SCR (Selective Catalytic Reduction) denitration ammonia spraying control method, system, medium and equipment

Abstract

The invention discloses an SCR (Selective Catalytic Reduction) denitration ammonia spraying control method, system, medium and equipment, and relates to the field of heat energy and power engineering. The method comprises the steps that historical operation parameters of the denitration system and the nitrogen oxide concentration of an outlet of the denitration system are obtained; determining the delay time of each type of data in the historical operation parameters, and reconstructing the historical operation parameters based on the delay time; determining a characteristic value corresponding to each type of data in the reconstructed historical operation parameters and a correlation degree between any two characteristic values; reconstructing the reconstructed historical operation parameters again based on the correlation degree to obtain a training set; taking the training set as input, taking the denitration system outlet nitrogen oxide concentration as output, and training a long-short-term memory network to obtain a prediction model; and inputting the collected operation parameters into the prediction model to obtain a nitrogen oxide prediction concentration, and calculating the ammonia injection amount according to the nitrogen oxide prediction concentration. According to the scheme, the prediction precision of the emission concentration of the nitrogen oxide of the coal-fired unit can be improved, and the ammonia spraying amount can be accurately controlled.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal energy and power engineering, and in particular to a SCR denitration and ammonia injection control method, system, medium and equipment. Background Art

[0002] With the advocacy of energy conservation and emission reduction, new energy power generation technology has been gradually applied in more and more fields, and the installed capacity of new energy power generation equipment represented by wind power and photovoltaic power generation has increased rapidly. However, due to the randomness and seasonality of new energy power generation, and the immaturity of existing energy storage technology, the power generation through new energy power generation technology is unstable. With the increasing penetration rate of new energy power generation in the power grid, coal-fired units have to undertake more and more frequent peak-shaving tasks and load cycling processes to ensure the stable operation of the power grid.

[0003] Coal-fired units will emit a large amount of pollutants during the power generation process, such as nitrogen oxides (NOx), fine particulate matter (PMs), etc. Among them, NOx is one of the main causes of acid rain and photochemical smog, causing damage to the ozone layer and causing harm to human respiratory system and cardiovascular system. In order to reduce the threat of NOx to the ecological environment and human health, the daily average emission of NOx is required to be controlled at 100mg / Nm 3 Some fields even require daily NOx emissions to be less than 50mg / Nm 3 .

[0004] At present, the Selective Catalytic Reduction (SCR) technology is widely used in the denitration system of coal-fired power plants due to its mature process and reliable technology. The denitration efficiency of the SCR denitration system is affected by many factors, such as: NOx concentration at the system inlet, flue gas temperature, flue gas flow rate, ammonia / NOx molar ratio, etc. Among them, the NOx concentration at the system inlet and the ammonia / NOx molar ratio are the key factors affecting the denitration efficiency of the SCR system.

[0005] However, the NOx concentration in the flue gas of coal-fired boilers changes constantly during transient processes such as peak load regulation and variable load. The existing SCR system is affected by the combustion process of the reactor and boiler, and the system inertia and delay are large. This leads to a mismatch between the amount of ammonia sprayed in the SCR system and the NOx concentration, as well as large amounts of NOx emissions or NH3 escape, which makes the prediction accuracy of the NOx emission concentration of coal-fired units during transient processes low, and further leads to the inability to accurately control the amount of ammonia sprayed. Summary of the invention

[0006] Based on this, it is necessary to provide an SCR denitrification ammonia injection control method, system, medium and equipment to address the above technical problems.

[0007] The present invention adopts the following technical solutions:

[0008] In a first aspect, the present invention provides a method for controlling ammonia injection in SCR denitration, characterized by comprising:

[0009] Obtain the historical operating parameters of the denitrification system and the nitrogen oxide concentration at the outlet of the denitrification system;

[0010] Determine the delay time of each type of data in the historical operation parameters by a dynamic time warping algorithm, and reconstruct the historical operation parameters based on the delay time; determine the eigenvalues ​​corresponding to each type of data in the reconstructed historical operation parameters by a mutual information algorithm, and determine the degree of correlation between any two eigenvalues ​​in the eigenvalues ​​corresponding to each type of data by using the Spearman rank correlation coefficient; reconstruct the reconstructed historical operation parameters again based on the degree of correlation to obtain a training data set;

[0011] Taking the training data set as input and the nitrogen oxide concentration at the outlet of the denitrification system as output, training the long short-term memory network to obtain a nitrogen oxide concentration prediction model;

[0012] The operating parameters of the denitrification system collected in real time are input into the nitrogen oxide concentration prediction model to obtain the predicted nitrogen oxide concentration, and the ammonia injection amount is calculated according to the predicted nitrogen oxide concentration.

[0013] Furthermore, the historical operating parameters include: coal feed rate, air feed rate, furnace temperature, boiler load, nitrogen oxide concentration, oxygen concentration, ammonia concentration and flue gas temperature at the inlet of the denitrification system, oxygen concentration, ammonia concentration and flue gas temperature at the outlet of the denitrification system, and the air pressure difference between the inlet and outlet of the denitrification system.

[0014] Furthermore, before determining the delay time of each type of data in the historical operating parameters by the dynamic time warping algorithm, it also includes screening the historical operating parameters for outliers based on the 3σ criterion and removing the screened outliers, specifically including:

[0015] Calculate the mean and standard deviation of various types of data in the historical operating parameters;

[0016] According to the mean and the standard deviation, data in the historical operating parameters whose difference with the mean or the standard deviation is greater than a preset threshold are eliminated.

[0017] Furthermore, the determining of the delay time of each type of data in the historical operating parameters by a dynamic time warping algorithm specifically includes:

[0018] Calculate the delay time of each type of data according to the Euclidean distance;

[0019] The Euclidean distance calculation formula is:

[0020] d(x i ,y j )=(x i -y j ) 2 ;

[0021] Among them, x i and j are the i-th and j-th elements in time series X and Y respectively.

[0022] Furthermore, the eigenvalues ​​corresponding to each type of data in the reconstructed historical operating parameters are determined by the mutual information algorithm, and the expression is:

[0023]

[0024] Among them, p(x,y) is the joint probability distribution of time series X and time series Y, and p(x) and p(y) represent the independent probability distributions of time series X and time series Y respectively.

[0025] Furthermore, the expression for determining the correlation between any two eigenvalues ​​among the eigenvalues ​​corresponding to each type of data using the Spearman rank correlation coefficient is:

[0026]

[0027] d i =R(x i )-R(y i );

[0028] Among them, ρ is the correlation between any two eigenvalues, d i is the level difference of the eigenvalues ​​corresponding to each type of data, n is the number of eigenvalues, R(x i ) and R(y i ) are x i and i level.

[0029] In a second aspect, the present invention provides an SCR denitration ammonia injection control system, the system comprising:

[0030] An acquisition module, used to acquire historical operating parameters of the denitration system and the nitrogen oxide concentration at the outlet of the denitration system;

[0031] A reconstruction module, used to determine the delay time of each type of data in the historical operation parameters through a dynamic time warping algorithm, and reconstruct the historical operation parameters based on the delay time; determine the eigenvalues ​​corresponding to each type of data in the reconstructed historical operation parameters through a mutual information algorithm, and use the Spearman rank correlation coefficient to determine the correlation between any two eigenvalues ​​in the eigenvalues ​​corresponding to each type of data; reconstruct the reconstructed historical operation parameters based on the correlation to obtain a training data set;

[0032] A training model is used to train a long short-term memory network using the training data set as input and the nitrogen oxide concentration at the outlet of the denitrification system as output to obtain a nitrogen oxide concentration prediction model;

[0033] The prediction module is used to input the operating parameters of the denitrification system collected in real time into the nitrogen oxide concentration prediction model to obtain the predicted concentration of nitrogen oxides, and calculate the ammonia injection amount according to the predicted concentration of nitrogen oxides.

[0034] The present invention provides a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the SCR denitration ammonia injection control method is implemented.

[0035] The present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the SCR denitration ammonia injection control method when executing the program.

[0036] At least one technical solution adopted in the present invention can achieve the following beneficial effects: the present invention determines the delay time of each type of data in the historical operating parameters through a dynamic time warping algorithm, and reconstructs the historical operating parameters based on the delay time, and then determines the eigenvalues ​​corresponding to each type of data in the reconstructed historical operating parameters through a mutual information algorithm, and uses the Spearman rank correlation coefficient to determine the correlation between any two eigenvalues ​​of the eigenvalues ​​corresponding to each type of data, and reconstructs the reconstructed historical operating parameters again based on the correlation degree, which can reduce the time delay of the SCR system in predicting the nitrogen oxide concentration according to the input data, and can obtain a variety of training data sets with high correlation and strong correlation with nitrogen oxide prediction from a large amount of data. Then, the long short-term memory network is trained with the training data set as input and the nitrogen oxide concentration at the outlet of the denitrification system as output to obtain a nitrogen oxide concentration prediction model; finally, the operating parameters of the denitrification system collected in real time are input into the nitrogen oxide concentration prediction model to obtain the predicted nitrogen oxide concentration, and the ammonia injection amount is calculated according to the predicted nitrogen oxide concentration. It is possible to obtain a variety of highly correlated data with strong correlation with the nitrogen oxide prediction from a large amount of data, and reduce the time delay of the SCR system in predicting the nitrogen oxide concentration according to the input data, thereby improving the SCR system's prediction accuracy and speed for the nitrogen oxide concentration, and then the amount of ammonia injection required for denitrification can be calculated in real time according to the predicted nitrogen oxide concentration, thereby achieving precise control of the ammonia injection amount. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings described herein are used to provide a further understanding of the present invention and constitute a part 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 an improper limitation of the present invention. In the drawings:

[0038] Figure 1 A flow chart of a SCR denitration ammonia injection control method provided by the present invention;

[0039] Figure 2 A schematic diagram of the nitrogen oxide concentration of the inlet flue of the SCR denitration ammonia injection control system provided by the present invention;

[0040] Figure 3 A schematic diagram of the nitrogen oxide concentration of the inlet flue of the SCR denitration ammonia injection control system after screening provided by the present invention;

[0041] Figure 4 A comparison chart of the predicted value and the actual value of the nitrogen oxide concentration of the inlet flue of the SCR denitration ammonia injection control system provided by the present invention;

[0042] Figure 5 A structural diagram of an SCR denitration and ammonia injection control system provided by the present invention;

[0043] Figure 6 A structural diagram of an SCR denitration ammonia injection control device provided by the present invention;

[0044] Figure 7 A schematic diagram of a computer device for implementing a SCR denitration ammonia injection control method provided by the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0046] The server mentioned in the present invention can be a server set on a business platform, or a device such as a desktop computer, a laptop computer, etc. that can execute the solution of the present invention. For the convenience of description, the following description is only based on the server as the execution subject. The following is a detailed description of the technical solutions provided by various embodiments of the present invention in conjunction with the accompanying drawings.

[0047] The present invention proposes a SCR denitration ammonia injection control method, which is implemented based on an SCR denitration ammonia injection control device. Figure 5 As shown, the device comprises:

[0048] Distributed control module 1, coal-fired boiler 2, selective oxidation-reduction denitrification module 3, machine learning module 8.

[0049] The distributed control module includes: a distributed control system server 1-1 and a data collector 1-2.

[0050] The selective redox denitration module includes: inlet sensors 4 - 1 and 4 - 2 , outlet sensors 5 - 1 and 5 - 2 , an ammonia injection grid 6 and a catalytic bed 7 .

[0051] The inlet sensor and the outlet sensor are symmetrically installed in the flues on both sides of the selective redox denitrification module.

[0052] The machine learning module includes: a machine learning server 8-1, a data preprocessing program 8-2 installed in the machine learning server, a real-time data acquisition and storage system 8-3 and a nitrogen oxide concentration prediction model 8-4.

[0053] The data collector in the distributed control module establishes real-time communication with the data preprocessing program, and the inlet sensor and the outlet sensor establish real-time communication with the real-time data acquisition and storage system.

[0054] In addition, in one or more embodiments of the present invention, the distributed control module is used to record and store historical operating parameters in the coal-fired boiler, and the inlet sensor is used to online monitor the inlet flue gas temperature, oxygen concentration, nitrogen oxide concentration and inlet pressure.

[0055] Specifically, Figure 1 As shown, a SCR denitration ammonia injection control method proposed by the present invention specifically includes the following steps:

[0056] S10: Obtain historical operating parameters of the denitrification system and the nitrogen oxide concentration at the outlet of the denitrification system.

[0057] In this embodiment, the historical operating parameters include: coal feed rate, air feed rate, furnace temperature, boiler load, nitrogen oxide concentration, oxygen concentration, ammonia concentration and flue gas temperature at the inlet of the denitrification system, oxygen concentration, ammonia concentration and flue gas temperature at the outlet of the denitrification system, and the air pressure difference between the inlet and outlet of the denitrification system.

[0058] S20: Determine the delay time of each type of data in the historical operating parameters through a dynamic time warping algorithm, and reconstruct the historical operating parameters based on the delay time; determine the eigenvalues ​​corresponding to each type of data in the reconstructed historical operating parameters through a mutual information algorithm, and use the Spearman rank correlation coefficient to determine the correlation between any two eigenvalues ​​corresponding to each type of data; reconstruct the reconstructed historical operating parameters again based on the correlation to obtain a training data set.

[0059] In this embodiment, due to the different sensor delays and boiler reaction lags of the distributed control system for different characteristic values, there is a time difference between different characteristic values ​​at the same time, which cannot be perfectly one-to-one correspondence, resulting in the model being unable to accurately reflect the NOx concentration at the SCR inlet. Therefore, it is necessary to determine the dynamic delay time of each characteristic value collected.

[0060] S30: Taking the training data set as input and the nitrogen oxide concentration at the outlet of the denitrification system as output, the long short-term memory network is trained to obtain a nitrogen oxide concentration prediction model.

[0061] In this embodiment, a long short-term memory (LSTM) model is trained. The model is suitable for processing time series data and can effectively capture and learn long-term dependencies in long time series data. In the prediction of NOx and control of ammonia injection in thermal power plants, the LSTM model can capture the complex temporal relationship between various parameters in coal combustion, such as the inlet flue gas temperature, coal supply, air supply, boiler load, etc. of the denitrification system and the NOx concentration at the inlet of the SCR system by learning historical operating parameters, so as to accurately predict the NOx concentration, and on this basis adjust the controllable parameters to control the ammonia injection amount. Ultimately, the efficiency and economy of the SCR denitrification system are improved.

[0062] First, the historical operating parameters of the coal-fired boiler are used as the training set to train the LSTM model. The model structure of LSTM consists of an input layer, two LSTM layers, a dropout layer and a fully connected layer. Among them, the Dropout layer is a regularization technique used in neural network training. Its core idea is to randomly discard a part of neurons in each training iteration. These discarded neurons will not participate in the parameter update in the current training iteration, thereby reducing the complexity of the network and reducing the risk of overfitting. The model uses the Adaptive Moment Estimator Optimizer (Adam) algorithm as the optimizer. The Adam algorithm uses the first-order momentum of the historical gradient to accelerate training and the second-order momentum of the historical gradient to adjust the learning rate of the connection weights in the network. Using the Mean Squared Error (MSE) as the loss function helps to minimize the difference between the predicted value and the true value.

[0063] After the model training is completed, the trained model is used to predict the new input feature values ​​to obtain the NOx concentration at the SCR inlet, and the ammonia injection amount is calculated according to the chemical equation 4NO+4NH3+O2→4N2+6H2O to achieve the effect of precise ammonia injection.

[0064] The invention in this embodiment is used to transform the SCR system of a supercritical once-through boiler in a power plant. The NOx data at the inlet of the SCR denitration system of the unit before the transformation during 7 days of continuous operation are collected as a training set. Figure 2 shown.

[0065] S40: Inputting the operating parameters of the denitrification system collected in real time into a nitrogen oxide concentration prediction model to obtain a predicted nitrogen oxide concentration, and calculating the ammonia injection amount according to the predicted nitrogen oxide concentration.

[0066] In this embodiment, the trained nitrogen oxide concentration prediction model is used to predict the nitrogen oxide concentration at the inlet of the SCR denitrification system during the peak load regulation process of the unit in real time. The results of the predicted value and the measured value are as follows: Figure 3 shown.

[0067] Depend on Figure 3 It can be seen that the present invention can effectively predict the nitrogen oxide concentration at the inlet of the SCR denitration system.

[0068] Finally, the amount of ammonia injection was calculated based on the predicted NOx concentration and the chemical reaction equation "4NO+4NH3+O2→4N2+6H2O". The difference in urea consumption of coal-fired boilers before and after the SCR denitrification system transformation was compared, and the urea consumption was collected 8 months before the transformation and 1 month after the transformation. The results are shown in Table 1.

[0069] Table 1 Urea consumption 8 months before and 1 month after transformation

[0070]

[0071] As shown in Table 1, before the transformation, the urea consumption per kilowatt-hour was 0.61 g / Nm 3 After the transformation, the urea consumption per kilowatt-hour is 0.55g / Nm 3 Therefore, the urea consumption per kWh of electricity was reduced by 16.40% after the transformation.

[0072] based on Figure 1 An SCR denitrification ammonia injection control method is shown. The present invention determines the delay time of each type of data in the historical operating parameters through a dynamic time warping algorithm, and reconstructs the historical operating parameters based on the delay time, and then determines the eigenvalues ​​corresponding to each type of data in the reconstructed historical operating parameters through a mutual information algorithm, and uses the Spearman rank correlation coefficient to determine the correlation between any two eigenvalues ​​of the eigenvalues ​​corresponding to each type of data, and reconstructs the reconstructed historical operating parameters again based on the correlation degree, which can reduce the time delay of the SCR system in predicting the nitrogen oxide concentration according to the input data, and can obtain multiple training data sets with high correlation and strong correlation with nitrogen oxide prediction from a large amount of data. Then, the long short-term memory network is trained with the training data set as input and the nitrogen oxide concentration at the outlet of the denitrification system as output to obtain a nitrogen oxide concentration prediction model; finally, the operating parameters of the denitrification system collected in real time are input into the nitrogen oxide concentration prediction model to obtain the predicted nitrogen oxide concentration, and the ammonia injection amount is calculated according to the predicted nitrogen oxide concentration. It is possible to obtain a variety of highly correlated data with strong correlation with the nitrogen oxide prediction from a large amount of data, and reduce the time delay of the SCR system in predicting the nitrogen oxide concentration according to the input data, thereby improving the SCR system's prediction accuracy and speed for the nitrogen oxide concentration, and then the amount of ammonia injection required for denitrification can be calculated in real time according to the predicted nitrogen oxide concentration, thereby achieving precise control of the ammonia injection amount.

[0073] When applying the SCR denitration ammonia injection control method provided by the present invention, it is not necessary to Figure 1 The steps are executed in the order shown. The specific execution order of the steps can be determined according to needs, and the present invention does not limit this.

[0074] In addition, in one or more embodiments of the present invention, before determining the delay time of each type of data in the historical operating parameters by the dynamic time warping algorithm, it also includes screening the historical operating parameters for outliers based on the 3σ criterion and removing the screened outliers, specifically including:

[0075] Calculate the mean and standard deviation of various types of data in historical operating parameters.

[0076] Based on the mean and standard deviation, remove the data whose difference with the mean or standard deviation is greater than the preset threshold.

[0077] In this embodiment, the formulas for calculating the mean and standard deviation of various types of data in the historical operating parameters are:

[0078]

[0079]

[0080] Among them, x i is each data point at which each data is collected, and N is the number of collected data points.

[0081] In this embodiment, an abnormal value refers to data whose difference with the value of each type of data in the historical operating parameters is greater than a preset threshold. Due to the failure or interference of the test device of the power plant, the historical operating parameters collected in the distributed control system usually contain abnormal values. The collected historical operating parameters are screened using the 3σ criterion, and the screened abnormal values ​​are replaced by the average value of the six data points before and after. If there are abnormal values ​​in the six points before and after, the abnormal values ​​of the overall data are used for replacement. Different types of data correspond to different preset thresholds, and the preset thresholds refer to the corresponding maximum and minimum values ​​of each type of data.

[0082] Optionally, determine the threshold of the outlier, and set the upper and lower thresholds of the outlier according to the 3σ criterion, with the upper limit = μ+3σ and the lower limit = μ-3σ. Check whether each data point in each type of data exceeds the upper and lower thresholds. If the data point xi satisfies xi>μ+3σ or xi<μ+3σ, it is considered an outlier.

[0083] Optionally, after removing the outliers, the various data in the historical operating parameters that have been removed are standardized based on the Min-Max normalization method. The magnitude differences between the data obtained at different measurement points in the SCR denitration system are large, and the dimensions of each data are also different, resulting in slow training of the neural network model and low accuracy. It is necessary to use the Min-Max normalization method to standardize the various characteristic value data in the sample set for the historical operating parameters that have been removed. As shown in formula (3):

[0084]

[0085] Among them, x is the original eigenvalue, x min is the minimum value of the feature, x max is the maximum value of the feature, and x' is the standardized feature value.

[0086] The scheme shown in this embodiment can improve the correlation between historical operating parameters and nitrogen oxide concentration prediction by eliminating abnormal values ​​in historical operating parameters, thereby improving the training effect of the nitrogen oxide concentration prediction model.

[0087] In addition, in one or more embodiments of the present invention, the delay time of each type of data in the historical operating parameters is determined by a dynamic time warping algorithm, specifically including:

[0088] The delay time of each type of data is calculated based on the Euclidean distance.

[0089] In this embodiment, due to the different sensor delays and boiler reaction lags of the distributed control system for different eigenvalue measurements, there is a time difference between different eigenvalues ​​at the same time, which cannot correspond perfectly one to one, resulting in the model being unable to accurately reflect the NOx concentration at the SCR inlet. Therefore, it is necessary to determine the dynamic time delay of each eigenvalue collected. The calculation is based on the dynamic time warping algorithm (Dynamic Time Warping, DTW), as follows:

[0090] The distance between time series points is measured according to the Euclidean distance, as shown in formula (4):

[0091] d(x i ,y j )=(x i -y j ) 2 (4)

[0092] Among them, x i and j are the i-th and j-th elements in the time series X and Y respectively.

[0093] Define a cumulative distance matrix D, whose element D(i,j) represents the minimum cumulative distance between the first i elements of X and the first j elements of Y. The recursive formula is shown in formula (5):

[0094] D(i,j)=d(x i ,y j )+min{D(i-1,j),D(i,j-1),D(i-1,j-1)} (5)

[0095] Boundary conditions:

[0096]

[0097] Among them, D = {x(k), y(k)} is the data set used for modeling, x(k) is the initial input eigenvalue data set, y(k) is the NOx concentration at the inlet of the SCR system, and k is the number of samples. The DTW distance is the value in the lower right corner of the cumulative distance matrix: DTW(X,Y) = D(N,M). The original data is reconstructed according to the calculated eigenvalues, that is, the eigenvalues ​​are aligned in time for the next step of training.

[0098] In addition, in one or more embodiments of the present invention, the eigenvalues ​​corresponding to each type of data in the reconstructed historical operating parameters are determined by a mutual information algorithm, and the expression is:

[0099]

[0100] Among them, p(x,y) is the joint probability distribution of time series X and time series Y, and p(x) and p(y) represent the independent probability distributions of time series X and time series Y respectively.

[0101] In this embodiment, the mutual information (MI) algorithm is based on the information entropy principle to test the correlation between time series X and time series Y. The correlation between two random variables can be expressed by MI as:

[0102]

[0103] I(X,Y)=H(X)+H(Y)-H(X,Y) (9)

[0104] Among them, H(X) and H(Y) represent the information entropy of X and Y respectively, n is the number of samples in the data set X, and p(xi) is the i The probability distribution of X, Y is the joint entropy of X and Y.

[0105] Specifically, the larger the mutual information value, the stronger the dependency between the two variables. A correlation coefficient MI between two variables greater than 0.4 indicates that the two variables are moderately or highly correlated. Therefore, from multiple eigenvalues, eigenvalues ​​with a correlation coefficient greater than 0.6 are selected as feature input values ​​for subsequent model calculations.

[0106] This embodiment preliminarily determines the characteristic input values ​​of the model based on the mutual information (MI) algorithm, which can achieve database dimensionality reduction, reduce the amount of calculation, avoid the reduction of the generalization ability of the prediction model, and improve the accuracy of NOx prediction.

[0107] In addition, in one or more embodiments of the present invention, the expression for determining the correlation between any two eigenvalues ​​among the eigenvalues ​​corresponding to each type of data using the Spearman rank correlation coefficient is:

[0108]

[0109] d i =R(x i )-R(y i ) (11)

[0110] Among them, ρ is the correlation between any two eigenvalues, d i is the eigenvalue corresponding to each type of data, n is the number of eigenvalues, R(x i ) and R(y i ) are x i and i level.

[0111] In this embodiment, ρ=1 indicates that there is a complete positive correlation between the two variables, ρ=-1 indicates that there is a complete negative correlation between the two variables, and ρ=0 indicates that there is no correlation between the two variables. After calculation, the characteristic input value of the model is determined.

[0112] The above is an SCR denitration ammonia injection control method provided by one or more embodiments of the present invention. Based on the same idea, the present invention also provides a corresponding SCR denitration ammonia injection control system, such as Figure 6 As shown, the system includes:

[0113] An acquisition module, used to acquire historical operating parameters of the denitration system and the nitrogen oxide concentration at the outlet of the denitration system;

[0114] A reconstruction module, used to determine the delay time of each type of data in the historical operation parameters through a dynamic time warping algorithm, and reconstruct the historical operation parameters based on the delay time; determine the eigenvalues ​​corresponding to each type of data in the reconstructed historical operation parameters through a mutual information algorithm, and use the Spearman rank correlation coefficient to determine the correlation between any two eigenvalues ​​in the eigenvalues ​​corresponding to each type of data; reconstruct the reconstructed historical operation parameters based on the correlation to obtain a training data set;

[0115] A training model is used to train a long short-term memory network using the training data set as input and the nitrogen oxide concentration at the outlet of the denitrification system as output to obtain a nitrogen oxide concentration prediction model;

[0116] The prediction module is used to input the operating parameters of the denitrification system collected in real time into the nitrogen oxide concentration prediction model to obtain the predicted concentration of nitrogen oxides, and calculate the ammonia injection amount according to the predicted concentration of nitrogen oxides.

[0117] Each module in an SCR denitration ammonia injection control device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each of the above modules.

[0118] The present invention also provides a computer-readable storage medium, which stores a computer program, which can be used to execute the Figure 1 A SCR denitrification ammonia injection control method is provided.

[0119] The present invention also provides Figure 7 The structural diagram of the computer device shown in FIG. Figure 7 As shown, at the hardware level, the computer device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and may also include other hardware required for the business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the Figure 1 A SCR denitrification ammonia injection control method is provided.

[0120] Those skilled in the art can understand that all or part of the processes in the embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0121] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of the present invention.

Claims

1. A SCR denitration ammonia injection control method, applied to a coal-fired boiler, characterized in that: include: Obtain the historical operating parameters of the denitrification system and the nitrogen oxide concentration at the outlet of the denitrification system; Determine the delay time of each type of data in the historical operation parameters by a dynamic time warping algorithm, and reconstruct the historical operation parameters based on the delay time; determine the eigenvalues ​​corresponding to each type of data in the reconstructed historical operation parameters by a mutual information algorithm, and determine the correlation between any two eigenvalues ​​in the eigenvalues ​​corresponding to each type of data by using the Spearman rank correlation coefficient; Reconstructing the reconstructed historical operating parameters again based on the correlation degree to obtain a training data set; Taking the training data set as input and the nitrogen oxide concentration at the outlet of the denitrification system as output, training the long short-term memory network to obtain a nitrogen oxide concentration prediction model; The operating parameters of the denitrification system collected in real time are input into the nitrogen oxide concentration prediction model to obtain the predicted nitrogen oxide concentration, and the ammonia injection amount is calculated according to the predicted nitrogen oxide concentration.

2. The SCR denitration ammonia injection control method according to claim 1, characterized in that: The historical operating parameters include: coal feed rate, air feed rate, furnace temperature, boiler load, nitrogen oxide concentration, oxygen concentration, ammonia concentration and flue gas temperature at the inlet of the denitrification system, oxygen concentration, ammonia concentration and flue gas temperature at the outlet of the denitrification system, and the air pressure difference between the inlet and outlet of the denitrification system.

3. The SCR denitration ammonia injection control method according to claim 1, characterized in that: Before determining the delay time of each type of data in the historical operating parameters by the dynamic time warping algorithm, the historical operating parameters are screened for outliers based on the 3σ criterion, and the screened outliers are removed, specifically including: Calculate the mean and standard deviation of various types of data in the historical operating parameters; According to the mean and the standard deviation, data in the historical operating parameters whose difference with the mean or the standard deviation is greater than a preset threshold are eliminated.

4. The SCR denitration ammonia injection control method according to claim 1, characterized in that: Determining the delay time of each type of data in the historical operating parameters by using a dynamic time warping algorithm specifically includes: Calculate the delay time of each type of data according to the Euclidean distance; The Euclidean distance calculation formula is: d(x i ,y j )=(x i -y j ) 2 ; Among them, x i and j are the i-th and j-th elements in time series X and Y respectively.

5. The SCR denitration ammonia injection control method according to claim 1, characterized in that: The mutual information algorithm is used to determine the eigenvalues ​​corresponding to each type of data in the reconstructed historical operating parameters, and its expression is: Among them, p(x,y) is the joint probability distribution of time series X and time series Y, and p(x) and p(y) represent the independent probability distributions of time series X and time series Y respectively.

6. The SCR denitration ammonia injection control method according to claim 1, characterized in that: The expression for determining the correlation between any two eigenvalues ​​among the eigenvalues ​​corresponding to each type of data using the Spearman rank correlation coefficient is: d i =R(x i )-R(y i ); Among them, ρ is the correlation between any two eigenvalues, d i is the level difference of the eigenvalues ​​corresponding to each type of data, n is the number of eigenvalues, R(x i ) and R(y i ) are x i and i level.

7. An SCR denitration ammonia injection control system, comprising: An acquisition module, used to acquire historical operating parameters of the denitration system and the nitrogen oxide concentration at the outlet of the denitration system; A reconstruction module, used to determine the delay time of each type of data in the historical operation parameters by a dynamic time warping algorithm, and reconstruct the historical operation parameters based on the delay time; determine the eigenvalues ​​corresponding to each type of data in the reconstructed historical operation parameters by a mutual information algorithm, and determine the correlation between any two eigenvalues ​​of the eigenvalues ​​corresponding to each type of data by using the Spearman rank correlation coefficient; Reconstructing the reconstructed historical operating parameters again based on the correlation degree to obtain a training data set; A training model is used to train a long short-term memory network using the training data set as input and the nitrogen oxide concentration at the outlet of the denitrification system as output to obtain a nitrogen oxide concentration prediction model; The prediction module is used to input the operating parameters of the denitrification system collected in real time into the nitrogen oxide concentration prediction model to obtain the predicted concentration of nitrogen oxides, and calculate the ammonia injection amount according to the predicted concentration of nitrogen oxides.

8. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the SCR denitration ammonia injection control method as claimed in claim 1 is implemented.

9. A computer device, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the SCR denitration and ammonia injection control method as claimed in claim 1 is implemented.

Images