These computer program codes can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the computer or other programmable data processing device, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on a computer, partially on a computer, as an independent software package, partially on a computer and partially on a remote computer, or entirely on a remote computer or server.

[0129] In the context of the present invention, computer program code or related data can be carried by any appropriate carrier to enable a device, apparatus, or processor to perform the various processes and operations described above. Examples of carriers include signals, computer-readable media, and the like. Examples of signals include electrical, optical, radio, acoustic, or other forms of propagated signals, such as carrier waves, infrared signals, and the like.

[0130] Those skilled in the art will appreciate that the units and algorithm steps of the various examples described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0131] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. An intelligent control method based on diesel engine after-treatment system emissions, characterized in that: The following steps are involved: Constructing a multi-head perceptron neural network model, the model comprising an input layer, a hidden layer, and an output layer, wherein the hidden layer has multiple layers, and starting from the second hidden layer, the neural network is divided into multiple channels, each of which predicts different parameters; The multi-head perceptron neural network model is based on a multi-layer perceptron model and draws on the multi-head attention mechanism to separate the relationships and influences between different variables, reduce mutual interference between variables, and maintain high stability even in the face of noisy data. The multi-headed perceptron neural network includes an input layer, three hidden layers, and an output layer, wherein the number of neurons in the hidden layers is 6, 9, and 9 respectively; The third and fourth layers of the neural network are connected separately, and the weighted input and activation value of each neuron in each segment of the fourth layer must be given separately; The weighted input and activation value of any neuron in the fourth layer connected to the first neuron in the output layer is: in, is the output value of the third layer of neurons in the neural network, at this time i=1, 2, 3; The weighted input and activation value of any neuron in the fourth layer connected to the second neuron in the output layer is: in, is the output value of the third layer of neurons in the neural network, at this time i=4, 5, 6; The weighted input and activation value of any neuron in the fourth layer connected to the third neuron in the output layer is: in, is the output value of the third layer of neurons in the neural network, at this time i=7, 8, 9; The weighted input and activation values ​​of each neuron in the output layer of the neural network are: ； Input the training data into the model to obtain the prediction results of diesel engine aftertreatment system emissions; Define the loss function, optimize the model parameters, train the model, and use the trained model to predict diesel engine aftertreatment system emissions; The prediction results of diesel engine aftertreatment system emissions include NO x Conversion rate, N2O concentration and NH3 concentration at the end of the post-treatment system; Adjust the relevant inputs of the after-treatment system based on the prediction results to achieve intelligent control of diesel engine after-treatment system emissions; Adjust the relevant inputs of the post-processing system based on the predicted results, specifically: The prediction results of the multi-head perceptron neural network model are transmitted to the ECU. The ECU makes timely adjustments based on the prediction results according to the preset control strategy and method. By controlling the throttle opening, changing the diesel engine rail pressure and controlling the post-injection, the airspeed, exhaust temperature and the NO2 ratio before the selective catalytic reduction converter are adjusted. The ammonia nitrogen ratio is controlled by controlling the urea injection amount. A method combining multi-head sensor neural network and feedback control is adopted, relying on the classic layout of the after-treatment system of DOC+DPF+SCR+ASC. A neural network model is built and trained using relevant laboratory data to predict the NOx conversion rate, N2O concentration at the end of the after-treatment system, and NH3 concentration. The relevant inputs of the after-treatment system are adjusted according to the predicted data information to control NOx, N2O and NH3.

2. The intelligent control method based on diesel engine after-treatment system emissions according to claim 1, characterized in that: The training data includes exhaust gas temperature, air velocity, ammonia nitrogen ratio and NO2 ratio before the selective catalytic reduction converter.

3. The intelligent control method based on diesel engine after-treatment system emissions according to claim 1, characterized in that: The root mean square error and coefficient of determination were used to evaluate the accuracy of the prediction results.

4. The intelligent control method based on diesel engine after-treatment system emissions according to claim 1, characterized in that: The multi-headed perceptron neural network model uses the error back propagation algorithm to adjust the weights and biases to reduce the error, and uses the gradient descent method to optimize the algorithm.

5. An intelligent control system based on diesel engine after-treatment system emissions using the method of claim 1, characterized in that: include: The model construction module is configured to: construct a multi-head perceptron neural network model, the model including an input layer, a hidden layer, and an output layer, the hidden layer having multiple layers, and starting from the second hidden layer, the neural network is divided into multiple channels, and the multiple channels respectively predict different parameters; The prediction module is configured to: input the training data into the model to obtain a prediction result of the diesel engine aftertreatment system emissions; The model training module is configured to: define a loss function, optimize model parameters, train the model, and use the trained model to predict emissions from a diesel engine aftertreatment system; The intelligent control module is configured to adjust relevant inputs of the after-treatment system according to the predicted results to achieve intelligent control of emissions from the diesel engine after-treatment system.

6. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 4 is completed.

7. A computer-readable storage medium, characterized in that Used to store computer instructions, which, when executed by a processor, complete the method according to any one of claims 1 to 4.

8. A computer program product, characterized in that The invention comprises a computer program, which is used to implement the method according to any one of claims 1 to 4 when executed by a processor.

Images