CN118286838B - Combustion-pollutant control full process intelligent regulation and control pollution reduction and carbon reduction method and system - Google Patents

Abstract

The invention discloses a method and a system for intelligently regulating, controlling, reducing pollution and reducing carbon in a whole flow of combustion-pollutant treatment, wherein the control method comprises the following steps: based on accurate prediction models of all controlled variables of a boiler combustion system of a coal-electricity/thermoelectric unit, predicted values of the controlled variables are obtained in advance, a control strategy of the boiler combustion system is formulated by combining the predicted values, key equipment adjustment of the combustion system is performed in advance, adjustment lag time is shortened, unit steam coal consumption and emission concentration of various pollutants are reduced, and pollution and carbon reduction of a source are realized; based on an accurate prediction model of a target controlled variable of the pollutant removal system, the change of the concentration of various pollutants at a boiler outlet is predicted in advance, the intelligent accurate regulation and control of the multi-device key parameter steps of the whole-flow multi-device in the boiler combustion-flue gas treatment process are combined, the ultra-high precision clamping control of the pollutant emission concentration under variable load/fuel is realized, the material consumption of a pollutant removal device is reduced in the whole-flow manner, and the effective reduction of CO ₂ emission is realized cooperatively.

Description

Combustion-pollutant control full process intelligent regulation and control pollution reduction and carbon reduction method and system

Technical Field

The present invention relates to the field of energy conservation and environmental protection, and in particular to a method and system for intelligently controlling pollution reduction and carbon reduction in the entire combustion-pollutant control process.

Background Art

In the ultra-low emission system, a certain amount of energy and materials need to be invested to efficiently remove pollutants from the flue gas. However, this will also lead to an increase in the operating energy consumption and operating costs of the ultra-low emission system. According to statistics, the plant power consumption rate of the ultra-low emission system can reach more than 2%. With the large-scale access to renewable energy such as wind energy and solar energy, and the blending of zero-carbon/low-carbon complex fuels, the load fluctuations of coal-fired units are frequent (30% to 100%), which puts forward new requirements for the stability, economy and adjustability of the operation of the ultra-low emission system. Therefore, how to achieve efficient and stable ultra-low emissions of coal-fired units under complex fuel blending and deep peak regulation, and reduce the energy consumption and material consumption of the system, and improve the stability, economy and reliability of the ultra-low emission system are problems that need to be solved urgently.

However, the conventional pollution reduction and carbon reduction control methods of existing boiler combustion and pollutant treatment systems are generally controlled by DCS systems, supplemented by manual control. After the operating conditions change, it is difficult to make timely adjustments and responses. The pollutant concentration fluctuates greatly under variable load/fuel conditions, which easily leads to large energy and material consumption. Manual adjustments by operators have slow response speed and delayed adjustment, and high requirements for operators (such as system familiarity, operating experience, etc.); and the control parameters are strongly coupled. After the operating conditions change, other parameters are not optimized and adjusted accordingly, and the overall changes in boiler operation are not considered, which is easy to cause energy waste and pollutant emissions exceeding the standard; at the same time, the operating parameters and outlet concentrations of pollutant removal equipment are affected by load fluctuations and changes in coal quality, and the level of manual control is limited. Sometimes, in order to meet ultra-low emission requirements, excessive materials are added or the operating power is increased during equipment operation, resulting in excessive investment costs, which is easy to cause resource waste and secondary pollution.

To this end, the present invention constructs an operation monitoring and database covering the boiler-ultra-low emission system, and establishes an intelligent control system with a four-layer structure including a boiler combustion-pollutant treatment system layer, a knowledge-data coupling modeling layer, a model parameter identification and optimization layer, and a cascade intelligent and precise control layer. A carbon pollution source reduction-ultra-low emission system intelligent control method of "model construction-global optimization-advanced control" is proposed, and a comprehensive platform for the intelligent control system for the coordinated removal of multiple pollutants in the entire process of combustion-flue gas pollutant removal is constructed. While achieving efficient and stable removal of flue gas pollutants, the energy consumption and material consumption of the system are reduced.

Summary of the invention

The present invention provides a method and system for pollution reduction and carbon reduction through intelligent control of the entire combustion-pollutant treatment process. Based on the idea of intelligent and precise control of multiple systems and multiple devices in the entire boiler combustion-flue gas treatment process, a "model construction-global optimization-advanced control" method for carbon pollution source reduction-ultra-low emission system intelligent control is proposed, which realizes the early and precise prediction of the target controlled variables of the boiler combustion system and the flue gas multi-pollutant removal system-precise control of key parameters and coordinated optimization of pollution reduction and carbon reduction in the entire process. Among them, based on the accurate prediction model of each controlled variable of the boiler combustion system, the predicted values of the controlled variables are obtained in advance, and the control strategy of the boiler combustion system is formulated based on the predicted values, and the key equipment of the combustion system is adjusted in advance, shortening the adjustment lag time, reducing unit steam coal consumption, and reducing pollutant emission concentrations, thereby achieving pollution reduction and carbon reduction at the source; based on the accurate prediction model of the target controlled variables of the pollutant removal system, the changes in pollutant concentration at the boiler outlet are predicted in advance, and the key parameters of multiple devices in the entire process of boiler combustion-flue gas treatment are intelligently and accurately controlled in a step-by-step manner to achieve ultra-high precision edge control of pollutant emission concentrations under variable loads/fuels, reduce the material consumption and energy consumption of pollutant removal devices throughout the entire process, and synergistically achieve an effective reduction in _CO2 emissions.

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

A combustion-pollutant control full-process intelligent control pollution reduction and carbon reduction system, the system comprising: a boiler combustion-pollutant control system layer, a knowledge-data coupling modeling layer, a model parameter identification and optimization layer, and a cascade intelligent and precise control layer;

The process of the combustion-pollutant control full-process intelligent regulation pollution reduction and carbon reduction system is as follows: at the knowledge-data coupling modeling layer, by using the pollutant generation and removal mechanism of the full-process device in the boiler combustion-pollutant control system layer, a generation and cascade removal mechanism model of pollutant concentrations in multiple sections of the boiler combustion-pollutant control full-process is established, and combined with the historical operation data and system operation prior knowledge in the boiler combustion-pollutant control system layer, a knowledge-data fusion-driven multi-section pollutant concentration prediction model is further constructed;

In the model parameter identification and optimization layer, the target controlled variable is used as the target of intelligent regulation, and an optimization problem is constructed. Then, according to the characteristics of the optimization problem, the PSO algorithm, WOA algorithm, particle swarm-gradient descent algorithm, and enumeration algorithm are used to identify and optimize the model parameters. At the same time, the optimization parameters are optimized and solved by combining offline mining and online iteration, so as to establish a key parameter prediction model driven by the process mechanism of boiler combustion-flue gas treatment process and multi-condition segmented machine learning;

Through the cascade intelligent precise control layer, the pollutant concentrations at the boiler outlet and multiple sections of the pollutant removal system are accurately predicted. Through the boiler low-carbon/zero-carbon fuel blending amount under different operating conditions, the frequency of boiler section fans (primary fans, secondary fans, induced draft fans), the ammonia injection amount in different areas of the denitrification device, the secondary voltage of different types of power supplies in different electric fields of the electrostatic precipitator, the different slurry circulating pump combinations of the wet flue gas desulfurization device and their circulating pump frequencies and slurry pH key control parameters, the cascade emission reduction effects of multiple pollutants such as particulate matter, _SO3 , and heavy metals along the entire flue gas process are studied. The corresponding relationship between the pollutant emission concentration at the outlet of different pollutant treatment devices and the key control parameters is obtained, and a pollutant removal multi-device control strategy with multi-objective coordination of energy consumption, material consumption, and pollutant emissions is established to achieve cascade optimization control of pollutants in the entire process and coordinated energy conservation and carbon reduction.

Preferably, the boiler combustion-pollutant treatment system layer is composed of a low-carbon/zero-carbon fuel blending system, a boiler combustion system, a denitrification device, an electrostatic precipitator/low-temperature electrostatic precipitator/electric bag composite precipitator, a wet desulfurization device, and a wet electrostatic precipitator;

Preferably, the cascade intelligent precision control layer is composed of a boiler combustion prediction control module, a denitration device prediction control module, an electrostatic precipitator prediction control module, a wet desulfurization device prediction control module, a wet electrostatic precipitator prediction control module and a full-process cascade collaborative optimization control module;

Preferably, the controlled variables of the boiler combustion system include: boiler oxygen content, furnace outlet pressure and outlet pollutant concentration; the controlled variables of the flue gas treatment device include total outlet particulate matter concentration, total outlet _SO2 concentration/desulfurization device outlet _SO2 concentration, total outlet/denitrification device outlet _NOx concentration and ammonia slip concentration, total outlet _SO3 concentration.

A method for reducing pollution and carbon emissions through intelligent regulation of the entire combustion-pollutant control process, the method comprising:

Based on the pollutant generation and removal mechanism, the operating parameters that affect the target controlled variables are analyzed and determined. The influencing parameters of the target controlled variables are used as input variables, and the target controlled variables are used as output variables. A prediction model for each controlled variable of the boiler combustion-flue gas treatment system is constructed, and each controlled variable is predicted separately to obtain the predicted value of each controlled variable in advance. According to the prediction results, a control module of the target controlled variable is constructed, and a control strategy for the boiler combustion system and the flue gas treatment device is formulated. The key influencing parameters that affect the target controlled variables are adjusted in advance to achieve accurate and stable control of the target controlled variables while reducing pollution and carbon emissions throughout the entire process.

Preferably, the boiler oxygen content, furnace outlet pressure and NOx concentration collaborative prediction control process is as follows:

Step Sa1, using the coal feed rate, primary fan frequency, and secondary fan frequency as input variables, and the oxygen content as the output variable, using the step response vector of the input variables to predict the change of the oxygen content in real time, and establishing an oxygen content concentration prediction model;

The oxygen content concentration prediction model expression is as follows:

Where: is the uncorrected total predicted value of oxygen content for the future P time domain lengths at time k; are the uncorrected total predicted values of oxygen content in P time domain lengths under the influence of coal feed rate, primary fan frequency and secondary fan frequency, and P is the rolling optimization time domain length; U(k-1) is the value of each variable in N time domain lengths before time k, and N is the model time domain; ΔU(k) is the control increment predicted value of each variable in the k time domain length for the future M moments, and M is the control time domain length; A ₀ and A are the dynamic matrices of each variable, which describe the influence of each input variable on the system response.

Step Sa2: To correct the errors caused by model mismatch and environmental interference, the module prediction value is corrected using real-time information. The feedback correction process is as follows:

Where: _Yc (k) is the corrected total predicted value at time k; H is the feedback correction coefficient; y(k) and _yc (k) are the actual measured value and predicted value of the oxygen content at time k, respectively.

Step Sa3, minimize the difference between the oxygen content target value and the predicted value to optimize the control, and the optimized performance index at time k is expressed in vector form as follows:

J(k)＝[Y _c (k)-Y _r (k)] ^T Q[Y _c (k)-Y _r (k)]+ΔU(k) ^T RΔU(k)

Where: J(k) is the optimization objective function; Y _r (k) is the target controlled variable control target value; Q and R are the target controlled variable prediction error weight matrix and the key parameter control weight matrix respectively.

make:

Step Sa4: Determine the control increment of the secondary fan according to the target value of oxygen content to achieve stable control of oxygen content, and perform real-time feedback correction according to the measured value of oxygen content. The output optimized control increment expression is:

ΔU(k)＝(A ^T QA+R) ^-1 A ^T Q{Y _r (k)-A ₀ U(k-1)-H[y(k)-y _c (k)]}

In step Sa5, the future frequency control instruction of the secondary fan outputted in step S4 is used as the input of the furnace negative pressure control module; under dynamic conditions, a linear regression model is established to analyze the relationship between the frequency change of the primary fan, the frequency change of the secondary fan and the frequency of the induced draft fan, and the frequency of the induced draft fan is adjusted in advance; under static conditions, the least squares method is used to identify the parameters of the PID controller, the initial parameter value of the PID controller is determined, and the control instruction of the induced draft fan is given in advance to ensure the stability of the furnace negative pressure.

Further preferably, a PSO or WOA optimization algorithm is used to determine the step response parameters of the prediction control module for identification.

Preferably, the denitration outlet NOx concentration and ammonia escape concentration prediction and control process is as follows:

Step Sb1, using boiler load, coal feed rate, air volume, and flue gas temperature as input variables, and using the NOx concentration at the furnace outlet, i.e., the denitration device inlet, as the output variable, to establish a NOx concentration prediction model based on the zoning and segmentation denitration device inlet;

Further preferably, the furnace outlet NOx concentration, i.e., the denitration prediction model expression is as follows:

Where: is the uncorrected total predicted value of the furnace outlet NOx concentration for the future P time domain lengths at time k; are the uncorrected predicted values of the NOx concentration at the furnace outlet under the influence of coal feed rate, air volume and flue gas temperature for P time domain lengths, and P is the rolling optimization time domain length; U(k-1) is the value of each variable N time domain lengths before time k, and N is the model time domain; ΔU(k) is the control increment predicted value of each variable in the k time domain length for the future M moments, and M is the control time domain length; A ₀ and A are the dynamic matrices of each variable, which describe the influence of each input variable on the system response.

Step Sb2, in order to correct the errors caused by model mismatch and environmental interference, the module prediction value is corrected using real-time information. The feedback correction process is as follows:

In the formula: Y _N (k) is the corrected total predicted value at time k; H is the feedback correction coefficient; y(k) and y _N (k) are the measured value and predicted value of the current furnace outlet NOx concentration at time k, i.e., the NOx concentration at the inlet of the denitrification device.

Step Sb3, in order to correct the stability deviation of the NOx concentration model prediction at the denitration device inlet, the difference between the predicted result of the NOx concentration at the denitration device inlet in a certain period in the future and the controlled target is calculated, and a specific coefficient is used for correction to achieve real-time and accurate prediction of the NOx concentration at the furnace outlet, i.e., the NOx concentration at the denitration device inlet;

Step Sb4, using boiler load, denitration area operating temperature, ammonia injection flow rate and denitration device inlet NOx concentration as input variables, and denitration device outlet NOx concentration as output variable, to establish a denitration device outlet NOx concentration prediction control model based on zoning and segmentation;

Step Sb5, adding the predicted value of the inlet NOx concentration, i.e., the predicted value of the furnace outlet NOx as a feedforward forecast to the full-operating-condition multi-parameter denitrification device prediction control module, and then outputting the optimized set value of the ammonia injection flow rate, the opening value of the ammonia injection regulating valve is calculated by the deviation between the measured value of the ammonia injection flow rate and the optimized value through the intelligent advanced controller, and a full-operating-condition multi-parameter coordination-cascade intelligent advanced control strategy is formulated to achieve stable control of the _NOx concentration and ammonia slip at the denitrification outlet, and perform real-time feedback correction according to the actual measured values of the _NOx concentration and ammonia slip at the denitrification outlet.

Preferably, the total outlet particulate matter and _SO3 concentration prediction and control process is as follows:

Step Sc1, based on the generation mechanism of particulate matter and SO ₃ in the boiler combustion process and the sorting and particle size distribution characteristics in the slag and flue gas, a flue gas particulate matter and SO ₃ generation concentration prediction model is established to realize the prediction of the particulate matter and SO ₃ concentration and mass distribution at the inlet of the dust removal device;

Step Sc2: Based on the electric field corona discharge mechanism in the electrostatic precipitator and the charged migration mechanism of particulate matter and SO ₃ , the turbulence and reflux multi-physics field enhanced particulate matter and SO ₃ capture mechanism in the wet flue gas desulfurization device, and the enhanced mechanism of condensation, agglomeration, charging and migration of fine particles and SO ₃ in the wet electrostatic precipitator, a mechanism model for predicting the concentration and mass distribution of particulate matter and SO 3 removal in the whole process of the inlet and outlet of the dust removal device, the wet flue gas desulfurization device and the wet electrostatic precipitator is constructed;

Step Sc3: based on the actual operation data, the mechanism model for predicting the concentration and mass distribution of particulate matter and _SO3 removal in the whole process established in step Sc1 and step Sc2 is revised, and the mechanism model for predicting the concentration of particulate matter and _SO3 driven by the process mechanism and multi-condition segmented machine learning is revised;

Preferably, the data correction model construction method includes a parameter identification method based on gradient descent + particle swarm algorithm and a long short-term memory neural network algorithm based on attention mechanism;

Step Sc4: Based on the process mechanism established in the above steps and the particulate matter and _SO3 concentration prediction model driven by the collaborative work of multi-operating condition segmented machine learning, the energy consumption, total outlet particulate matter and SO3 concentration of the electrostatic precipitator/low-temperature electrostatic precipitator/electric bag composite precipitator and wet electrostatic precipitator are used as the targets of intelligent control, and the energy consumption and particulate matter and _{SO3 concentration} emission optimization problem is constructed. According to the characteristics of the optimization problem, the particle swarm algorithm or particle swarm-gradient descent algorithm is used to solve it, so as to obtain the optimal secondary voltage setting method of the electrostatic precipitator/low-temperature electrostatic precipitator/electric bag composite precipitator and wet electrostatic precipitator under different operating conditions, and construct an intelligent control strategy for different types of power supplies in the electric field and chamber of the particulate matter electrostatic removal device.

Further preferably, the coal-fired flue gas particulate matter/SO ₃ generation-removal model involves a denitrification device, an electrostatic precipitator/low-temperature electrostatic precipitator/electric bag composite dust removal device, a wet electrostatic precipitator, and a wet desulfurization coordinated dust removal device;

Further preferably, the energy consumption optimization problem is expressed as follows:

Where n is the total number of electric fields of the electrostatic precipitator/low-temperature electrostatic precipitator/electric bag composite precipitator; _Wf is the power of the fth electric field; m is the total number of electric fields of the wet electrostatic precipitator; _Wi is the power of the i-th electric field; and c _limit are the predicted value and limit of the total outlet particulate matter/ _SO3 emission concentration; U _f,min and U _f,max are the minimum secondary voltage and maximum secondary voltage of the f-th electric field; U _i,min and U _i,max are the minimum secondary voltage and maximum secondary voltage of the i-th electric field.

Further optimization, in the process of building an intelligent control strategy, in order to ensure the stability and convergence of the algorithm, it is necessary to predict the changing trend of the controlled variables over a period of time, so as to implement a rolling optimization strategy. In addition, it is also necessary to perform online correction of the prediction model according to the actual operation conditions on site. The specific steps are as follows:

Step Sc401, setting the constraint condition of the outlet pollutant concentration particulate matter/SO ₃ ;

Step Sc402: To ensure the relative stability of the outlet pollutant concentration and to cope with the rapid mutation of the working conditions, the error between the reference trajectory and the predicted value is added to the optimization target. The reference trajectory setting needs to be smaller than the outlet concentration limit in the above formula.

Step Sc403: Based on step Sc402, to ensure that the control process can converge to the vicinity of the target value, a terminal error needs to be added to the optimization target;

Step Sc404: In the feedback correction module of conventional model predictive control, the model is usually corrected by taking the actual value of the outlet particulate matter concentration measured at the next moment as the model error. In this paper, since the outlet pollutant concentration mark measurement value itself has a certain amount of noise, the sliding average error is used as the model error, where the sliding average window length is the same as the prediction time domain.

Preferably, the total outlet _SO2 concentration prediction and control process is as follows:

Step Sd1, based on the generation of SO ₂ in the boiler combustion process and the SO ₂ /SO ₃ conversion mechanism in the denitrification device, a prediction model for the generation concentration of SO ₂ in the flue gas at the inlet of the desulfurization device is established based on a machine learning algorithm;

Step Sd2, based on the principle of multi-absorbent SO ₂ removal process in the desulfurization device, a multi-absorbent SO ₂ removal process mechanism model is constructed;

Step Sd3: Based on historical operation data, the PSO algorithm is used to identify the correction parameters of the coal-fired flue gas _SO2 generation-removal mechanism model, and further a _SO2 removal process data correction model is constructed based on the LSTM network;

Step Sd4: Based on the coal-fired flue gas _SO2 generation-removal model established in the above steps, the outlet _SO2 concentration and the desulfurization operation pH are used as the targets of intelligent control, and an optimization problem is constructed. Then, according to the characteristics of the optimization problem, the enumeration method, particle swarm algorithm or particle swarm-gradient descent algorithm are used to solve it, so as to obtain the optimal circulation pump and its frequency adjustment strategy of the wet desulfurization device under different operating conditions;

Further preferably, the desulfurization device optimization problem is expressed as follows:

min cost(s _A ,s _B ,s _C ,s _D ...s _n )＝s _A p _A +s _B p _B +s _C p _C +s _D p _D +...+s _n p _n

Wherein, s _A , s _B , s _C , s _D ... s _E are the operating status of the circulation pumps A, B, C, D, n. When the circulation pump is turned on, the operating status is 1; when the circulation pump is turned off, the operating status is 0. _{p A} , p _B , p _C , p _D ... p _n are the rated powers of the circulation pumps A, B, C, D ... n. _{c outlet} is the actual outlet SO ₂ concentration, c _{outlet, target} is the target outlet SO ₂ concentration.

Further optimization, in the process of building an intelligent control strategy, in order to ensure the stability and convergence of the algorithm, it is necessary to predict the changing trend of the controlled variables over a period of time, so as to implement a rolling optimization strategy. In addition, it is also necessary to perform online correction of the prediction model according to the actual operation conditions on site. The specific steps are as follows:

Step Sc401, set the constraint condition of outlet pollutant concentration SO ₂ , expressed as follows:

Among them, r _p,i is the number of circulating pumps and the frequency weight coefficient; p(t+i) is the sum of the slurry volume of each circulating pump; r _e,j is the excess weight coefficient; sgn(·) is the sign function; w _limit (t+j) is the upper limit of the outlet pollutant concentration SO ₂ emission; M is the control time domain; P is the prediction time domain;

Step Sc402: To ensure the relative stability of the outlet pollutant concentration and to cope with the rapid mutation of the working conditions, the error between the reference trajectory and the predicted value is added to the optimization target. The reference trajectory setting needs to be smaller than the outlet concentration limit in the above formula, which is expressed as follows:

Among them, q _j is the tracking weight coefficient; w _t (t+j) is the emission target;

Step Sc403: Based on step Sc402, in order to ensure that the control process can converge to the vicinity of the target value, the terminal error needs to be added to the optimization target. Therefore, the final rolling optimization problem can be expressed as follows:

Among them, min J(t) is the objective function at time t; q _j is the tracking weight coefficient; w _t (t+j) is the emission target; r _p,i is the number of circulating pumps and the frequency weight coefficient; p(t+i) is the sum of the slurry volume of each circulating pump; _re,j is the excess weight coefficient; sgn(·) is the sign function; w _limit (t+j) is the upper limit of the outlet pollutant concentration SO ₂ emission; V _f (·) is the terminal error function; χ(P) is the control quantity calculated at the last P moments in the prediction time domain; χ _s is the control quantity after steady-state optimization; y _s is the outlet concentration after steady-state optimization.

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

(1) Based on the accurate prediction model of each controlled variable of the boiler combustion system, the predicted values of the controlled variables are obtained in advance, and the control strategy of the boiler combustion system is formulated based on the predicted values. The key equipment of the combustion system is adjusted in advance, the adjustment lag time is shortened, and the unit power generation/steam coal consumption is reduced to achieve source pollution reduction and carbon reduction; based on the accurate prediction model of the target controlled variables of the pollutant removal system, the change of pollutant concentration at the boiler outlet is predicted in advance, and the key parameters of multiple devices in the whole process of boiler combustion-flue gas treatment are intelligently and accurately controlled in a step-by-step manner to achieve ultra-high precision edge control of pollutant emission concentration under variable load/fuel, reduce the material consumption and energy consumption of the pollutant removal device in the whole process, and synergistically achieve an effective reduction in _CO2 emissions.

(2) The present invention reduces unit steam coal consumption and pollutant emissions through the whole process of coupling the key parameter regulation of the boiler combustion system with the key parameter regulation of the pollutant removal device, achieving a unit steam/kWh coal consumption reduction of more than 1.5%, a total outlet particulate matter concentration fluctuation of ≤±0.2mg/ ^m3 , a _SO2 concentration fluctuation of ≤±0.5mg/ ^m3 , and a _NOx concentration fluctuation of ≤±0.5mg/ ^m3 , achieving ultra-high precision edge control of pollutant emission concentration under variable load and variable fuel conditions, and collaboratively achieving _SO3 emission concentration of ≤1mg/ ^m3 and total emission concentration of five heavy metals of mercury/arsenic/lead/cadmium/chromium of <20μg/ ^m3 . At the same time, the overall power consumption of the flue gas treatment system is reduced by more than 20%, and energy conservation and carbon reduction are synergistically achieved. The ammonia consumption is reduced by more than 20%, and the limestone consumption is reduced by more than 5%, achieving an effective reduction in _CO2 emissions, and realizing the cascade pollution reduction and carbon reduction synergistic optimization of the whole process.

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 schematic diagram of a process flow of a combustion-pollutant control full-process intelligent regulation pollution reduction and carbon reduction system provided by an embodiment of the present invention;

FIG2 is a schematic diagram of global optimization of coordinated removal of multiple pollutants in a boiler-ultra-low emission system according to an embodiment of the present invention;

FIG3 is a logic diagram of coordinated control of boiler oxygen content and furnace outlet pressure controlled variables provided by an embodiment of the present invention;

FIG4 is a comparison of the actual operation effects of the collaborative control model provided by an embodiment of the present invention and the original control;

FIG5 is a schematic diagram of a multi-parameter coordinated predictive control strategy for a denitration device under all operating conditions provided by an embodiment of the present invention.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

The purpose of the present invention is to provide a method and system for intelligent control of pollution reduction and carbon reduction in the entire process of combustion-pollutant control. Through the prediction model of each controlled variable, each controlled variable is predicted respectively, a control strategy for the boiler combustion system is formulated, and equipment adjustments are made in advance, thereby shortening the adjustment lag time and achieving the effect of energy conservation and emission reduction.

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.

Embodiment 1

As shown in Figures 1 and 2, the present invention provides a combustion-pollutant control full-process intelligent control pollution reduction and carbon reduction system, the system comprising: a boiler combustion-pollutant control system layer, a knowledge-data coupling modeling layer, a model parameter identification and optimization layer, and a cascade intelligent and precise control layer;

The process of the combustion-pollutant control full-process intelligent regulation pollution reduction and carbon reduction system is as follows: at the knowledge-data coupling modeling layer, by using the pollutant generation and removal mechanism of the full-process device in the boiler combustion-pollutant control system layer, a generation and cascade removal mechanism model of pollutant concentrations in multiple sections of the boiler combustion-pollutant control full-process is established, and combined with historical operation data and advanced experience knowledge in the boiler combustion-pollutant control system layer, a knowledge-data fusion-driven multi-section pollutant concentration prediction model is further constructed;

In the model parameter identification and optimization layer, the target controlled variable is used as the target of intelligent regulation, and an optimization problem is constructed. Then, according to the characteristics of the optimization problem, the PSO algorithm, WOA algorithm, particle swarm-gradient descent algorithm, and enumeration algorithm are used to identify and optimize the model parameters. At the same time, the optimization parameters are optimized and solved by combining offline mining and online iteration, so as to establish a key parameter prediction model driven by the process mechanism of boiler combustion-flue gas treatment process and multi-condition segmented machine learning;

Through the cascade intelligent precise control layer, the pollutant concentrations at the boiler outlet and multiple sections of the pollutant removal system are accurately predicted. By analyzing the boiler low-carbon/zero-carbon fuel blending amount under different operating conditions (high load, medium load, low load, sudden load increase, sudden load decrease, etc.), the boiler section fan (primary fan, secondary fan, induced draft fan) frequency, the ammonia injection amount in different areas of the denitrification device, the secondary voltage of different types of power supplies in different electric fields of the electrostatic precipitator, the different slurry circulating pump combinations of the wet flue gas desulfurization device and their circulating pump frequencies and slurry pH key control parameters, the cascade emission reduction effects of multiple pollutants such as particulate matter, _SO3 , and heavy metals along the entire flue gas process are analyzed. The corresponding relationship between the pollutant emission concentration at the outlet of different pollutant treatment devices and the key control parameters is obtained, and a pollutant removal multi-device control strategy with multi-objective coordination of energy consumption, material consumption, and pollutant emissions is established to achieve cascade optimization control of pollutants in the entire process and coordinated energy conservation and carbon reduction.

The boiler combustion-pollutant treatment system layer is composed of a low-carbon/zero-carbon fuel blending system, a boiler combustion system, a denitrification device, an electrostatic precipitator/low-temperature electrostatic precipitator/electric bag composite precipitator, a wet desulfurization device, and a wet electrostatic precipitator;

The cascade intelligent precision control layer is composed of a boiler combustion prediction control module, a denitrification device prediction control module, an electrostatic precipitator prediction control module, a wet desulfurization device prediction control module, a wet electrostatic precipitator prediction control module and a full-process cascade collaborative optimization control module;

The controlled variables of the boiler combustion system include: boiler oxygen content, furnace outlet pressure and NOx concentration; the controlled variables of the flue gas treatment device include total outlet particulate matter concentration, total outlet _SO2 concentration, i.e., _SO2 concentration at the outlet of the desulfurization device, total outlet, i.e., _NOx concentration and ammonia escape concentration at the outlet of the denitrification device, and total outlet _SO3 concentration.

Embodiment 2

The present invention also provides a method for intelligently regulating pollution reduction and carbon reduction in the entire combustion-pollutant treatment process, the method comprising:

Based on the pollutant generation and removal mechanism, the operating parameters that affect the target controlled variables are analyzed and determined, and the influencing parameters of the target controlled variables are used as input variables, and the target controlled variables are used as output variables to construct prediction models for each controlled variable of the boiler combustion-flue gas treatment system; each controlled variable is predicted separately, and the predicted value of each controlled variable is obtained in advance. According to the prediction results, the control module of the target controlled variable is constructed, and the control strategy of the boiler combustion system and the flue gas treatment device is formulated. The key influencing parameters that affect the target controlled variables are adjusted in advance to achieve accurate and stable control of the target controlled variables. At the same time, the ultra-low emission system of a coal-fired power plant usually includes a boiler combustion system, a catalytic denitrification device SCR, an electrostatic precipitator ESP, a wet desulfurization device WFGD, and a wet electrostatic precipitator WESP removal equipment. During operation, each system has different degrees of synergistic effects on NOx, SO ₂ , PM, SO ₃ and Hg pollutants. Therefore, on the basis of considering the synergistic removal effect, a boiler combustion-pollutant removal system control strategy with multi-objective coordination of energy consumption, material consumption and pollutant emissions was established to achieve coordinated optimization of the entire operation process of the ultra-low emission system and provide guidance for the key parameter operation of key equipment for boiler combustion and pollutant removal.

As shown in FIG2 , as a preferred implementation method, the implementation process of predictive control of oxygen content, furnace outlet pressure and NOx of a circulating fluidized bed boiler is as follows:

Step Sa1, using the coal feed rate, primary fan frequency, and secondary fan frequency as input variables, and the boiler oxygen content as the output variable, using the step response vector of the input variables to predict the change of oxygen content in real time, and establishing an oxygen content concentration prediction model;

The oxygen content concentration prediction model expression is as follows:

Where: is the uncorrected total predicted value of oxygen content for the future P time domain lengths at time k; are the uncorrected total predicted values of oxygen content in P time domain lengths under the influence of coal feed rate, primary fan frequency and secondary fan frequency, and P is the rolling optimization time domain length; U(k-1) is the value of each variable in N time domain lengths before time k, and N is the model time domain; ΔU(k) is the control increment predicted value of each variable in the k time domain length for the future M moments, and M is the control time domain length; A ₀ and A are the dynamic matrices of each variable, which describe the influence of each input variable on the system response.

Step Sa2, in order to correct the errors caused by model mismatch and environmental interference, the module prediction value is corrected using real-time information. The feedback correction process is as follows:

Where: _Yc (k) is the corrected total predicted value at time k; H is the feedback correction coefficient; y(k) and _yc (k) are the actual measured value and predicted value of the oxygen content at time k, respectively.

Step Sa3, minimize the difference between the oxygen content target value and the predicted value to optimize the control, and the optimized performance index at time k is expressed in vector form as follows:

J(k)＝[Y _c (k)-Y _r (k)] ^T Q[Y _c (k)-Y _r (k)]+ΔU(k) ^T RΔU(k) (6)

Where: J(k) is the optimization objective function; Y _r (k) is the target controlled variable control target value; Q and R are the target controlled variable prediction error weight matrix and the key parameter control weight matrix respectively.

make:

Step Sa4, determine the control increment of the secondary fan according to the oxygen content target value, realize the stable control of the oxygen content, and perform real-time feedback correction according to the measured value of the oxygen content, and output the optimized control increment expression as follows:

ΔU(k)＝(A ^T QA+R) ^-1 A ^T Q{Y _r (k)-A ₀ U(k-1)-H[y(k)-y _c (k)]} (8)

In step Sa5, the future frequency control instruction of the secondary fan outputted in step S4 is used as the input of the furnace negative pressure control module; under dynamic conditions, a linear regression model is established to analyze the relationship between the frequency change of the primary fan, the frequency change of the secondary fan and the frequency of the induced draft fan, and the frequency of the induced draft fan is adjusted in advance; under static conditions, the least squares method is used to identify the parameters of the PID controller, the initial parameter value of the PID controller is determined, and the control instruction of the induced draft fan is given in advance to ensure the stability of the furnace negative pressure.

Under steady-state conditions, when the output furnace outlet pressure is within the set range, there is no need to adjust the induced draft fan frequency. When the furnace outlet pressure exceeds the set range, the induced draft fan frequency is corrected through PID. The correction formula is as follows:

Where: u(k) is the change in the frequency of the induced draft fan output by the PID controller; e(k) is the difference between the target value and the measured value of the furnace outlet pressure.

Under dynamic conditions, the frequency relationship between the primary fan, secondary fan and induced draft fan is analyzed by fitting to determine the frequency control increment of the induced draft fan. The fitting formula is as follows:

ΔL ₌ 0555 18ΔL _primary + 0.602 37ΔL _secondary + -0001 35L ₊ 0.013 63

In the formula: _ΔLprimary is the change in primary fan frequency; _ΔLsecondary is the change in secondary fan frequency; _{Linduced is} the measured value of induced draft fan frequency.

As shown in Figure 3, it is a comparison chart of the actual operation effect of the coordinated control based on oxygen content and furnace outlet pressure and the original control. Under the original control, the fluctuation range of flue gas oxygen content is 1.8% to 3.0%, and the distribution is relatively discrete; under the coordinated predictive control of the present invention, the fluctuation range of oxygen content is 2.3% to 2.8%, and the distribution is more concentrated. The deviation between the actual measured value of oxygen content and the set target value is within the range of ±0.25%.

The boiler predictive control module is used to predict the secondary fan frequency control amount that will be output in the future, and the coordinated control of the induced draft fan frequency is realized in advance in combination with the actual operating conditions of the boiler. Under the original control, the furnace outlet pressure fluctuation range is -180Pa～+105Pa, and the standard deviation is 44.61Pa; under the collaborative predictive control of the present invention, the furnace outlet pressure fluctuation range is -110Pa～-10Pa, and the standard deviation is 12.86Pa, of which 99% of the measured values of the furnace outlet pressure are within the set target value ±45Pa. Compared with the original control, under the collaborative control, the secondary fan and the induced draft fan have the characteristics of small amplitude and multi-frequency adjustment, the adjustment is more timely, the oxygen content and the furnace outlet pressure fluctuation amplitude are reduced, and the current surge problem caused by the large adjustment of the fan can also be avoided; under the collaborative control, the original content of the furnace outlet NOx concentration, _SO2 , particulate matter concentration, and heavy metal pollutant concentration are further reduced.

Compared with the original control, under the coordinated control, the coal consumption per unit steam output of the boiler is saved by more than 1.6%, and the power consumption of the fan per unit steam output is reduced by more than 2%.

As shown in FIG4 , as a preferred embodiment, the denitration outlet NOx concentration and ammonia slip controlled variable predictive control process is as follows:

Step Sb1, using boiler load, coal feed rate, air volume, and flue gas temperature as input variables, and using the NOx concentration at the furnace outlet, i.e., the denitration device inlet, as the output variable, to establish a NOx concentration prediction model based on the zoning and segmentation denitration device inlet;

The furnace outlet NOx concentration prediction model expression is as follows:

Where: is the uncorrected total predicted value of the furnace outlet NOx concentration for the future P time domain lengths at time k; are the uncorrected predicted values of the NOx concentration at the furnace outlet under the influence of coal feed rate, air volume and flue gas temperature for P time domain lengths, and P is the rolling optimization time domain length; U(k-1) is the value of each variable N time domain lengths before time k, and N is the model time domain; ΔU(k) is the control increment predicted value of each variable in the k time domain length for the future M moments, and M is the control time domain length; A ₀ and A are the dynamic matrices of each variable, which describe the influence of each input variable on the system response.

Step Sb2, using real-time information to correct the module prediction value, the feedback correction process is as follows:

In the formula: Y _N (k) is the corrected total predicted value at time k; H is the feedback correction coefficient; y(k) and y _N (k) are the measured value and predicted value of the current furnace outlet NOx concentration at time k, i.e., the NOx concentration at the inlet of the denitrification device.

Step Sb3: In order to correct the stable deviation of the model prediction of the NOx concentration at the inlet of the denitrification device, the difference between the predicted result of the NOx concentration at the inlet of the denitrification device in a certain period in the future and the controlled target is calculated, and a specific coefficient is used for correction to achieve real-time and accurate prediction of the NOx concentration at the furnace outlet, i.e., the NOx concentration at the inlet of the denitrification device.

Step Sb4, using boiler load, denitrification area operating temperature, ammonia injection flow rate and denitrification device inlet NOx concentration as input variables, and denitrification device outlet NOx concentration as output variable, establish a zone-based segmented denitrification device outlet NOx concentration prediction control model.

Step Sb5, adding the predicted value of the inlet NOx concentration as a feedforward forecast to the denitrification device predictive control module, outputting the optimized set value of the ammonia injection flow rate, and the opening value of the ammonia injection regulating valve is calculated by the advanced controller based on the deviation between the measured value of the ammonia injection flow rate and the optimized value, formulating a full-condition multi-parameter coordination-cascade advanced control strategy to achieve stable control of the _NOx concentration and ammonia slip at the denitrification outlet, and performing real-time feedback correction based on the measured values of the _NOx concentration and ammonia slip at the denitrification outlet.

As a preferred implementation, the total outlet particulate matter and _SO3 concentration prediction and control process is as follows:

Step Sc1: Based on the generation mechanism of particulate matter and SO ₃ in the boiler combustion process and the sorting and particle size distribution characteristics in the slag and flue gas, a flue gas particulate matter and SO ₃ generation concentration prediction model is established to realize the prediction of the particulate matter and SO ₃ concentration and mass distribution at the dust removal device inlet;

Step Sc2: Based on the electric field corona discharge mechanism in the electrostatic precipitator and the charged migration mechanism of particulate matter and SO ₃ , the turbulence and reflux multi-physics field enhanced particulate matter and SO ₃ capture mechanism in the wet flue gas desulfurization device, and the enhanced mechanism of condensation, agglomeration, charging and migration of fine particles and SO ₃ in the wet electrostatic precipitator, a mechanism model for predicting the concentration and mass distribution of particulate matter and SO ₃ removal in the whole process of the inlet and outlet of the dust removal device, the wet flue gas desulfurization device and the wet electrostatic precipitator is constructed;

Step Sc3: based on the actual operation data, the mechanism model for predicting the concentration and mass distribution of particulate matter and _SO3 removal in the whole process established in step Sc1 and step Sc2 is modified to establish a particulate matter and _SO3 concentration prediction model driven by the process mechanism and multi-condition segmented machine learning;

The data correction model construction method includes a parameter identification method based on gradient descent + particle swarm algorithm and a long short-term memory neural network algorithm based on attention mechanism;

Step Sc4: Based on the process mechanism established in the above steps and the particle and _SO3 concentration prediction model driven by the multi-operating condition segmented machine learning, the energy consumption, total outlet particle and SO3 concentration of the electrostatic precipitator/low-temperature electrostatic precipitator/electric bag composite precipitator and wet electrostatic precipitator are used as the targets of intelligent control, and the energy consumption and particle and _{SO3 concentration} emission optimization problem is constructed. Then, according to the characteristics of the optimization problem, the particle swarm algorithm or the particle swarm-gradient descent algorithm is used to solve it, so as to obtain the optimal secondary voltage setting mode of the electrostatic precipitator/low-temperature electrostatic precipitator/electric bag composite precipitator and wet electrostatic precipitator under different operating conditions, and the intelligent control strategy of different types of power supplies in the electric field and chamber area of the particle electrostatic removal device is constructed;

The energy consumption optimization problem is expressed as follows:

Where n is the total number of electric fields of the electrostatic precipitator/low-temperature electrostatic precipitator/electric bag composite precipitator; _Wf is the power of the fth electric field; m is the total number of electric fields of the wet electrostatic precipitator; _Wi is the power of the i-th electric field; and c _limit are the predicted value and limit of the total outlet particulate matter/SO3 emission concentration; U _f,min and U _f,max are the minimum secondary voltage and maximum secondary voltage of the f-th electric field; U _i,min and U _i,max are the minimum secondary voltage and maximum secondary voltage of the ith electric field.

In the process of building intelligent control strategies, in order to ensure the stability and convergence of the algorithm, it is necessary to predict the changing trend of the controlled variables over a period of time and implement a rolling optimization strategy. In addition, it is also necessary to perform online correction of the prediction model based on the actual operating conditions on site.

The comparison results before and after the application of the present invention on the electrostatic precipitator show that: in manual control, in order to ensure that the outlet concentration is stable and up to standard, each electric field basically operates at the highest operating voltage. At this time, the voltage and current of each electric field are controlled by flashover. At this time, not only the overall energy consumption of the electrostatic precipitator is high, but the frequent occurrence of flashover will also cause great harm to the safe operation of the precipitator, and seriously cause damage to the anode plate. In addition, since flashover often faces instantaneous voltage drop and current rise, it will also cause unstable outlet particle concentration. In manual control, the outlet concentration of the precipitator can fluctuate up to ±5mg/m ^3. After adopting the intelligent control method of the present invention, the energy consumption of the precipitator is significantly reduced. Under similar working conditions, the energy consumption is reduced to 200-800kW. At the same time, since the operating voltage is far away from the flashover voltage, under intelligent control, the number of flashovers of each electric field is reduced to 0. At the same time, it also means that the voltage and current of each electric field are more stable, so the outlet concentration is also more stable, only ±2mg/m ³ .

Further application of the full-process cascade collaborative optimization control module of the present invention shows that the power comparison effect before and after optimization shows that the total concentration of particulate matter and sulfur trioxide emissions at the total discharge port is lower than 1 mg/m ³ , and the fluctuation of particulate matter and sulfur trioxide emission concentrations is ≤±0.2 mg/m ^3. At the same time, compared with experience, the operating energy consumption can be reduced by 20% after the application of the present invention, and can be reduced by 42.0% compared with the maximum power operation; because the operating voltage is far from the flashover voltage, under intelligent control, the flashover times of each electric field are reduced to 0, which also means that the voltage and current of each electric field are more stable.

As a preferred implementation, the total outlet _SO2 concentration controlled variable predictive control process is as follows:

Step Sd1, based on the generation of SO ₂ in the boiler combustion process and the SO ₂ /SO ₃ conversion mechanism in the denitrification device, a prediction model for the generation concentration of SO ₂ in the flue gas at the inlet of the desulfurization device is established based on a machine learning algorithm;

Step Sd2, based on the principle of multi-absorbent SO ₂ removal process in the desulfurization device, a multi-absorbent SO ₂ removal process mechanism model is constructed;

Step Sd3: Based on historical operation data, the PSO algorithm is used to identify the correction parameters of the coal-fired flue gas _SO2 generation-removal mechanism model, and further a _SO2 removal process data correction model is constructed based on the LSTM network;

Step Sd4: Based on the coal-fired flue gas _SO2 generation-removal model established in the above steps, the outlet _SO2 concentration and the desulfurization operation pH are used as the targets of intelligent control, and an optimization problem is constructed. Then, according to the characteristics of the optimization problem, the enumeration method, particle swarm algorithm or particle swarm-gradient descent algorithm are used to solve it, so as to obtain the optimal circulation pump and its frequency adjustment strategy of the wet desulfurization device under different operating conditions;

Further preferably, the desulfurization device optimization problem is expressed as follows:

min cost(s _A , s _B , s _C , s _D ...s _n )=s _A p _A +s _B p _B +s _C p _C +s _D p _D +...s _n p _n

Wherein, s _A , s _B , s _C , s _D ... s _E are the operating status of the circulation pumps A, B, C, D, n. When the circulation pump is turned on, the operating status is 1; when the circulation pump is turned off, the operating status is 0. _{p A} , p _B , p _C , p _D ... p _n are the rated powers of the circulation pumps A, B, C, D ... n. _{c outlet} is the actual outlet SO ₂ concentration, c _{outlet, target} is the target outlet SO ₂ concentration.

In the process of building intelligent control strategies, in order to ensure the stability and convergence of the algorithm, it is necessary to predict the changing trend of the controlled variables over a period of time, so as to implement a rolling optimization strategy. In addition, it is also necessary to perform online correction of the prediction model according to the actual operation conditions on site. The specific steps are as follows:

Step Sc401, set the constraint condition of outlet pollutant concentration SO ₂ , expressed as follows:

Among them, r _p,i is the number of circulating pumps and the frequency weight coefficient; p(t+i) is the sum of the slurry volume of each circulating pump; r _e,j is the excess weight coefficient; sgn(·) is the sign function; w _limit (t+kj) is the upper limit of the outlet pollutant concentration SO ₂ emission; M is the control time domain; P is the prediction time domain;

Step Sc402: To ensure the relative stability of the outlet pollutant concentration and to cope with the rapid mutation of the working conditions, the error between the reference trajectory and the predicted value is added to the optimization target. The reference trajectory setting needs to be smaller than the outlet concentration limit in the above formula, which is expressed as follows:

Among them, q _j is the tracking weight coefficient; w _t (t+j) is the emission target;

Step Sc403: Based on step Sc402, in order to ensure that the control process can converge to the vicinity of the target value, the terminal error needs to be added to the optimization target. Therefore, the final rolling optimization problem can be expressed as follows:

Among them, min J(t) is the objective function at time t; q _j is the tracking weight coefficient; w _t (t+j) is the emission target; r _p,i is the number of circulating pumps and the frequency weight coefficient; p(t+i) is the sum of the slurry volume of each circulating pump; _re,j is the excess weight coefficient; sgn(·) is the sign function; w _limit (t+j) is the upper limit of the outlet pollutant concentration SO ₂ emission; V _f (·) is the terminal error function; χ(P) is the control quantity calculated at the last P moments in the prediction time domain; χ _s is the control quantity after steady-state optimization; y _s is the outlet concentration after steady-state optimization.

After the present invention is applied, when the outlet _SO2 concentration target is set to 35mg/ ^m3 , compared with before optimization, the energy consumption can be reduced by about 40%, the _SO2 concentration stably meets the target setting value and the fluctuation is ≤±0.5mg/ ^m3 , and the limestone consumption is reduced by more than 5%.

As a preferred implementation method, based on the idea of multi-device cascade intelligent and precise control of the whole process of boiler combustion-flue gas treatment, a cascade collaborative optimization and variable operating condition precision method of boiler combustion-multi-pollutant removal system is further constructed, and the pollutant concentration distribution of each section and the pollutant removal efficiency distribution of key devices are obtained. Then, the ammonia/urea spray gun valve control instructions of the denitrification device, the multi-field, multi-channel, and multi-type power supply control instructions of the electrostatic precipitator, the limestone slurry circulation pump control instructions of the desulfurization device, and the multi-field, multi-channel, and multi-type power supply control instructions of the wet electrostatic precipitator are given in advance along the flue gas process. Through the coupling of the key parameter control of the boiler combustion system with the key parameter control of the pollutant removal device, the unit steam coal consumption and pollutant emissions are reduced in the whole process, and the high-precision edge control of the pollutant emission concentration under variable load and variable fuel conditions is realized, and the SO ₃ emission concentration is ≤1mg/m ³ and the total emission concentration of five heavy metals of mercury/arsenic/lead/cadmium/chromium is <20μg/m ^3. At the same time, the overall power consumption of the flue gas treatment system is reduced by more than 20%, and energy conservation and carbon reduction are achieved in a coordinated manner. The ammonia consumption is reduced by more than 20%, and the limestone consumption is reduced by more than 5%, achieving an effective reduction in _CO2 emissions and realizing coordinated optimization of cascade pollution reduction and carbon reduction in the entire process.

Embodiment 3

In order to verify the effectiveness of the results of the present invention, an industrial verification study on the application of the intelligent control system for pollution reduction and carbon reduction in the whole process of combustion-pollutant treatment was carried out with a 220t/h thermal power unit as the object. The operating effects of the original control and the results of the present invention were compared. After the application of the results of the present invention, compared with the original control, the coal consumption per unit steam production of the boiler was saved by 0.38t/h, and the power consumption of the fan per unit steam production (primary fan, secondary fan, boiler section induced draft fan) was reduced by 17.73kWh; the nitrogen oxide removal device saved about 0.043t/h of ammonia water consumption, and the air resistance was reduced by about 200Pa, which was converted into induced draft fan power consumption saving of 24kWh/h; the particulate matter removal device saved about 26kW of power consumption; the sulfur oxide removal device circulation pump saved about 70kW of power consumption, the oxidation fan saved about 18kW of power consumption, and the system resistance was reduced by about 200Pa, which was converted into induced draft fan power saving of 24kW. By statistically analyzing the operating data of the invention for one year, compared with the original control, the coal consumption per unit steam output of the boiler was saved by more than 1.6%, the power consumption of the fan per unit steam output was reduced by more than 2%, the average consumption of denitrification ammonia water was reduced by about 40%, the energy consumption of the electric bag dust removal system was reduced by about 35%, the energy consumption of the desulfurization circulating pump was reduced by about 25%, and the energy consumption of the oxidation fan was reduced by about 30%. It is estimated that a single ultra-low emission intelligent control system can save about 3.8088 million yuan in direct operating costs each year.

Table 1 Calculation table of annual cost savings of the invention in a 220t/h application

In terms of environmental benefits, the results of the present invention can achieve a significant reduction in material consumption and energy consumption while ensuring that the emission concentrations of NOx, _SO2 , and particulate matter meet the ultra-low emission requirements at all times, thus achieving the goals of energy saving, consumption reduction, and emission reduction. After long-term operation verification, the main pollutants are stably discharged at ultra-low levels, the fluctuation range of the outlet pollutant concentration is significantly reduced, and the fluctuation range of the NOx concentration at the outlet of the nitrogen oxide removal device and the _SO2 concentration at the outlet of the sulfur oxide removal device are reduced by more than 70%. A single 220t/h boiler combustion system and pollutant treatment system can save about 2923t of standard coal per year (calculated according to the standard coal saving coefficient of 0.302gce/MWh), and it is estimated that the amount of carbon dioxide can be reduced by about 8221t per year. While significantly improving the operating stability and adjustability of the boiler combustion system and the pollutant treatment system and reducing the operating cost, it also achieves the coordinated reduction of carbon dioxide emissions.

In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the method disclosed in the embodiment, since it corresponds to the system disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

This article uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only used to help understand the method and core ideas of the present invention. At the same time, for those skilled in the art, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims (5)

1. The full-flow intelligent regulation pollution reduction and carbon reduction method for combustion-pollutant treatment is characterized by comprising the following steps of:

Based on a pollutant generation and removal mechanism, analyzing and determining operation parameters affecting target controlled variables, taking the influence parameters of the target controlled variables as input variables and the target controlled variables as output variables, and constructing each controlled variable prediction model of the boiler combustion-flue gas treatment system;

predicting each controlled variable respectively, and obtaining predicted values of each controlled variable in advance;

constructing a control module of a target controlled variable according to the prediction result, formulating a control strategy of a boiler combustion system and a flue gas treatment device, and adjusting key influence parameters influencing the target controlled variable in advance;

Based on the synergistic effect of the removal equipment in the pollutant treatment system on NOx, SO ₂、PM、SO₃ and Hg pollutants, a boiler combustion-pollutant removal system regulation strategy with multi-objective synergistic effect of energy consumption, material consumption and pollutant emission is established, and the real-time adjustment and optimization of key parameters of key equipment for boiler combustion and pollutant removal are realized;

the controlled variables of the boiler combustion system include: boiler oxygen content, furnace outlet pressure, outlet pollutant concentration;

The controlled variables of the flue gas treatment device comprise total discharge port particulate matter concentration, total discharge port SO ₂ concentration/desulfurization device outlet SO ₂ concentration, total discharge port/denitration device outlet NO _x concentration, ammonia escape concentration and total discharge port SO ₃ concentration;

the boiler oxygen content, hearth outlet pressure and NOx concentration collaborative prediction control process comprises the following steps:

Step Sa1, using a coal feeding amount, a primary fan frequency and a secondary fan frequency as input variables, using oxygen content as output variables, and utilizing a step response vector of the input variables to predict the change of the oxygen content in real time, so as to establish an oxygen content concentration prediction model;

The oxygen content concentration prediction model expression is as follows:

Wherein: The total predicted value of the oxygen content of P time domain lengths in the future is not corrected at the moment k; the total predicted values of the oxygen content of P time domain lengths under the influence of the coal supply amount, the primary fan frequency and the secondary fan frequency are respectively uncorrected, and P is the rolling optimization time domain length; u (k-1) is the value of N time domain lengths before the k moment of each variable, and N is the model time domain; Δu (k) is a control increment predicted value of each variable in k time domain length to M time instants in future, and M is a control time domain length; a ₀, A is the dynamic matrix of each variable, describing the influence of each input variable on the system response;

Step Sa2, for correcting errors caused by model mismatch and environmental interference, the model predicted value is corrected by using real-time information, and the feedback correction process is as follows:

Wherein: y _c (k) is the total predicted value after k time correction; h is a feedback correction coefficient; y (k) and y _c (k) are respectively the current oxygen content actual measurement value and the predicted value at the moment k;

Step Sa3, the difference between the oxygen content target value and the predicted value is minimized to optimize the control, and the k time optimizing performance index is expressed as follows in a vector form:

J(k)＝[Y_c(k)-Yr(k)]^TQ[Y_c(k)-Y_r(k)]+ΔU(k)^TRΔU(k) (6)

Wherein: j (k) is an optimization objective function; y _r (k) is a target controlled variable control target value; q, R are respectively a target controlled variable prediction error weight matrix and a key parameter control weight matrix;

and (3) making:

Step Sa4, determining a control increment of the secondary air blower according to the target value of the oxygen content, realizing stable control of the oxygen content, performing real-time feedback correction according to the actual measurement value of the oxygen content, and outputting an optimized control increment expression as follows:

ΔU(k)＝(A^TQA+R)^-1A^TQ{Y_r(k)-A₀U(k-1)-H[y(k)-y_c(k)]} (8)

Step Sa5, taking the future frequency regulation and control instruction of the secondary air blower output in the step S4 as the input quantity of the hearth negative pressure control module; under the dynamic working condition, the relation between the primary fan frequency variation and the secondary fan frequency variation and the induced draft fan frequency is analyzed by establishing a linear regression model, and the induced draft fan frequency is adjusted in advance; under the static working condition, the least square method is utilized to conduct parameter identification on the PID controller, the initial parameter value of the PID controller is determined, a draught fan control instruction is given in advance, and the negative pressure stability of the hearth is ensured;

Under the steady-state working condition, when the output hearth outlet pressure is in a set range, the frequency of the induced draft fan is not required to be regulated, and when the hearth outlet pressure exceeds the set range, the frequency of the induced draft fan is corrected through PID; the correction formula is as follows:

Wherein: u (k) is the frequency variation of the output induced draft fan of the PID controller; e (k) is the difference between the target value and the measured value of the furnace outlet pressure;

Under the dynamic working condition, the relationship among the frequencies of the primary fan, the secondary fan and the induced draft fan is analyzed by fitting, the frequency regulation increment of the induced draft fan is determined, and the fitting formula is as follows:

ΔL _{Guiding device} ＝0.555 18ΔL _Once-through +0.602 37ΔL _{Secondary time} +-0.001 35L _{Guiding device} +0.013 63

Wherein: Δl _Once-through is the primary fan frequency variation; Δl _{Secondary time} is the secondary fan frequency variation; l _{Guiding device} is the actual measurement value of the frequency of the induced draft fan;

the NOx concentration and ammonia slip prediction control process of the outlet of the denitration device is as follows:

Step Sb1, taking the load of a boiler, the coal feeding amount, the air quantity and the flue gas temperature as input variables, taking the NOx concentration at the outlet of a hearth, namely the inlet of a denitration device, as output variables, and establishing a partition-based sectional denitration device inlet NOx concentration prediction model;

the furnace outlet NOx concentration prediction model expression is as follows:

Wherein: For the time k, P time domain length hearths in future an uncorrected total predicted value of outlet NOx concentration; The predicted values of the NOx concentration of the outlet of the hearth with P time domain lengths under the influence of the coal supply quantity, the air quantity and the flue gas temperature are respectively uncorrected, and P is the rolling optimized time domain length; u (k-1) is the value of N time domain lengths before the k moment of each variable, and N is the model time domain; Δu (k) is a control increment predicted value of each variable in k time domain length to M time instants in future, and M is a control time domain length; a ₀, A is the dynamic matrix of each variable, describing the influence of each input variable on the system response;

In step Sb2, for correcting errors caused by model mismatch and environmental interference, the model predicted value is corrected by using real-time information, and the feedback correction process is as follows:

wherein: y _N (k) is the total predicted value after k time correction; h is a feedback correction coefficient; y (k) and y _N (k) are respectively the actual measurement value and the predicted value of the NOx concentration at the current hearth outlet at the moment k, namely the NOx concentration at the inlet of the denitration device;

Step Sb3, calculating a difference value between a predicted result of the NOx concentration at the inlet of the denitration device and a controlled target in a certain period of time in the future in order to correct the stable deviation of the NOx concentration model at the inlet of the denitration device, and correcting by adopting a coefficient to realize real-time and accurate prediction of the NOx concentration at the outlet of a hearth, namely the NOx concentration at the inlet of the denitration device;

step Sb4, taking the boiler load, the operating temperature of a denitration region, the ammonia injection flow and the NOx concentration at the inlet of the denitration device as input variables, and taking the NOx concentration at the outlet of the denitration device as output variables, and establishing a prediction control model based on the NOx concentration at the outlet of the partition and segmentation denitration device;

And step Sb5, adding an inlet NOx concentration predicted value, namely a furnace outlet NOx predicted value, into a denitration device prediction control module as a feedforward prediction, further outputting an optimized set value of ammonia injection flow, calculating an opening value of an ammonia injection regulating valve by an intelligent advanced controller according to deviation between a measured value of the ammonia injection flow and the optimized value, formulating a full-working-condition multi-parameter coordination-cascade intelligent advanced control strategy to realize stable control of denitration outlet NOx concentration and ammonia escape, and carrying out real-time feedback correction according to actual measurement values of denitration outlet NOx concentration and ammonia escape.

2. The full-flow intelligent regulation pollution abatement and carbon reduction method for combustion-pollutant treatment according to claim 1, wherein the total exhaust particulate matter and SO ₃ concentration predictive control process is as follows:

Step Sc1, based on the generation mechanism of particulate matters and SO ₃ in the boiler combustion process and the arrangement and particle size distribution characteristics of slag and flue gas, establishing a flue gas particulate matter and SO ₃ generation concentration prediction model, and realizing the prediction of the concentration and mass distribution of the particulate matters and SO ₃ at the inlet of the dust removal device;

sc2, based on an electric field corona discharge mechanism in the electrostatic dust collector and a particle and SO ₃ charge migration mechanism, turbulent flow, backflow multi-physical field strengthening particle and SO ₃ trapping mechanism in the wet desulphurization device, and a strengthening mechanism of condensation, agglomeration, charge and migration of fine particles and SO ₃ in the wet electrostatic dust collector, SO as to construct a mechanism model for predicting the concentration and mass distribution of the whole process of removing the particle and SO3 at the inlet and outlet of the dust collector, the wet desulphurization device and the wet electrostatic dust collector;

Step Sc3, correcting a mechanism model of the prediction of the concentration and the mass distribution of the particulate matters and SO ₃ removal overall process established in the step Sc1 and the step Sc2 based on actual operation data, and correcting a mechanism model established by a mechanism model of the particulate matters and SO ₃ concentration prediction model cooperatively driven by establishing a process mechanism and multi-station sectional machine learning;

the data correction model construction method comprises a parameter identification method based on a gradient descent+particle swarm algorithm and a long-term and short-term memory neural network algorithm based on an attention mechanism;

Sc4, a process mechanism established based on the step and multi-station sectional machine learning collaborative driven particulate matter and SO ₃ concentration prediction model, adopting the energy consumption of an electrostatic precipitator/a low-temperature electrostatic precipitator/an electric bag composite precipitator and the energy consumption of a wet electrostatic precipitator, the total exhaust particulate matter and SO ₃ concentration as intelligent regulation targets, constructing an energy consumption and particulate matter and SO ₃ concentration emission optimization problem, and solving by adopting a particle swarm algorithm or a particle swarm-gradient descent algorithm according to the characteristics of the optimization problem, thereby obtaining the optimal secondary voltage setting mode of the electrostatic precipitator/the low-temperature electrostatic precipitator/the electric bag composite precipitator and the wet electrostatic precipitator under different operation conditions, and constructing the intelligent regulation strategy of different types of power supplies of the particulate matter electrostatic removal device in the electric field and the partition chamber region.

3. The full-flow intelligent regulation pollution abatement and carbon reduction method for combustion-pollutant treatment according to claim 1, wherein the total discharge SO ₂ concentration predictive control process is as follows:

Sd1, based on SO ₂ in the generation of the boiler combustion process and SO ₂/SO₃ conversion mechanism in the denitration device, establishing a flue gas SO ₂ at the inlet of the desulfurization device based on a machine learning algorithm to generate a concentration prediction model;

Sd2, constructing a multi-absorbent SO ₂ removal process mechanism model based on a multi-absorbent SO ₂ removal process principle in the desulfurization device;

Sd3, performing correction parameter identification on a coal-fired flue gas SO ₂ generation-removal mechanism model by adopting a PSO algorithm based on historical operation data, and further constructing a SO ₂ removal process data correction model based on an LSTM network;

And Sd4, generating and removing a model based on the SO ₂ of the coal-fired flue gas established in the step, adopting the concentration of the outlet SO ₂ and the desulfurization operation pH as intelligent regulation targets, constructing an optimization problem, and solving by adopting an enumeration method, a particle swarm algorithm or a particle swarm-gradient descent algorithm according to the characteristics of the optimization problem, thereby obtaining the optimal circulating pump of the wet desulfurization device and a frequency regulation strategy thereof under different operation conditions.

4. The method for intelligently controlling pollution reduction and carbon reduction by complete flow of combustion-pollutant treatment according to claim 3, wherein in order to ensure stability and convergence of an algorithm in the process of constructing an intelligent control strategy, a rolling optimization strategy is implemented by predicting a change trend of a controlled variable in a period of time, and in addition, on-line correction is required to be carried out on a prediction model according to actual running conditions of a site, and the specific steps are as follows:

Step Sc401, setting constraints on the outlet contaminant concentration SO ₂, is expressed as follows:

Wherein r _p,i is the number of circulating pumps and the frequency weight coefficient; p (t+i) is the sum of the slurry amounts of the circulating pumps; r _e,j is an superscalar weight coefficient; sgn (·) is a sign function; w _limit (t+j) is the upper emission limit of the outlet contaminant concentration SO ₂; m is a control time domain; p is the prediction time domain;

In step Sc402, in order to ensure the relative stability of the outlet pollutant concentration, to cope with the rapid abrupt change of the working condition, an error between the reference trajectory and the predicted value is added to the optimization target, and the reference trajectory is set to be smaller than the outlet concentration limit value in the above formula, which is expressed as follows:

Wherein q _j is a tracking weight coefficient; w _t (t+j) is the emission target;

in step Sc403, on the basis of step Sc402, in order to ensure that the control process can converge to the vicinity of the target value, a terminal error needs to be added to the optimization target, so the final rolling optimization problem can be expressed as follows:

wherein minJ (t) is an objective function at time t; q _j is a tracking weight coefficient; w _t (t+j) is the emission target; r _p,i is the number of circulating pumps and the frequency weight coefficient; p (t+i) is the sum of the slurry amounts of the circulating pumps; r _e,j is an superscalar weight coefficient; sgn (·) is a sign function; w _limit (t+j) is the upper emission limit of the outlet contaminant concentration SO ₂; v _f (·) is the terminal error function; χ (P) is a control amount calculated when predicting the last P moments of the time domain; χ _s is the control after steady state optimization; y _d is the outlet concentration after steady state optimization.

5. A system for implementing the combustion-pollutant abatement full-process intelligent regulation and control pollution abatement and carbon reduction as set forth in claims 1-4, said system comprising:

A boiler combustion-pollutant treatment system layer, a knowledge-data coupling modeling layer, a model parameter identification optimizing layer and a cascade intelligent accurate regulation layer;

The boiler combustion-pollutant treatment system layer consists of a low-carbon/zero-carbon fuel blending system, a boiler combustion system, a denitration device, an electrostatic dust collection device, a low-temperature electric dust collection device/electric bag composite dust collection device, a wet desulfurization device and a wet electrostatic dust collection device;

The step intelligent accurate regulation and control layer consists of a boiler combustion prediction control module, a denitration device prediction control module, an electric dust collector prediction control module, a wet desulphurization device prediction control module, a wet electric dust collector prediction control module and a full-flow step collaborative optimization control module;

The flow of the combustion-pollutant treatment full-flow intelligent regulation pollution-reducing and carbon-reducing system is as follows, a knowledge-data coupling modeling layer is used for establishing a boiler combustion-pollutant treatment full-flow multi-section pollutant concentration generation and step removal mechanism model based on a pollutant generation and removal mechanism of a full-flow device in a boiler combustion-pollutant treatment system layer, and a knowledge-data fusion-driven multi-section pollutant concentration prediction model is further established by combining historical operation data and advanced experience knowledge in the boiler combustion-pollutant treatment system layer;

In the model parameter identification optimization layer, taking a target controlled variable as an intelligent regulation target, constructing an optimization problem, carrying out model parameter identification and optimization solution by adopting a PSO algorithm, a WOA algorithm, a particle swarm-gradient descent algorithm and an enumeration algorithm according to the characteristics of the optimization problem, and simultaneously carrying out rolling optimization solution on the optimization solution parameters by combining offline excavation and online iteration, thereby establishing a key parameter prediction model driven by the process mechanism of the boiler combustion-flue gas treatment process and multi-working-condition segmentation machine learning in a cooperative manner;

The method comprises the steps of accurately predicting the concentration of multi-section pollutants of a boiler outlet and a pollutant removal system through a cascade intelligent accurate regulation layer, establishing a pollutant removal multi-device control strategy by combining the pollutant discharge concentration of the outlet of a different pollutant treatment device with the key regulation parameters under different working conditions, namely the boiler low-carbon/zero-carbon fuel blending combustion quantity, the fan frequency of a boiler section, the ammonia injection quantity of different areas of a denitration device, the operation secondary voltage of different electric field types of power supplies of an electric dust collector, different circulating pump combinations of slurry of a wet desulphurization device and the key regulation parameters of the circulating pump frequency and the slurry pH of the slurry, and realizing the full-process pollutant optimal control and the collaborative energy-saving carbon reduction of the full-process pollutants along the flue gas cascade.