CN112418284B - A control method and system for a SCR denitration system in a full-operating power station - Google Patents

Abstract

The invention discloses a control method and a control system of an SCR denitration system of an all-condition power station, wherein the control method comprises the following steps: adding denitration cost into an optimization objective function, adopting a predictive control structure, and carrying out model establishment and control quantity optimization by combining a neural network and a genetic algorithm, thereby realizing the optimal control of the ammonia injection quantity. The method can well overcome the defects of large inertia and large delay of the SCR denitration system, improve the response speed of ammonia injection quantity control to unit load change and improve the dynamic regulation quality of the SCR denitration system.

Description

A control method and system for a SCR denitration system in a full-operating power station

Technical Field

The present invention relates to the technical field of coal-fired power plant optimization control, and in particular to a control method and system for a SCR denitration system of a full-operation power plant.

Background Art

In the process of energy structure transformation, the large-scale access of new energy power to the grid has put forward operational flexibility requirements for coal-fired units. Rapid deep load change means that the operating conditions of the unit change rapidly over a large range. The change of boiler operating conditions will increase the fluctuation of NO _x produced by combustion, which undoubtedly increases the difficulty of the unit to achieve ultra-low NO _x emissions. SCR (Selective Catalytic Reduction) - Selective catalytic reduction is currently the most mature and widely used flue gas denitrification technology in the world. SCR uses reducing agents such as NH ₃ to selectively react with NO _x in flue gas under the action of catalysts to generate non-toxic and non-polluting N ₂ and H ₂ O. SCR denitrification is the current mainstream flue gas denitrification technology. Its reaction is a complex physical and chemical process. A large amount of ammonia injection can reduce NO _x emissions, but it will increase economic costs and lead to increased ammonia slip, affecting the safe operation of the unit.

At present, there are two main types of SCR control systems used on site. One is the fixed molar ratio control method, which is an open-loop control method and cannot meet the ultra-low emission requirements; the other is the fixed outlet _NOx concentration control method. At present, the cascade PID control system is mostly used on site, and its parameters are adjusted according to the reactor characteristics under the design (rated) working conditions. When the unit working conditions change significantly, if the SCR system cannot be effectively controlled, it will cause low denitrification efficiency or serious ammonia escape, making it difficult to achieve optimal control, which is extremely unfavorable to the economic and environmental protection operation of thermal power plants. Therefore, it is often difficult to achieve good control effects using open-loop control and cascade PID control methods.

Therefore, how to optimize the control of the denitrification system and achieve economical operation of the unit while ensuring compliance with emission standards is an urgent problem to be solved in coal-fired power plants.

Summary of the invention

The purpose of the present invention is to provide a control method and system for a SCR denitration system in a power station under full-operation conditions, so as to optimize the control of the denitration system and realize economic operation of the unit while ensuring compliance with emission standards.

To achieve the above object, the present invention provides the following solutions:

A control method for a SCR denitration system of a power station under full working conditions, the control method comprising the following steps:

Obtain historical operating data of the SCR denitrification system;

A clustering algorithm is used to divide the historical operation data into multiple load intervals to obtain the historical operation data of multiple load intervals;

The neural network model is trained by using the historical operation data of each load interval to obtain the SCR denitration system output prediction model of each load interval; the SCR denitration system output prediction model is used to predict the outlet NOx concentration and outlet ammonia slip of the SCR denitration system at each prediction time point in the prediction time period according to the current working state data of the SCR denitration system and the ammonia injection amount at each control time point in the prediction time period; the working state data includes the unit load, _{the NOx} concentration and the flue gas flow rate at the inlet of the SCR denitration system;

According to the load instruction of the SCR denitration system, an output prediction model of the SCR denitration system in the load interval where the unit load corresponding to the load instruction is located is obtained;

According to the SCR denitration system output prediction model of the load interval where the unit load corresponding to the load instruction is located, a genetic algorithm is used to determine the ammonia injection amount that optimizes the multi-objective optimization function as the optimal ammonia injection amount;

The SCR destocking system is controlled according to the optimal ammonia injection amount.

Optionally, the neural network model includes an input layer, an output layer and a hidden layer, the input layer includes 4 neurons, the output layer includes 2 neurons, and the output layer includes 5 neurons.

Optionally, the SCR denitration system output prediction model in the load interval where the unit load corresponding to the load instruction is located uses a genetic algorithm to determine the ammonia injection amount that optimizes the multi-objective optimization function as the optimal ammonia injection amount, specifically including:

The ammonia injection amount at each control time point in the prediction period is taken as an individual to initialize the population of the genetic algorithm;

Determine whether the valve opening of the SCR denitration system corresponding to the ammonia injection amount at each control time point in each individual in the population is between the lower limit value and the upper limit value of the valve opening, and obtain a first determination result;

If the first judgment result indicates no, the ammonia injection amount in the individual of the SCR denitration system whose valve opening is less than the lower limit of the valve opening is replaced by the ammonia injection amount corresponding to the lower limit of the valve opening, and the ammonia injection amount in the individual of the SCR denitration system whose valve opening is greater than the upper limit of the valve opening is replaced by the ammonia injection amount corresponding to the upper limit of the valve opening;

Input each individual in the population into the SCR denitrification system output prediction model to obtain the predicted outlet _NOx concentration and predicted outlet ammonia slip at each prediction time point in the prediction time period corresponding to each individual;

Using the formula Correct the predicted outlet _NOx concentration at each prediction time point in the prediction time period corresponding to each individual to obtain the corrected predicted outlet _NOx concentration at each prediction time point in the prediction time period corresponding to each individual; wherein, is the predicted outlet _NOx concentration after correction at _time k+i·s1, is the predicted outlet _NOx concentration at time k+i·s ₁ , is the predicted outlet _NOx concentration at time k, is the actual outlet _NOx concentration of the SCR denitrification system at time k, i represents the i-th prediction time point, s ₁ is the prediction step length, and r is the correction coefficient;

Calculate the first objective function value and the second objective function value of each individual according to the predicted outlet _NOx concentration, the predicted outlet ammonia slip amount and the corrected predicted outlet _NOx concentration at each predicted time point within the prediction time period corresponding to each individual using the first objective function and the second objective function respectively;

Determine the individual with the smallest first objective function among the individuals in the population that satisfy the second objective function as the optimal individual of the Lth iteration, and set the individual with the larger first objective function value between the optimal individual of the Lth iteration and the global optimal individual of the L-1th iteration as the global optimal individual of the Lth iteration;

Determine whether the number of iterations is greater than an iteration number threshold, and obtain a second determination result;

If the second judgment result indicates no, then the value of the number of iterations is increased by 1, and the population is updated by the inheritance, mutation and recombination methods in the genetic algorithm, and the step of "determining whether the valve opening of the SCR denitrification system corresponding to the ammonia injection amount at each control time point in each individual in the population is between the lower limit and the upper limit of the valve opening to obtain the first judgment result" is returned;

If the second judgment result indicates yes, the global optimal individual of the Lth iteration is output as the optimal ammonia injection amount.

Optionally, the first objective function is:

Where J ₁ represents the first objective function value, is the predicted outlet ammonia slip at _time k+i·s1, _M2 is the liquid ammonia price, P is the number of predicted time points in the prediction time period, Q _gas is the flue gas flow rate, is the oxygen content of flue gas, _M1 is the sewage fee price, is the ammonia injection amount at _time k+j·s2, j is the jth control time point, _s2 is the control step size, M is the number of control time points in the prediction time period, N is the power generation of the unit, _M3 is the electricity price subsidy price, _ω1 is the first weight coefficient, and _ω2 is the second weight coefficient;

Optionally, the second objective function is:

Among them, J ₂ is the second objective function value, P is the number of prediction time points in the prediction time period, r(k+i·s ₁ ) is the expected value of the outlet NO _x concentration at moment k+i·s ₁ , ||Δu(k+j·s ₂ )|| is the difference between the ammonia injection amount at moment k+j·s ₂ and the expected ammonia injection amount, j is the jth control time point, s ₂ is the control step size, M is the number of control time points in the prediction time period, and ω ₃ is the third weight coefficient.

A control system for a full-operation power station SCR denitration system, the control system comprising:

A historical operation data acquisition module is used to obtain historical operation data of the SCR denitration system;

A clustering module is used to divide the historical operation data into multiple load intervals by using a clustering algorithm to obtain the historical operation data of the multiple load intervals;

A training module is used to train a neural network model using historical operation data of each load interval to obtain an output prediction model of the SCR denitration system in each load interval; the output prediction model of the SCR denitration system is used to predict the outlet NOx concentration and outlet ammonia slip of the SCR denitration system at each predicted time point in the predicted time period according to the current working state data of the SCR denitration system and the injection amount of ammonia at each control time point in the predicted time period; the working state data includes unit load, NOx concentration at the inlet of the SCR denitration system and flue gas flow;

An SCR denitration system output prediction model selection module is used to obtain an SCR denitration system output prediction model for a load interval where the unit load corresponding to the load instruction is located according to the load instruction of the SCR denitration system;

The optimal ammonia injection amount determination module is used to determine the ammonia injection amount that optimizes the multi-objective optimization function according to the SCR denitration system output prediction model of the load interval where the unit load corresponding to the load instruction is located, and use a genetic algorithm to determine the optimal ammonia injection amount as the optimal ammonia injection amount;

A control module is used to control the SCR destocking system according to the optimal ammonia injection amount.

Optionally, the neural network model includes an input layer, an output layer and a hidden layer, the input layer includes 4 neurons, the output layer includes 2 neurons, and the output layer includes 5 neurons.

Optionally, the optimal ammonia injection amount determination module specifically includes:

The initialization submodule is used to initialize the population of the genetic algorithm by taking the ammonia injection amount at each control time point in the prediction time period as an individual;

The first judgment submodule is used to judge whether the valve opening of the SCR denitration system corresponding to the ammonia injection amount at each control time point of each individual in the population is between the lower limit value and the upper limit value of the valve opening, and obtain a first judgment result;

an individual updating submodule, for replacing the amount of ammonia injection in the individual whose valve opening of the SCR denitration system is less than the lower limit of the valve opening with the amount of ammonia injection corresponding to the lower limit of the valve opening, and replacing the amount of ammonia injection in the individual whose valve opening of the SCR denitration system is greater than the upper limit of the valve opening with the amount of ammonia injection corresponding to the upper limit of the valve opening if the first judgment result indicates no;

The prediction submodule is used to input each individual in the population into the SCR denitration system output prediction model to obtain the predicted outlet NOx concentration and predicted outlet ammonia slip at each prediction time point in the prediction time period corresponding to each individual;

Correction submodule, used to use the formula Correct the predicted outlet NOx concentration at each prediction time point in the prediction time period corresponding to each individual to obtain the corrected predicted outlet NOx concentration at each prediction time point in the prediction time period corresponding to each individual; wherein, is the predicted outlet NOx concentration after correction at time k+i·s ₁ , is the predicted outlet NOx concentration at time k+i·s ₁ , is the predicted outlet NOx concentration at time k, is the actual outlet NOx concentration of the SCR denitrification system at time k, i represents the i-th prediction time point, s ₁ is the prediction step length, and r is the correction coefficient;

An objective function value calculation submodule, for calculating the first objective function value and the second objective function value of each individual according to the predicted outlet NOx concentration, the predicted outlet ammonia slip amount and the corrected predicted outlet NOx concentration at each prediction time point within the prediction time period corresponding to each individual, using the first objective function and the second objective function respectively;

The optimal individual determination submodule is used to determine the individual with the smallest first objective function among the individuals in the population that satisfy the second objective function as the optimal individual of the Lth iteration, and set the individual with the larger first objective function value between the optimal individual of the Lth iteration and the global optimal individual of the L-1th iteration as the global optimal individual of the Lth iteration;

A second judgment submodule is used to judge whether the number of iterations is greater than an iteration number threshold, and obtain a second judgment result;

A population update submodule, for, if the second judgment result indicates no, increasing the value of the number of iterations by 1, updating the population by adopting the inheritance, mutation and recombination methods in the genetic algorithm, and returning to the step of "judging whether the valve opening of the SCR denitration system corresponding to the ammonia injection amount at each control time point of each individual in the population is between the lower limit and the upper limit of the valve opening, and obtaining the first judgment result";

The optimal ammonia injection amount output submodule is used to output the global optimal individual of the Lth iteration as the optimal ammonia injection amount if the second judgment result indicates yes.

Optionally, the first objective function is:

Where J ₁ represents the first objective function value, is the predicted outlet ammonia slip at _time k+i·s1, _M2 is the liquid ammonia price, P is the number of predicted time points in the prediction time period, Q _gas is the flue gas flow rate, is the oxygen content of flue gas, _M1 is the sewage fee price, is the ammonia injection amount at _time k+j·s2, j is the jth control time point, _s2 is the control step size, M is the number of control time points in the prediction time period, N is the power generation of the unit, _M3 is the electricity price subsidy price, _ω1 is the first weight coefficient, and _ω2 is the second weight coefficient;

Optionally, the second objective function is:

Among them, J ₂ is the second objective function value, P is the number of prediction time points in the prediction time period, r(k+i·s ₁ ) is the expected value of the outlet NOx concentration at moment k+i·s ₁ , ||Δu(k+j·s ₂ )|| is the difference between the ammonia injection amount at moment k+j·s ₂ and the expected ammonia injection amount, j is the jth control time point, s ₂ is the control step size, M is the number of control time points in the prediction time period, and ω ₃ is the third weight coefficient.

According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

The present invention discloses a control method and system for a full-operation power station SCR denitration system. The control method includes: adding denitration cost to an optimization objective function, adopting a predictive control structure, combining a neural network and a genetic algorithm to establish a model and optimize the control quantity, thereby realizing the optimization control of the ammonia injection amount. By constructing a neural network model to predict the future trend of the outlet NO _x concentration of the SCR reactor, by adopting a genetic algorithm to optimize and determine the ammonia injection flow rate of the SCR denitration system, and adjusting the valve opening of the ammonia injection valve, the method can better overcome the shortcomings of the large inertia and large delay of the SCR denitration system, improve the response speed of the ammonia injection amount control to the load change of the unit, and improve the dynamic regulation quality of the SCR denitration system.

A multi-objective control method is adopted. The predictive control takes into account the upper and lower limits of the valve opening, the emission concentration of the outlet _NOx concentration, ammonia escape, and the economic indicators of the denitrification system, so as to avoid the influence of the system performance on the saturation of the actuator. On the basis of ensuring that the outlet _NOx concentration reaches the target value, the amount of ammonia injection can be minimized, and the operating cost and ammonia escape rate can be effectively reduced.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

FIG1 is a flow chart of a control method for a full-operation power station SCR denitration system provided by the present invention;

FIG2 is a schematic diagram showing the principle of a control method for a SCR denitration system of a power station under full operating conditions provided by the present invention;

FIG. 3 is a schematic diagram of the principle of using a genetic algorithm to determine the optimal ammonia injection amount for a multi-objective optimization function provided by the present invention.

FIG4 is a flow chart of the present invention for using a genetic algorithm to determine the optimal ammonia injection amount for the multi-objective optimization function.

DETAILED DESCRIPTION

The purpose of the present invention is to provide a control method and system for a SCR denitration system in a power station under full-operation conditions, so as to optimize the control of the denitration system and realize economic operation of the unit while ensuring compliance with emission standards.

In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

It should be noted that the terms "including" and "having" and any variations thereof in the embodiments of the present invention and the accompanying drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

As shown in FIG. 1-2 , the present invention provides a control method for a SCR denitration system of a power station under full operating conditions, and the control method comprises the following steps:

Step 101, obtaining historical operation data of the SCR denitration system.

The historical operation data include the load of the unit, NOx concentration at the inlet of the SCR denitrification system, flue gas flow, ammonia injection amount, NOx concentration at the outlet of the SCR denitrification system and ammonia slip.

The frequency of collecting historical operation data in the present invention is once every 2 minutes.

Step 101 : using a clustering algorithm to divide historical operation data into multiple load intervals, and obtaining historical operation data of multiple load intervals.

The historical data is divided into high load interval, medium load interval and low load interval through clustering algorithm according to the load interval.

Step 102, respectively train the neural network model using the historical operation data of each load interval to obtain the SCR denitration system output prediction model for each load interval; the SCR denitration system output prediction model is used to predict the outlet NOx concentration and outlet ammonia slip of the SCR denitration system at each prediction time point within the prediction time period according to the current working status data of the SCR denitration system and the injection amount of ammonia at each control time point within the prediction time period; the working status data includes the unit load, _{the NOx} concentration at the inlet of the SCR denitration system, and the flue gas flow rate.

The neural network models are trained separately according to the historical data of different load ranges. During the operation of the control algorithm, different neural network models are selected according to the real-time input data of the SCR denitrification system. This can better adapt to the background of large-scale changes in the unit operating conditions and significantly improve the accuracy of model prediction.

The neural network model is an LSTM neural network, including an input layer, an output layer and a hidden layer. The input layer includes 4 neurons, the output layer includes 2 neurons, and the output layer includes 5 neurons.

The present invention also performs normalization preprocessing on the historical operation data before training. During training, the neural network model training times are set to 1000 times, the learning rate is 0.05, and the minimum error is 10 ^-3 .

Step 103 , according to the load instruction of the SCR denitration system, an output prediction model of the SCR denitration system in the load interval where the unit load corresponding to the load instruction is located is obtained.

Step 104, according to the SCR denitration system output prediction model in the load interval where the unit load corresponding to the load instruction is located, a genetic algorithm is used to determine the ammonia injection amount that optimizes the multi-objective optimization function as the optimal ammonia injection amount.

The multi-objective optimization function of the present invention takes into account the outlet _NOx , ammonia slip and related economic costs under the premise that the SCR outlet _NOx concentration in the denitration control system meets the requirements, and realizes the safe and economical operation of the unit under the premise of meeting the emission standards. The upper and lower limits of the valve opening are added to the constraint objective function.

The present invention performs online rolling optimization in the prediction time domain, and uses a genetic algorithm to optimize and determine the control quantity sequence in the objective function, so that the prediction output of the neural network model can be as close to the expected values of the outlet _NOx concentration and ammonia slip as possible.

Compared with the traditional iterative algorithm, the genetic algorithm of the present invention can well avoid falling into the local minimum trap and the "dead loop" phenomenon, and is a global optimization algorithm.

As shown in FIGS. 3 and 4 , the SCR denitration system output prediction model in the load interval corresponding to the load instruction is used in step 103, and a genetic algorithm is used to determine the ammonia injection amount that optimizes the multi-objective optimization function as the optimal ammonia injection amount, which specifically includes:

Step 401, taking the ammonia spraying amount at each control time point in the prediction time period as an individual, initialize the population of the genetic algorithm; the present invention sets the population size of the genetic algorithm to 300, the maximum number of iterations (iteration number threshold) to 100, the mating probability to 0.85, and the mutation probability to 0.2.

Step 402, determining whether the valve opening of the SCR denitration system corresponding to the ammonia injection amount at each control time point in each individual in the population is between the lower limit and the upper limit of the valve opening, and obtaining a first determination result.

Step 403, if the first judgment result indicates no, the amount of ammonia sprayed in the individual of the SCR denitration system whose valve opening is less than the lower limit of the valve opening is replaced by the amount of ammonia sprayed corresponding to the lower limit of the valve opening, and the amount of ammonia sprayed in the individual of the SCR denitration system whose valve opening is greater than the upper limit of the valve opening is replaced by the amount of ammonia sprayed corresponding to the upper limit of the valve opening. The present invention takes into account both the upper and lower limits of the valve opening, the emission concentration of the outlet _NOx concentration, ammonia slip, and the economic index of the denitration system, avoids affecting the system performance due to the saturation of the actuator, and can minimize the amount of ammonia sprayed on the basis of ensuring that the outlet _NOx concentration reaches the target value, thereby effectively reducing the operating cost and the ammonia slip rate.

Step 404, input each individual in the population into the SCR denitration system output prediction model to obtain the predicted outlet _NOx concentration and the predicted outlet ammonia slip at each prediction time point within the prediction time period corresponding to each individual. The present invention performs a denormalization process on the output value of the SCR denitration system output prediction model to obtain the predicted value of the outlet NOx concentration and ammonia slip of the SCR denitration system at each prediction time point.

Step 405, using the formula Correct the predicted outlet _NOx concentration at each prediction time point in the prediction time period corresponding to each individual to obtain the corrected predicted outlet _NOx concentration at each prediction time point in the prediction time period corresponding to each individual; wherein, is the predicted outlet _NOx concentration after correction at _time k+i·s1, is the predicted outlet _NOx concentration at time k+i·s ₁ , is the predicted outlet _NOx concentration at time k, is the actual outlet NO _x concentration of the SCR denitration system at time k, i represents the i-th prediction time point, s ₁ is the prediction step, and r is the correction coefficient. The number of prediction time points of the present invention is 10, and the number of control time points is 3. The present invention: at time k, the amount of ammonia sprayed obtained by optimization is recorded as u(k-1), and the actual output y _m (k-1) of the system and the output y(k-1) of the prediction model can be obtained. Correspondingly, y _m (k) and y(k+1) can be obtained through u(k). Because there is inevitably a deviation between the prediction model and the actual system, the deviation between the actual output y _m (k) at time k and the model output y(k) at time k is regarded as an estimated value of the prediction error at time k, and it is used as a feedback correction signal to compensate the prediction model output y(k+i) at time k+i, that is, the prediction value y _p (k+i) after feedback correction is: y _p (k+i)=y(k+i)+r(y _m (k)-y(k)). The feedback correction link takes into account the model prediction error at the previous moment, which can improve the model prediction accuracy to a certain extent, thereby improving the control quality of predictive control.

Step 406, according to the predicted outlet _NOx concentration, the predicted outlet ammonia slip amount and the corrected predicted outlet _NOx concentration at each prediction time point within the prediction time period corresponding to each individual, the first objective function and the second objective function are used to calculate the first objective function value and the second objective function value of each individual.

Step 407, determining the individual with the smallest first objective function among the individuals in the population that satisfy the second objective function as the optimal individual of the Lth iteration, and setting the individual with the larger first objective function value between the optimal individual of the Lth iteration and the global optimal individual of the L-1th iteration as the global optimal individual of the Lth iteration;

Step 408, determine whether the number of iterations is greater than an iteration number threshold, and obtain a second determination result.

Step 409, if the second judgment result indicates no, then increase the value of the number of iterations by 1, and update the population by the inheritance, mutation and recombination methods in the genetic algorithm, and return to the step of "determining whether the valve opening of the SCR denitrification system corresponding to the ammonia injection amount at each control time point in each individual in the population is between the lower limit and the upper limit of the valve opening, and obtaining the first judgment result".

Step 410: If the second judgment result indicates yes, output the global optimal individual of the Lth iteration as the optimal ammonia injection amount.

Wherein, the first objective function is:

Where J ₁ represents the first objective function value, is the predicted outlet ammonia slip at _time k+i·s1, _M2 is the liquid ammonia price, P is the number of predicted time points in the prediction time period, Q _gas is the flue gas flow rate, is the oxygen content of flue gas, _M1 is the sewage fee price, is the ammonia injection amount at _time k+j·s2, j is the jth control time point, _s2 is the control step size, M is the number of control time points in the prediction time period, N is the power generation of the unit, _M3 is the electricity price subsidy price, _ω1 is the first weight coefficient, and _ω2 is the second weight coefficient;

The second objective function is:

Among them, J ₂ is the second objective function value, P is the number of prediction time points in the prediction time period, r(k+i·s ₁ ) is the expected value of the outlet NO _x concentration at moment k+i·s ₁ , ||Δu(k+j·s ₂ )|| is the difference between the ammonia injection amount at moment k+j·s ₂ and the expected ammonia injection amount, j is the jth control time point, s ₂ is the control step size, M is the number of control time points in the prediction time period, and ω ₃ is the third weight coefficient.

The objective function of the present invention includes ammonia escape, flue gas flow, flue gas oxygen content, nitrogen oxide emission, emission fee price, ammonia flow, liquid ammonia price, unit power generation, and electricity price subsidy price.

Step 105: Control the SCR destocking system according to the optimal ammonia injection amount.

The present invention also provides a control system for a full-operation power station SCR denitration system, the control system comprising:

The historical operation data acquisition module is used to obtain the historical operation data of the SCR denitrification system.

The clustering module is used to divide the historical operation data into multiple load intervals by using a clustering algorithm to obtain the historical operation data of the multiple load intervals.

The training module is used to train the neural network model using the historical operation data of each load interval to obtain the SCR denitration system output prediction model for each load interval; the SCR denitration system output prediction model is used to predict the outlet NOx concentration and outlet ammonia slip of the SCR denitration system at each prediction time point within the prediction time period according to the current working status data of the SCR denitration system and the injection amount of ammonia at each control time point within the prediction time period; the working status data includes the unit load, _{the NOx} concentration at the inlet of the SCR denitration system and the flue gas flow rate.

The neural network model includes an input layer, an output layer and a hidden layer, wherein the input layer includes 4 neurons, the output layer includes 2 neurons, and the output layer includes 5 neurons.

The SCR denitration system output prediction model selection module is used to obtain the SCR denitration system output prediction model for the load interval where the unit load corresponding to the load instruction is located according to the load instruction of the SCR denitration system.

The optimal ammonia injection amount determination module is used to determine the ammonia injection amount that optimizes the multi-objective optimization function according to the SCR denitration system output prediction model in the load range where the unit load corresponding to the load instruction is located, using a genetic algorithm as the optimal ammonia injection amount.

The optimal ammonia injection amount determination module specifically includes: an initialization submodule, which is used to initialize the population of the genetic algorithm with the ammonia injection amount at each control time point within the prediction time period as an individual; a first judgment submodule, which is used to judge whether the valve opening of the SCR denitration system corresponding to the ammonia injection amount at each control time point in each individual in the population is between the lower limit value and the upper limit value of the valve opening, and obtain a first judgment result; an individual update submodule, which is used to replace the ammonia injection amount in the individual whose valve opening of the SCR denitration system is less than the lower limit value of the valve opening with the ammonia injection amount corresponding to the lower limit value of the valve opening, and replace the ammonia injection amount in the individual whose valve opening of the SCR denitration system is greater than the upper limit value of the valve opening with the ammonia injection amount corresponding to the upper limit value of the valve opening; a prediction submodule, which is used to input each individual in the population into the SCR denitration system output prediction model, and obtain the predicted outlet _NOx concentration and predicted outlet ammonia slip at each prediction time point within the prediction time period corresponding to each individual; a correction submodule, which is used to use the formula Correct the predicted outlet _NOx concentration at each prediction time point in the prediction time period corresponding to each individual to obtain the corrected predicted outlet _NOx concentration at each prediction time point in the prediction time period corresponding to each individual; wherein, is the predicted outlet _NOx concentration after correction at _time k+i·s1, is the predicted outlet _NOx concentration at time k+i·s ₁ , is the predicted outlet _NOx concentration at time k, is the actual outlet _NOx concentration of the SCR denitrification system at time k, i represents the i-th prediction time point, _s1 is the prediction step, and r is the correction coefficient; the objective function value calculation submodule is used to calculate the predicted outlet _NOx concentration, predicted outlet ammonia slip and corrected predicted outlet NOx at each prediction time point in the prediction time period corresponding to each individual. _x concentration, respectively using the first objective function and the second objective function to calculate the first objective function value and the second objective function value of each individual; an optimal individual determination submodule, used to determine the individual with the smallest first objective function among the individuals in the population that meet the second objective function as the optimal individual of the Lth iteration, and set the individual with the larger first objective function value between the optimal individual of the Lth iteration and the global optimal individual of the L-1th iteration as the global optimal individual of the Lth iteration; a second judgment submodule, used to judge whether the number of iterations is greater than the iteration number threshold, and obtain a second judgment result; a population update submodule, used to increase the value of the number of iterations by 1 if the second judgment result indicates no, update the population by the inheritance, mutation and recombination method in the genetic algorithm, and return to the step of "judging whether the valve opening of the SCR denitrification system corresponding to the ammonia injection amount at each control time point in each individual in the population is between the lower limit and the upper limit of the valve opening, and obtain the first judgment result"; an optimal ammonia injection amount output submodule, used to output the global optimal individual of the Lth iteration as the optimal ammonia injection amount if the second judgment result indicates yes.

A control module is used to control the SCR destocking system according to the optimal ammonia injection amount.

Wherein, the first objective function is:

Where J ₁ represents the first objective function value, is the predicted outlet ammonia slip at _time k+i·s1, _M2 is the liquid ammonia price, P is the number of predicted time points in the prediction time period, Q _gas is the flue gas flow rate, is the oxygen content of flue gas, _M1 is the sewage fee price, is the amount of ammonia injection at _time k+j·s2, j is the jth control time point, _s2 is the control step, M is the number of control time points in the prediction time period, N is the power generation of the unit, _M3 is the electricity price subsidy price, _ω1 is the first weight coefficient, and _ω2 is the second weight coefficient.

The second objective function is:

Among them, J ₂ is the second objective function value, P is the number of prediction time points in the prediction time period, r(k+i·s ₁ ) is the expected value of the outlet NO _x concentration at moment k+i·s ₁ , ||Δu(k+j·s ₂ )|| is the difference between the ammonia injection amount at moment k+j·s ₂ and the expected ammonia injection amount, j is the jth control time point, s ₂ is the control step size, M is the number of control time points in the prediction time period, and ω ₃ is the third weight coefficient.

According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

1. This method can better overcome the shortcomings of large inertia and large delay of the SCR denitrification system, improve the response speed of ammonia injection control to unit load changes, and improve the dynamic regulation quality of the SCR denitrification system;

2. The multi-objective control method is adopted. The predictive control takes into account the upper and lower limits of the valve opening, the valve action rate limit, the emission concentration of the outlet _NOx concentration, ammonia escape and the economic indicators of the denitrification system, so as to avoid the influence of the system performance due to the saturation of the actuator. On the basis of ensuring that the outlet _NOx concentration reaches the target value, the amount of ammonia injection can be reduced as much as possible, effectively reducing the operating cost and the ammonia escape rate;

3. The predictive control technology of the present invention has low model requirements, is easy to calculate online, and has good control effect.

The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

This article uses specific examples to illustrate the principles and implementation methods of the invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. 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.

Claims (8)

1. A control method for a SCR denitration system in a full-operation power station, characterized in that the control method comprises the following steps: Obtain historical operating data of the SCR denitrification system; A clustering algorithm is used to divide the historical operation data into multiple load intervals to obtain the historical operation data of multiple load intervals; The neural network model is trained by using the historical operation data of each load interval to obtain the SCR denitration system output prediction model of each load interval; the SCR denitration system output prediction model is used to predict the outlet NOx concentration and outlet ammonia slip of the SCR denitration system at each prediction time point in the prediction time period according to the current working state data of the SCR denitration system and the ammonia injection amount at each control time point in the prediction time period; the working state data includes the unit load, _{the NOx} concentration and the flue gas flow rate at the inlet of the SCR denitration system; According to the load instruction of the SCR denitration system, an output prediction model of the SCR denitration system in the load interval where the unit load corresponding to the load instruction is located is obtained; According to the SCR denitration system output prediction model of the load interval where the unit load corresponding to the load instruction is located, a genetic algorithm is used to determine the ammonia injection amount that optimizes the multi-objective optimization function as the optimal ammonia injection amount; Controlling the SCR destocking system according to the optimal ammonia injection amount; The output prediction model of the SCR denitration system in the load interval where the unit load corresponding to the load instruction is located uses a genetic algorithm to determine the ammonia injection amount that optimizes the multi-objective optimization function as the optimal ammonia injection amount, specifically including: The ammonia injection amount at each control time point in the prediction period is taken as an individual to initialize the population of the genetic algorithm; Determine whether the valve opening of the SCR denitration system corresponding to the ammonia injection amount at each control time point in each individual in the population is between the lower limit value and the upper limit value of the valve opening, and obtain a first determination result; If the first judgment result indicates no, the ammonia injection amount in the individual of the SCR denitration system whose valve opening is less than the lower limit of the valve opening is replaced by the ammonia injection amount corresponding to the lower limit of the valve opening, and the ammonia injection amount in the individual of the SCR denitration system whose valve opening is greater than the upper limit of the valve opening is replaced by the ammonia injection amount corresponding to the upper limit of the valve opening; Input each individual in the population into the SCR denitrification system output prediction model to obtain the predicted outlet _NOx concentration and predicted outlet ammonia slip at each prediction time point in the prediction time period corresponding to each individual; Using the formula Correct the predicted outlet _NOx concentration at each prediction time point in the prediction time period corresponding to each individual to obtain the corrected predicted outlet _NOx concentration at each prediction time point in the prediction time period corresponding to each individual; wherein, is the predicted outlet _NOx concentration after correction at _time k+i·s1, is the predicted outlet _NOx concentration at time k+i·s ₁ , is the predicted outlet _NOx concentration at time k, is the actual outlet _NOx concentration of the SCR denitrification system at time k, i represents the i-th prediction time point, s ₁ is the prediction step length, and r is the correction coefficient; Calculate the first objective function value and the second objective function value of each individual according to the predicted outlet _NOx concentration, the predicted outlet ammonia slip amount and the corrected predicted outlet _NOx concentration at each predicted time point within the prediction time period corresponding to each individual using the first objective function and the second objective function respectively; Determine the individual with the smallest first objective function among the individuals in the population that satisfy the second objective function as the optimal individual of the Lth iteration. When the first objective function value of the optimal individual of the Lth iteration is greater than or equal to the first objective function value of the global optimal individual of the L-1th iteration, set the optimal individual of the Lth iteration as the global optimal individual of the Lth iteration; otherwise, set the global optimal individual of the L-1th iteration as the global optimal individual of the Lth iteration. Determine whether the number of iterations is greater than an iteration number threshold, and obtain a second determination result; If the second judgment result indicates no, then the value of the number of iterations is increased by 1, and the population is updated by the inheritance, mutation and recombination methods in the genetic algorithm, and the step of "determining whether the valve opening of the SCR denitrification system corresponding to the ammonia injection amount at each control time point in each individual in the population is between the lower limit and the upper limit of the valve opening to obtain the first judgment result" is returned; If the second judgment result indicates yes, the global optimal individual of the Lth iteration is output as the optimal ammonia injection amount. 2. The control method of the SCR denitrification system of a full-operating power station according to claim 1 is characterized in that the neural network model includes an input layer, an output layer and a hidden layer, the input layer includes 4 neurons, the output layer includes 2 neurons, and the output layer includes 5 neurons. 3. The control method of the SCR denitration system of a power station under full working conditions according to claim 1, characterized in that the first objective function is: Where J ₁ represents the first objective function value, is the predicted outlet ammonia slip at _time k+i·s1, _M2 is the liquid ammonia price, P is the number of predicted time points in the prediction time period, Q _gas is the flue gas flow rate, is the oxygen content of flue gas, _M1 is the sewage fee price, is the amount of ammonia injection at _time k+j·s2, j is the jth control time point, _s2 is the control step, M is the number of control time points in the prediction time period, N is the power generation of the unit, _M3 is the electricity price subsidy price, _ω1 is the first weight coefficient, and _ω2 is the second weight coefficient. 4. The control method of the SCR denitration system of a power station under full working conditions according to claim 1, characterized in that the second objective function is: Among them, J ₂ is the second objective function value, P is the number of prediction time points in the prediction time period, r(k+i·s ₁ ) is the expected value of the outlet NO _x concentration at moment k+i·s ₁ , ||Δu(k+j·s ₂ )|| is the difference between the ammonia injection amount at moment k+j·s ₂ and the expected ammonia injection amount, j is the jth control time point, s ₂ is the control step size, M is the number of control time points in the prediction time period, and ω ₃ is the third weight coefficient. 5. A control system for a full-operation power station SCR denitrification system, characterized in that the control system comprises: A historical operation data acquisition module is used to obtain historical operation data of the SCR denitration system; A clustering module is used to divide the historical operation data into multiple load intervals by using a clustering algorithm to obtain the historical operation data of the multiple load intervals; A training module is used to train a neural network model using historical operation data of each load interval to obtain an output prediction model of the SCR denitration system in each load interval; the output prediction model of the SCR denitration system is used to predict the outlet NOx concentration and outlet ammonia slip of the SCR denitration system at each predicted time point in the predicted time period according to the current working state data of the SCR denitration system and the ammonia injection amount at each control time point in the predicted time period; the working state data includes unit load, NOx concentration at the inlet of the SCR denitration system and flue gas flow; An SCR denitration system output prediction model selection module is used to obtain an SCR denitration system output prediction model for a load interval where the unit load corresponding to the load instruction is located according to the load instruction of the SCR denitration system; The optimal ammonia injection amount determination module is used to determine the ammonia injection amount that optimizes the multi-objective optimization function according to the SCR denitration system output prediction model of the load interval where the unit load corresponding to the load instruction is located, and use a genetic algorithm to determine the optimal ammonia injection amount as the optimal ammonia injection amount; A control module, used for controlling the SCR destocking system according to the optimal ammonia injection amount; The optimal ammonia injection amount determination module specifically includes: The initialization submodule is used to initialize the population of the genetic algorithm by taking the ammonia injection amount at each control time point in the prediction time period as an individual; The first judgment submodule is used to judge whether the valve opening of the SCR denitration system corresponding to the ammonia injection amount at each control time point of each individual in the population is between the lower limit value and the upper limit value of the valve opening, and obtain a first judgment result; an individual updating submodule, for replacing the amount of ammonia injection in the individual whose valve opening of the SCR denitration system is less than the lower limit of the valve opening with the amount of ammonia injection corresponding to the lower limit of the valve opening, and replacing the amount of ammonia injection in the individual whose valve opening of the SCR denitration system is greater than the upper limit of the valve opening with the amount of ammonia injection corresponding to the upper limit of the valve opening if the first judgment result indicates no; The prediction submodule is used to input each individual in the population into the SCR denitration system output prediction model to obtain the predicted outlet NOx concentration and predicted outlet ammonia slip at each prediction time point in the prediction time period corresponding to each individual; Correction submodule, used to use the formula Correct the predicted outlet NOx concentration at each prediction time point in the prediction time period corresponding to each individual to obtain the corrected predicted outlet NOx concentration at each prediction time point in the prediction time period corresponding to each individual; wherein, is the predicted outlet NOx concentration after correction at time k+i·s ₁ , is the predicted outlet NOx concentration at time k+i·s ₁ , is the predicted outlet NOx concentration at time k, is the actual outlet NOx concentration of the SCR denitrification system at time k, i represents the i-th prediction time point, s ₁ is the prediction step length, and r is the correction coefficient; An objective function value calculation submodule, for calculating the first objective function value and the second objective function value of each individual according to the predicted outlet NOx concentration, the predicted outlet ammonia slip amount and the corrected predicted outlet NOx concentration at each prediction time point within the prediction time period corresponding to each individual, using the first objective function and the second objective function respectively; The optimal individual determination submodule is used to determine the individual with the smallest first objective function among the individuals in the population that satisfy the second objective function as the optimal individual of the Lth iteration. When the first objective function value of the optimal individual of the Lth iteration is greater than or equal to the first objective function value of the global optimal individual of the L-1th iteration, the optimal individual of the Lth iteration is set as the global optimal individual of the Lth iteration; otherwise, the global optimal individual of the L-1th iteration is set as the global optimal individual of the Lth iteration. A second judgment submodule is used to judge whether the number of iterations is greater than an iteration number threshold, and obtain a second judgment result; A population update submodule, for, if the second judgment result indicates no, increasing the value of the number of iterations by 1, updating the population by adopting the inheritance, mutation and recombination methods in the genetic algorithm, and returning to the step of "judging whether the valve opening of the SCR denitration system corresponding to the ammonia injection amount at each control time point of each individual in the population is between the lower limit and the upper limit of the valve opening, and obtaining the first judgment result"; The optimal ammonia injection amount output submodule is used to output the global optimal individual of the Lth iteration as the optimal ammonia injection amount if the second judgment result indicates yes. 6. The control system of the SCR denitrification system of a full-operating power station according to claim 5 is characterized in that the neural network model includes an input layer, an output layer and a hidden layer, the input layer includes 4 neurons, the output layer includes 2 neurons, and the output layer includes 5 neurons. 7. The control system of the SCR denitration system of a power station under full working conditions according to claim 5, characterized in that the first objective function is: Where J ₁ represents the first objective function value, is the predicted outlet ammonia slip at _time k+i·s1, _M2 is the liquid ammonia price, P is the number of predicted time points in the prediction time period, Q _gas is the flue gas flow rate, is the oxygen content of flue gas, _M1 is the sewage fee price, is the amount of ammonia injection at _time k+j·s2, j is the jth control time point, _s2 is the control step, M is the number of control time points in the prediction time period, N is the power generation of the unit, _M3 is the electricity price subsidy price, _ω1 is the first weight coefficient, and _ω2 is the second weight coefficient. 8. The control system of the SCR denitration system of a power station under full working conditions according to claim 5, characterized in that the second objective function is: Among them, J ₂ is the second objective function value, P is the number of prediction time points in the prediction time period, r(k+i·s ₁ ) is the expected value of the outlet NO _x concentration at moment k+i·s ₁ , ||Δu(k+j·s ₂ )|| is the difference between the ammonia injection amount at moment k+j·s ₂ and the expected ammonia injection amount, j is the jth control time point, s ₂ is the control step size, M is the number of control time points in the prediction time period, and ω ₃ is the third weight coefficient.