An ammonia injection control method for an SCR denitration system based on a random distribution control algorithm

Abstract

The present application relates to the technical field of SCR denitration system, and particularly relates to a kind of ammonia injection control method of SCR denitration system based on random distribution control algorithm.The method comprises the following steps: S100, real-time acquisition of SCR denitration system outlet NOx multi-point sampling data and SCR denitration system ammonia injection grid each partition ammonia injection amount;S200, according to the NOx multi-point sampling data and each partition ammonia injection amount obtained in step S100, estimate the probability distribution of SCR denitration system outlet NOx concentration;S300, the known SCR denitration system outlet NOx concentration probability density function is represented;S400, the next moment of SCR denitration system outlet NOx concentration probability density function corresponding is predicted;S500, the predicted output information is used to combine the feedback correction of SCR denitration system outlet NOx concentration probability distribution;S600, according to the concentration correction, the ammonia injection output control of SCR denitration system ammonia injection grid is adjusted in real time, and the random distribution control amount output of SCR denitration system is calculated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of SCR denitration system, and particularly relates to an ammonia injection control method for an SCR denitration system based on a stochastic distribution control algorithm. BACKGROUND

[0002] Thermal power generation plays a core role in energy supply in China, however, the nitrogen oxides (NOx) released in its combustion process are the main air pollutants, which have a long-term negative impact on air quality and further pose a threat to human health. China has implemented strict regulatory measures for air pollutant emissions of thermal power generating units. However, due to the nonlinear characteristics and hysteresis of the SCR flue gas denitration system, it is extremely difficult to accurately control the emission of NOx.

[0003] The existing control method generally uses PID to measure the NOx concentration at a certain point in the flue gas through a sensor, and then adjusts the supply amount of ammonia gas to control the emission amount of NOx, which has a serious hysteresis. In recent years, a method of predictive control has been used to control and adjust the ammonia injection amount, but the selection of the control feedback amount is still a single-point value of the SCR denitration system outlet NOx concentration, which leads to large and uneven fluctuations of the SCR denitration system outlet NOx concentration, and the denitration effect is not ideal. SUMMARY

[0004] The purpose of the present application is to provide an ammonia injection control method for an SCR denitration system based on a stochastic distribution control algorithm, which uses the outlet cross-section NOx concentration PDF as feedback and uses the stochastic distribution control algorithm to better realize zoned ammonia injection adjustment, thereby reducing the waste of ammonia gas resources while ensuring that the emissions meet the standards.

[0005] The present application adopts the following technical solution: an ammonia injection control method for an SCR denitration system based on a stochastic distribution control algorithm, comprising:

[0006] S100: acquiring real-time sampling values of multiple NOx concentration measuring points at the outlet of the SCR denitration system and ammonia injection amounts of each zone of the ammonia injection grid of the SCR denitration system;

[0007] S200: estimating the probability distribution of the NOx concentration at the outlet of the SCR denitration system according to the NOx multi-point sampling data and the ammonia injection amounts of each zone obtained in step S100;

[0008] S300: representing the probability density function of the NOx concentration at the outlet of the SCR denitration system;

[0009] S400: predicting the probability distribution of the NOx concentration at the outlet of the SCR denitration system at the next moment;

[0010] S500: feeding back and correcting the predicted output information in combination with the probability distribution of the NOx concentration at the outlet of the SCR denitration system.

[0011] S600: Real-time adjustment of the ammonia injection grid ammonia output control of the SCR denitration system according to the concentration correction, and calculation of the random distribution control output of the SCR denitration system.

[0012] In some embodiments, the sampling values ​​of multiple NOx concentration measurement points at the outlet of the SCR denitrification system are , the ammonia injection control amount of n ammonia injection grid partitions at time k in the SCR denitrification system at the corresponding time point is .

[0013] In some embodiments, step S200 includes:

[0014] S201: Obtaining the influence weights of the equidistant outlet NOx concentration nodes to be determined based on the NOx concentration sampling values ​​of multiple measurement points at the outlet of the SCR denitration system obtained in step S100;

[0015] S202: Based on the influence weights, the NOx concentrations of the nodes at the outlets to be determined are obtained by fitting:

[0016]

[0017]

[0018] ;

[0019] S203: Estimated SCR denitration system outlet NOx concentration k time point data , and obtain its probability density function by performing kernel density estimation using the following formula;

[0020]

[0021] in, is the kernel function, and h is the bandwidth.

[0022] In some embodiments, step S201 includes:

[0023] Assume that the SCR denitrification reactor has q fixed outlet NOx concentration sampling points and p equally spaced outlet NOx concentration nodes to be determined. Let the coordinates of the i-th fixed outlet NOx concentration sampling point be ,in The node of NOx concentration at the outlet of SCR denitrification reactor is ,in , and their distances from each fixed outlet NOx concentration sampling point are , find the NOx concentration sampling points of each fixed outlet for the equidistant outlet NOx concentration nodes to be calculated The influence weight is

[0024]

[0025] in, is the width parameter of the distance-related Gaussian function.

[0026] In some embodiments, in step S203, the bandwidth is calculated using the following formula:

[0027]

[0028] in, To estimate the standard deviation of the data, the kernel function used is the Gaussian kernel function, as shown below;

[0029]

[0030] in, is the bandwidth smoothing parameter, which affects the shape of the distribution.

[0031] In some embodiments, step S300 includes:

[0032] S301: Establish an xy coordinate system for representing the probability density function, where the x-axis in the coordinate system represents the range of NOx concentration at the outlet of the SCR denitration system, and the y-axis represents the probability density of NOx concentration at the outlet of the SCR denitration system at various values. Select a set of appropriate coordinate points on the x-axis. , As base coordinates;

[0033] S302: By substituting the selected base coordinates into In this paper, we can get a set of probability density state vectors:

[0034]

[0035] in, is the current control output;

[0036] S303: Use the probability density state vector to approximate the output probability density function;

[0037]

[0038] Among them, the m-dimensional vector represents the probability density state vector.

[0039] In some embodiments, step S400 takes the following approach:

[0040] S401: Based on the genetic algorithm, a set of appropriate basis functions are selected to approximate the probability density function of the NOx concentration at the outlet of the SCR denitrification system, and the probability density function is converted into the form of the sum of the basis functions and their weighted products;

[0041] S402: Using the basis function weights of the probability density function of the NOx concentration at the outlet of the SCR denitration system at time k and the ammonia injection control amount of the ammonia injection grid as input, and the basis function weights of the probability density function of the NOx concentration at the outlet of the SCR denitration system at time k+1 as output, a random weight neural network is trained. The basis function weights of the probability density function of the NOx concentration at the outlet of the SCR denitration system at the current time and the ammonia injection control amount of the ammonia injection grid are input to obtain the basis function weights of the probability density function of the NOx concentration at the outlet of the SCR denitration system at the next time:

[0042] ,

[0043] S403: Obtaining a predicted probability density function of NOx concentration at the outlet of the SCR denitration system at the next moment;

[0044]

[0045] in, , , For the l The weights corresponding to the basis functions are is the selected basis function.

[0046] S404: The probability density function of the NOx concentration at the outlet of the SCR denitrification system at the next moment is approximated as ,Right now

[0047] .

[0048] In some embodiments, in step S401,

[0049] The basis function is selected using the Gaussian basis function, and its expression is:

[0050] ,

[0051] in, and Represent the center value and width of the i-th Gaussian basis function respectively.

[0052] In some embodiments, step S500 includes:

[0053] S501: determine the error between the actual output SCR denitration reactor outlet NOx concentration probability distribution density and the predicted output SCR denitration reactor outlet NOx concentration probability distribution density at the k moment as:

[0054] ;

[0055] S502: use the error to feedback correct the predicted output outlet NOx concentration, and the corrected predicted output is:

[0056]

[0057] wherein, is a correction coefficient, and the value range of the correction coefficient is .

[0058] In some embodiments, step S600 comprises:

[0059] S601: convert the SCR denitration system outlet NOx concentration expected probability density function into a random probability distribution state vector ;

[0060] S602: calculate the difference between the expected outlet NOx concentration at the k+1 moment and the corrected predicted outlet NOx:

[0061]

[0062] S603: construct a performance index function for minimizing the SCR denitration system outlet NOx concentration difference:

[0063]

[0064] ;

[0065] S604: considering the resource saving effect of the whole SCR denitration system, construct a performance index function for minimizing the ammonia injection amount:

[0066]

[0067]

[0068] S605: optimization solution, obtain the control output that satisfies the performance index function of the NOx concentration difference and the performance index function of the ammonia injection amount minimization, that is:

[0069]

[0070] .

[0071] Compared with the prior art, the present application has the following beneficial effects:

[0072] The present invention provides an ammonia injection control method for an SCR denitration system that utilizes random distribution control theory for control decisions. This method achieves precise ammonia injection optimization control for each zone of the ammonia injection grid in the SCR denitration system, optimizing ammonia injection according to a given desired outlet NOx concentration distribution. The present invention uses feedback information based on the outlet NOx concentration distribution of the entire SCR denitration system. By adjusting the ammonia injection amount for each zone of the ammonia injection grid using a random distribution control algorithm, the present invention controls the outlet NOx concentration distribution of the entire SCR denitration system, ensuring that the outlet NOx concentration of the entire cross-section meets emission standards while reducing ammonia waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 It is a system working block diagram of the present invention;

[0074] Figure 2 This is a diagram of a system example of the present invention;

[0075] Figure 3 The distribution of NOx sampling points and estimation points at the outlet of the SCR denitrification system;

[0076] Figure 4 This is the point estimation error contour map of the NOx two-dimensional concentration field at the outlet of the SCR denitrification system;

[0077] Figure 5 is the outlet NOx concentration PDF of the SCR denitrification system at a certain moment;

[0078] Figure 6 is the PDF state vector representation of the NOx concentration at the outlet of the SCR denitrification system;

[0079] Figure 7 The probability distribution of outlet NOx concentration under different control outputs of the SCR denitrification system;

[0080] Figure 8 PDF comparison of expected, initial and final outlet NOx concentrations of the SCR denitrification system;

[0081] Figure 9 PDF change of NOx concentration at the outlet of SCR denitrification system. DETAILED DESCRIPTION

[0082] In order to make the technical solutions and advantages of the embodiments of the present application more clearly understood, the exemplary embodiments of the present application are further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, and are not an exhaustive list of all the embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other unless they conflict.

[0083] The ammonia injection control method of the SCR denitrification system based on the random distribution control algorithm mainly includes the following steps:

[0084] S100: Real-time sampling values ​​of multiple NOx concentration measurement points at the outlet of the SCR denitration system and the amount of ammonia sprayed in each partition of the ammonia spray grid of the SCR denitration system, including sampling values ​​of multiple NOx concentration measurement points at the outlet of the SCR denitration system , and the ammonia injection control amount of n ammonia injection grid partitions at time k in the SCR denitrification system at the corresponding time point .

[0085] S200: Estimate the NOx concentration probability distribution at the outlet of the SCR denitration system based on the acquired information. The estimation is specifically implemented in the following manner.

[0086] S201: Sampling the NOx concentration at multiple measurement points at the outlet of the SCR denitration system obtained in step S100 , estimate the equidistant sampling points at the outlet of the SCR denitrification system as follows:

[0087] As attached Figure 3 As shown, there are q fixed outlet NOx concentration sampling points and p equidistant outlet NOx concentration nodes on one side of the SCR denitrification reactor. Let the coordinates of the i-th fixed outlet NOx concentration sampling point be ,in ; The SCR denitrification reactor outlet NOx concentration nodes to be calculated are ,in , and their distances from each fixed outlet NOx concentration sampling point are , we can get the NOx concentration sampling points of each fixed outlet for the equidistant outlet NOx concentration node The influence weight of

[0088]

[0089]

[0090] in, is the width parameter of the distance-related Gaussian function.

[0091] S202: Based on the influence weight, the NOx concentration of each equally spaced outlet NOx concentration node to be determined can be expressed as

[0092]

[0093]

[0094] ;

[0095] S203: Estimated SCR denitration system outlet NOx concentration k time point data The probability density function (PDF) is obtained by kernel density estimation using the following formula, as shown in the attached figure. Figure 5 shown.

[0096]

[0097]

[0098] in, is the kernel function, h is the bandwidth, and the bandwidth is calculated using the following formula.

[0099]

[0100] in, To estimate the standard deviation of the data, the kernel function is a Gaussian kernel function, as shown below.

[0101]

[0102] in, is the bandwidth smoothing parameter, which affects the shape of the distribution.

[0103] S300: Using the state vector to represent the known NOx concentration probability density function at the outlet of the SCR denitration system, this is specifically achieved in the following manner:

[0104] S301: Establish an xy coordinate system for representing the probability density function, where the x-axis in the coordinate system represents the range of NOx concentration at the outlet of the SCR denitration system, and the y-axis represents the probability density of NOx concentration at the outlet of the SCR denitration system at various values. Select a set of appropriate coordinate points on the x-axis. , As base coordinates;

[0105] S302: By substituting the selected base coordinates into In this paper, we can get a set of probability density state vectors:

[0106]

[0107] in, is the current control output.

[0108] S303: Use the probability density state vector to approximate the output probability density function (PDF).

[0109]

[0110] Among them, the m-dimensional vector represents the probability density state vector.

[0111] S400: Predict the probability distribution of NOx concentration at the outlet of the SCR denitration system at the next moment, which is specifically achieved by the following method:

[0112] S401: Based on the genetic algorithm, a set of appropriate basis functions are selected to approximate the probability density function of the NOx concentration at the outlet of the SCR denitrification system, and the probability density function is converted into the form of the sum of the basis functions and their weighted products;

[0113] The basis function is selected using the Gaussian basis function, and its expression is:

[0114] ,

[0115] in, and Represent the center value and width of the i-th Gaussian basis function respectively.

[0116] S402: Using the basis function weights of the probability density function of the NOx concentration at the outlet of the SCR denitration system at time k and the ammonia injection control amount of the ammonia injection grid as input, and the basis function weights of the probability density function of the NOx concentration at the outlet of the SCR denitration system at time k+1 as output, a random weight neural network is trained. The basis function weights of the probability density function of the NOx concentration at the outlet of the SCR denitration system at the current time and the ammonia injection control amount of the ammonia injection grid are input to obtain the basis function weights of the probability density function of the NOx concentration at the outlet of the SCR denitration system at the next time:

[0117] ,

[0118] S403: Obtaining a predicted probability density function of NOx concentration at the outlet of the SCR denitration system at the next moment;

[0119]

[0120] in, , , For the l The weights corresponding to the basis functions are is the selected basis function.

[0121] S404: The probability density function of the NOx concentration at the outlet of the SCR denitrification system at the next moment is approximated as ,Right now

[0122] .

[0123] S500: Feedback correction is performed using the predicted output information combined with the probability distribution of NOx concentration at the outlet of the SCR denitration system. This is specifically achieved in the following manner:

[0124] S501: Determine the error between the probability distribution density of NOx concentration at the outlet of the SCR denitration reactor actually output at the kth moment and the probability distribution density of NOx concentration at the outlet of the SCR denitration reactor predicted to be:

[0125] ;

[0126] S502: The error is used to perform feedback correction on the predicted outlet NOx concentration. The corrected predicted output is:

[0127]

[0128] in, is the correction factor, and its value range is .

[0129] S600: Real-time adjustment of the ammonia output control of the SCR denitration system's ammonia grid is performed based on the concentration correction, and the random distribution control output of the SCR denitration system is calculated. This is achieved specifically in the following manner:

[0130] S601: Convert the expected probability density function (PDF) of NOx concentration at the outlet of the SCR denitrification system into a random probability distribution state vector ;

[0131] S602: Calculate the difference between the expected outlet NOx concentration at time k+1 and the corrected predicted outlet NOx concentration:

[0132]

[0133] S603: Construct a performance index function that minimizes the difference in NOx concentration at the outlet of the SCR denitration system:

[0134]

[0135] ;

[0136] S604: Considering the resource saving effect of the entire SCR denitrification system, a performance index function for minimizing the amount of ammonia injection is constructed:

[0137]

[0138]

[0139] S605: Optimize and solve to obtain the control output that satisfies both the performance index function of NOx concentration difference and the performance index function of minimizing the amount of ammonia injection, namely:

[0140] .

[0141] .

[0142] The difference between the expected outlet NOx concentration distribution and the predicted outlet NOx concentration distribution at the k+1th moment is used as input, and the optimization algorithm is used to determine the ammonia injection control amount of the SCR denitrification system ammonia injection grid. A performance index that satisfies both the optimal outlet cross-section NOx concentration distribution and the relatively low ammonia injection consumption is constructed, and the optimization algorithm is used to optimize the performance index to obtain the final system output control amount, that is, the ammonia injection amount of each partition of the ammonia injection grid. Finally, the comparison of the expected, initial and final outlet NOx concentration PDF of the SCR denitrification system before and after control is shown in the attached figure. Figure 8 The change of NOx concentration PDF at the outlet of SCR denitrification system during iterative optimization control is shown in the attached figure. Figure 9 shown.

[0143] The overall process of the SCR denitrification system ammonia injection control method based on random distribution control algorithm proposed in this patent is as shown in the attached Figure 1 As shown in the figure, the error feedback adjustment process based on the predicted outlet NOx concentration probability distribution and the expected outlet NOx concentration probability distribution in real time is shown in the attached figure. Figure 2 As shown. The present invention estimates the probability distribution of NOx concentration at the outlet of the SCR denitration system through a random distribution algorithm, establishes the relationship between the ammonia injection amount of each partition of the ammonia injection grid in the SCR denitration system and the probability distribution of the outlet NOx concentration, and timely adjusts the ammonia injection amount of each partition of the ammonia injection grid according to the real-time change of the NOx concentration at the outlet cross section. The NOx concentration emission at the outlet of the SCR denitration system not only meets the requirements for the expected distribution of NOx concentration at the outlet of the SCR denitration system, but also achieves low and uniform emissions while effectively reducing unnecessary resource loss, which has important practical value.

[0144] Matters not covered by the present invention are known technologies.

[0145] The above embodiments are intended only to illustrate the technical concepts and features of the present invention. Their purpose is to enable those skilled in the art to understand the contents of the present invention and implement them accordingly. They are not intended to limit the scope of protection of the present invention. Any equivalent changes or modifications made in accordance with the spirit of the present invention are intended to be covered by the scope of protection of the present invention.

Claims

1. A method for controlling ammonia injection in an SCR denitrification system based on a random distribution control algorithm, characterized in that: include: S100: Real-time acquisition of the NOx concentration sampling points at the outlet of the SCR denitration system and the ammonia injection amount of each zone of the ammonia injection grid of the SCR denitration system; The sampling values ​​of multiple NOx concentration measurement points at the outlet of the SCR denitrification system are , the ammonia injection control amount of n ammonia injection grid partitions at time k in the SCR denitrification system at the corresponding time point is ; S2 00: Estimate the NOx concentration probability distribution at the outlet of the SCR denitration system based on the NOx multi-point sampling data and the ammonia injection amount of each partition obtained in step S100; Step S200 includes: S201: Obtaining the influence weights of the equidistant outlet NOx concentration nodes to be determined based on the NOx concentration sampling values ​​of multiple measurement points at the outlet of the SCR denitration system obtained in step S100; Step S201 includes: Assume that the SCR denitrification reactor has q fixed outlet NOx concentration sampling points and p equally spaced outlet NOx concentration nodes to be determined. Let the coordinates of the i-th fixed outlet NOx concentration sampling point be ,in The node of NOx concentration at the outlet of SCR denitrification reactor is ,in , and their distances from each fixed outlet NOx concentration sampling point are , find the NOx concentration sampling points of each fixed outlet for the equidistant outlet NOx concentration nodes to be calculated The influence weight is in, is the width parameter of the distance-related Gaussian function; S202: Based on the influence weights, the NOx concentrations of the nodes at the outlets to be determined are obtained by fitting: ； S203: Estimated SCR denitration system outlet NOx concentration k time point data , and obtain its probability density function by performing kernel density estimation using the following formula; in, is the kernel function, h is the bandwidth; In step S203, the bandwidth is calculated using the following formula: in, To estimate the standard deviation of the data, the kernel function used is the Gaussian kernel function, as shown below; in, is the smoothing parameter of the bandwidth, which affects the shape of the distribution; S300: expressing the probability density function of NOx concentration at the outlet of the SCR denitration system; Step S300 includes: S301: Establish an xy coordinate system for representing the probability density function, where the x-axis in the coordinate system represents the range of NOx concentration at the outlet of the SCR denitration system, and the y-axis represents the probability density of NOx concentration at the outlet of the SCR denitration system at various values. Select a set of appropriate coordinate points on the x-axis. , As base coordinates; S302: By substituting the selected base coordinates into In this paper, we can get a set of probability density state vectors: in, is the current control output; S303: Use the probability density state vector to approximate the output probability density function; Among them, the m-dimensional vector represents the probability density state vector; S400: Predicting the probability distribution of NOx concentration at the outlet of the SCR denitrification system at the next moment; Step S400 adopts the following method: S401: Based on the genetic algorithm, a set of appropriate basis functions are selected to approximate the probability density function of the NOx concentration at the outlet of the SCR denitrification system, and the probability density function is converted into the form of the sum of the basis functions and their weighted products; S402: Using the basis function weights of the probability density function of the NOx concentration at the outlet of the SCR denitration system at time k and the ammonia injection control amount of the ammonia injection grid as input, and the basis function weights of the probability density function of the NOx concentration at the outlet of the SCR denitration system at time k+1 as output, a random weight neural network is trained. The basis function weights of the probability density function of the NOx concentration at the outlet of the SCR denitration system at the current time and the ammonia injection control amount of the ammonia injection grid are input to obtain the basis function weights of the probability density function of the NOx concentration at the outlet of the SCR denitration system at the next time: ， S403: Obtaining a predicted probability density function of NOx concentration at the outlet of the SCR denitration system at the next moment; in, , , For the l The weights corresponding to the basis functions are is the selected basis function; S404: The probability density function of the NOx concentration at the outlet of the SCR denitrification system at the next moment is approximated as ,Right now ； S500: Using the predicted output information combined with the probability distribution of NOx concentration at the outlet of the SCR denitrification system for feedback correction; S600: Real-time adjustment of the ammonia injection grid ammonia output control of the SCR denitration system according to the concentration correction, and calculation of the random distribution control output of the SCR denitration system.

2. The ammonia injection control method for an SCR denitrification system based on a random distribution control algorithm according to claim 1 is characterized in that: In step S401, The basis function is selected using the Gaussian basis function, and its expression is: ， in, and Represent the center value and width of the i-th Gaussian basis function respectively.

3. The ammonia injection control method for an SCR denitrification system based on a random distribution control algorithm according to claim 1, characterized in that: The step S500 includes: S501: Determine the error between the probability distribution density of NOx concentration at the outlet of the SCR denitration reactor actually output at the kth moment and the probability distribution density of NOx concentration at the outlet of the SCR denitration reactor predicted to be: ； S502: The error is used to perform feedback correction on the predicted outlet NOx concentration. The corrected predicted output is: in, is the correction factor, and its value range is .

4. The ammonia injection control method for an SCR denitrification system based on a random distribution control algorithm according to claim 1 is characterized in that: The step S600 includes: S601: Convert the expected probability density function of NOx concentration at the outlet of the SCR denitrification system into a random probability distribution state vector ; S602: Calculate the difference between the expected outlet NOx concentration at time k+1 and the corrected predicted outlet NOx concentration: S603: Construct a performance index function that minimizes the difference in NOx concentration at the outlet of the SCR denitration system: ； S604: Considering the resource saving effect of the entire SCR denitrification system, a performance index function for minimizing the amount of ammonia injection is constructed: S605: Optimize and solve to obtain the control output that satisfies both the performance index function of NOx concentration difference and the performance index function of minimizing the amount of ammonia injection, namely: 。

Images