A method and device for predicting nitrogen oxide emissions based on a genetic algorithm and a convolution network

Abstract

The application provides a nitrogen oxide emission prediction method and device based on a genetic algorithm and a convolution network. The method is based on historical operation data, converts variable selection, time lag characteristics of variable influence, and a prediction model optimization process into an integrated function optimization problem, optimizes parameters of each link by using a genetic algorithm, determines a model structure according to a function solution space corresponding to an optimal solution, solves a local optimal solution problem caused by variable selection and model optimization due to measurement time lag, and enables the prediction model to more fully mine NOx generation rules, thereby providing more accurate support for NOx removal control.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of coal-fired thermal power generation, and particularly relates to a nitrogen oxide emission prediction method, device, equipment and storage medium based on a genetic algorithm and a convolutional network. BACKGROUND

[0002] In the power production in China, the energy structure dominated by coal determines that coal-fired thermal power generation occupies a dominant position. With the development of society, environmental protection is paid more and more attention, and the NOx emission standard of thermal power plants has also risen to a new height. This promotes the thermal power plant to have higher requirements for the quality of denitration control. At present, the SCR (selective catalytic reduction) method is generally used in large thermal power generating units. In order to reduce the emission of NOx, avoid the blockage of the air preheater caused by excessive ammonia injection, and avoid the emission of NOx exceeding the standard caused by too little reducing agent, it is necessary to measure and monitor the emission concentration of nitrogen oxides in real time, and to optimize the control of the denitration system.

[0003] The flue gas automatic monitoring system (CEMS) currently in wide use has many shortcomings. CEMS needs regular offline maintenance, which is labor-intensive, and its measurement value is always accurate and effective in actual operation. The CEMS standard uses the extraction method for measurement, which requires a long measurement lag time, resulting in a large lag characteristic of the feedback of the control system and affecting the control accuracy.

[0004] In order to overcome the above-mentioned shortcomings, data-driven prediction models are widely used in the soft measurement and control optimization of flue gas NOx, such as linear regression, support vector machine, naive Bayes, and recurrent neural network. Most of the existing machine learning methods cannot learn the continuous influence in the time dimension. Although the recurrent neural network has strong non-linear and time dimension learning ability, it does not have a good solution to the delay of model data. SUMMARY

[0005] The present application provides a nitrogen oxide emission prediction method, device, equipment and storage medium based on a genetic algorithm and a convolutional network, which aims to solve the variable selection and model optimization problems caused by measurement data lag in the prior art.

[0006] To this end, the first object of the present application is to provide a nitrogen oxide emission prediction method based on a genetic algorithm and a convolutional network, comprising:

[0007] constructing a nitrogen oxide emission prediction model;

[0008] The nitrogen oxide emission prediction model comprises a variable selection hyperparameter setting module, a variable lag hyperparameter setting module, a convolution network regression model hyperparameter setting module and a parameter optimization module; the parameter optimization module performs parameter optimization on the variable selection hyperparameters set by the variable selection hyperparameter setting module, the variable lag hyperparameters set by the variable lag hyperparameter setting module and the convolution network regression model hyperparameters set by the convolution network regression model hyperparameter setting module based on a genetic algorithm, and determines the nitrogen oxide emission prediction model based on the optimization result;

[0009] The historical boiler operation data is obtained, the boiler abnormal operation data is screened, and the nitrogen oxide emission prediction model is trained by using the screened boiler abnormal operation data as training data;

[0010] The boiler operation data is obtained in real time, the trained nitrogen oxide emission prediction model is input, and the output result is used as the nitrogen oxide emission prediction result.

[0011] The historical boiler operation data is obtained, the boiler abnormal operation data is screened, and the nitrogen oxide emission prediction model is trained by using the screened boiler abnormal operation data as training data;

[0012] In the step of screening the boiler abnormal operation data, the following steps are included:

[0013] The historical boiler operation data is divided into a plurality of equal intervals along a time axis, and the fluctuation amplitude of each type of data in the historical boiler operation data in any equal interval is calculated;

[0014] If the fluctuation amplitude of the corresponding data meets the preset condition, the corresponding type of data in the corresponding equal interval is taken as the boiler abnormal operation data;

[0015] The screening is repeated for a plurality of times to obtain first sample data.

[0016] The variable selection hyperparameter setting module is used to construct a variable combination mapping dictionary; specifically including:

[0017] The model-related variables of the nitrogen oxide emission prediction model are defined as basic variables and optimization variables; wherein the basic variables include boiler load Steam, coal supply amount Coal, total air volume Wind and flue gas oxygen content O2, and the optimization variables include primary air volume Wind1, secondary air volume Wind2, furnace temperature BoilerTem, flue gas volume Gas and flue gas temperature GasTem;

[0018] The model-related variable selection strategy is basic variables + optimization variables; according to combination recursion, the variable combination type of the model is determined, and the variable combination mapping dictionary is established;

[0019] According to the variable combination mapping dictionary, determine the variable selection hyperparameter, and perform feature filtering on the first sample data to obtain second sample data.

[0020] The variable lag hyperparameter setting module is used to eliminate the time delay characteristics generated in the boiler operation production process.

[0021] If the current time is t, and the variable lag parameter in the second sample data is t1, it indicates that the variables at times [t-t1, t] will affect the generation of NOx at the current time, and the variables at times [t-t1, t] are used as input data.

[0022] If the current time is t, and the predicted variable NOx has a measurement delay t2, it indicates that the current NOx measurement value is actually the production data at time t-t2, and the second sample data needs to be shifted, i.e., the NOx data at time t corresponds to the data produced at time t-t2.

[0023] The variable lag hyperparameter is established to represent the lag parameter set of each variable type in the second sample data. The second sample data is preprocessed according to the independent variable and the delay parameter of the predicted variable to obtain sample data for training the model.

[0024] The convolutional network regression model includes:

[0025] The first convolutional layer, the second convolutional layer, the pooling layer, the Flatten layer, the Dropout layer, the fully connected layer, and the output layer.

[0026] The first convolutional layer is a linear activation function, and the second convolutional layer is a nonlinear activation function.

[0027] The set of convolutional network regression model hyperparameters includes: the number of first convolutional kernels, the size of the first convolutional layer, the number of second convolutional kernels, the size of the second convolutional layer, the parameters of the nonlinear activation function, and the pooling parameters.

[0028] In the parameter optimization module based on the genetic algorithm for parameter optimization, the steps include:

[0029] The optimization function of the convolutional network regression model is used as the optimization target function of the genetic algorithm.

[0030] Determine the optimization space of the variable selection hyperparameter, the variable lag hyperparameter, and the convolutional network regression model hyperparameter.

[0031] For each parameter, randomly select an initial population in the optimization space, calculate the fitness value according to the optimization target function, and if the termination condition is met, output the optimal solution to determine the final prediction model, otherwise update the parameter population and recalculate the fitness value. Iterative operation is performed until the termination condition is met.

[0032] The second object of the present application is to provide a nitrogen oxide emission prediction device based on a genetic algorithm and a convolutional network, comprising:

[0033] A model construction module is configured to construct a nitrogen oxide emission prediction model.

[0034] The nitrogen oxide emission prediction model comprises a variable selection hyperparameter setting module, a variable lag hyperparameter setting module, a convolutional network regression model hyperparameter setting module, and a parameter optimization module.

[0035] A model training module is configured to obtain historical boiler operation data, filter abnormal boiler operation data as training data, and train the nitrogen oxide emission prediction model.

[0036] A prediction module is configured to obtain real-time boiler operation data, input the trained nitrogen oxide emission prediction model, and output the result as a nitrogen oxide emission prediction result.

[0037] The third object of the present application is to provide an electronic device comprising at least one processor and a memory communicatively connected to the at least one processor.

[0038] The fourth object of the present application is to provide a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the steps in the method according to the foregoing technical solutions.

[0039] Compared with the prior art, the nitrogen oxide emission prediction method based on a genetic algorithm and a convolutional network provided by the present application converts the variable selection, the time lag characteristic of variable influence, and the prediction model optimization process into a comprehensive function optimization problem based on historical operation data. BRIEF DESCRIPTION OF DRAWINGS

[0040] The present application and / or additional aspects and advantages will become apparent and appreciated from a consideration of the following detailed description, taken in conjunction with the accompanying drawings, in which:

[0041] Figure 1 is a flowchart of a nitrogen oxide emission prediction method based on a genetic algorithm and a convolutional network provided by the present application.

[0042] Figure 2 is a logic diagram of a nitrogen oxide emission prediction method based on a genetic algorithm and a convolutional network provided by the present application.

[0043] Figure 3 is a logic diagram of a genetic algorithm in a nitrogen oxide emission prediction method based on a genetic algorithm and a convolutional network provided by the present application.

[0044] Figure 4 is a model prediction effect diagram in a nitrogen oxide emission prediction method based on a genetic algorithm and a convolutional network provided by the present application.

[0045] Figure 5 is a structural diagram of a nitrogen oxide emission prediction device based on a genetic algorithm and a convolutional network provided by the present application.

[0046] Figure 6 is a structural diagram of a non-transitory computer readable storage medium provided by the present application. DETAILED DESCRIPTION

[0047] Embodiments of the present application are described in detail below with reference to the attached drawings, examples of which are shown in the drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present application, and cannot be understood as limiting the present application.

[0048] As shown in Figure 1 is a nitrogen oxide emission prediction method based on a genetic algorithm and a convolutional network provided by the present application, comprising:

[0049] S110: constructing a nitrogen oxide emission prediction model;

[0050] The nitrogen oxide emission prediction model comprises a variable selection hyperparameter setting module, a variable lag hyperparameter setting module, a convolutional network regression model hyperparameter setting module, and a parameter optimization module. The parameter optimization module performs parameter optimization on the variable selection hyperparameters set by the variable selection hyperparameter setting module, the variable lag hyperparameters set by the variable lag hyperparameter setting module, and the convolutional network regression model hyperparameters set by the convolutional network regression model hyperparameter setting module based on the genetic algorithm, and determines the nitrogen oxide emission prediction model based on the optimization result.

[0051] S120: Obtain historical boiler operation data, screen boiler abnormal operation data as training data to train the nitrogen oxide emission prediction model.

[0052] S130: Real-time obtain boiler operation data, input the trained nitrogen oxide emission prediction model, and output the result as the nitrogen oxide emission prediction result.

[0053] In the present application, the historical boiler operation data is the data generated by the boiler operation in a specified time interval, at least including coal supply, primary air volume, secondary air volume, flue gas oxygen content, boiler load, furnace temperature, flue gas volume, flue gas temperature, and desulfurization reactor inlet NOx concentration.

[0054] The logic of the method of the present application is shown in Figure 2 .

[0055] Specifically, in the step of screening the boiler abnormal operation data, the step includes:

[0056] Divide the historical boiler operation data into several equal intervals along the time axis, and calculate the fluctuation amplitude of each type of data in the historical boiler operation data in any equal interval;

[0057] The fluctuation amplitude v is

[0058]

[0059] If the fluctuation amplitude of the corresponding data meets the preset condition, the corresponding type of data in the corresponding equal interval is taken as the boiler abnormal operation data;

[0060] For example, the boiler load fluctuation amplitude v_steam is greater than or equal to 0.03; the coal supply fluctuation amplitude v_coals is greater than or equal to 0.03; the flue gas oxygen content fluctuation amplitude v_o2 is greater than or equal to 0.05; and the other types of boiler abnormal operation data are set according to the actual situation.

[0061] Repeat the screening multiple times to obtain the first sample data D1(L1, L2, L3……Ln).

[0062] The variable selection hyperparameter setting module is used to construct a variable combination mapping dictionary; specifically including:

[0063] Define the model-related variables of the nitrogen oxide emission prediction model as basic variables and optimization variables; wherein the basic variables include boiler load Steam, coal supply Coal, total air volume Wind, and flue gas oxygen content O2, and the optimization variables include primary air volume Wind1, secondary air volume Wind2, furnace temperature BoilerTem, flue gas volume Gas, and flue gas temperature GasTem.

[0064] The model-related variable selection strategy is a basic variable + an optimization variable; according to combination recursion, the variable combination type of the model is determined, and a variable combination mapping dictionary is established.

[0065] The combination recursion formula is expressed as The model has 2 n variable combinations.

[0066] The variable combination mapping dictionary is expressed as:

[0067] Dict = {1: basic variable

[0068] 2: basic variable + optimization variable 1

[0069] 3: basic variable + optimization variable 2

[0070] …

[0071] 2 n : basic variable + optimization variable 1 + optimization variable 1 … optimization variable n}.

[0072] According to the variable combination mapping dictionary, the variable selection hyperparameters are determined, expressed as: params1 = [1, 2, 3 … 2 n ];

[0073] When the variable parameter is i, the mapping dictionary Dict corresponding to the variable is used to perform feature filtering on the sample data D1 to obtain the second sample data D2 (D21, D22, …, D2n).

[0074] Due to the time delay characteristics in the boiler operation production process, the coal, wind and other media are sent into the furnace for combustion, and there is inertia and pure lag characteristics between the generation of flue gas. The variable lag hyperparameter setting module is used to eliminate the time delay characteristics generated in the boiler operation production process.

[0075] Wherein, if the current is t time, the variable lag parameter in the second sample data is t1, which indicates that the variables at [t-t1, t] time will affect the generation of NOx at the current time, and the variables at [t-t1, t] time are all input data.

[0076] If the current is t time, the predicted variable NOx has a measurement delay t2, which indicates that the current NOx measurement value is actually the production data at t-t2 time, and the second sample data needs to be shifted, that is, the NOx data at t time corresponds to the data produced at t-t2 time.

[0077] The variable lag hyperparameter is established, which represents the lag parameter set of each variable type in the second sample data. According to the delay parameters of independent variables and predicted variables, the second sample data is preprocessed to obtain sample data for training the model.

[0078] Establish variable lag hyperparameters params2 = [T1, T2, T3 …… Tn, Tnox].

[0079] The convolutional network is a kind of feedforward neural network containing convolution calculation and having a deep structure, and is one of the representative algorithms of deep learning, and has a representation learning ability to perform translation invariant classification on input information according to a hierarchical mechanism, so that it has better learning ability for sequence data.

[0080] The convolutional network regression model comprises:

[0081] The first convolutional layer: linear activation function (number of convolution kernels k11, size of convolution kernel k12) ;

[0082] The second convolutional layer: nonlinear activation function (number of convolution kernels k21, size of convolution kernel k22, nonlinear activation function parameter fun) ;

[0083] The pooling layer: (pooling parameter pool, including maximum pooling and average pooling) ;

[0084] The Flatten layer: the multi-dimensional data is processed into one-dimensional data;

[0085] The Dropout layer: the neurons are randomly inactivated in proportion;

[0086] The fully connected layer: (linear activation function) ;

[0087] The output layer.

[0088] The loss function of the convolutional network regression model takes the mean absolute error, and the formula is:

[0089]

[0090] Where y' is the predicted value, y is the actual value, and n is the data amount of the test set.

[0091] The activation function comprises a linear activation function linear, a nonlinear activation function tanh, relu and sigmoid; wherein,

[0092] linear (x) = x

[0093]

[0094]

[0095]

[0096] The set parameters in the network structure form a model hyperparameter params3=[k11, k12, k21, k22, fun, pool], wherein k11 and k21 are in the optimization space [1:32], k12 and k22 are in the optimization space [1:10], the activation function is in the optimization space [linear, tanh, relu, sigmoid], and the pooling mode is in the optimization space [maximum pooling, average pooling].

[0097] The sample data as the model training set is divided into K parts, (K-1) parts of which are used to train the model, and the remaining 1 part of data is used to evaluate the quality of the model; the present application uses k-fold cross-validation to train the model through the neural network mechanism, and sets the optimization function as the average value of the average absolute error on different test sets:

[0098]

[0099] Wherein, mae k is the average absolute error of the model on the kth test set.

[0100] Genetic algorithm (GA) is a self-adaptive random search heuristic algorithm designed according to the evolution law of organisms in nature, and its basic framework is to simulate the evolution and heredity of organisms in nature to solve problems based on natural selection law and genetic theory, which is a kind of self-organizing and self-adaptive artificial intelligence technology, and has been widely used in complex function system optimization, machine learning, system identification, fault diagnosis, classification system, controller design, neural network design, adaptive filter design, etc.

[0101] In the parameter optimization module based on genetic algorithm, the steps include:

[0102] The optimization function of the convolutional network regression model is used as the optimization target function of the genetic algorithm; the aforementioned optimization function fitness is the optimization target function in the genetic algorithm;

[0103] The optimization space of the variable selection hyperparameter, the variable lag hyperparameter and the convolutional network regression model hyperparameter is determined;

[0104] That is,

[0105] The variable selection hyperparameter params1=[1, 2, 3 … 2 n ]、

[0106] The variable lag hyperparameter params2=[T1, T2, T3 … Tn, Tnox]、

[0107] The model hyperparameter params3=[k11, k12, k21, k22, fun, pool] is optimized.

[0108] For each parameter, an initial population is randomly selected in the optimization space, the fitness value is calculated according to the optimization objective function, if the termination condition is met, the optimal solution is output to determine the final prediction model, otherwise the parameter population is updated, the fitness value is recalculated, and the iteration operation is performed until the termination condition is met. The genetic algorithm process is shown in the accompanying Figure 3 .

[0109] In summary, the NOx prediction method combining genetic algorithm and convolutional neural network provided in the embodiment converts the variable selection, time delay characteristics of variable influence and prediction model optimization process into a comprehensive function optimization problem, optimizes the parameters of each link using genetic algorithm, determines the model structure according to the function solution space corresponding to the optimal solution, solves the local optimal solution problem caused by variable selection and model optimization due to measurement time lag, enables the prediction model to more fully tap the NOx generation law, and further provides more accurate support for denitration control. The prediction effect is shown in the accompanying Figure 4 .

[0110] As shown in the accompanying Figure 5 , the application also provides a nitrogen oxide emission prediction device based on genetic algorithm and convolutional network, comprising:

[0111] The model construction module 310 is used to construct a nitrogen oxide emission prediction model.

[0112] The nitrogen oxide emission prediction model comprises a variable selection hyperparameter setting module, a variable lag hyperparameter setting module, a convolutional network regression model hyperparameter setting module and a parameter optimization module. The parameter optimization module optimizes the variable selection hyperparameters set by the variable selection hyperparameter setting module, the variable lag hyperparameters set by the variable lag hyperparameter setting module, and the convolutional network regression model hyperparameters set by the convolutional network regression model hyperparameter setting module based on genetic algorithm, and determines the nitrogen oxide emission prediction model based on the optimization results.

[0113] The model training module 320 is used to obtain historical boiler operation data, filter boiler abnormal operation data as training data, and train the nitrogen oxide emission prediction model.

[0114] The prediction module 330 is used to obtain real-time boiler operation data, input the trained nitrogen oxide emission prediction model, and output the result as the nitrogen oxide emission prediction result.

[0115] To achieve the embodiments, the application further provides an electronic device, comprising: at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform each step in the nitrogen oxide emission prediction method of the foregoing technical solutions.

[0116] As shown in Figure 6 the non-transitory computer readable storage medium includes a memory 810 of instructions executable by a processor 820 for performing the method. Optionally, the storage medium can be a non-transitory computer readable storage medium, for example, the non-transitory computer readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0117] To achieve the embodiments, the application further provides a non-transitory computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the nitrogen oxide emission prediction according to the embodiments of the application.

[0118] In the description of the present specification, the description referring to the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples without contradiction.

[0119] In addition, the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, for example, two, three, etc., unless otherwise explicitly specified.

[0120] Any processes or methods described in the flowcharts or otherwise described herein can be understood as representing modules, segments, or portions of code that include one or more executable instructions for implementing specific logic functions (or steps) and / or can be implemented entirely in hardware. The steps or functions of the preferred embodiments of the application can be embodied in software, firmware, or hardware, and can be embodied within one or more processors or other structures that execute instructions to perform the steps or functions of the preferred embodiments of the application. The preferred embodiments of the application are not limited in this regard.

[0121] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be embodied in computer-executable instructions or code that can be implemented by or in connection with an instruction execution system, apparatus, or device such as a computer-based system, processor- containing system, or other system that can fetch the instructions from a non-transitory computer-readable medium and execute the instructions. For purposes of this specification, a "computer-readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer-readable medium can be a non-transitory computer-readable medium. The non-transitory computer-readable medium can be a computer-readable medium that is not a transitory propagating signal. The non-transitory computer-readable medium can be a computer-readable medium that is tangible rather than a propagating signal per se. More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium can even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example, via optical scanning of the paper or other medium, then compiled, interpreted, or otherwise processed in a suitable manner, if necessary, and stored in a computer memory.

[0122] It should be understood that aspects of the application can be implemented in hardware, software, firmware, or combinations thereof. In the described embodiments, various steps or methods can be implemented in software or firmware that is stored in memory and executed by suitable instruction execution systems. As such, in hardware implementations, any of the following technologies, or combinations thereof, can be used: discrete logic circuitry having logic gates for implementing logic functions upon data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and so forth.

[0123] Those skilled in the art of the present technology can understand that all or part of the steps carried out by the method of the embodiment can be instructed by a program to complete the relevant hardware, and the program can be stored in a computer readable storage medium. When the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0124] In addition, each functional unit in each embodiment of the present application can be integrated into one processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The integrated module can be realized in the form of hardware or in the form of a software functional module. The integrated module, if realized in the form of a software functional module and sold or used as an independent product, can also be stored in a computer readable storage medium.

[0125] The storage medium mentioned can be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it should be understood that the embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the embodiments within the scope of the present application.

Claims

1. A method for predicting nitrogen oxide emissions based on a genetic algorithm and a convolutional network, characterized by, The method comprises the following steps: constructing a nitrogen oxide emission prediction model; The nitrogen oxide emission prediction model comprises a variable selection hyperparameter setting module, a variable lag hyperparameter setting module, a convolutional network regression model hyperparameter setting module, and a parameter optimization module. The parameter optimization module optimizes the variable selection hyperparameters set by the variable selection hyperparameter setting module, the variable lag hyperparameters set by the variable lag hyperparameter setting module, and the convolutional network regression model hyperparameters set by the convolutional network regression model hyperparameter setting module based on a genetic algorithm. The nitrogen oxide emission prediction model is determined based on the optimization result. The variable selection hyperparameter setting module is used to construct a variable combination mapping dictionary. The variable lag hyperparameter setting module is used to eliminate the time delay characteristics generated in the boiler operation process. Historical boiler operation data is obtained, and abnormal boiler operation data is screened as training data to train the nitrogen oxide emission prediction model. Real-time boiler operation data is obtained, and the trained nitrogen oxide emission prediction model is inputted to output the result as a nitrogen oxide emission prediction result. In the parameter optimization step based on the genetic algorithm, the following steps are included: The optimization function of the convolutional network regression model is used as the optimization target function of the genetic algorithm. The optimization space of the variable selection hyperparameters, the variable lag hyperparameters, and the convolutional network regression model hyperparameters is determined. For each parameter, an initial population is randomly selected in the optimization space. The fitness value is calculated according to the optimization target function. If the termination condition is met, the optimal solution is outputted to determine the final prediction model. Otherwise, the parameter population is updated, the fitness value is recalculated, and the iteration operation is performed until the termination condition is met. 2.The method of claim 1, wherein, The historical boiler operation data is the data generated by the boiler operation within a specified time interval, including at least coal supply, primary air volume, secondary air volume, flue gas oxygen content, boiler load, furnace temperature, flue gas volume, flue gas temperature, and desulfurization reactor inlet NOx concentration.

3. The method of claim 2, wherein the genetic algorithm and convolutional network-based prediction of nitrogen oxide emissions is based on a genetic algorithm and convolutional network-based prediction of nitrogen oxide emissions from a combustion engine. In the step of screening abnormal boiler operation data, the following steps are included: The historical boiler operation data is divided into several equal intervals along the time axis, and the fluctuation amplitude of each type of data in the historical boiler operation data in any equal interval is calculated. If the fluctuation amplitude of the corresponding data meets the preset condition, the corresponding type of data in the corresponding equal interval is taken as the abnormal boiler operation data. Repeat the screening multiple times to obtain the first sample data.

4. The method of claim 3, wherein the genetic algorithm and convolutional network-based prediction of nitrogen oxide emissions is based on a genetic algorithm and convolutional network-based prediction of nitrogen oxide emissions from a combustion engine. The variable selection hyperparameter setting module is used to construct a variable combination mapping dictionary. Specifically, the following steps are included: Define the model-related variables of the nitrogen oxide emission prediction model as basic variables and optimization variables. The basic variables include boiler load Steam, coal supply Coal, total air volume Wind, and flue gas oxygen content O2. The optimization variables include primary air volume Wind1, secondary air volume Wind2, furnace temperature BoilerTem, flue gas volume Gas, and flue gas temperature GasTem. The model-related variable selection strategy is basic variables + optimization variables. According to combination recursion, the variable combination type of the model is determined, and the variable combination mapping dictionary is established. According to the variable combination mapping dictionary, determine the variable selection hyperparameter, and perform feature filtering on the first sample data to obtain second sample data.

5. The genetic algorithm and convolutional network-based method for predicting nitrogen oxide emissions according to claim 4, characterized in that ； Wherein, if the current is t time, the variable lag parameter in the second sample data is t1, then it indicates that the variables at [t-t1, t] time will affect the generation of NOx at the current time, and the variables at [t-t1, t] time are all used as input data; If the current is t time, the predicted variable NOx has a measurement delay t2, then it indicates that the current NOx measurement value is actually the production data at t-t2 time, and the second sample data needs to be shifted, that is, the NOx data at t time corresponds to the data produced at t-t2 time; The variable lag hyperparameter is established to represent the lag parameter set of each variable type in the second sample data, and the second sample data is preprocessed according to the independent variable and the delay parameter of the predicted variable to obtain sample data for training the model. 6.The method of claim 5, wherein, The convolutional network regression model comprises: The first convolutional layer, the second convolutional layer, the pooling layer, the Flatten layer, the Dropout layer, the fully connected layer and the output layer; Wherein, the first convolutional layer is a linear activation function, and the second convolutional layer is a nonlinear activation function; Then set the convolutional network regression model hyperparameters, including: the first convolutional kernel number, the first convolutional layer size, the second convolutional kernel number, the second convolutional layer size, the nonlinear activation function parameter and the pooling parameter.

7. A nitrogen oxide emission prediction device based on a genetic algorithm and a convolution network, characterized by, Comprise: The model construction module is used to construct a nitrogen oxide emission prediction model. The nitrogen oxide emission prediction model comprises a variable selection hyperparameter setting module, a variable lag hyperparameter setting module, a convolutional network regression model hyperparameter setting module and a parameter optimization module; wherein, the parameter optimization module is based on a genetic algorithm to optimize the variable selection hyperparameters set by the variable selection hyperparameter setting module, the variable lag hyperparameters set by the variable lag hyperparameter setting module, and the convolutional network regression model hyperparameters set by the convolutional network regression model hyperparameter setting module, and determines the nitrogen oxide emission prediction model based on the optimization result, the variable selection hyperparameter setting module is used to construct a variable combination mapping dictionary, and the variable lag hyperparameter setting module is used to eliminate the delay characteristics generated in the boiler operation production process; in the parameter optimization step based on the genetic algorithm, comprising: The optimization function of the convolutional network regression model is used as the optimization target function of the genetic algorithm; Determine the optimization space of the variable selection hyperparameter, the variable lag hyperparameter and the convolutional network regression model hyperparameter; Randomly select an initial population for each parameter in the optimization space, calculate the fitness value according to the optimization target function, if the termination condition is met, output the optimal solution to determine the final prediction model, otherwise update the parameter population and recalculate the fitness value, and iterate until the termination condition is met; The model training module is used to obtain historical boiler operation data, filter the abnormal boiler operation data as training data, and train the nitrogen oxide emission prediction model. The prediction module is configured to acquire the boiler operation data in real time, input the trained nitrogen oxide emission prediction model, and output a result as a nitrogen oxide emission prediction result.

8. An electronic device, comprising: at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps in the method of any one of claims 1-6.

9. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are configured to enable the computer to perform the steps in the method of any one of claims 1-6.

Images