Dedusting system coupling flue gas denitration integrated reaction control method

Abstract

The invention discloses a dust removal system coupling flue gas denitration integrated reaction control method, and relates to the field of industrial flue gas pollution abatement, multi-dimensional data such as flue gas temperature, NOx concentration and the like are collected through 50Hz frequency on the basis of multi-field parameter dynamic perception, and a 3 * 5 * 7 * 10 order parameter tensor set is generated through multi-scale fusion; on the basis, a fractional order coupling reaction kinetic model containing 28 time-varying parameters is constructed, and denitration and dust removal coupling characteristics are accurately described. The optimal control parameters are solved through an improved variational algorithm by taking the denitration standard reaching and the lowest energy consumption as targets, and a delay compensation coefficient is introduced to generate a cooperative instruction to drive multiple actuators to act. And finally, forming a closed loop through db4 wavelet basis five-stage decomposition and a composite correction formula, and updating model parameters in real time. In the application of a 300MW unit, the ammonia escape is controlled within 5ppm, the dust emission is less than or equal to 8mg / Nm, the energy consumption is reduced by 22%, and the device can adapt to load fluctuation and meet the GB13223-2011 standard.

Description

Technical Field

[0001] This invention relates to the field of industrial flue gas pollution control, specifically to an integrated reaction control method for dust removal systems coupled with flue gas denitrification. Background Technology

[0002] Nitrogen oxides (NOx) and particulate matter in industrial flue gas are core sources of air pollution, and their emission control has become a rigid requirement for achieving the "dual carbon" target and environmental regulations. According to the current "Emission Standard of Air Pollutants for Thermal Power Plants" (GB13223-2011), the NOx emission limit for coal-fired power plants in key areas must be ≤50 mg / Nm³. 3 Dust emission limit ≤10mg / Nm 3 Stringent standards are driving the upgrading of flue gas treatment technology towards higher efficiency and synergy. Currently, the industrial sector generally adopts a series treatment mode of "denitrification + dust removal", with the mainstream configuration being a combination of SCR / SNCR denitrification units and electrostatic precipitators and bag filters. However, this mode has gradually exposed many technical bottlenecks in actual operation.

[0003] The most prominent problem with existing technologies is the strong independence of the systems and the lack of coordinated control. Denitrification and dust removal units often employ segmented control strategies. The former relies on preset ammonia injection rates and catalyst temperature parameters, while the latter adjusts its operation through fixed electric field strength or filter bag pressure differential. There is a lack of parameter linkage mechanism between the two. This separate control mode is prone to parameter conflicts: for example, while increasing the electric field strength of the electrostatic precipitator can enhance dust removal, it can interfere with the charge distribution on the surface of the denitrification catalyst, leading to decreased activity; and when excessive ammonia is injected to ensure denitrification meets standards, the escaped ammonia reacts with SO3 in the flue gas to form highly viscous NH4HSO4, which can clog the dust collector filter bags and air preheater, often exceeding the safe threshold of 10 ppm.

[0004] Poor model adaptability and delayed response to operating conditions further exacerbate the control challenges. Traditional systems often employ integer-order reaction kinetic models, which can only describe single reaction processes under steady-state conditions and cannot accurately characterize the unsteady coupling characteristics under flue gas composition fluctuations and load changes. When industrial kilns switch fuels or adjust production capacity, sudden changes in parameters such as flue gas flow rate and temperature can widen the deviation between model predictions and actual responses. Relying on PID controllers or manual adjustments makes it difficult to achieve real-time parameter optimization, often resulting in short-term excessive emissions or energy waste.

[0005] In summary, faced with complex and ever-changing industrial conditions and stringent environmental requirements, existing "distributed control" flue gas treatment technologies can no longer meet the core demands of "high efficiency and coordination, low consumption and stability." Developing an integrated technology capable of multi-parameter coupled modeling, dynamic optimization control, and adaptive adjustment has become crucial to solving the current industry pain points. Summary of the Invention

[0006] The purpose of this invention is to provide an integrated reaction control method for dust removal systems coupled with flue gas denitrification. This method involves a multi-field parameter detection module that collects multi-field, multi-dimensional parameters such as flue gas temperature and flow fields, and generates a dynamic parameter tensor set through multi-scale coupling and fusion. Based on this, a fractional-order coupled reaction dynamic model containing intermediate product paths is constructed, outputting a time-varying characteristic coefficient matrix. Then, the optimal control parameter tensor, including dimensions such as electric field strength, is solved using a multi-constraint adaptive dynamic optimization objective function and an improved variational algorithm. Combined with an actuator response delay model, multi-order collaborative control commands are generated to drive the actuator module's actions. Finally, a feedback adjustment module collects the multi-scale error between the calculated terminal parameters and the model's predicted values, updates the model coefficients using a wavelet decomposition-fractional integral composite correction formula, and feeds it back, forming an adaptive closed loop to achieve efficient and coordinated operation of dust removal and denitrification.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for integrated reaction control of dust removal system coupled with flue gas denitrification, characterized by comprising the following steps:

[0009] S1: Real-time acquisition of multi-field and multi-dimensional parameters of flue gas and generation of dynamic parameter tensor set through multi-scale coupling and fusion formula;

[0010] S2: Construct a fractional-order coupled reaction kinetic model containing intermediate product paths based on the parameter tensor set and output the time-varying characteristic coefficient matrix;

[0011] S3: Based on the time-varying characteristic coefficient matrix, establish an adaptive dynamic optimization objective function with multiple constraint terms and solve the optimal control parameter tensor using an improved variational algorithm;

[0012] S4: Generate a multi-order cooperative control instruction sequence based on the optimal control parameter tensor and the actuator response delay model;

[0013] S5: Real-time acquisition of multi-dimensional effect parameters of the reaction terminal and multi-scale error of model prediction values, adjustment of the time-varying characteristic coefficient matrix of the fractional-order coupled reaction dynamics model through wavelet decomposition-fractional integral composite correction formula and feedback to step S2 to form adaptive closed-loop control.

[0014] Step S1 involves real-time acquisition of multi-field, multi-dimensional parameters of flue gas and generation of a dynamic parameter tensor set using a multi-scale coupling and fusion formula. Specifically, the multi-field, multi-dimensional parameters include: a temperature field (temperature at reactor inlet, middle section, and outlet), a flow field, an electric field, a NOx concentration field, and a dust concentration field; the acquisition frequency is 50 Hz; and the multi-scale coupling and fusion formula is as follows:

[0015] ;

[0016] In the formula, in the formula, It is a 3×5×7×10 order dynamic parameter tensor. For scale marking, yes The 5×7 weight matrix at this scale was obtained through offline optimization using the particle swarm optimization algorithm. yes The original parameter matrix of the 7×10 order at the scale has one feature parameter in each row, including flue gas temperature, flow velocity modulus, pressure, humidity, NOx concentration, dust resistivity, and catalytic activity factor. Each column corresponds to the measured data of one sampling period. For Kronecker product, yes The 3×5 order field gradient tensor at the scale contains the spatial first-order partial derivatives of the parameters. For Hadamard product, yes The attenuation coefficient at the scale is taken as: , , , For the duration of the reaction, It is a 3×1 electric field intensity vector. It is a 3×1 flue gas velocity vector. For vector cross product, It is the Frobenius norm. It is a 3×5×7×10 order disturbance compensation tensor, generated based on historical operational errors during training. Let be the error function. This is the fluctuation value of the pressure difference between the system's inlet and outlet. It is the standard deviation of pressure difference fluctuation.

[0017] Step S2 constructs a fractional-order coupled reaction kinetic model containing intermediate product pathways based on the parameter tensor set and outputs a time-varying characteristic coefficient matrix. Specifically, it involves the reversible reaction from NO2 to N2O4 and the formation of intermediate states catalyzed on the dust surface. The time-varying characteristic coefficient matrix includes two reaction pathways, three scales, multiple field parameters, and 28 time-varying parameters related to the catalyst state. The formula for the fractional-order coupled reaction kinetic model is as follows:

[0018] ;

[0019] In the formula, It is the Caputo fractional derivative operator. The order of the fractional derivative is obtained by the least squares method. It is the NOx concentration. It is the dust concentration. This is the catalyst activity factor, with a value of 1 indicating full catalyst activity and 0 indicating complete deactivation. It is a divergence operator. The flue gas velocity vector is 3×1. It is the main reaction pathway identifier. It is the first The rate constant of the main reaction pathway, It is the first The NOx reaction order for each pathway was obtained by fitting experimental data. It is the first The dust catalytic order of each path was determined by fitting experimental data. It is the electric field strength, i.e., in step S1 The modulus, It is the first The electric field enhancement index of the path, This is the Arrhenius exponent term. It is the first Activation energy of the pathway It is the gas constant, with a fixed value of 8.314. It is the flue gas temperature. It is a catalyst poisoning factor.

[0020] Step S3 establishes an adaptive dynamic optimization objective function with multiple constraints based on the time-varying characteristic coefficient matrix and solves for the optimal control parameter tensor using an improved variational algorithm. The optimal control parameter tensor includes four control dimensions: electric field strength, ammonia injection rate, electrode spacing, and induced draft fan frequency. The formula for the adaptive dynamic optimization objective function is:

[0021] ;

[0022] In the formula, in the formula, It is about optimizing the objective function value. It is the time-varying weight of the denitrification index, expressed as: ,in As the initial weights, , , These are the Fourier coefficients. For the fundamental frequency, , , This is the phase angle, and its weight is dynamically adjusted over time to adapt to the requirements of denitrification priority due to fluctuations in operating conditions. It is the time-varying weight of the dust removal index, expressed as: ,in As the initial weights, , , The coefficients of the sine series are... , , The phase angle, yes At any given time, the NOx concentration at the reactor outlet This is the NOx export concentration limit. yes Constant time of reactor outlet dust concentration It is the dust outlet concentration limit. It is the time-varying weight of energy consumption, expressed as: ,in As the initial weights, This is a time-varying factor for energy consumption weighting, which dynamically increases over time. yes Real-time energy consumption of the system, including the total power consumption of high-voltage power supply, induced draft fan, and ammonia injection pump equipment; This is the system's rated energy consumption. It is the time-varying weight of catalyst loss, expressed as: ,in As the initial weights, hyperbolic secant function , As a characteristic time, this weight is higher in the early stage and gradually decreases in the later stage, adapting to the normal aging process of the catalyst. yes Catalyst aging rate at any time.

[0023] The adaptive dynamic optimization objective function formula in step S3 is constrained as follows:

[0024] ;

[0025] In the formula, in the formula, It refers to the flue gas temperature inside the reactor, within a specified range. Corresponding to actual temperature , It is the Reynolds number of the flue gas velocity, constrained. To avoid excessive turbulence that could lead to catalyst wear and reduced dust removal efficiency; yes The catalyst deactivation rate function at time intervals, constrained by it This indicates that the maximum permissible deactivation of the catalyst is 20%. It is the rate of change of electric field voltage, constrained. This can prevent sudden voltage spikes from causing unstable corona discharge and damage to equipment insulation. This is the theoretical ammonia injection rate. This refers to the actual amount of ammonia injected, to avoid insufficient ammonia injection leading to low denitrification efficiency, or excessive ammonia injection leading to ammonia escape.

[0026] Step S4 generates a multi-order cooperative control command sequence based on the optimal control parameter tensor and the actuator response delay model. The actuator includes a high-voltage power supply, an ammonia injection valve group, a plate regulating motor, and an induced draft fan frequency converter. The formula for generating the multi-order cooperative control command sequence is as follows:

[0027] ;

[0028] In the formula, It is the first The control command vector for future time steps is 4×1 in dimension and includes the electric field intensity adjustment. Ammonia injection volume adjustment Adjustment amount of electrode spacing Expectorant fan frequency adjustment ; It is a control order identifier. It is a control cycle. It is the first 4×4 control gain matrix of order, It is a 4×10 order optimal control parameter tensor, which includes four types of parameters: electric field strength, ammonia injection rate, electrode spacing, and induced draft fan frequency. Each type of parameter contains the optimal value for 10 sampling periods. It is a 4×10 order real-time detection parameter tensor. It is the delay time constant of the actuator. It is the delay compensation coefficient. It is a 4×1 order reference control command vector. It is a 4×10-order nominal parameter tensor. It is a 4×1 order parameter deviation standard deviation vector. It is the integral time variable. It is a 4×4 dynamic compensation kernel function matrix. yes Real-time parameter change rate tensor It is a dynamic compensation item.

[0029] Step S5 involves real-time acquisition of the multi-dimensional effect parameters of the reaction terminal and the multi-scale error between the predicted values ​​and the model. The time-varying characteristic coefficient matrix of the fractional-order coupled reaction kinetic model is adjusted using a wavelet decomposition-fractional-integral composite correction formula and fed back to step S2 to form adaptive closed-loop control. The multi-dimensional effect parameters of the reaction terminal include outlet NOx concentration, dust concentration, system energy consumption, and catalyst activity. The wavelet decomposition-fractional-integral composite correction formula is as follows:

[0030] ;

[0031] In the formula, in the formula, It is the time-varying characteristic coefficient correction amount, used to update the time-varying parameters of the fractional-order dynamic model in step S2; This is the total correction factor, with a value of 0.05. This is a wavelet decomposition level identifier, with values ​​from 1 to 5 corresponding to 5 levels of wavelet decomposition. They are wavelet basis numbers within the same decomposition level. It is the first The number of nodes in the layer wavelet decomposition satisfies , It is the first Layer The decomposition coefficient operator of the wavelet basis is adopted, using the db4 wavelet basis. It is a 4×10 order terminal feedback effect parameter tensor, which includes four types of parameters: outlet NOx concentration, dust concentration, system energy consumption, and catalyst activity. Each type of parameter contains measured values ​​for 10 sampling periods. It is the 4×10 order model prediction parameter tensor, and The dimensions are consistent, including the predicted values ​​of the four types of parameters calculated by the model; It is a multi-scale error tensor. It is the Riemann-Liouville fractional derivative operator. It is the first Layer The order of the fractional derivatives corresponding to each wavelet basis is adaptively adjusted according to the decomposition level. It is a 28×1 order initial characteristic coefficient matrix, which is the initial parameter of the dynamic model in step S2.

[0032] An integrated dust removal system coupled with flue gas denitrification reaction device is characterized by comprising a flue gas conveying module, a multi-field parameter detection module, a coupled reaction main module, a central control module, a multi-dimensional execution module, and a feedback adjustment module, which are sequentially linked to form a closed-loop control system. Specifically: the flue gas conveying module includes an inlet flue, a flow regulating valve, an induced draft fan, and a flow equalization device connected in sequence, with the outlet of the flow equalization device connected to the coupled reaction main module; the multi-field parameter detection module is located in the inlet flue, the reaction main module, and the outlet flue, and includes temperature sensing and flow field testing units, transmitting raw parameters to the central control module via industrial Ethernet; the coupled reaction main module is a horizontal reactor, with its inner edge... The airflow direction includes electrostatic precipitator, denitrification catalysis, and reaction homogenization units. The electrostatic precipitator unit includes a corona electrode plate, a dust collection electrode plate, and a spacing adjustment mechanism. The denitrification catalysis unit consists of a staged catalyst layer and a matching ammonia injection grid. The central control module includes a data processing, model calculation, and optimization control unit, which implements the methods described in claims 1-7 to generate instructions. The multi-dimensional execution module includes a high-voltage power supply, ammonia injection control, and other units, which are respectively connected to the electrostatic precipitator unit and the ammonia injection grid component, and execute the instructions of the central control module. The feedback adjustment module includes a terminal parameter acquisition and error correction unit, which acquires the outlet parameters and calculates the error with the model prediction value, and generates a model correction amount through a composite correction formula to feed back to the central control module.

[0033] A non-transitory computer-readable storage medium storing computer instructions that, when executed by at least one processor, cause the at least one processor to perform the method as described in any one of claims 1 to 7.

[0034] A computer program product comprising a computing program stored on a non-transitory computer-readable storage medium, the computer program including program instructions that, when executed by a computer, cause the computer to perform the method as described in any one of claims 1 to 7.

[0035] The specific mechanism of the system is as follows: First, the flue gas conveying module uniformly introduces the flue gas to be treated into the horizontal coupled reactor body through the inlet flue, flow regulating valve and flow equalization device, avoiding local reaction efficiency deviations caused by uneven flow velocity; at the same time, the multi-field parameter detection module collects temperature field, flow field, electric field, NOx concentration field and dust concentration field data of reactor inlet, middle section and outlet in real time at a frequency of 50Hz. Through multi-scale coupling fusion formula, the original parameters of different scales are transformed into a 3×5×7×10 order dynamic parameter tensor set, eliminating the dimensional differences and measurement deviations of multi-field parameters, and providing accurate input for subsequent model construction.

[0036] Secondly, based on the aforementioned parameter tensor set, the central control module constructs a fractional-order coupled reaction kinetic model containing intermediate product pathways. On the one hand, it considers the reversible reaction between NO2 and N2O4 and the generation process of intermediate states catalyzed on the dust surface. On the other hand, it describes the "non-integer-order dynamic characteristics" of the coupled reaction through the Caputo fractional derivative operator. The final output includes a time-varying coefficient matrix containing 28 time-varying parameters, covering the rate constants, reaction orders, electric field enhancement exponents, catalyst activity, poisoning factors, etc. of the two main reaction pathways, accurately characterizing the dynamic influence of factors such as temperature, electric field, and dust concentration on the coupled reaction.

[0037] Subsequently, the system establishes a multi-constraint adaptive dynamic optimization objective function based on the time-varying characteristic coefficient matrix, with the goal of "achieving denitrification standards, minimizing energy consumption, and minimizing catalyst loss". The improved variational algorithm is used to solve for the 4×10-order optimal control parameter tensor, which includes electric field strength, ammonia injection rate, electrode spacing, and induced draft fan frequency.

[0038] Next, to address the response delay issue of the actuators, the system combines the actuator delay model and uses a multi-level collaborative control command sequence generation formula and dynamic compensation integral term to transform the optimal control parameters into 4×1-dimensional multi-level future control commands, which drive the electrostatic precipitator, denitrification catalytic unit, and induced draft fan respectively, thereby achieving coordinated action of multiple actuators and avoiding response imbalance caused by command lag.

[0039] Finally, the feedback adjustment module collects the NOx concentration, dust concentration, system energy consumption, and catalyst activity parameters at the reaction terminal outlet in real time, calculates the multi-scale error between these parameters and the model predictions, and generates time-varying characteristic coefficient corrections through wavelet decomposition-fractional integral composite correction formulas. These corrections are then fed back to the fractional coupled reaction kinetic model to update its coefficient matrix, forming an adaptive closed loop. This ensures that the system maintains efficient and coordinated operation of dust removal and denitrification under varying operating conditions such as flue gas composition and load fluctuations.

[0040] Compared with the prior art, the beneficial effects of the present invention are:

[0041] 1. This invention constructs a fractional-order coupled reaction kinetic model containing intermediate product pathways, covering the reversible NO2-N2O4 reaction and the dust catalytic intermediate state. It couples multiple field parameters, such as temperature, electric field, and flow field, into a 3×5×7×10 order dynamic parameter tensor, thereby achieving parameter synergistic optimization of the two major functions, avoiding efficiency loss caused by separate control, and improving the overall processing efficiency.

[0042] 2. This invention introduces the Caputo fractional derivative operator and 28 time-varying parameter matrices to accurately characterize the dynamic effects of electric field enhancement, dust catalysis, and catalyst poisoning. Combined with wavelet decomposition-fractional integral composite correction, the model prediction error is adjusted in real time according to the changes in operating conditions, thereby improving the control accuracy under complex operating conditions.

[0043] 3. This invention simultaneously optimizes four major objectives—denitrification compliance, high dust removal efficiency, minimum energy consumption, and minimum catalyst loss—through an adaptive dynamic optimization objective function with multiple constraints. It solves for the optimal control parameters using an improved variational algorithm, achieving a multi-dimensional balance of "high efficiency, energy saving, and low consumption."

[0044] 4. When generating multi-level collaborative control commands, this invention introduces a delay compensation coefficient and a dynamic compensation integral term to offset the effects of execution delay; at the same time, through 50Hz high-frequency multi-field parameter acquisition and adaptive closed loop, it responds to flue gas composition and load fluctuations in real time, avoiding the instability problem of traditional open-loop / semi-closed-loop systems. Attached Figure Description

[0045] Figure 1 This is a flowchart of an integrated reaction control method for dust removal system coupled with flue gas denitrification according to the present invention; Detailed Implementation

[0046] The technical solutions of the present invention will now be described in detail with reference to the accompanying drawings.

[0047] like Figure 1 As shown, a method for integrated reaction control of dust removal system coupled with flue gas denitrification is characterized by the following steps:

[0048] S1: Real-time acquisition of multi-field and multi-dimensional parameters of flue gas and generation of dynamic parameter tensor set through multi-scale coupling and fusion formula;

[0049] S2: Construct a fractional-order coupled reaction kinetic model containing intermediate product paths based on the parameter tensor set and output the time-varying characteristic coefficient matrix;

[0050] S3: Based on the time-varying characteristic coefficient matrix, establish an adaptive dynamic optimization objective function with multiple constraint terms and solve the optimal control parameter tensor using an improved variational algorithm;

[0051] S4: Generate a multi-order cooperative control instruction sequence based on the optimal control parameter tensor and the actuator response delay model;

[0052] S5: Real-time acquisition of multi-dimensional effect parameters of the reaction terminal and multi-scale error of model prediction values, adjustment of the time-varying characteristic coefficient matrix of the fractional-order coupled reaction dynamics model through wavelet decomposition-fractional integral composite correction formula and feedback to step S2 to form adaptive closed-loop control.

[0053] Step S1 involves real-time acquisition of multi-field, multi-dimensional parameters of flue gas and generation of a dynamic parameter tensor set using a multi-scale coupling and fusion formula. Specifically, the multi-field, multi-dimensional parameters include: a temperature field (temperature at reactor inlet, middle section, and outlet), a flow field, an electric field, a NOx concentration field, and a dust concentration field; the acquisition frequency is 50 Hz; and the multi-scale coupling and fusion formula is as follows:

[0054] ;

[0055] In the formula, in the formula, It is a 3×5×7×10 order dynamic parameter tensor. For scale marking, yes The 5×7 weight matrix at this scale was obtained through offline optimization using the particle swarm optimization algorithm. yes The original parameter matrix of the 7×10 order at the scale has one feature parameter in each row, including flue gas temperature, flow velocity modulus, pressure, humidity, NOx concentration, dust resistivity, and catalytic activity factor. Each column corresponds to the measured data of one sampling period. For Kronecker product, yes The 3×5 order field gradient tensor at the scale contains the spatial first-order partial derivatives of the parameters. For Hadamard product, yes The attenuation coefficient at the scale is taken as: , , , For the duration of the reaction, It is a 3×1 electric field intensity vector. It is a 3×1 flue gas velocity vector. For vector cross product, It is the Frobenius norm. It is a 3×5×7×10 order disturbance compensation tensor, generated based on historical operational errors during training. Let be the error function. This is the fluctuation value of the pressure difference between the system's inlet and outlet. It is the standard deviation of pressure difference fluctuation.

[0056] Step S2 constructs a fractional-order coupled reaction kinetic model containing intermediate product pathways based on the parameter tensor set and outputs a time-varying characteristic coefficient matrix. Specifically, it involves the reversible reaction from NO2 to N2O4 and the formation of intermediate states catalyzed on the dust surface. The time-varying characteristic coefficient matrix includes two reaction pathways, three scales, multiple field parameters, and 28 time-varying parameters related to the catalyst state. The formula for the fractional-order coupled reaction kinetic model is as follows:

[0057] ;

[0058] In the formula, It is the Caputo fractional derivative operator. The order of the fractional derivative is obtained by the least squares method. It is the NOx concentration. It is the dust concentration. This is the catalyst activity factor, with a value of 1 indicating full catalyst activity and 0 indicating complete deactivation. It is a divergence operator. The flue gas velocity vector is 3×1. It is the main reaction pathway identifier. It is the first The rate constant of the main reaction pathway, It is the first The NOx reaction order for each pathway was obtained by fitting experimental data. It is the first The dust catalytic order of each path was determined by fitting experimental data. It is the electric field strength, i.e., in step S1 The modulus, It is the first The electric field enhancement index of the path, This is the Arrhenius exponent term. It is the first Activation energy of the pathway It is the gas constant, with a fixed value of 8.314. It is the flue gas temperature. It is a catalyst poisoning factor.

[0059] Step S3 establishes an adaptive dynamic optimization objective function with multiple constraints based on the time-varying characteristic coefficient matrix and solves for the optimal control parameter tensor using an improved variational algorithm. The optimal control parameter tensor includes four control dimensions: electric field strength, ammonia injection rate, electrode spacing, and induced draft fan frequency. The formula for the adaptive dynamic optimization objective function is:

[0060] ;

[0061] In the formula, in the formula, It is about optimizing the objective function value. It is the time-varying weight of the denitrification index, expressed as: ,in As the initial weights, , , These are the Fourier coefficients. For the fundamental frequency, , , This is the phase angle, and its weight is dynamically adjusted over time to adapt to the requirements of denitrification priority due to fluctuations in operating conditions. It is the time-varying weight of the dust removal index, expressed as: ,in As the initial weights, , , The coefficients of the sine series are... , , The phase angle, yes At any given time, the NOx concentration at the reactor outlet This is the NOx export concentration limit. yes Constant time of reactor outlet dust concentration It is the dust outlet concentration limit. It is the time-varying weight of energy consumption, expressed as: ,in As the initial weights, This is a time-varying factor for energy consumption weighting, which dynamically increases over time. yes Real-time energy consumption of the system, including the total power consumption of high-voltage power supply, induced draft fan, and ammonia injection pump equipment; This is the system's rated energy consumption. It is the time-varying weight of catalyst loss, expressed as: ,in As the initial weights, hyperbolic secant function , As a characteristic time, this weight is higher in the early stage and gradually decreases in the later stage, adapting to the normal aging process of the catalyst. yes Catalyst aging rate at any time.

[0062] The adaptive dynamic optimization objective function formula in step S3 is constrained as follows:

[0063] ;

[0064] In the formula, in the formula, It refers to the flue gas temperature inside the reactor, within a specified range. Corresponding to actual temperature , It is the Reynolds number of the flue gas velocity, constrained. To avoid excessive turbulence that could lead to catalyst wear and reduced dust removal efficiency; yes The catalyst deactivation rate function at time intervals, constrained by it This indicates that the maximum permissible deactivation of the catalyst is 20%. It is the rate of change of electric field voltage, constrained. This can prevent sudden voltage spikes from causing unstable corona discharge and damage to equipment insulation. This is the theoretical ammonia injection rate. This refers to the actual amount of ammonia injected, to avoid insufficient ammonia injection leading to low denitrification efficiency, or excessive ammonia injection leading to ammonia escape.

[0065] Step S4 generates a multi-order cooperative control command sequence based on the optimal control parameter tensor and the actuator response delay model. The actuator includes a high-voltage power supply, an ammonia injection valve group, a plate regulating motor, and an induced draft fan frequency converter. The formula for generating the multi-order cooperative control command sequence is as follows:

[0066] ;

[0067] In the formula, It is the first The control command vector for future time steps is 4×1 in dimension and includes the electric field intensity adjustment. Ammonia injection volume adjustment Adjustment amount of electrode spacing Expectorant fan frequency adjustment ; It is a control order identifier. It is a control cycle. It is the first 4×4 control gain matrix of order, It is a 4×10 order optimal control parameter tensor, which includes four types of parameters: electric field strength, ammonia injection rate, electrode spacing, and induced draft fan frequency. Each type of parameter contains the optimal value for 10 sampling periods. It is a 4×10 order real-time detection parameter tensor. It is the delay time constant of the actuator. It is the delay compensation coefficient. It is a 4×1 order reference control command vector. It is a 4×10-order nominal parameter tensor. It is a 4×1 order parameter deviation standard deviation vector. It is the integral time variable. It is a 4×4 dynamic compensation kernel function matrix. yes Real-time parameter change rate tensor It is a dynamic compensation item.

[0068] Step S5 involves real-time acquisition of the multi-dimensional effect parameters of the reaction terminal and the multi-scale error between the predicted values ​​and the model. The time-varying characteristic coefficient matrix of the fractional-order coupled reaction kinetic model is adjusted using a wavelet decomposition-fractional-integral composite correction formula and fed back to step S2 to form adaptive closed-loop control. The multi-dimensional effect parameters of the reaction terminal include outlet NOx concentration, dust concentration, system energy consumption, and catalyst activity. The wavelet decomposition-fractional-integral composite correction formula is as follows:

[0069] ;

[0070] In the formula, in the formula, It is the time-varying characteristic coefficient correction amount, used to update the time-varying parameters of the fractional-order dynamic model in step S2; This is the total correction factor, with a value of 0.05. This is a wavelet decomposition level identifier, with values ​​from 1 to 5 corresponding to 5 levels of wavelet decomposition. They are wavelet basis numbers within the same decomposition level. It is the first The number of nodes in the layer wavelet decomposition satisfies , It is the first Layer The decomposition coefficient operator of the wavelet basis is adopted, using the db4 wavelet basis. It is a 4×10 order terminal feedback effect parameter tensor, which includes four types of parameters: outlet NOx concentration, dust concentration, system energy consumption, and catalyst activity. Each type of parameter contains measured values ​​for 10 sampling periods. It is the 4×10 order model prediction parameter tensor, and The dimensions are consistent, including the predicted values ​​of the four types of parameters calculated by the model; It is a multi-scale error tensor. It is the Riemann-Liouville fractional derivative operator. It is the first Layer The order of the fractional derivatives corresponding to each wavelet basis is adaptively adjusted according to the decomposition level. It is a 28×1 order initial characteristic coefficient matrix, which is the initial parameter of the dynamic model in step S2.

[0071] An integrated dust removal system coupled with flue gas denitrification reaction device is characterized by comprising a flue gas conveying module, a multi-field parameter detection module, a coupled reaction main module, a central control module, a multi-dimensional execution module, and a feedback adjustment module, which are sequentially linked to form a closed-loop control system. Specifically: the flue gas conveying module includes an inlet flue, a flow regulating valve, an induced draft fan, and a flow equalization device connected in sequence, with the outlet of the flow equalization device connected to the coupled reaction main module; the multi-field parameter detection module is located in the inlet flue, the reaction main module, and the outlet flue, and includes temperature sensing and flow field testing units, transmitting raw parameters to the central control module via industrial Ethernet; the coupled reaction main module is a horizontal reactor, with its inner edge... The airflow direction includes electrostatic precipitator, denitrification catalysis, and reaction homogenization units. The electrostatic precipitator unit includes a corona electrode plate, a dust collection electrode plate, and a spacing adjustment mechanism. The denitrification catalysis unit consists of a staged catalyst layer and a matching ammonia injection grid. The central control module includes a data processing, model calculation, and optimization control unit, which implements the methods described in claims 1-7 to generate instructions. The multi-dimensional execution module includes a high-voltage power supply, ammonia injection control, and other units, which are respectively connected to the electrostatic precipitator unit and the ammonia injection grid component, and execute the instructions of the central control module. The feedback adjustment module includes a terminal parameter acquisition and error correction unit, which acquires the outlet parameters and calculates the error with the model prediction value, and generates a model correction amount through a composite correction formula to feed back to the central control module.

[0072] A non-transitory computer-readable storage medium storing computer instructions that, when executed by at least one processor, cause the at least one processor to perform the method as described in any one of claims 1 to 7.

[0073] A computer program product comprising a computing program stored on a non-transitory computer-readable storage medium, the computer program including program instructions that, when executed by a computer, cause the computer to perform the method as described in any one of claims 1 to 7.

[0074] This implementation uses the flue gas treatment system of a 300MW coal-fired unit. The original SCR denitrification and electrostatic precipitator control system of this unit had problems such as ammonia slip exceeding 10ppm and large fluctuations in dust emissions. After modification, this control method is used in combination with a horizontal coupled reactor. The specific process is as follows: The induced draft fan introduces the boiler outlet flue gas into the system through the inlet flue. The flow regulating valve and flow equalization device stabilize the flue gas velocity within a Reynolds number of 45,000. A multi-parameter detection module synchronously collects temperature, flow rate, modulus, pressure, humidity, NOx concentration, dust resistivity, and catalytic activity factors at a frequency of 50Hz from the inlet flue, the middle section of the reactor, and the outlet flue. Temperature is the primary focus of monitoring. Point data ensures the reactor temperature is maintained between 30℃ and 150℃. The collected raw data undergoes multi-scale coupling and fusion processing. In this process, a 5×7 order weight matrix obtained through offline optimization using the particle swarm optimization algorithm is calculated with the original parameter matrices at each scale. This is combined with a 3×5 order field gradient tensor containing the first-order spatial partial derivatives, and three-level attenuation coefficients of 0.02, 0.05, and 0.1 are used. Furthermore, an interference compensation tensor trained based on historical data is incorporated to correct for the influence of inlet and outlet pressure difference fluctuations. Finally, a 3×5×7×10 order dynamic parameter tensor set is generated and input into the central control module. The central control module constructs a fractional-order coupled reaction kinetic model based on this tensor set. The model focuses on the reversible reaction of NO2 to N2O4 and the surface catalytic process of dust. The fractional derivative order, adapted to the current operating conditions, is identified using the least squares method. The output consists of a matrix of 28 time-varying parameters, including information on two reaction paths, three scales, and catalyst state. Subsequently, with the core objectives of an outlet NOx concentration not exceeding 50 mg / Nm³ and a dust concentration not exceeding 10 mg / Nm³, an optimization function is established combining the system's rated energy consumption and catalyst aging rate. Under the constraints of an electric field voltage change rate not exceeding 500 V / s and actual ammonia injection controlled at 0.8 to 1.2 times the theoretical value, an improved variational algorithm is used to solve the model. The system employs a 4×10-order optimal control parameter tensor, including electric field strength, ammonia injection rate, electrode spacing, and induced draft fan frequency. Considering the response delay of actuators such as the high-voltage power supply and ammonia injection valve group, a delay compensation coefficient is introduced to correct the optimal parameters, generating a 4×1-dimensional multi-order collaborative control command. This command drives the electrostatic precipitator to adjust the electrode spacing and electric field strength, the denitrification unit to precisely control the ammonia injection rate through the ammonia injection grid, and the induced draft fan to adjust its frequency according to the command to stabilize the flue gas velocity. The feedback adjustment module continuously collects data on outlet NOx concentration, dust concentration, real-time energy consumption, and catalyst activity. The error between these data and the model predictions is processed through a 5-level decomposition using a db4 wavelet basis and then corrected by a total correction coefficient of 0.The composite correction formula of 05 generates parameter correction values, updates 28 time-varying parameters of the kinetic model in real time, and feeds them back to the model construction stage to form a closed loop. After the modification, the system operates stably, with ammonia slip concentration controlled below 5 ppm, dust emissions stabilized below 8 mg / Nm³, unit flue gas treatment energy consumption reduced by 22% compared to before the modification, and catalyst aging rate reduced by 30%. Even under fluctuating operating conditions where the unit load increases from 60% to 100%, all indicators still meet the requirements of GB13223-2011 standard.

Claims

1. A method for integrated reaction control of dust removal system coupled with flue gas denitrification, characterized in that, Includes the following steps: S1: Real-time acquisition of multi-field and multi-dimensional parameters of flue gas and generation of dynamic parameter tensor set through multi-scale coupling and fusion formula; S2: Construct a fractional-order coupled reaction kinetic model containing intermediate product paths based on the parameter tensor set and output the time-varying characteristic coefficient matrix; S3: Based on the time-varying characteristic coefficient matrix, establish an adaptive dynamic optimization objective function with multiple constraint terms and solve the optimal control parameter tensor using an improved variational algorithm; S4: Generate a multi-order cooperative control instruction sequence based on the optimal control parameter tensor and the actuator response delay model; S5: Real-time acquisition of multi-dimensional effect parameters of the reaction terminal and multi-scale error of model prediction values, adjustment of the time-varying characteristic coefficient matrix of the fractional-order coupled reaction dynamics model through wavelet decomposition-fractional integral composite correction formula and feedback to step S2 to form adaptive closed-loop control.

2. The integrated reaction control method for dust removal system coupled with flue gas denitrification according to claim 1, characterized in that, Step S1 involves real-time acquisition of multi-field, multi-dimensional parameters of flue gas and generation of a dynamic parameter tensor set using a multi-scale coupling and fusion formula. Specifically, the multi-field, multi-dimensional parameters include: a temperature field (temperature at reactor inlet, middle section, and outlet), a flow field, an electric field, a NOx concentration field, and a dust concentration field; the acquisition frequency is 50 Hz; and the multi-scale coupling and fusion formula is as follows: ； In the formula, in the formula, It is a 3×5×7×10 order dynamic parameter tensor. For scale marking, yes The 5×7 weight matrix at this scale was obtained through offline optimization using the particle swarm optimization algorithm. yes The original parameter matrix of the 7×10 order at the scale has one feature parameter in each row, including flue gas temperature, flow velocity modulus, pressure, humidity, NOx concentration, dust resistivity, and catalytic activity factor. Each column corresponds to the measured data of one sampling period. For Kronecker product, yes The 3×5 order field gradient tensor at the scale contains the spatial first-order partial derivatives of the parameters. For Hadamard product, yes The attenuation coefficient at the scale is taken as: , , , For the duration of the reaction, It is a 3×1 electric field intensity vector. It is a 3×1 flue gas velocity vector. For vector cross product, It is the Frobenius norm. It is a 3×5×7×10 order disturbance compensation tensor, generated based on historical operational errors during training. Let be the error function. This is the fluctuation value of the pressure difference between the system's inlet and outlet. It is the standard deviation of pressure difference fluctuation.

3. The integrated reaction control method for dust removal system coupled with flue gas denitrification according to claim 1, characterized in that, Step S2 constructs a fractional-order coupled reaction kinetic model containing intermediate product pathways based on the parameter tensor set and outputs a time-varying characteristic coefficient matrix. Specifically, it involves the reversible reaction from NO2 to N2O4 and the formation of intermediate states catalyzed on the dust surface. The time-varying characteristic coefficient matrix includes two reaction pathways, three scales, multiple field parameters, and 28 time-varying parameters related to the catalyst state. The formula for the fractional-order coupled reaction kinetic model is as follows: ； In the formula, It is the Caputo fractional derivative operator. The order of the fractional derivative is obtained by the least squares method. It is the NOx concentration. It is the dust concentration. This is the catalyst activity factor, with a value of 1 indicating full catalyst activity and 0 indicating complete deactivation. It is a divergence operator. The flue gas velocity vector is 3×1. It is the main reaction pathway identifier. It is the first The rate constant of the main reaction pathway, It is the first The NOx reaction order for each pathway was obtained by fitting experimental data. It is the first The dust catalytic order of each path was determined by fitting experimental data. It is the electric field strength, i.e., in step S1 The modulus, It is the first The electric field enhancement index of the path, This is the Arrhenius exponent term. It is the first Activation energy of the pathway It is the gas constant, with a fixed value of 8.

314. It is the flue gas temperature. It is a catalyst poisoning factor.

4. The integrated reaction control method for dust removal system coupled with flue gas denitrification according to claim 1, characterized in that, Step S3 establishes an adaptive dynamic optimization objective function with multiple constraints based on the time-varying characteristic coefficient matrix and solves for the optimal control parameter tensor using an improved variational algorithm. The optimal control parameter tensor includes four control dimensions: electric field strength, ammonia injection rate, electrode spacing, and induced draft fan frequency. The formula for the adaptive dynamic optimization objective function is: ； In the formula, in the formula, It is about optimizing the objective function value. It is the time-varying weight of the denitrification index. It is the time-varying weight of the dust removal index. yes At any given time, the NOx concentration at the reactor outlet This is the NOx export concentration limit. yes Constant time of reactor outlet dust concentration It is the dust outlet concentration limit. It is the time-varying weight of energy consumption. yes Real-time energy consumption of the system, including the total power consumption of high-voltage power supply, induced draft fan, and ammonia injection pump equipment; This is the system's rated energy consumption. It is the time-varying weight of catalyst loss. yes Catalyst aging rate at any time.

5. The integrated reaction control method for dust removal system coupled with flue gas denitrification according to claim 4, characterized in that, The adaptive dynamic optimization objective function formula in step S3 is constrained as follows: ； In the formula, in the formula, It refers to the flue gas temperature inside the reactor, within a specified range. Corresponding to actual temperature , It is the Reynolds number of the flue gas velocity, constrained. To avoid excessive turbulence that could lead to catalyst wear and reduced dust removal efficiency; yes The catalyst deactivation rate function at time intervals, constrained by it This indicates that the maximum permissible deactivation of the catalyst is 20%. It is the rate of change of electric field voltage, constrained. This can prevent sudden voltage spikes from causing unstable corona discharge and damage to equipment insulation. This is the theoretical ammonia injection rate. This refers to the actual amount of ammonia injected, to avoid insufficient ammonia injection leading to low denitrification efficiency, or excessive ammonia injection leading to ammonia escape.

6. The integrated reaction control method for dust removal system coupled with flue gas denitrification according to claim 1, characterized in that, Step S4 generates a multi-order cooperative control command sequence based on the optimal control parameter tensor and the actuator response delay model. The actuator includes a high-voltage power supply, an ammonia injection valve group, a plate regulating motor, and an induced draft fan frequency converter. The formula for generating the multi-order cooperative control command sequence is as follows: ； In the formula, It is the first The control command vector for future time steps is 4×1 in dimension and includes the electric field intensity adjustment. Ammonia injection volume adjustment Adjustment amount of electrode spacing Expectorant fan frequency adjustment ; It is a control order identifier. It is a control cycle. It is the first 4×4 control gain matrix of order, It is a 4×10 order optimal control parameter tensor, which includes four types of parameters: electric field strength, ammonia injection rate, electrode spacing, and induced draft fan frequency. Each type of parameter contains the optimal value for 10 sampling periods. It is a 4×10 order real-time detection parameter tensor. It is the delay time constant of the actuator. It is the delay compensation coefficient. It is a 4×1 order reference control command vector. It is a 4×10-order nominal parameter tensor. It is a 4×1 order parameter deviation standard deviation vector. It is the integral time variable. It is a 4×4 dynamic compensation kernel function matrix. yes Real-time parameter change rate tensor It is a dynamic compensation item.

7. The integrated reaction control method for dust removal system coupled with flue gas denitrification according to claim 1, characterized in that, Step S5 involves real-time acquisition of the multi-dimensional effect parameters of the reaction terminal and the multi-scale error between the predicted values ​​and the model. The time-varying characteristic coefficient matrix of the fractional-order coupled reaction kinetic model is adjusted using a wavelet decomposition-fractional-integral composite correction formula and fed back to step S2 to form adaptive closed-loop control. The multi-dimensional effect parameters of the reaction terminal include outlet NOx concentration, dust concentration, system energy consumption, and catalyst activity. The wavelet decomposition-fractional-integral composite correction formula is as follows: ； In the formula, in the formula, It is the time-varying characteristic coefficient correction amount, used to update the time-varying parameters of the fractional-order dynamic model in step S2; This is the total correction factor, with a value of 0.

05. This is a wavelet decomposition level identifier, with values ​​from 1 to 5 corresponding to 5 levels of wavelet decomposition. They are wavelet basis numbers at the same decomposition level. It is the first The number of nodes in the layer wavelet decomposition satisfies , It is the first Layer The decomposition coefficient operator of the wavelet basis is adopted, using the db4 wavelet basis. It is a 4×10 order terminal feedback effect parameter tensor, which includes four types of parameters: outlet NOx concentration, dust concentration, system energy consumption, and catalyst activity. Each type of parameter contains measured values ​​for 10 sampling periods. It is the 4×10 order model prediction parameter tensor, and The dimensions are consistent, including the predicted values ​​of the four types of parameters calculated by the model; It is a multi-scale error tensor. It is the Riemann-Liouville fractional derivative operator. It is the first Layer The order of the fractional derivatives corresponding to each wavelet basis is adaptively adjusted according to the decomposition level. It is a 28×1 order initial characteristic coefficient matrix, which is the initial parameter of the dynamic model in step S2.

8. An integrated dust removal system coupled with flue gas denitrification reaction device, characterized in that, The system comprises a flue gas conveying module, a multi-field parameter detection module, a coupled reaction main module, a central control module, a multi-dimensional execution module, and a feedback adjustment module, all interconnected to form a closed-loop control system. Specifically: The flue gas conveying module includes an inlet flue, a flow regulating valve, an induced draft fan, and a flow equalization device connected in sequence, with the outlet of the flow equalization device connected to the coupled reaction main module; the multi-field parameter detection module is located in the inlet flue, the reaction main module, and the outlet flue, and includes temperature sensing and flow field testing units, transmitting raw parameters to the central control module via industrial Ethernet; the coupled reaction main module is a horizontal reactor, with electrostatic precipitators, denitrification catalysts, and other components installed along the flue gas flow direction. The reaction homogenization unit and the electrostatic precipitator unit include corona plates, dust collection plates, and a spacing adjustment mechanism. The denitrification catalytic unit consists of a staged catalyst layer and a matching ammonia injection grid. The central control module includes a data processing, model calculation, and optimization control unit, which implements the method described in claims 1-7 to generate instructions. The multi-dimensional execution module includes a high-voltage power supply, ammonia injection control, and other units, which are respectively connected to the electrostatic precipitator unit and the ammonia injection grid component, and execute the instructions of the central control module. The feedback adjustment module includes a terminal parameter acquisition and error correction unit, which acquires the outlet parameters and calculates the error with the model prediction value, and generates a model correction amount through a composite correction formula to feed back to the central control module.

9. A non-transitory computer-readable storage medium storing computer instructions that, when executed by at least one processor, cause the at least one processor to perform the method as described in any one of claims 1 to 7.

10. A computer program product comprising a computing program stored on a non-transitory computer-readable storage medium, the computer program including program instructions that, when executed by a computer, cause the computer to perform the method as described in any one of claims 1 to 7.

Images