Intelligent ammonia spraying control method and system

Abstract

The invention relates to an intelligent ammonia spraying control method and system, and belongs to the technical field of ammonia spraying control. The method comprises the following steps: acquiring historical ammonia spraying data, and constructing an outlet nitrogen oxide concentration prediction model through a Bayesian optimization CNN-LSSVM algorithm according to the historical ammonia spraying data; monitoring ammonia spraying data of the denitration system in real time, obtaining the outlet nitrogen oxide concentration through the outlet nitrogen oxide concentration prediction model according to the ammonia spraying data, and outputting an ammonia spraying control signal through neural network self-adaptive control according to the outlet nitrogen oxide concentration, the valve opening degree of an ammonia injection system is controlled according to the ammonia injection control signal, and automatic ammonia injection amount control is achieved, so that emission of nitric oxide is effectively controlled, the denitration efficiency is improved, meanwhile, excessive ammonia injection is avoided, ammonia escape is reduced, and environmental pollution and operation cost are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of ammonia injection control, and in particular relates to an intelligent ammonia injection control method and system. Background Art

[0002] Nitrogen oxides are one of the main components of air pollution, which can cause environmental problems such as acid rain and photochemical smog, and are harmful to human health. In order to control the emission of nitrogen oxides, coal-fired power plants often use a denitrification system for catalytic reduction, using ammonia as a reducing agent to reduce nitrogen oxides. However, there are certain difficulties in controlling the injection amount of ammonia: excessive injection of ammonia will cause unreacted ammonia to escape into the atmosphere, causing secondary pollution and waste of resources, and insufficient ammonia injection will reduce the denitrification efficiency, resulting in substandard nitrogen oxide emissions. During the operation of the denitrification system, the outlet nitrogen oxide concentration fluctuates greatly, and the emission concentration frequently exceeds the standard, making it difficult to stably achieve ultra-low emission standards. Therefore, excessive ammonia injection is often used to ensure that the outlet nitrogen oxide emission concentration meets the standard, which increases the operating cost of the denitrification system.

[0003] At present, PID control method is commonly used for ammonia injection control, which adjusts the ammonia injection amount by monitoring the deviation between the outlet nitrogen oxide concentration and the set value. However, due to the nonlinearity and hysteresis of the denitrification system, PID control is difficult to accurately adjust the ammonia injection amount.

[0004] Therefore, there is an urgent need to develop an intelligent ammonia injection control method to accurately control the injection amount of ammonia to achieve efficient denitrification while minimizing ammonia slip. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention proposes an intelligent ammonia injection control method and system.

[0006] The purpose of the present invention can be achieved through the following technical solutions: An intelligent ammonia injection control method, comprising: Acquire historical ammonia injection data, wherein the historical ammonia injection data includes historical total air volume, historical inlet flue gas temperature, historical ammonia injection amount, historical inlet nitrogen oxide concentration, and historical outlet nitrogen oxide concentration; According to the historical ammonia injection data, a prediction model for outlet nitrogen oxide concentration is constructed by using a Bayesian optimization CNN-LSSVM algorithm; real-time monitoring of ammonia injection data, and obtaining an outlet nitrogen oxide concentration through the outlet nitrogen oxide concentration prediction model according to the ammonia injection data, wherein the ammonia injection data includes a total air volume, an inlet flue gas temperature, an ammonia injection amount, and an inlet nitrogen oxide concentration; An ammonia injection control signal is outputted through a neural network adaptive control according to the outlet nitrogen oxide concentration, and a valve opening of an ammonia injection system is controlled according to the ammonia injection control signal.

[0007] Preferably, constructing an outlet nitrogen oxide concentration prediction model based on the historical ammonia injection data by using a Bayesian optimization CNN-LSSVM algorithm includes: Obtaining historical ammonia injection standard data through normalization processing according to the historical ammonia injection data, and constructing an ammonia injection data set according to the historical ammonia injection standard data; Extracting an ammonia injection feature vector set through CNN according to the ammonia injection data set; The outlet nitrogen oxide concentration prediction model is obtained through LSSVM training and testing according to the ammonia injection feature vector set.

[0008] Preferably, the extracting an ammonia injection feature vector set through CNN according to the ammonia injection data set comprises: For the ammonia injection data set, a convolution kernel is used through a convolution layer to generate an ammonia injection data feature map, some neurons are randomly discarded through a Dropout layer to prevent overfitting, and an ammonia injection feature vector is extracted through the activations function, and finally the ammonia injection feature vector set is obtained by combining them.

[0009] Preferably, the outlet nitrogen oxide concentration prediction model obtained by LSSVM training and testing according to the ammonia injection feature vector set includes: Obtaining an ammonia injection feature vector training set and an ammonia injection feature vector test set by randomly dividing the ammonia injection feature vector set; According to the ammonia injection feature vector training set, an initial model for predicting outlet nitrogen oxide concentration is obtained by LSSVM training; According to the initial model for predicting the outlet nitrogen oxide concentration, a combination of LSSVM hyperparameters is selected by Bayesian optimization combined with five-fold cross validation; According to the LSSVM hyperparameter combination, the outlet nitrogen oxide concentration prediction model is obtained by retraining the LSSVM; The performance of the outlet nitrogen oxide concentration prediction model is evaluated by using an ammonia injection evaluation index according to the ammonia injection feature vector test set.

[0010] Preferably, the ammonia injection evaluation index is the ammonia injection mean absolute error and the ammonia injection determination coefficient, and the mathematical expression is: , , Among them, E is the average absolute error of ammonia injection, R 2 is the ammonia injection determination coefficient, n is the number of samples in the ammonia injection feature vector test set, y i is the actual value of the nitrogen oxide concentration at the outlet of the i-th ammonia injection test sample, is the predicted value of the nitrogen oxide concentration at the outlet of the i-th ammonia injection test sample, and y is the average nitrogen oxide concentration at the outlet of the ammonia injection test sample.

[0011] Preferably, the outputting of the ammonia injection control signal through a neural network adaptive control according to the outlet nitrogen oxide concentration comprises: Constructing a neural network ammonia injection controller, wherein the neural network ammonia injection controller is a three-layer neural network structure; Preset learning rate and momentum factor; Preset expected reference model; By comparing the outlet nitrogen oxide concentration value with the expected reference model output value, an ammonia injection error is calculated; The neural network ammonia injection controller receives the ammonia injection error. The neural network adaptively adjusts the weighting coefficient through the back propagation method and the gradient descent method to minimize the ammonia injection error. It iterates continuously until the ammonia injection error is less than the preset ammonia injection error threshold or the number of iterations reaches the preset maximum number of iterations. The training process ends and the ammonia injection control signal is output.

[0012] Preferably, the input layer of the neural network ammonia injection controller includes three nodes, which respectively receive the current outlet nitrogen oxide concentration error value, the outlet nitrogen oxide concentration error change, and the outlet nitrogen oxide concentration system setting value; the hidden layer of the neural network ammonia injection controller includes six nodes; and the output layer of the neural network ammonia injection controller includes one node for outputting an ammonia injection control signal.

[0013] An intelligent ammonia injection control system using the above-mentioned intelligent ammonia injection control method comprises an ammonia injection data acquisition module, an outlet nitrogen oxide concentration prediction model module, an outlet nitrogen oxide concentration real-time prediction module, and an adaptive ammonia injection control module; The ammonia injection data acquisition module is used to collect the total air volume, inlet flue gas temperature, ammonia injection amount, and inlet nitrogen oxide concentration of the denitrification system; The outlet nitrogen oxide concentration prediction model module is used to construct an outlet nitrogen oxide concentration prediction model according to the historical ammonia injection data through the Bayesian optimization CNN-LSSVM algorithm; The outlet nitrogen oxide concentration real-time prediction module is used to monitor the total air volume, inlet flue gas temperature, ammonia injection amount, inlet nitrogen oxide concentration of the denitrification system in real time and output the outlet nitrogen oxide concentration through the outlet nitrogen oxide concentration prediction model; The adaptive ammonia injection control module is used to output an ammonia injection control signal through a neural network adaptive control according to the outlet nitrogen oxide concentration, and control the valve opening of the ammonia injection system according to the ammonia injection control signal to achieve automatic ammonia injection amount control.

[0014] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned intelligent ammonia injection control method when executing the program.

[0015] A storage medium containing computer executable instructions, which are used to execute the above-mentioned intelligent ammonia injection control method when executed by a computer processor.

[0016] The beneficial effects of the present invention are: (1) The outlet nitrogen oxide concentration prediction model is constructed through the Bayesian optimization CNN-LSSVM algorithm, and the ammonia injection data is monitored in real time. The outlet nitrogen oxide concentration is obtained through the outlet nitrogen oxide concentration prediction model based on the ammonia injection data. The ammonia injection amount is adjusted through the neural network adaptive control according to the predicted outlet nitrogen oxide concentration. Automatically adjusting the ammonia injection amount can more effectively control nitrogen oxide emissions and improve denitrification efficiency. At the same time, excessive ammonia injection can be avoided, thereby reducing ammonia escape, environmental pollution and operating costs.

[0017] (2) By constructing an outlet nitrogen oxide concentration prediction model, the complex nonlinear relationship between the ammonia injection data and the outlet nitrogen oxide concentration can be accurately modeled. The LSSVM hyperparameter combination is selected through Bayesian optimization combined with five-fold cross validation, which reduces the error of the outlet nitrogen oxide concentration prediction model and can provide more accurate feedforward knowledge for the neural network ammonia injection controller, alleviating the nonlinear and hysteresis problems existing in the denitrification system.

[0018] (3) The neural network ammonia injection controller receives the output of the outlet nitrogen oxide concentration prediction model, and the neural network adaptively adjusts the weight to minimize the error between the outlet nitrogen oxide concentration prediction value and the preset reference model output value, outputs the ammonia injection control signal, and controls the valve opening of the ammonia injection system according to the ammonia injection control signal. It can adaptively adjust the ammonia injection amount to reduce ammonia slip while meeting the outlet nitrogen oxide emission standard. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0020] Figure 1 The figure is a flow chart of an intelligent ammonia injection control method of the present invention. DETAILED DESCRIPTION

[0021] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0022] See also Figure 1 , an intelligent ammonia injection control method, comprising: S1: Acquire historical ammonia injection data, wherein the historical ammonia injection data includes historical total air volume, historical inlet flue gas temperature, historical ammonia injection amount, historical inlet nitrogen oxide concentration, and historical outlet nitrogen oxide concentration; S2: constructing an outlet nitrogen oxide concentration prediction model based on the historical ammonia injection data by using a Bayesian optimization CNN-LSSVM algorithm; S3: real-time monitoring of ammonia injection data, and obtaining the outlet nitrogen oxide concentration through the outlet nitrogen oxide concentration prediction model according to the ammonia injection data, wherein the ammonia injection data includes total air volume, inlet flue gas temperature, ammonia injection amount, and inlet nitrogen oxide concentration; S4: outputting an ammonia injection control signal through a neural network adaptive control according to the outlet nitrogen oxide concentration, and controlling the valve opening of the ammonia injection system according to the ammonia injection control signal.

[0023] In this embodiment, the historical ammonia injection data is derived from the historical ammonia injection records of the denitrification system.

[0024] In this embodiment, the construction of the outlet nitrogen oxide concentration prediction model based on the historical ammonia injection data by using the Bayesian optimization CNN-LSSVM algorithm is specifically implemented by the following steps: S201: obtaining historical ammonia injection standard data through normalization processing according to the historical ammonia injection data, and constructing an ammonia injection data set according to the historical ammonia injection standard data; S202: extracting an ammonia injection feature vector set through CNN according to the ammonia injection data set; Specifically, for the ammonia injection data set, the ammonia injection data feature map is first generated by using the convolution kernel through the convolution layer, and then some neurons are randomly discarded through the Dropout layer to prevent overfitting, and finally the ammonia injection feature vector is extracted through the activations function, and the ammonia injection feature vector set is obtained by combining them.

[0025] S203: obtaining an outlet nitrogen oxide concentration prediction model through LSSVM training and testing according to the ammonia injection feature vector set; S203-1: obtaining an ammonia injection feature vector training set and an ammonia injection feature vector test set by randomly dividing the ammonia injection feature vector set; S203-2: obtaining an initial model for predicting outlet nitrogen oxide concentration by LSSVM training according to the ammonia injection feature vector training set; S203-3: selecting a LSSVM hyperparameter combination according to the initial model for predicting the outlet nitrogen oxide concentration by Bayesian optimization combined with five-fold cross validation; In this embodiment, the optimized LSSVM hyperparameter combination is a regularization parameter and a kernel function parameter; specifically, the negative value of the mean absolute error of ammonia injection is used as the target maximum optimization, the LSSVM hyperparameter is used as input, the proxy model of the hyperparameter is initialized, the Gaussian process is used to approximate the objective function, the proxy model is trained through the initial hyperparameter points, and the exploration and utilization are balanced by the expected improvement of the acquisition function. In each iterative step, the acquisition function values ​​of all potential hyperparameter points are calculated using the current proxy model, and the hyperparameter combination that maximizes the acquisition function value is selected. The objective function is evaluated on the selected hyperparameter combination and the proxy model is retrained to reflect the new information. The iteration is repeated until the preset maximum number of iterations is completed, and the optimal hyperparameter combination predicted by the proxy model is output, which is the LSSVM hyperparameter combination.

[0026] S203-4: Obtaining an outlet nitrogen oxide concentration prediction model by retraining the LSSVM according to the LSSVM hyperparameter combination; S203-5: Evaluating the performance of the outlet nitrogen oxide concentration prediction model by using an ammonia injection evaluation index according to the ammonia injection feature vector test set; The ammonia injection evaluation index is the ammonia injection mean absolute error and the ammonia injection determination coefficient, and the mathematical expression is: , , Among them, E is the average absolute error of ammonia injection, R 2 is the ammonia injection determination coefficient, n is the number of samples in the ammonia injection feature vector test set, y i is the actual value of the nitrogen oxide concentration at the outlet of the i-th ammonia injection test sample, is the predicted value of the nitrogen oxide concentration at the outlet of the i-th ammonia injection test sample, and y is the average nitrogen oxide concentration at the outlet of the ammonia injection test sample.

[0027] It should be noted that, in this embodiment, LSSVM performs regression prediction in the feature space through a Gaussian kernel function to obtain the outlet nitrogen oxide concentration prediction result.

[0028] In this embodiment, the ammonia injection data is monitored in real time, and the outlet nitrogen oxide concentration is obtained through the outlet nitrogen oxide concentration prediction model based on the ammonia injection data to provide feedforward knowledge for the neural network ammonia injection controller.

[0029] Specifically, an anemometer is used in the flue to measure the total air volume passing through the flue in real time, a temperature sensor is installed at the flue entrance to measure the inlet flue gas temperature, a flow meter is installed in the ammonia injection system to measure the injection amount of ammonia, and a nitrogen oxide analyzer is installed at the flue entrance to measure the inlet nitrogen oxide concentration.

[0030] In this embodiment, the outputting of the ammonia injection control signal through the neural network adaptive control according to the outlet nitrogen oxide concentration is specifically implemented by the following steps: S401: constructing a neural network ammonia injection controller, wherein the neural network ammonia injection controller is a three-layer neural network structure; Specifically, the input layer of the neural network ammonia injection controller includes three nodes, which respectively receive the current outlet nitrogen oxide concentration error value, the outlet nitrogen oxide concentration error change, and the outlet nitrogen oxide concentration system setting value; the hidden layer of the neural network ammonia injection controller includes six nodes; and the output layer of the neural network ammonia injection controller includes one node for outputting an ammonia injection control signal.

[0031] S402: Preset learning rate and momentum factor; S403: Preset an expected reference model; S404: Calculating an ammonia injection error by comparing the outlet nitrogen oxide concentration value with the expected reference model output value; S405: The neural network ammonia injection controller receives the ammonia injection error, and the neural network adaptively adjusts the weighting coefficient through the back propagation method and the gradient descent method to minimize the ammonia injection error. It continues to iterate until the ammonia injection error is less than the preset ammonia injection error threshold or the number of iterations reaches the preset maximum number of iterations. The training process ends and the ammonia injection control signal is output to adjust the ammonia injection amount.

[0032] In this embodiment, the neural network ammonia injection controller uses the outlet nitrogen oxide concentration prediction model as a feedforward identifier. The neural network ammonia injection controller receives the outlet nitrogen oxide concentration prediction value output by the feedforward identifier, minimizes the error between the outlet nitrogen oxide concentration prediction value and the preset reference model output value by adaptively adjusting the weight, outputs an ammonia injection control signal, and controls the valve opening of the ammonia injection system according to the ammonia injection control signal, thereby timely adjusting the ammonia injection amount to reduce ammonia slip while meeting the outlet nitrogen oxide emission standard.

[0033] An intelligent ammonia injection control system includes an ammonia injection data acquisition module, an outlet nitrogen oxide concentration prediction model module, an outlet nitrogen oxide concentration real-time prediction module, and an adaptive ammonia injection control module; The ammonia injection data acquisition module is used to collect the total air volume, inlet flue gas temperature, ammonia injection amount, and inlet nitrogen oxide concentration of the denitrification system; The outlet nitrogen oxide concentration prediction model module is used to construct an outlet nitrogen oxide concentration prediction model according to the historical ammonia injection data through the Bayesian optimization CNN-LSSVM algorithm; The outlet nitrogen oxide concentration real-time prediction module is used to monitor the total air volume, inlet flue gas temperature, ammonia injection amount, inlet nitrogen oxide concentration of the denitrification system in real time and output the outlet nitrogen oxide concentration through the outlet nitrogen oxide concentration prediction model; The adaptive ammonia injection control module is used to output an ammonia injection control signal through a neural network adaptive control according to the outlet nitrogen oxide concentration, and control the valve opening of the ammonia injection system according to the ammonia injection control signal to achieve automatic ammonia injection amount control.

[0034] The computer storage medium of the embodiment of the present invention may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device.

[0035] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0036] The program code included on the computer readable medium can be transmitted with any appropriate medium, including but not limited to wireless, electric wire, optical cable, RF, etc., or any suitable combination of the above. The computer program code for performing the operation of the present invention can be written in one or more programming languages ​​or their combinations, and the programming language includes object-oriented programming languages-such as Java, Smalltalk, C++, and also includes conventional procedural programming languages-such as "C" language or similar programming languages. The program code can be executed completely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on the remote computer, or completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0037] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. An intelligent ammonia injection control method, characterized in that: include: Acquire historical ammonia injection data, wherein the historical ammonia injection data includes historical total air volume, historical inlet flue gas temperature, historical ammonia injection amount, historical inlet nitrogen oxide concentration, and historical outlet nitrogen oxide concentration; According to the historical ammonia injection data, a prediction model for outlet nitrogen oxide concentration is constructed by using a Bayesian optimization CNN-LSSVM algorithm; real-time monitoring of ammonia injection data, and obtaining an outlet nitrogen oxide concentration through the outlet nitrogen oxide concentration prediction model according to the ammonia injection data, wherein the ammonia injection data includes a total air volume, an inlet flue gas temperature, an ammonia injection amount, and an inlet nitrogen oxide concentration; An ammonia injection control signal is outputted through a neural network adaptive control according to the outlet nitrogen oxide concentration, and a valve opening of an ammonia injection system is controlled according to the ammonia injection control signal.

2. The intelligent ammonia injection control method according to claim 1, characterized in that: The constructing of the outlet nitrogen oxide concentration prediction model by using the Bayesian optimization CNN-LSSVM algorithm according to the historical ammonia injection data includes: Obtaining historical ammonia injection standard data through normalization processing according to the historical ammonia injection data, and constructing an ammonia injection data set according to the historical ammonia injection standard data; Extracting an ammonia injection feature vector set through CNN according to the ammonia injection data set; The outlet nitrogen oxide concentration prediction model is obtained through LSSVM training and testing according to the ammonia injection feature vector set.

3. The intelligent ammonia injection control method according to claim 2, characterized in that: The extracting of the ammonia injection feature vector set by CNN according to the ammonia injection data set comprises: For the ammonia injection data set, a convolution kernel is used through a convolution layer to generate an ammonia injection data feature map, some neurons are randomly discarded through a Dropout layer to prevent overfitting, and an ammonia injection feature vector is extracted through the activations function, and finally the ammonia injection feature vector set is obtained by combining them.

4. The intelligent ammonia injection control method according to claim 2, characterized in that: The outlet nitrogen oxide concentration prediction model obtained by LSSVM training and testing according to the ammonia injection feature vector set includes: Obtaining an ammonia injection feature vector training set and an ammonia injection feature vector test set by randomly dividing the ammonia injection feature vector set; According to the ammonia injection feature vector training set, an initial model for predicting outlet nitrogen oxide concentration is obtained by LSSVM training; According to the initial model for predicting the outlet nitrogen oxide concentration, a combination of LSSVM hyperparameters is selected by Bayesian optimization combined with five-fold cross validation; According to the LSSVM hyperparameter combination, the outlet nitrogen oxide concentration prediction model is obtained by retraining the LSSVM; The performance of the outlet nitrogen oxide concentration prediction model is evaluated by using an ammonia injection evaluation index according to the ammonia injection feature vector test set.

5. The intelligent ammonia injection control method according to claim 4, characterized in that: The ammonia injection evaluation index is the ammonia injection mean absolute error and the ammonia injection determination coefficient, and the mathematical expression is: ， ， Among them, E is the average absolute error of ammonia injection, R 2 is the ammonia injection determination coefficient, n is the number of samples in the ammonia injection feature vector test set, y i is the actual value of the nitrogen oxide concentration at the outlet of the i-th ammonia injection test sample, is the predicted value of the nitrogen oxide concentration at the outlet of the i-th ammonia injection test sample, and y is the average nitrogen oxide concentration at the outlet of the ammonia injection test sample.

6. The intelligent ammonia injection control method according to claim 1, characterized in that: The outputting of the ammonia injection control signal through the neural network adaptive control according to the outlet nitrogen oxide concentration comprises: Constructing a neural network ammonia injection controller, wherein the neural network ammonia injection controller is a three-layer neural network structure; Preset learning rate and momentum factor; Preset expected reference model; By comparing the outlet nitrogen oxide concentration value with the expected reference model output value, an ammonia injection error is calculated; The neural network ammonia injection controller receives the ammonia injection error. The neural network adaptively adjusts the weighting coefficient through the back propagation method and the gradient descent method to minimize the ammonia injection error. It iterates continuously until the ammonia injection error is less than the preset ammonia injection error threshold or the number of iterations reaches the preset maximum number of iterations. The training process ends and the ammonia injection control signal is output.

7. The intelligent ammonia injection control method according to claim 6, characterized in that: The input layer of the neural network ammonia injection controller includes three nodes, which respectively receive the current outlet nitrogen oxide concentration error value, the outlet nitrogen oxide concentration error change, and the outlet nitrogen oxide concentration system setting value. The hidden layer of the neural network ammonia injection controller includes six nodes. The output layer of the neural network ammonia injection controller includes one node for outputting an ammonia injection control signal.

8. An intelligent ammonia injection control system using the intelligent ammonia injection control method according to claim 1, comprising an ammonia injection data acquisition module, an outlet nitrogen oxide concentration prediction model module, an outlet nitrogen oxide concentration real-time prediction module, and an adaptive ammonia injection control module, characterized in that: The ammonia injection data acquisition module is used to collect the total air volume, inlet flue gas temperature, ammonia injection amount, and inlet nitrogen oxide concentration of the denitrification system; The outlet nitrogen oxide concentration prediction model module is used to construct an outlet nitrogen oxide concentration prediction model according to the historical ammonia injection data through the Bayesian optimization CNN-LSSVM algorithm; The outlet nitrogen oxide concentration real-time prediction module is used to monitor the total air volume, inlet flue gas temperature, ammonia injection amount, inlet nitrogen oxide concentration of the denitrification system in real time and output the outlet nitrogen oxide concentration through the outlet nitrogen oxide concentration prediction model; The adaptive ammonia injection control module is used to output an ammonia injection control signal through a neural network adaptive control according to the outlet nitrogen oxide concentration, and control the valve opening of the ammonia injection system according to the ammonia injection control signal to achieve automatic ammonia injection amount control.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the intelligent ammonia injection control method as described in any one of claims 1-7 is implemented.

10. A storage medium containing computer executable instructions, characterized in that: The computer executable instructions, when executed by a computer processor, are used to execute the intelligent ammonia injection control method as described in any one of claims 1-7.

Images