CN105629736A - Data-driven thermal power generation unit SCR denitration disturbance suppression prediction control method - Google Patents

Abstract

The invention discloses a data-driven thermal power generation unit SCR denitration disturbance suppression prediction control method. The method can estimate influence of unknown disturbance to the system with a selective catalytic reduction (SCR) flue gas denitration system as a controlled object, opening signals of an ammonia injection valve as system control input and reactor outlet nitrogen oxide concentration as system output, on the basis of a subspace identification method and by utilizing system input/output data; model prediction precision is improved by utilizing observation disturbance values; and under the condition of not damaging prediction control optimality, disturbance is overcome actively, and better control quality is realized.

Description

Data-driven thermal power generating unit SCR denitration disturbance suppression prediction control method

Technical Field

The invention belongs to the technical field of thermal automatic control, and particularly relates to a data-driven thermal power generating unit SCR denitration disturbance suppression predictive control method.

Background

With the emphasis on improving the quality of atmospheric environment and protecting the ecological environment, the promotion of denitration is brought into the key project of energy conservation and emission reduction of the twelve-five national planning, and the latest national emission standard requires the nitrogen oxide NO of a coal-fired power plant_xThe discharge concentration should be lower than 100mg/m³. NO in the flue gas is sprayed by ammonia_xA Selective Catalytic Reduction (SCR) denitration system for reducing nitrogen into harmless nitrogen and water vapor is mainstream equipment for realizing nitrogen emission reduction of a thermal power unit and is a process object needing important monitoring in the running process of the unit.

Because the SCR denitration technology relates to a series of chemical reactions, nitrogen oxide NO is discharged_xThe concentration controlled object has larger inertia, so that the traditional control method is difficult to achieve a satisfactory control effect. In recent years, predictive control algorithms have achieved certain success in power station SCR denitration control applications. However, the predictive control algorithm deals with the problems such as the smoke flow, the temperature and the nitrogen oxide NO_xWhen disturbances such as concentration change, equipment wear, faults, etc. occur, the control effect is not ideal due to the lack of estimation of the undetectable disturbances. At present, most predictive controllers mainly solve system constraint and inertia, and active interference resistance is not considered through estimation disturbance. There is also a method of combining disturbance observation technology with predictive control, but it generally adopts a simple method of directly compensating for the control action by a disturbance estimator, destroying the optimality of predictive control, and reducing the control quality.

The data-driven disturbance suppression prediction control is based on a subspace identification method, the influence of unknown disturbance on a system is estimated by using input and output data of the system, the model prediction precision is improved by using a disturbance estimation value, and the disturbance effect is actively overcome on the premise of ensuring the optimal performance of the prediction control.

The invention fully utilizes the idea of predictive control, and one-time suboptimal solution is carried out on each step to obtain the optimal opening degree of the ammonia injection valve. Simulation results show that compared with a general predictive control algorithm, the algorithm disclosed by the invention can more effectively inhibit the undetectable disturbance and maintain the nitrogen oxide NO at the outlet of the reactor_xThe concentration is around the set value. When no immeasurable disturbance exists, the algorithm is equivalent to a common predictive control algorithm, and has better set value tracking and adjusting performance.

Disclosure of Invention

The purpose of the invention is as follows: in order to overcome the defects in the prior art, the invention provides a data-driven thermal power generating unit SCR denitration disturbance suppression predictive control method which can effectively suppress the undetectable disturbance in the process on the premise of not damaging the optimality of predictive control and improve the regulation quality of a denitration system.

The technical scheme is as follows: in order to achieve the purpose, the technical scheme adopted by the invention comprises the following steps:

step 1, switching the SCR denitration system to a manual state under a stable operation state (all parameter variables of the SCR denitration system are basically kept unchanged), exciting the SCR denitration system by taking an ammonia spraying valve opening signal u as input, and obtaining nitrogen oxide NO at the outlet of a reactor_xOpen loop response data of the concentration y, selecting a sampling period Ts, and constructing a subspace matrix of a pure lag-free part of the SCR denitration system by utilizing a subspace identification methodAndthe subscript w represents past data, u represents input, and the superscript dob represents disturbance observations;

step 2, utilizing the subspace matrixAndestimating the input estimation value at the current moment through the input and output data of the SCR denitration system operation at the past moment

Step 3, inputting the estimated value of the current timeAnd the actual input value u_kComparing, filtering the obtained signals by a filter Q(s) to obtain a disturbance estimation value equivalent to the input end

Step 4, switching the SCR denitration system to a manual state in a stable operation state, and designing an opening signal u of the ammonia spraying valve_pcAnd artificially adding a disturbance signal d to excite an SCR denitration system to obtain nitrogen oxide NO at the outlet of the reactor_xConcentration y_pcOpen loop response data and disturbance estimation signal

Step 5, inFor the amplification of the input, nitrogen oxides NO at the reactor outlet_xConcentration y_pcFor the output of the SCR denitration system, a subspace identification method in the step 1 is adopted to obtain a subspace matrixAndcalculating the nitrogen oxide NO at the outlet of the reactor in a future period_xConcentration of

Step 6, according to the nitrogen oxide NO at the outlet of the reactor in a period of time in the future_xConcentration ofCalculating to obtain an optimal ammonia injection valve opening signal u_optAnd the catalyst is used for an SCR denitration system.

The step 1 comprises the following steps:

step 1-1, continuously obtaining pure lag-removed output data Y and input data U from the 0 th time to the 2N + j-2 th time, and respectively arranging the pure lag-removed output data Y and the input data U into a Hankel matrix form:

wherein N is the number of rows of the matrix, N is greater than the order of the SCR denitration system, j is the number of columns of the matrix, the larger the matrix is, the better the matrix is, Y and U respectively represent a Hankel matrix formed by output and input data, and Y and U respectively represent Hankel matrices formed by output and input data^fAnd Y^pFuture data and past data, U, representing output data, respectively^fAnd U^pFuture data and past data, y, representing input data, respectively_jDenotes the jth output data, u_jRepresents the jth input data, the superscripts f and p represent the future and past, respectively, and the subscripts 0, 1.. and 2N + j-2 represent the number of data;

step 1-2, let W^p＝[(Y^p)^T(U^p)^T]^TThe following matrix is subjected to QR decomposition:

$\left[\begin{array}{c}{W}^p\\ U^f\\ Y^f\end{array}\right]=\left[\begin{array}{ccc}{R}_{11}& 0& 0\\ R_{21}& R_{22}& 0\\ R_{31}& R_{32}& R_{33}\end{array}\right]\left[\begin{array}{c}{Q}_1\\ Q_2\\ Q_3\end{array}\right],$

obtaining a matrix L:

step 1-3, obtaining a matrix L_w＝L(:,1:N(m+l))，L_uL (: N (m + L) +1: end), m being the input variable dimension, L being the output variable dimension, L (: 1: N (m + L)) representing the first N (m + L) columns of the matrix L, L (: N (m + L) +1: end) representing all columns of the matrix L following the N (m + L) +1 columns;

step 1-4, obtaining a subspace matrix

In step 2, the input estimation value of the current time kWherein y is_kIs the output of the SCR denitration system at the current moment k,the output and input data combinations of the SCR denitration system at the past N moments have the value of N which is equal to the number of rows N of the matrix in the step 1-1, the output data of the SCR denitration system at N past moments,and tau is input data of the SCR denitration system at the past N moments, and tau is the pure lag time of the SCR denitration system.

In step 5, the following formula is adopted to calculate the nitrogen oxide NO at the outlet of the reactor in a future period_xConcentration of

WhereinThe output and the amplification input data of past N moments (the moment is the concept in discrete control, each sampling time is called a moment, if the sampling time is 5s, namely, data acquisition is carried out every 5s, one moment is 5s) of the SCR denitration system are combined,

inputting data for the past N times of the SCR denitration system,

for the future N₂The amplification input data at each time point is,

${\tilde{u}}_f={\left[\begin{array}{cccc}{\tilde{u}}_{k+1}^T& {\tilde{u}}_{k+2}^T& \mathrm{...}& {\tilde{u}}_{k+N_2}^T\end{array}\right]}^T.$

the step 6 comprises the following steps:

step 6-1, calculating a performance index function J by adopting the following formula:

wherein Q is_fAnd R_fIs a weight matrix for adjusting the quality of input/output control,r_fis the future N₁Instantaneous reactor outlet nitrogen oxides NO_xA sequence of set values of the concentration,

r_k+1,...,respectively representing the time k +1 to k + N₁Nitrogen at the momentOxide NO_xThe concentration set point. ,

is the future N₁Instantaneous reactor outlet nitrogen oxides NO_xThe sequence of the concentration pre-estimated value,

respectively representing the time k +1 to k + N₁Time of day nitrogen oxide NO_xThe estimated value of the concentration is estimated,

Δu_fis the future N₂Opening signal sequence of ammonia injection valve at any momentAn increment of (d);

step 6-2, limiting the amplitude of an opening signal u of an ammonia spraying valve of the SCR denitration system (u)_min,u_max) And an incremental constraint (Δ u)_min,Δu_max) As follows:

u_min,u_maxrespectively representing the minimum value and the maximum value, delta u, of an ammonia injection valve opening signal u_min,Δu_maxThe minimum increment and the maximum increment of the ammonia injection valve opening signal u are respectively shown, subscript min, max respectively show the minimum and the maximum, and delta shows the increment.

6-3, substituting the formula (1) into the formula (2) at each sampling moment, and minimizing the performance index function J under the condition of satisfying the formulas (3) and (4) to obtain the optimal control increment sequence input increment

Step 6-4, extracting an optimal control increment sequence delta u_fFirst block increment in the meter Δ u_k+1And is acted on by the control of the current time u_kAdding and calculating the optimal ammonia injection valve opening signal u_k+1：

u_k+1＝u_k+Δu_k+1。

Subspace matrix for observing system disturbancesAndand a subspace matrix for predicting a future output of the systemAndis obtained by off-line identification. And (3) repeating the steps 2, 3, 5 and 6 to realize continuous control during online operation.

Sampling period T in step 1_sCan use the empirical rule T95/T_sSelecting 5-15, wherein T95 is the adjusting time of the transition process rising to 95%; the filter in the step 3 is a low-pass filter with the steady-state gain of 1; step 6 predictive control parameter Q_f,R_f,N₁,N₂The method can be artificially selected according to factors such as performance quality, calculation time and the like in the actual control process.

Has the advantages that: compared with the prior art, the thermal power generating unit SCR denitration data drive disturbance suppression prediction control method provided by the invention has the following advantages: based on data-driven design, the influence of workload and modeling error of the traditional predictive control modeling process is reduced; the method has the advantages that general prediction control is convenient to process constrained and large-inertia objects; the method has more excellent anti-interference capability, and can quickly estimate and eliminate the influence of suppression disturbance on the system on the premise of ensuring the optimality of predictive control; be applied to power station SCR deNOx systems and can effectively restrain interference, ensure export nitrogen oxide NO_xThe concentration is near the set value; meanwhile, the method has the same set value tracking and adjusting capacity as a common predictive controller, and the operation level of the SCR denitration system is generally improved. Compared with a general predictive control algorithm, the data-driven disturbance suppression predictive control method adopted by the invention can suppress the undetectable disturbance more quickly and timely and maintain the nitrogen oxide NO_xThe concentration is stabilized near the set value; compared with the conventional disturbance compensation prediction control technology, the method can suppress disturbance without destroying the optimal performance of prediction control. When no immeasurable disturbance exists, the method is equivalent to a common predictive control algorithm, and has better tracking and adjusting performance. In addition, because the method is completely based on data, the influence of a modeling process and modeling errors which are complicated in common prediction control can be effectively avoided.

Drawings

The foregoing and/or other advantages of the invention will become further apparent from the following detailed description of the invention when taken in conjunction with the accompanying drawings.

FIG. 1 is a block diagram of the system architecture of the present invention;

FIG. 2 shows the present invention (solid line) with the normal disturbance compensation predictive control (dashed line) and the general predictive control (dotted line) at the reactor inlet NO_xA control effect comparison graph (a dot-dash line is a set value) under the concentration step disturbance and the sine disturbance;

FIG. 3 is a graph comparing the control effect of the present invention (solid line) with the normal disturbance compensation prediction control (dotted line) and the general prediction control (dotted line) under the ammonia injection side step disturbance and the sinusoidal disturbance (the dot-dash line is the set value);

Detailed Description

The control method is applied to a simulation model of an SCR denitration system of a certain 600MW thermal power generating unit, and the control aim is to ensure that nitrogen oxide NO at the outlet of a reactor meets the input constraint condition_xThe concentration follows the set value.

The invention discloses a data-driven thermal power unit SCR denitration disturbance suppression predictive control method, which is based on a subspace identification method, utilizes system input and output data to estimate the influence of unknown disturbance on a system, utilizes a disturbance estimation value to improve model prediction precision, actively suppresses the disturbance action on the premise of not damaging the predictive control optimality, has the same set value tracking and adjusting capacity as a common predictive controller under the disturbance-free condition of an algorithm, and generally improves the adjusting quality of an SCR denitration system. The method comprises the following steps:

(1) under the stable operation state, the design is changed once in 30 seconds, the opening input signal u of the ammonia spraying valve lasts for 30000 seconds, the system is excited, and a series of nitrogen oxides NO at the outlet of the reactor are obtained_xOpen loop response data of concentration y. Selecting a sampling period T_s30s, constructing a subspace matrix without a pure lag part of the system by using a subspace identification methodAndthe method comprises the following specific steps:

11) 1000 sets of successively obtained pure lag-removed output data Y and input data U are arranged in a Hankel matrix form (2N + j-2 ═ 1000), respectively:

wherein N is the number of matrix lines, and N is 10; j is the number of matrix columns, the larger the matrix column number is, the better the matrix column number is, Y and U respectively represent Hankel matrix formed by output data and input data, and Y is^fAnd Y^pFuture data and past data, U, representing output data, respectively^fAnd U^pFuture data and past data, y, representing input data, respectively_jDenotes the jth output data, u_jRepresents the jth input data, the superscripts f and p represent the future and past, respectively, and the subscripts 0, 1.. and 2N + j-2 represent the number of data;

12) let W^p＝[(Y^p)^T(U^p)^T]^TThe following matrix is subjected to QR decomposition:

$\left[\begin{array}{c}{W}^p\\ U^f\\ Y^f\end{array}\right]=\left[\begin{array}{ccc}{R}_{11}& 0& 0\\ R_{21}& R_{22}& 0\\ R_{31}& R_{32}& R_{33}\end{array}\right]\left[\begin{array}{c}{Q}_1\\ Q_2\\ Q_3\end{array}\right],$

a matrix L is obtained which is,

13) obtain the matrix L_w＝L(:,1:N(m+l))，L_uL (: N (m + L) +1: end). m is 1, m is the input variable dimension, L is 1, L is the input output variable dimension, L (: 1: N (m + L)) represents the first N (m + L) columns of L, L (: N (m + L) +1: end) represents all columns of L following the N (m + L) +1 columns;

14) subspace matrix

(2) Estimating an input estimation value at the current time k by using the subspace matrix and through input and output data of the SCR denitration system operation at the past time:wherein y is_kIs the output of the system at the current moment,the method is characterized in that the method is a combination of output and input data of the system at N past moments, wherein tau is 15s which is the pure lag time of an SCR denitration system and can be obtained through visual estimation of a step response test;

(3) comparing the input estimation value of the current time with the actual input value, and selecting the filter Q(s) ═ 1/(5s +1)⁴Filtering the obtained signal to obtain a disturbance estimation value equivalent to the input end;

(4) under the stable operation state, designing an opening signal u of an ammonia injection valve_pcArtificially adding a disturbance signal d to excite a system to obtain a series of nitrogen oxides NO at the outlet of the reactor_xOpen loop response data y of concentration_pc；

(5) To be provided withFor amplification of the input, y_pcObtaining a subspace matrix for system output based on the subspace identification method in the step (1)Andfor nitrogen oxides NO at the outlet of the reactor in a future period_xAnd (4) estimating the concentration:

${\hat{y}}_f={\tilde{l}}_w{\tilde{w}}_p+{\tilde{l}}_u{\tilde{u}}_f---\left(1\right)$

whereinFor the output data and augmented input data combination of the system at the past N times,for the future N₂The amplification input at each time point is input,in this example, take N₂＝10；

(6) Calculating an optimal ammonia injection valve opening signal u, and taking a performance index functional formula of the formula (2):

$J={({\hat{y}}_f-r_f)}^T{Q}_f({\hat{y}}_f-r_f)+{\mathrm{\Δu}}_f^T{R}_f{\mathrm{\Δu}}_f---\left(2\right)$

wherein,is a weight matrix for adjusting the quality of input/output control,is the future N₁Instantaneous reactor outlet NO_xA sequence of set values is set, and,is the future N₁Instantaneous reactor outlet nitrogen oxides NO_xThe estimated value sequence can be described by formula (1), and N is taken₁＝10，Δu_fIs the future N₂Opening signal sequence of ammonia injection valve at any momentThe increment of (c).

Considering the amplitude constraint (u) of the opening signal u of the ammonia injection valve of the SCR denitration system_min＝0,u_max1) and incremental constraint (Δ u)_min＝-0.01/s,Δu_max0.01/s)：

$\left[\begin{array}{c}I\\ I\\ .\\ .\\ .\\ I\end{array}\right]{\mathrm{\Δu}}_{min}\≤{\mathrm{\Δu}}_f\≤\left[\begin{array}{c}I\\ I\\ .\\ .\\ .\\ I\end{array}\right]{\mathrm{\Δu}}_{max}---\left(4\right)$

Substituting (1) into the performance index formula (2) at each sampling moment, and minimizing (2) under the condition of satisfying constraints (3) and (4) to obtain the optimal control increment sequence input incrementExtracting an optimal control increment sequence delta u_fFirst block in the meter Δ u_k+1And is acted on by the control of the current time u_kAdding and calculating the optimal ammonia injection valve opening signal u_k+1：

u_k+1＝u_k+Δu_k+1(5)

And applied to an SCR denitration system.

(7) Subspace matrix for stationary observation system disturbancesAndand a subspace matrix for predicting a future output of the systemAndand (5) repeating the steps (2) (3) (5) (6) to realize continuous control.

In this embodiment, in order to compare the control effects of the data-driven disturbance prediction control method, the conventional disturbance compensation prediction control method, and the general prediction control method in the present invention, two sets of simulation tests are performed: 1) the initial output of the SCR denitration system is stabilized at 60mg/m³At t 10m, the outlet nitrogen oxide NOx setpoint is from 60mg/m³The change is 100mg/m³At t 25m and t 60m, the nitrogen oxides NO at the reactor inlet due to variations in the coal type and combustion conditions_xRespectively generating unknown step disturbance d in concentration_{inlet_NOx}＝20mg/m³And a sinusoidal disturbance d_{inlet_NOx}20sin (0.2 (t-60)); 2) the initial output of the SCR denitration system is stabilized at 120mg/m³At t 10m, nitrogen oxide NO is dischargedx is set to be 120mg/m³The change is 70mg/m³When t is 35m and t is 65-75m, the ammonia injection valve generates unknown step disturbance d respectively due to faults_u0.2 and a slope disturbance d_u＝0.2-0.04×(t-65)。

As shown in FIGS. 2 and 3, nitrogen oxide NO is generated when there is NO unknown disturbance_xUnder the condition that the outlet concentration set value is increased or reduced in step, the optimization control effect curve of the denitration system is basically overlapped with the conventional disturbance compensation prediction control curve and the general prediction control curve, and the denitration system has the same set value tracking and adjusting capacity as a common prediction controller. When different types of unknown disturbances occur, the optimization control method can eliminate the influence of the disturbances, maintain the concentration of the outlet nitrogen oxides NOx at a set value, and meanwhile compared with the conventional disturbance suppression prediction control, the method has a faster and timely disturbance suppression effect and improves the operation quality of the SCR denitration system.

The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.

Claims (6)

1. The data-driven thermal power generating unit SCR denitration disturbance suppression prediction control method is characterized by comprising the following steps of:

step 1, switching an SCR denitration system to a manual state in a stable operation state, taking an ammonia spraying valve opening signal u as input, exciting the SCR denitration system, and obtaining nitrogen oxide NO at an outlet of a reactor_xOpen loop response data of the concentration y, selecting a sampling period Ts, and constructing a subspace matrix of a pure lag-free part of the SCR denitration system by utilizing a subspace identification methodAnd

step 2, utilizing the subspace matrixAndestimating the input estimation value of the current time k through the input and output data of the SCR denitration system operation at the past time

Step 3, inputting the estimated value of the current timeAnd the actual input value u_kComparing, filtering the obtained signals by a filter Q(s) to obtain a disturbance estimation value equivalent to the input end

Step 4, switching the SCR denitration system to a manual state in a stable operation state, and designing an opening signal u of the ammonia spraying valve_pcAdding a disturbance signal d to excite an SCR denitration system to obtain nitrogen oxide NO at the outlet of the reactor_xConcentration y_pcOpen loop response data and disturbance estimation signal

Step 5, inFor the amplification of the input, nitrogen oxides NO at the reactor outlet_xConcentration y_pcFor the output of the SCR denitration system, a subspace identification method in the step 1 is adopted to obtain a subspace matrixAndcalculating the nitrogen oxide NO at the outlet of the reactor in a future period_xConcentration of

Step 6, according to the nitrogen oxide NO at the outlet of the reactor in a period of time in the future_xConcentration ofCalculating to obtain an optimal ammonia injection valve opening signal u_optAnd the catalyst is used for an SCR denitration system.

2. The method of claim 1, wherein: the step 1 comprises the following steps:

step 1-1, continuously obtaining pure lag-removed output data Y and input data U from the 0 th time to the 2N + j-2 th time, and respectively arranging the pure lag-removed output data Y and the input data U into a Hankel matrix form:

$Y=\left[\frac{Y^p}{Y^f}\right]=\left[\frac{\begin{array}{cccc}{y}_0& y_1& ...& y_{j-1}\\ y_1& y_2& ...& y_j\\ ...& ...& ...& ...\\ y_{N-1}& y_N& ...& y_{N+j-2}\end{array}}{\begin{array}{cccc}{y}_N& y_{N+1}& ...& y_{N+j-1}\\ y_{N+1}& y_{N+2}& ...& y_{N+j}\\ ...& ...& ...& ...\\ y_{2N-1}& y_{2N}& ...& y_{2N+j-2}\end{array}}\right],$

$U=\left[\frac{U^p}{U^f}\right]=\left[\frac{\begin{array}{cccc}{u}_0& u_1& ...& u_{j-1}\\ u_1& u_2& ...& u_j\\ ...& ...& ...& ...\\ u_{N-1}& u_N& ...& u_{N+j-2}\end{array}}{\begin{array}{cccc}{u}_N& u_{N+1}& ...& u_{N+j-1}\\ u_{N+1}& u_{N+2}& ...& u_{N+j}\\ ...& ...& ...& ...\\ u_{2N-1}& u_{2N}& ...& u_{2N+j-2}\end{array}}\right],$

wherein N is the number of rows of the matrix, N is greater than the order of the SCR denitration system, j is the number of columns of the matrix, Y and U respectively represent a Hankel matrix formed by output and input data, and Y^fAnd Y^pFuture data and past data, U, representing output data, respectively^fAnd U^pFuture data and past data, y, representing input data, respectively_jDenotes the jth output data, u_jRepresents the jth input data;

step 1-2, let W^p＝[(Y^p)^T(U^p)^T]^TThe following matrix is subjected to QR decomposition:

$\left[\begin{array}{c}{W}^p\\ U^f\\ Y^f\end{array}\right]=\left[\begin{array}{ccc}{R}_{11}& 0& 0\\ R_{21}& R_{22}& 0\\ R_{31}& R_{32}& R_{33}\end{array}\right]\left[\begin{array}{c}{Q}_1\\ Q_2\\ Q_3\end{array}\right],$

obtaining a matrix L:

step 1-3, obtaining a matrix L_w＝L(:,1:N(m+l))，L_uL (: N (m + L) +1: end), m being the input variable dimension, L being the output variable dimension, L (: 1: N (m + L)) representing the first N (m + L) columns of the matrix L, L (: N (m + L) +1: end) representing all columns of the matrix L following the N (m + L) +1 columns;

step 1-4, obtaining a subspace matrix

3. The method of claim 2, wherein: in step 2, the input estimation value of the current time kWherein y is_kIs the output of the SCR denitration system at the current moment k,the output and input data combinations of the SCR denitration system at the past N moments have the value of N which is equal to the number of rows N of the matrix in the step 1-1, the output data of the SCR denitration system at N past moments,and tau is input data of the SCR denitration system at the past N moments, and tau is the pure lag time of the SCR denitration system.

4. The method of claim 3, wherein the filter in step 3 is a low pass filter with a steady state gain of 1.

5. The method of claim 4, wherein in step 5, the reactor outlet nitrogen oxides NO for the future period of time is calculated using the following formula_xConcentration of

WhereinThe output data and the amplification input data of the SCR denitration system at the past N moments are combined,

inputting data for the past N times of the SCR denitration system,

for the future N₂The amplification input data at each time point is,

${\tilde{u}}_f={\left[\begin{array}{cccc}{\tilde{u}}_{k+1}^T,& {\tilde{u}}_{k+2}^T& ...& {\tilde{u}}_{k+N_2}^T\end{array}\right]}^T.$

6. the method of claim 5, wherein step 6 comprises the steps of:

step 6-1, calculating a performance index function J by adopting the following formula:

wherein Q is_fAnd R_fIs a weight matrix for adjusting the quality of input/output control,

$Q_f=Q_f^T>0,R_f=R_f^T>0,$

r_fis the future N₁Instantaneous reactor outlet nitrogen oxides NO_xA sequence of set values of the concentration,

$r_f={\left[\begin{array}{cccc}{r}_{k+1}^T& r_{k+2}^T& ...& r_{k+N_1}^T\end{array}\right]}^T,$

respectively representing the time k +1 to k + N₁Time of day nitrogen oxide NO_xThe concentration of the water is set to a value,

is the future N₁Instantaneous reactor outlet nitrogen oxides NO_xThe sequence of the concentration pre-estimated value,

${\hat{y}}_f={\left[\begin{array}{cccc}{\hat{y}}_{k+1}^T& {\hat{y}}_{k+2}^T& ...& {\hat{y}}_{k+N_1}^T\end{array}\right]}^T,$

respectively representing the time k +1 to k + N₁Time of day nitrogen oxide NO_xThe estimated value of the concentration is estimated,

Δu_fis the future N₂Opening signal sequence of carved ammonia spraying valveAn increment of (d);

step 6-2, spraying ammonia by the SCR denitration systemAmplitude constraint of valve opening signal u (u)_min,u_max) And an incremental constraint (Δ u)_min,Δu_max) As follows:

wherein u is_min,u_maxRespectively representing the minimum value and the maximum value, delta u, of an ammonia injection valve opening signal u_min,Δu_maxRespectively representing the minimum increment and the maximum increment of an ammonia injection valve opening signal u;

6-3, substituting the formula (1) into the formula (2) at each sampling moment, and minimizing the performance index function J under the condition of satisfying the formulas (3) and (4) to obtain an optimal control increment sequence delta u_f：

${\mathrm{\Δu}}_f={\left[\begin{array}{cccc}{\mathrm{\Δu}}_{k+1}^T& {\mathrm{\Δu}}_{k+2}^T& ...& {\mathrm{\Δu}}_{k+N_2}^T\end{array}\right]}^T;$

Step 6-4, extracting an optimal control increment sequence delta u_fFirst block increment in the meter Δ u_k+1And is acted on by the control of the current time u_kAdding and calculating the optimal ammonia injection valve opening signal u_k+1：

u_k+1＝u_k+Δu_k+1。