CN112580250A - Thermal power generating unit denitration system based on deep learning and optimization control method - Google Patents
Thermal power generating unit denitration system based on deep learning and optimization control method Download PDFInfo
- Publication number
- CN112580250A CN112580250A CN202011261792.XA CN202011261792A CN112580250A CN 112580250 A CN112580250 A CN 112580250A CN 202011261792 A CN202011261792 A CN 202011261792A CN 112580250 A CN112580250 A CN 112580250A
- Authority
- CN
- China
- Prior art keywords
- control
- unit
- data
- nox concentration
- nox
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000005457 optimization Methods 0.000 title claims abstract description 35
- 238000013135 deep learning Methods 0.000 title claims abstract description 25
- QGZKDVFQNNGYKY-UHFFFAOYSA-N Ammonia Chemical compound N QGZKDVFQNNGYKY-UHFFFAOYSA-N 0.000 claims abstract description 137
- 229910021529 ammonia Inorganic materials 0.000 claims abstract description 67
- 238000002347 injection Methods 0.000 claims abstract description 58
- 239000007924 injection Substances 0.000 claims abstract description 58
- 239000011159 matrix material Substances 0.000 claims abstract description 20
- 230000004044 response Effects 0.000 claims abstract description 15
- 238000012843 least square support vector machine Methods 0.000 claims abstract description 12
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 claims description 34
- 239000003546 flue gas Substances 0.000 claims description 34
- 230000008569 process Effects 0.000 claims description 27
- 238000006243 chemical reaction Methods 0.000 claims description 26
- 239000002245 particle Substances 0.000 claims description 26
- 239000003245 coal Substances 0.000 claims description 25
- 230000006870 function Effects 0.000 claims description 20
- 238000002485 combustion reaction Methods 0.000 claims description 19
- 230000008859 change Effects 0.000 claims description 14
- 238000005070 sampling Methods 0.000 claims description 14
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims description 13
- 239000001301 oxygen Substances 0.000 claims description 13
- 229910052760 oxygen Inorganic materials 0.000 claims description 13
- 238000005516 engineering process Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 12
- 230000009471 action Effects 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000005259 measurement Methods 0.000 claims description 9
- 239000000243 solution Substances 0.000 claims description 9
- 238000011217 control strategy Methods 0.000 claims description 8
- 230000002068 genetic effect Effects 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 238000012546 transfer Methods 0.000 claims description 6
- 238000012886 linear function Methods 0.000 claims description 5
- 229910000069 nitrogen hydride Inorganic materials 0.000 claims description 5
- 238000005096 rolling process Methods 0.000 claims description 4
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 230000006978 adaptation Effects 0.000 claims description 3
- 230000001149 cognitive effect Effects 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 239000000446 fuel Substances 0.000 claims description 3
- 230000009191 jumping Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000005507 spraying Methods 0.000 abstract description 3
- MWUXSHHQAYIFBG-UHFFFAOYSA-N Nitric oxide Chemical compound O=[N] MWUXSHHQAYIFBG-UHFFFAOYSA-N 0.000 description 15
- 239000003054 catalyst Substances 0.000 description 9
- 238000004891 communication Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 230000007613 environmental effect Effects 0.000 description 5
- XKMRRTOUMJRJIA-UHFFFAOYSA-N ammonia nh3 Chemical group N.N XKMRRTOUMJRJIA-UHFFFAOYSA-N 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010531 catalytic reduction reaction Methods 0.000 description 2
- 239000003638 chemical reducing agent Substances 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000004134 energy conservation Methods 0.000 description 2
- 238000006479 redox reaction Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- IJGRMHOSHXDMSA-UHFFFAOYSA-N Atomic nitrogen Chemical compound N#N IJGRMHOSHXDMSA-UHFFFAOYSA-N 0.000 description 1
- 238000003915 air pollution Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000003197 catalytic effect Effects 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000003795 desorption Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 229910001873 dinitrogen Inorganic materials 0.000 description 1
- 244000144992 flock Species 0.000 description 1
- 230000002431 foraging effect Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000006722 reduction reaction Methods 0.000 description 1
- 238000001179 sorption measurement Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B01—PHYSICAL OR CHEMICAL PROCESSES OR APPARATUS IN GENERAL
- B01D—SEPARATION
- B01D53/00—Separation of gases or vapours; Recovering vapours of volatile solvents from gases; Chemical or biological purification of waste gases, e.g. engine exhaust gases, smoke, fumes, flue gases, aerosols
- B01D53/34—Chemical or biological purification of waste gases
- B01D53/74—General processes for purification of waste gases; Apparatus or devices specially adapted therefor
- B01D53/86—Catalytic processes
- B01D53/8621—Removing nitrogen compounds
- B01D53/8625—Nitrogen oxides
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B01—PHYSICAL OR CHEMICAL PROCESSES OR APPARATUS IN GENERAL
- B01D—SEPARATION
- B01D53/00—Separation of gases or vapours; Recovering vapours of volatile solvents from gases; Chemical or biological purification of waste gases, e.g. engine exhaust gases, smoke, fumes, flue gases, aerosols
- B01D53/34—Chemical or biological purification of waste gases
- B01D53/74—General processes for purification of waste gases; Apparatus or devices specially adapted therefor
- B01D53/86—Catalytic processes
- B01D53/8696—Controlling the catalytic process
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/25—Design optimisation, verification or simulation using particle-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B01—PHYSICAL OR CHEMICAL PROCESSES OR APPARATUS IN GENERAL
- B01D—SEPARATION
- B01D2251/00—Reactants
- B01D2251/20—Reductants
- B01D2251/206—Ammonium compounds
- B01D2251/2062—Ammonia
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Chemical & Material Sciences (AREA)
- Environmental & Geological Engineering (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Analytical Chemistry (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Oil, Petroleum & Natural Gas (AREA)
- General Chemical & Material Sciences (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Evolutionary Biology (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Physiology (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Genetics & Genomics (AREA)
- Treating Waste Gases (AREA)
- Feedback Control In General (AREA)
Abstract
The thermal power generating unit denitration system based on deep learning comprises a NOx concentration target value setting unit, a dynamic matrix control DMC control unit, a PID control unit, an ammonia injection valve, an SCR reactor, a hearth outlet NOx concentration prediction module unit and an intelligent feedforward controller unit. The optimization control method provided by the invention is used for carrying out data modeling on the basis of historical data and real-time data of target power plant operation, taking boiler side adjustable parameters as input, taking SCR outlet NOx concentration measured values as output, and constructing a prediction model by utilizing a least square support vector machine algorithm. According to the thermal power generating unit denitration system and the optimization control method, accurate ammonia spraying control of the SCR system is achieved, the problems of slow response time, large parameter fluctuation and the like of the system are solved, and the control quality of the system is ensured.
Description
Technical Field
The invention relates to a thermal power generating unit denitration system and an optimization control method, in particular to a thermal power generating unit denitration system based on deep learning and an optimization control method.
Background
The nitrogen oxide is one of the main pollution sources of air pollution, wherein the nitrogen oxide discharged by thermal power plants accounts for about 35-40% of the national nitrogen oxide discharge. The energy conservation and emission reduction of the thermal power generating unit conforms to the development trend of energy conservation and environmental protection at present, along with the increasing awareness of environmental protection of people, the treatment of the flue gas of the thermal power generating unit is paid social attention, and NOx is one of the main harmful substances of the flue gas.
Under the large background, the environmental protection department of China has developed a series of schemes to inhibit the emission of NOx in the thermal power plant, and the scheme stipulates that the content of NOx in each cubic meter of flue gas cannot exceed 50mg/Nm3This presents a great challenge to the environmentally friendly operation of thermal power plants. The denitration method of SCR (selective catalytic reduction denitration technology) becomes a mainstream technology of a denitration system of a domestic large-scale thermal power generating unit in recent years due to the advantage of higher denitration efficiency, and is adopted in denitration schemes of most domestic thermal power generating units due to maturity and high feasibility. In addition, with the large-scale grid connection of new energy such as wind power, photovoltaic and the like with strong volatility in China, the thermal power generating unit needs to perform rapid and deep variable load operation to improve the operation elasticity of the thermal power generating unit so as to improve the consumption efficiency of the new energy.
The NOx concentration in flue gas can be greatly changed when a thermal power unit operates due to the fact that load is lifted and lowered rapidly and on a large scale, the control difficulty is increased for a denitration control system, although a denitration task can be completed to a certain extent by a traditional SCR denitration control system, along with further improvement of environmental protection and operation elasticity requirements of the thermal power unit, the flue gas emission of the thermal power unit during operation needs to be guaranteed to reach the standard and meet economic requirements, the control system is more and more prominent in importance, the control quality of the thermal power unit is directly related to denitration operation cost and environmental protection index assessment, and therefore, the accurate SCR system model is established, and the control system is reasonable in design so as to achieve optimization of ammonia injection amount.
The SCR denitration system is mainly influenced by factors such as coal quality, reaction temperature, catalyst activity, ammonia injection flow and flue gas flow. The ammonia injection flow is a main controllable factor, the excessive ammonia injection amount can cause the NOx concentration at the SCR outlet to exceed the standard, and the excessive ammonia injection amount can increase the ammonia escape amount to cause NH3Water and SO2Reacting under the action of catalyst to generate NH4HSO4And (NH)4)2SO4The catalyst adheres to the surface of the catalyst, the activity of the catalyst is seriously influenced, and the air preheater is also blocked, so that the differential pressure of the air preheater is large. Therefore, in order to ensure safe, stable and efficient operation of the SCR denitration system, a reasonable control strategy is particularly important.
Because the SCR denitration system is closely coupled with the combustion of the hearth, the flue gas state at the outlet of the hearth determines the flue gas parameters of the inlet of the SCR system, including the flue gas temperature, the flue gas flow rate, the flue gas NOx content and the like, and the change of the parameters can influence the SCR denitration reaction process and the denitration efficiency. The state of the flue gas at the outlet of the furnace chamber is influenced by the operation condition of the boiler, including load and adjustable input parameters of the boiler. Under the background of large-scale renewable energy grid connection in the power industry at present, a thermal power generating unit frequently participates in peak shaving, and the peak shaving depth is larger. Therefore, the operation condition of the boiler frequently changes in a large range, so that the flue gas parameters at the outlet of the hearth frequently change. The detection of the flue gas parameter signals at the inlet and the outlet of the SCR system is lagged, and the reaction process of the SCR reactor is greatly delayed, so that the traditional PID control cannot obtain satisfactory control effect on the SCR denitration process which has the characteristics of complex reaction mechanism, nonlinearity, large delay, multivariable coupling and the like. Excessive ammonia injection can cause increased ammonia slip, leading to scaling and corrosion of the air preheater, and too little ammonia injection can cause the outlet NOx emission to fail the national emission requirements.
Disclosure of Invention
In order to overcome the defects of the technical problems, the invention provides a deep learning-based fire power unit denitration system and an optimization control method.
The thermal power generating unit denitration system based on deep learning comprises a NOx concentration target value setting unit, a dynamic matrix control DMC control unit, a PID control unit, an ammonia injection valve, an SCR reactor, a hearth outlet NOx concentration prediction module unit and an intelligent feedforward controller unit; the method is characterized in that: the input end of the furnace outlet NOx concentration prediction module unit is connected with boiler load, total air quantity, coal mill coal feeding quantity, total combustion quantity, oxygen quantity, secondary air and over-fire air signals, the NOx concentration prediction signal output by the furnace outlet NOx concentration prediction module unit is connected with the input end of the intelligent feedforward controller unit, and the input end of the intelligent feedforward controller unit is connected with an SCR inlet NOx concentration measurement value signal and an SCR outlet NOx concentration measurement value signal; the input end of the SCR reactor is connected with reaction temperature, flue gas flow and a furnace outlet NOx signal;
the NOx concentration target value output by the NOx concentration target value setting unit and the SCR outlet NOx concentration measured value are subjected to subtraction operation and then input to the input end of the dynamic matrix control DMC control unit; after the output signal of the dynamic matrix control DMC control unit is added and operated with the feedforward control quantity output by the intelligent feedforward controller unit, the output signal of the dynamic matrix control DMC control unit is subtracted and operated with the output signal of the ammonia injection valve and then input to the input end of the PID control unit, the output signal of the PID control unit is connected with the ammonia injection valve, and the opening degree of the ammonia injection valve is controlled; the output of the ammonia injection valve is connected with the input of the SCR reactor.
The optimal control method of the thermal power generating unit denitration system based on deep learning comprises the steps of firstly, carrying out data modeling on the basis of historical data and real-time data of target power plant operation, taking boiler side adjustable parameters as input, taking a NOx concentration measured value at an SCR outlet as output, and constructing a prediction model by utilizing a least square support vector machine algorithm, wherein the prediction model is used for constructing an intelligent feedforward controller unit in an ammonia injection amount control strategy; then, constructing a cascade feedback control structure by taking a dynamic matrix control DMC control unit as a main controller and a PID as a secondary controller; in the operation process, the intelligent feedforward controller unit outputs a feedforward control signal in real time according to the parameter change of the boiler side, quickly responds to the change of the working condition of the unit, and forms an optimized control strategy of the ammonia injection amount of the SCR system together with feedback control to realize quick and accurate control of the ammonia injection amount; optimizing input parameters by self-learning technologies such as a genetic algorithm, a particle swarm algorithm and the like, and establishing a NOx emission prediction model based on a least square support vector machine to perform system rolling optimization.
The invention discloses an optimal control method of a thermal power generating unit denitration system based on deep learning, wherein an intelligent feedforward controller unit comprises a hearth outlet NOx concentration prediction module and a feedforward control quantity conversion module, the hearth outlet NOx concentration prediction module predicts the hearth outlet NOx concentration in real time according to boiler side input parameters, and the feedforward control quantity conversion module converts a measured value of the hearth outlet NOx concentration in real time into a feedforward control quantity;
the feedforward control quantity conversion module converts the predicted value of the concentration of NOx at the outlet of the hearth into the variable quantity of the opening of the valve in real time, and the variable quantity is used as a feedforward control quantity, and the conversion formula is as follows:
Δy＝k·Δn+b (2)
wherein, Δ n is the valve opening variation, Δ cNOxIs the variation of the concentration of NOx at the outlet of the furnace, which is the difference between the real-time predicted value of the concentration of NOx at the outlet of the furnace and the actual value of the concentration of NOx at the current moment, and the unit is mg/Nm3(ii) a Q is the flue gas flow in m3H; eta is denitration efficiency; mNH3、MNOxAre each NH3And the molar mass of NOx; Δ y is the amount of change in ammonia injection amount; and a linear function is formed between the ammonia injection amount and the valve opening, k and b are fitting coefficients in the function, and the calculated ammonia injection amount variation delta y is subjected to inverse calculation to obtain the valve opening variation delta n.
According to the thermal power unit denitration system optimization control method based on deep learning, data construction of the hearth outlet NOx concentration prediction module unit is realized through the following steps:
a) boiler side input parameters are used as input variables of the model, and the concentration of NOx at the outlet of the hearth is used as an output variable of the model; selecting a unit load, a total air quantity and a total fuel quantity to represent a unit combustion characteristic, wherein a coal feeding amount primary air quantity of a coal mill influences boiler combustion, a secondary air door opening represents a secondary air distribution mode to influence combustion, an overfire air door opening represents a 2-layer overfire air influence, primary air and secondary air respectively influence the early-stage and later-stage combustion conditions of pulverized coal, the overfire air enables the lower part of a hearth to form a reducing atmosphere, the outlet NOx content is reduced, the influence of the furnace feeding oxygen quantity is described by using the flue gas oxygen content, and a boiler outlet NOx concentration prediction model is established; the boiler side input parameters comprise boiler load, total air quantity, coal feeding quantity of a coal mill, total combustion quantity, oxygen quantity, secondary air and over-fire air;
b) selecting stable historical operating data with a certain time length from the historical data of the unit, preprocessing the data, selecting the historical operating data within the range of 160 MW-330 MW of unit load, with the sampling interval of 5s, after eliminating abnormal data, taking 40000 groups of data as samples, wherein 36000 groups of data are training data, and the rest 4000 groups of data are test data, and according to the formula: x ═ 2x-xmax-xmin)/(xmax-xmin) Normalizing data to [ -1, 1]After obtaining the output, the formula x ═ x' (x) is usedmax-xmin)+xmax+xmin]And/2, carrying out inverse normalization.
c) Training and modeling the data by adopting a least square support vector machine algorithm, and verifying the model accuracy; establishing linear transfer function models on a plurality of load points through step response tests on the plurality of load points, wherein the model of the intermediate load can be calculated through the established linear transfer function models on the adjacent load points by an interpolation method; the ammonia injection valve has a nonlinear relation between the opening and the ammonia injection flow, and a nonlinear function between the valve opening and the ammonia flow is fitted by a conventional data fitting method under the condition of the pressure before the ammonia injection valve.
The invention discloses an optimal control method of a thermal power generating unit denitration system based on deep learning, wherein a predictive control optimization algorithm of a dynamic matrix control DMC control unit is as follows:
the DMC controller unit adopts DMC prediction control optimizing selection range [ Uk, min, Uk, max ], wherein Uk, min is max [ u (k-1) -delta max, UL ]; uk, max ═ min [ u (k-1) + Δ max, UH ], max (), min () are large and small functions, respectively; k represents the kth sampling moment, and if the sampling period is T, u (k) represents u (kT), namely the control action of the kT moment;
the predicted value output by the process at each future sampling moment is as follows:
in the formula: y (k) represents the process output at time kT, ξ is a perturbation, G () is a non-linear function fitted based on experimental data, m and n are the input and output orders of the non-linear process, assuming the future control action of the process is fixed, assuming u is constantF(ii) a The future disturbance of the process is fixed, namely: ξ (k) ═ ξ (k + i), i ═ 1, 2.., N;
the performance index J of the predictive control is defined as:
in the formula: r is a set value of the control system; n is the maximum prediction step number in the future, and lambda is the weight coefficient for controlling increment; the control component ui (i ═ 0, 1.., L) is determined to be the closest to the optimal control component uopControl component u ofMThe performance index JM corresponding to the performance index JM meets the following requirements:
and (3) on the basis of taking the step response of the controlled object as a model, implementing step change on the ammonia injection amount by the SCR reactor, and obtaining a step response curve of the outlet NOx concentration.
The optimal control method of the thermal power unit denitration system based on deep learning disclosed by the invention adopts a particle group optimization algorithm to obtain an optimal solution, and the process is as follows:
1) firstly, initializing a population size m, the number' N of optimized variables, an optimized algebra t, the range of an optimized variable X, a forgetting factor w, a learning factor c and the like;
2) the fitness Q of the initial population is expressed as:
in the formula: y (i) is the ith output test data collected; y isd(i) Outputting data for the identification under the action of the collected ith input; m is the number of the collected test data;
3) optimal position of particle i: xbesti＝(xi1,xi2,...,xiN) Adaptation value QbestiThe best position is represented as:
4) velocity vinAnd position xinThe updating is performed according to the following formula:
xin(t+1)＝xin(t)+vin(t+1) (8)
in the formula: c. C1Is a cognitive factor; c. C2Is a social factor, the value of which is a non-negative constant; i is 1, 2, …, m; n is 1, 2, …, N; t is the tth generation; r is1、r2Are independent pseudo-random numbers, whose values are taken from [0, 1 ]]Uniformly distributing;
is the inertial weight; xbestinSearching the optimal position of the particle i for the nth time; xbestgnOptimizing the optimal position of the particle swarm experience for the nth time;
5) calculating the new fitness of each particle in the population;
6) comparison XbestinAnd XbestgnIf the value is superior, the value is replaced again;
7) judging whether the requirements of precision and evolution algebra are met, and if so, jumping out of the loop; if not, the step 4) is carried out, and the program is continuously executed until the parameter is met, so that the parameter optimization is realized.
The invention has the beneficial effects that: according to the thermal power unit denitration system based on deep learning and the optimization control method, the traditional control method is used for reference according to the operation condition of the SCR denitration system of the power plant, on the basis of researching a field engineering actual control system, the relevant SCR denitration system control theory is combined, field historical data is collected, on the basis of keeping a field actual control algorithm unchanged, an intelligent identification method is adopted, advanced control technologies such as a deep self-learning technology, a prediction control technology and a parameter self-setting technology are exerted, a mathematical model of the SCR denitration system is established, a highly-engineered human-simulated intelligent algorithm is adopted, accurate ammonia injection control of the SCR system is realized, the problems of slow system response time, large parameter fluctuation and the like are solved, and the control quality of the system is ensured. Through the practical application of the project, the fluctuation amount of NOx is effectively reduced, the response speed of a control system is improved, the ammonia gas usage amount is effectively reduced, the equipment health level is kept, the service life of a unit is prolonged, the optimal operation working condition is kept, the economy of the unit is improved, obvious economic benefits are achieved, and the stable and economic operation of an SCR denitration system is ensured.
Drawings
FIG. 1 is a basic flow diagram of a prior art SCR denitration system;
FIG. 2 is a schematic diagram of PID control adopted by a conventional SCR denitration system;
FIG. 3 is a schematic diagram of the deep learning theory-based method of the present invention;
fig. 4 is a schematic diagram of a thermal power generating unit denitration system based on deep learning.
In the figure: the method comprises the following steps of 1NOx concentration target value setting unit, 2 dynamic matrix control DMC control unit, 3PID control unit, 4 ammonia injection valve, 5SCR reactor, 6 valve opening degree constraint unit, 7 furnace outlet NOx concentration prediction module unit and 8 intelligent feedforward controller unit.
Detailed Description
The invention is further described with reference to the following figures and examples.
As shown in fig. 1, a basic flow chart of the existing SCR denitration system is given, the SCR flue gas denitration system is arranged between the economizer and the air preheater, the flue gas temperature in the region is usually 300-500 ℃, and the catalyst activity of the SCR flue gas denitration system is stronger, which is more beneficial for the oxidation-reduction reaction of ammonia gas and NOx. The SCR flue gas denitration system mainly comprises an ammonia spraying grid, a guide plate, a rectifier, a catalyst, an SCR reactor and the like. The flue gas is led out from the horizontal flue at the outlet of the economizer, enters the vertical joint of the ascending flue at the inlet of the SCR reactor and is mixed with ammonia gas sprayed from the ammonia spraying grid, and is vertically downward through the rectifier, and NOx in the flue gas is converted into nitrogen gas and water through the oxidation-reduction reaction of a denitration catalyst arranged in the SCR reactor, so that the aim of flue gas denitration is fulfilled. Mainly comprises the following reactions:
4NO+4NH3+O2→4N2+6H2O (1)
6NO+4NH3→5N2+6H2O (2)
6NO2+8NH3→7N2+12H2O (3)
2NO2+4NH3+O2→3N2+6H2O (4)
about 95% of NOx in flue gas at the tail part of a boiler in a generator set of a coal power plant exists in the form of NO, so the influence of NO is mainly considered when SCR denitration is researched. The boiler flue gas oxygen content is typically above 2.5%, which is present and in excess for the SCR denitration reaction. Therefore, equation (1) is the most dominant reaction in the SCR flue gas denitration process.
As shown in FIG. 2, a schematic diagram of PID control adopted by the prior SCR denitration system is provided, which showsThe basic control strategy is: the concentration of nitrogen oxides at the inlet and the outlet of an SCR system is collected, and then PID feedback control is carried out. The set value under the control mode is ammonia nitrogen molar ratio or denitration efficiency, and the control system calculates NH according to the current flue gas flow, the concentration of NOx at an SCR inlet and the set ammonia nitrogen molar ratio3Flow demand, and finally NH is adjusted by changing the opening degree of an ammonia valve through flow PID3Actual flow, this control approach approximates open loop control. The ammonia nitrogen molar ratio is an index for evaluating the economy of the SCR denitration system, and NH is added under the same denitration efficiency3The higher the/NOx molar ratio, the more reducing agent is required and the higher the cost of the denitration operation. In fact due to denitration of the controlled object (NH)3Flow → chimney inlet NOx concentration) is close to 3 minutes, the whole response process is as long as ten minutes and several minutes, which is a typical large-lag controlled object, and the SCR denitration process itself is a complex nonlinear chemical reaction process, and as the catalytic agent is continuously consumed, the dynamic characteristic of the controlled process of denitration can be greatly changed. Therefore, it is difficult to achieve a desired control quality using the conventional PID control scheme. There are mainly the following problems:
1. the inlet and outlet nitrogen oxide concentration sampling pipelines are relatively long, so that the measurement purity delay is relatively large, the measurement value generally has 2-3 minutes delay, and particularly, the inlet NOx concentration is the most direct factor influencing the denitration efficiency.
2. The existing system adopts a sampling measurement mode, so that the measurement data cannot represent the average concentration of the whole section.
3. By adopting PID control, PID parameters are not changed after initial setting, so that when the load operation condition of the unit changes, the denitration system is not adjusted in time, and standard-exceeding emission is easy to occur.
4. In order to ensure the standard emission, the set value of the PID parameter is generally lower, which causes the ammonia injection amount to be overlarge, the overshoot of the control system is larger, the response rate of the system is lower, the reducing agent is wasted, and the risk of subsequent equipment blockage is increased.
The SCR denitration system has the characteristics of complex reaction mechanism, large inertia, nonlinearity and the like, when a denitration control model of the thermoelectric generator set is established, due to the fact that the physical phenomena of adsorption and desorption exist in the reactor and the chemical reaction of catalytic reduction, the concentration, the temperature and the catalyst activity in the reactor can affect the reaction, the input and the output of the model have the characteristics of nonlinearity and strong coupling, and the establishment of a mechanism model capable of accurately reflecting the nonlinear relation between the input and the output of the model is very difficult. However, a DCS (distributed control system) system stores massive operation historical data, wherein massive useful information is contained, so that an SCR system model can be established by using an experimental modeling method, and a more accurate calculation result is obtained.
As shown in fig. 4, a schematic diagram of a thermal power generating unit denitration system based on deep learning of the present invention is provided, which is composed of a NOx concentration target value setting unit 1, a dynamic matrix control DMC control unit 2, a PID control unit 3, an ammonia injection valve 4, an SCR reactor 5, a furnace outlet NOx concentration prediction module unit 7 and an intelligent feedforward controller unit 8, wherein an input end of the furnace outlet NOx concentration prediction module unit is connected with boiler load, total air volume, coal mill coal feeding amount, total combustion amount, oxygen amount, secondary air and over-fire air signals, a NOx concentration prediction signal output by the furnace outlet NOx concentration prediction module unit is connected with an input end of the intelligent feedforward controller unit, and an input end of the intelligent feedforward controller unit is connected with an SCR inlet NOx concentration measurement value signal and an SCR outlet NOx concentration measurement value signal; the input end of the SCR reactor is connected with a reaction temperature, a flue gas flow and a furnace outlet NOx signal.
The NOx concentration target value output by the NOx concentration target value setting unit (1) and the SCR outlet NOx concentration measured value are subjected to subtraction operation and then input to the input end of the dynamic matrix control DMC control unit; after the output signal of the dynamic matrix control DMC control unit is added and operated with the feedforward control quantity output by the intelligent feedforward controller unit, the output signal of the dynamic matrix control DMC control unit is subtracted and operated with the output signal of the ammonia injection valve and then input to the input end of the PID control unit, the output signal of the PID control unit is connected with the ammonia injection valve, and the opening degree of the ammonia injection valve is controlled; the output of the ammonia injection valve is connected with the input of the SCR reactor.
The invention discloses an optimal control method of a thermal power generating unit denitration system based on deep learning, which is realized by the following steps:
firstly, performing data modeling on the basis of historical data and real-time data of target power plant operation, taking boiler side adjustable parameters as input, taking a NOx concentration measured value at an SCR outlet as output, and constructing a prediction model by utilizing a least square support vector machine algorithm, wherein the prediction model is used for constructing an intelligent feedforward controller unit in an ammonia injection amount control strategy; then, constructing a cascade feedback control structure by taking a dynamic matrix control DMC control unit as a main controller and a PID as a secondary controller; in the operation process, the intelligent feedforward controller unit outputs a feedforward control signal in real time according to the parameter change of the boiler side, quickly responds to the change of the working condition of the unit, and forms an optimized control strategy of the ammonia injection amount of the SCR system together with feedback control to realize quick and accurate control of the ammonia injection amount; optimizing input parameters by self-learning technologies such as a genetic algorithm, a particle swarm algorithm and the like, and establishing a NOx emission prediction model based on a least square support vector machine to carry out system rolling optimization.
An intelligent algorithm is adopted to establish a prediction model of NOx emission of the coal-fired unit, comparative analysis is carried out, and accurate ammonia injection control is realized by adopting a prediction control technology on the basis of prediction. The denitration system is subjected to model identification through system history and real-time operation data, the operation characteristics of a closed-loop system are better reflected, meanwhile, a humanoid intelligent algorithm with strong engineering performance is adopted, model parameters are optimized by using a genetic algorithm on the basis that a boiler NOx emission model is established by using a least square support vector machine LS-SVM, and the model parameters are compared and analyzed with the real-time performance and the accuracy of data modeling. Optimizing the input parameters by the genetic algorithm and the particle swarm algorithm, and analyzing and comparing the convergence speed and accuracy of the optimized NOx prediction model parameters; a prediction model of NOx emission is established by adopting a least square support vector machine LS-SVM, and the accurate control and rolling optimization of ammonia injection amount are realized by utilizing a prediction control technology.
The intelligent feedforward controller unit comprises a hearth outlet NOx concentration prediction module and a feedforward control amount conversion module, wherein the hearth outlet NOx concentration prediction module predicts the hearth outlet NOx concentration in real time according to boiler side input parameters, and the feedforward control amount conversion module converts a measured value of the hearth outlet NOx concentration in real time into a feedforward control amount;
the feedforward control quantity conversion module converts the predicted value of the concentration of NOx at the outlet of the hearth into the variable quantity of the opening of the valve in real time, and the variable quantity is used as a feedforward control quantity, and the conversion formula is as follows:
Δy＝k·Δn+b (2)
wherein, Δ n is the valve opening variation, Δ cNOxIs the variation of the concentration of NOx at the outlet of the furnace, which is the difference between the real-time predicted value of the concentration of NOx at the outlet of the furnace and the actual value of the concentration of NOx at the current moment, and the unit is mg/Nm3(ii) a Q is the flue gas flow in m3H; eta is denitration efficiency;
MNOxare each NH3And the molar mass of NOx; Δ y is the amount of change in ammonia injection amount; and a linear function is formed between the ammonia injection amount and the valve opening, k and b are fitting coefficients in the function, and the calculated ammonia injection amount variation delta y is subjected to inverse calculation to obtain the valve opening variation delta n.
The data construction of the furnace outlet NOx concentration prediction module unit (7) is realized by the following steps:
a) boiler side input parameters are used as input variables of the model, and the concentration of NOx at the outlet of the hearth is used as an output variable of the model; selecting a unit load, a total air quantity and a total fuel quantity to represent a unit combustion characteristic, wherein a coal feeding amount primary air quantity of a coal mill influences boiler combustion, a secondary air door opening represents a secondary air distribution mode to influence combustion, an overfire air door opening represents a 2-layer overfire air influence, primary air and secondary air respectively influence the early-stage and later-stage combustion conditions of pulverized coal, the overfire air enables the lower part of a hearth to form a reducing atmosphere, the outlet NOx content is reduced, the influence of the furnace feeding oxygen quantity is described by using the flue gas oxygen content, and a boiler outlet NOx concentration prediction model is established; the boiler side input parameters comprise boiler load, total air quantity, coal feeding quantity of a coal mill, total combustion quantity, oxygen quantity, secondary air and over-fire air;
b) selecting stable historical operating data with a certain time length from the historical data of the unit, preprocessing the data, selecting the historical operating data within the range of 160 MW-330 MW of unit load, with the sampling interval of 5s, after eliminating abnormal data, taking 40000 groups of data as samples, wherein 36000 groups of data are training data, and the rest 4000 groups of data are test data, and according to the formula: x ═ 2x-xmax-xmin)/(xmax-xmin) Normalizing data to [ -1, 1]After obtaining the output, the formula x ═ x' (x) is usedmax-xmin)+xmax+xmin]And/2, carrying out inverse normalization.
c) Training and modeling the data by adopting a least square support vector machine algorithm, and verifying the model accuracy; establishing linear transfer function models on a plurality of load points through step response tests on the plurality of load points, wherein the model of the intermediate load can be calculated through the established linear transfer function models on the adjacent load points by an interpolation method; the ammonia injection valve has a nonlinear relation between the opening and the ammonia injection flow, and a nonlinear function between the valve opening and the ammonia flow is fitted by a conventional data fitting method under the condition of the pressure before the ammonia injection valve.
The predictive control optimization algorithm of the dynamic matrix control DMC control unit (2) is as follows:
the DMC controller unit adopts DMC prediction control optimizing selection range [ Uk, min, Uk, max ], wherein Uk, min is max [ u (k-1) -delta max, UL ]; uk, max ═ min [ u (k-1) + Δ max, UH ], max (), min () are large and small functions, respectively; k represents the kth sampling moment, and if the sampling period is T, u (k) represents u (kT), namely the control action of the kT moment;
the predicted value output by the process at each future sampling moment is as follows:
in the formula: y (k) represents the process output at time kT, ξ is a perturbation, G () is a non-linear function fitted based on experimental data, m and n are the input and output orders of the non-linear process, assuming the future control action of the process is fixed, assuming u is constantF(ii) a The future disturbance of the process is fixed, namely: ξ (k) ═ ξ (k + i), i ═ 1, 2.., N;
the performance index J of the predictive control is defined as:
in the formula: r is a set value of the control system; n is the maximum prediction step number in the future, and lambda is the weight coefficient for controlling increment; the control component ui (i ═ 0, 1.., L) is determined to be the closest to the optimal control component uopControl component u ofMThe performance index JM corresponding to the performance index JM meets the following requirements:
and (3) on the basis of taking the step response of the controlled object as a model, implementing step change on the ammonia injection amount by the SCR reactor, and obtaining a step response curve of the outlet NOx concentration.
By applying the self-learning setting technology of the control parameters, the self-adaptive capacity of the thermal power generating unit denitration system controller is improved. The system adopts a genetic algorithm and a particle swarm algorithm to optimize input parameters, analyzes and compares the convergence speed and accuracy of NOx prediction model parameters, corrects various characteristic parameters closely related to a control system in the running of a unit in real time, and calculates various control parameters in a loop of the control system in real time according to the characteristic parameters, so that the whole system is always in an online learning state, the control performance is continuously close to an optimal target, and the response speed and the capacity of resisting external interference of the system are improved.
Aiming at the problems, the optimal solution is obtained by adopting a Particle Swarm Optimization (PSO) algorithm in the project. The PSO algorithm is an optimization algorithm for simulating foraging of a bird flock, belongs to one of evolutionary algorithms, is based on a random solution, considers an optimal solution of an optimization problem as a particle in a solution space, and determines the optimal solution by continuously searching a high-dimensional space. During the search process, each particle has a fitness corresponding to it through the calculation of a fitness function. Meanwhile, the particle determines the searching direction and step length through the speed function, and then the particle continuously searches in a solution space along with the current optimal particle, so that the PSO algorithm has the advantages of simplicity, easiness in implementation and few adjusting parameters.
The optimal solution is obtained by adopting a particle swarm optimization algorithm, and the process is as follows:
1) firstly, initializing a population size m, the number' N of optimized variables, an optimized algebra t, the range of an optimized variable X, a forgetting factor w, a learning factor c and the like;
2) the fitness Q of the initial population is expressed as:
in the formula: y (i) is the ith output test data collected; y isd(i) Outputting data for the identification under the action of the collected ith input; m is the number of the collected test data;
3) optimal position of particle i: xbesti＝(xi1,xi2,...,xiN) Adaptation value QbestiThe best position is represented as:
4) velocity vinAnd position xinThe updating is performed according to the following formula:
xin(t+1)＝xin(t)+vin(t+1) (8)
in the formula: c. C1Is a cognitive factor; c. C2Is a social factor, the value of which is a non-negative constant; i is 1, 2, …, m; n is 1, 2, …, N; t is the tth generation; r is1、r2Are independent pseudo-random numbers, whose values are taken from [0, 1 ]]Uniformly distributing;
is the inertial weight; xbestinSearching the optimal position of the particle i for the nth time; xbestgnOptimizing the optimal position of the particle swarm experience for the nth time;
5) calculating the new fitness of each particle in the population;
6) comparison XbestinAnd XbestgnIf the value is superior, the value is replaced again;
7) judging whether the requirements of precision and evolution algebra are met, and if so, jumping out of the loop; if not, the step 4) is carried out, and the program is continuously executed until the parameter is met, so that the parameter optimization is realized.
The implementation of the scheme is realized based on a denitration optimization control system, the core of the thermal power generating unit denitration system based on deep learning is a high-performance server, parameters (boiler operation parameters such as unit load, primary air volume, secondary air volume, total coal volume and coal mill operation state) required by calculation are obtained from a DCS through a data acquisition card, and after calculation, the parameters are returned to the original DCS through an AO/DO output module, so that closed-loop control is realized. The controller is communicated with the original DCS system in communication modes such as ModBus 485 and the like, and can be in two-way communication with a domestic mainstream DCS system. The denitration optimization controller and the DCS field controller can realize undisturbed switching.
The optimization controller mainly comprises a system core algorithm computing module, an I/O input/output module, a communication module and the like, wherein the system communication module is mainly responsible for realizing data acquisition communication with the DCS; the core algorithm module mainly realizes calculation and optimization of various control algorithms. When the system runs, firstly, the running data is collected through the system communication module and then input into the core algorithm module, and the core algorithm module is output to the DCS through the I/O, so that closed-loop control is realized.
The thermal power unit denitration system based on deep learning is comprehensive multipurpose software integrating modeling, debugging and facing to a thermodynamic system object, an algorithm library is constructed by using a computer programming language mixed programming algorithm with the help of Visual Stadio software on the basis of a computer operating system, and a system model function can be modified and expanded conveniently for a user. The thermal power generating unit denitration system based on deep learning is complete in function, high in platform integration level and small in occupied space of resources, and can run on Windows98 and above operating system versions.
Claims (6)
1. A thermal power generating unit denitration system based on deep learning comprises a NOx concentration target value setting unit (1), a dynamic matrix control DMC control unit (2), a PID control unit (3), an ammonia injection valve (4), an SCR reactor (5), a hearth outlet NOx concentration prediction module unit (7) and an intelligent feedforward controller unit (8); the method is characterized in that: the input end of the furnace outlet NOx concentration prediction module unit is connected with boiler load, total air quantity, coal mill coal feeding quantity, total combustion quantity, oxygen quantity, secondary air and over-fire air signals, a NOx concentration prediction signal output by the furnace outlet NOx concentration prediction module unit is connected with the input end of the intelligent feedforward controller unit, and the input end of the intelligent feedforward controller unit is connected with an SCR inlet NOx concentration measured value signal and an SCR outlet NOx concentration measured value signal; the input end of the SCR reactor is connected with a reaction temperature, a flue gas flow and a furnace outlet NOx signal;
the NOx concentration target value output by the NOx concentration target value setting unit (1) and the SCR outlet NOx concentration measured value are subjected to subtraction operation and then input to the input end of a Dynamic Matrix Control (DMC) control unit; after the output signal of the dynamic matrix control DMC control unit is added and operated with the feedforward control quantity output by the intelligent feedforward controller unit, the output signal of the dynamic matrix control DMC control unit is subtracted and operated with the output signal of the ammonia injection valve and then input to the input end of the PID control unit, the output signal of the PID control unit is connected with the ammonia injection valve, and the opening degree of the ammonia injection valve is controlled; the output of the ammonia injection valve is connected with the input of the SCR reactor.
2. The thermal power generating unit denitration system optimization control method based on deep learning of claim 1 is realized by the following steps:
firstly, performing data modeling on the basis of historical data and real-time data of target power plant operation, taking boiler side adjustable parameters as input, taking a NOx concentration measured value at an SCR outlet as output, and constructing a prediction model by utilizing a least square support vector machine algorithm, wherein the prediction model is used for constructing an intelligent feedforward controller unit in an ammonia injection amount control strategy; then, constructing a cascade feedback control structure by taking a dynamic matrix control DMC control unit as a main controller and a PID as a secondary controller; in the operation process, the intelligent feedforward controller unit outputs a feedforward control signal in real time according to the parameter change of the boiler side, quickly responds to the change of the working condition of the unit, and forms an optimized control strategy of the ammonia injection amount of the SCR system together with feedback control to realize quick and accurate control of the ammonia injection amount; optimizing the input parameters by using self-learning technologies such as a genetic algorithm, a particle swarm algorithm and the like, and establishing a NOx emission prediction model based on a least square support vector machine to perform system rolling optimization.
3. The thermal power generating unit denitration system optimization control method based on deep learning of claim 2, wherein the intelligent feedforward controller unit comprises a furnace outlet NOx concentration prediction module and a feedforward control amount conversion module, the furnace outlet NOx concentration prediction module predicts the furnace outlet NOx concentration in real time according to boiler side input parameters, and the feedforward control amount conversion module converts the real-time prediction value of the furnace outlet NOx concentration measurement value into a feedforward control amount;
the feedforward control quantity conversion module converts the predicted value of the concentration of NOx at the outlet of the hearth into the variable quantity of the opening of the valve in real time, and the variable quantity is used as a feedforward control quantity, and the conversion formula is as follows:
Δy＝k·Δn+b (2)
wherein, Δ n is the valve opening variation, Δ cNOxIs the variation of the concentration of NOx at the outlet of the hearth, is the difference between the real-time predicted value of the concentration of the NOx at the outlet of the hearth and the actual value of the concentration of the NOx at the current moment, and has the unit of mg/Nm3(ii) a Q is the flue gas flow in m3H; eta is denitration efficiency;
MNOxare each NH3And the molar mass of NOx; Δ y is the amount of change in ammonia injection amount; and a linear function is formed between the ammonia injection amount and the valve opening, k and b are fitting coefficients in the function, and the calculated ammonia injection amount variation delta y is subjected to inverse calculation to obtain the valve opening variation delta n.
4. The thermal power generating unit denitration system optimization control method based on deep learning of claim 2 or 3, wherein the data construction of the furnace outlet NOx concentration prediction module unit (7) is realized by the following steps:
a) boiler side input parameters are used as input variables of the model, and the concentration of NOx at the outlet of the hearth is used as an output variable of the model; selecting a unit load, a total air quantity and a total fuel quantity to represent a unit combustion characteristic, wherein a coal feeding quantity primary air quantity of a coal mill has an influence on boiler combustion, a secondary air door opening represents a secondary air distribution mode to influence combustion, a burnout air door opening represents a 2-layer burnout air influence, primary air and secondary air respectively influence the early-stage and later-stage combustion conditions of pulverized coal, the burnout air enables the lower part of a hearth to form a reducing atmosphere, the outlet NOx content is reduced, the influence of the furnace feeding oxygen quantity is described by using the flue gas oxygen content, and a boiler outlet NOx concentration prediction model is established; the boiler side input parameters comprise boiler load, total air quantity, coal feeding quantity of a coal mill, total combustion quantity, oxygen quantity, secondary air and over-fire air;
b) selecting stable historical operation data with a certain time length from the historical data of the unit, preprocessing the data, and selecting the range of 160 MW-330 MW of unit loadThe sampling interval of the internal historical operating data is 5s, after abnormal data is removed, 40000 groups of data are used as samples, wherein 36000 groups of data are training data, and 4000 groups of data are left as testing data, and according to a formula: x ═ 2x-xmax-xmin)/(xmax-xmin) Normalizing data to [ -1, 1]After obtaining the output, the formula x ═ x' (x) is usedmax-xmin)+xmax+xmin]And/2, carrying out reverse normalization.
c) Training and modeling the data by adopting a least square support vector machine algorithm, and verifying the accuracy of the model; establishing linear transfer function models on a plurality of load points through step response tests on the plurality of load points, wherein the model of the intermediate load can be obtained through the established linear transfer function models on the adjacent load points through calculation by an interpolation method; the ammonia injection valve has a nonlinear relation between the opening and the ammonia injection flow, and a nonlinear function between the valve opening and the ammonia flow is fitted by a conventional data fitting method under the condition of the pressure before the ammonia injection valve.
5. The thermal power generating unit denitration system optimization control method based on deep learning of claim 2 or 3, wherein the prediction control optimization algorithm of the dynamic matrix control DMC control unit (2) is as follows:
the DMC controller unit adopts DMC prediction control optimizing selection range [ Uk, min, Uk, max ], wherein Uk, min is max [ u (k-1) -delta max, UL ]; uk, max ═ min [ u (k-1) + Δ max, UH ], max (), min () are large and small functions, respectively; k represents the kth sampling moment, and if the sampling period is T, u (k) represents u (kT), namely the control action of the kT moment;
the predicted value output by the process at each future sampling moment is as follows:
in the formula: y (k) represents the process output at time kT, ξ is a perturbation, G () is a nonlinear function based on a fitting of experimental data, m and n are the input and output orders of the nonlinear process, assuming that the future control action of the process is fixed, assuming u is constantF(ii) a The future disturbance of the process is fixed, namely: ξ (k) ═ ξ (k + i), i ═ 1, 2.., N;
the performance index J of the predictive control is defined as:
in the formula: r is a set value of the control system; n is the maximum prediction step number in the future, and lambda is the weight coefficient of the control increment; the closest optimal control action u is found out of the individual control components ui (i ═ 0, 1.., L)opControl component u ofMThe performance index JM corresponding to the performance index JM meets the following requirements:
and (3) on the basis of taking the step response of the controlled object as a model, implementing step change on the ammonia injection amount by the SCR reactor, and obtaining a step response curve of the outlet NOx concentration.
6. The thermal power generating unit denitration system optimization control method based on deep learning of claim 2 or 3, characterized in that the particle swarm optimization algorithm is adopted to obtain the optimal solution, and the process is as follows:
1) firstly, initializing a population size m, the number' N of optimized variables, an optimized algebra t, the range of an optimized variable X, a forgetting factor w, a learning factor c and the like;
2) the fitness Q of the initial population is expressed as:
in the formula: y (i) is the ith output test data collected; y isd(i) Outputting data for the identification under the action of the ith input; m is the number of the collected test data;
3) optimal position of particle i: xbesti＝(xi1,xi2,...,xiN) Adaptation value QbestiThe best position is represented as:
4) velocity vinAnd position xinThe updating is performed according to the following formula:
xin(t+1)＝xin(t)+vin(t+1) (8)
in the formula: c. C1Is a cognitive factor; c. C2Is a social factor, the value of which is a non-negative constant; i is 1, 2, …, m; n is 1, 2, …, N; t is the tth generation; r is1、r2Are mutually independent pseudo-random numbers whose values obey [0, 1 ]]Uniformly distributing;
is the inertial weight; xbestinSearching the optimal position of the particle i for the nth time; xbestgnOptimizing the optimal position of the particle swarm experience for the nth time;
5) calculating the new fitness of each particle in the population;
6) comparison XbestinAnd XbestgnIf the value is superior, the value is replaced again;
7) judging whether the requirements of precision and evolution algebra are met, and if so, jumping out of the loop; if not, the step 4) is carried out, and the program is continuously executed until the parameter is met, so that the parameter optimization is realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011261792.XA CN112580250A (en) | 2020-11-12 | 2020-11-12 | Thermal power generating unit denitration system based on deep learning and optimization control method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011261792.XA CN112580250A (en) | 2020-11-12 | 2020-11-12 | Thermal power generating unit denitration system based on deep learning and optimization control method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112580250A true CN112580250A (en) | 2021-03-30 |
Family
ID=75122858
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011261792.XA Pending CN112580250A (en) | 2020-11-12 | 2020-11-12 | Thermal power generating unit denitration system based on deep learning and optimization control method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112580250A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113082969A (en) * | 2021-04-07 | 2021-07-09 | 唐锦婷 | Exhaust gas treatment system |
| CN113426264A (en) * | 2021-07-15 | 2021-09-24 | 国电环境保护研究院有限公司 | Intelligent operation control method and control platform for flue gas purification island |
| CN113433911A (en) * | 2021-06-30 | 2021-09-24 | 浙江大学 | Denitration device ammonia injection accurate control system and method based on concentration accurate prediction |
| CN113592163A (en) * | 2021-07-22 | 2021-11-02 | 汉谷云智(武汉)科技有限公司 | Method and equipment for predicting concentration of nitrogen oxide at inlet of SCR (Selective catalytic reduction) denitration reactor |
| CN113847609A (en) * | 2021-09-29 | 2021-12-28 | 华南理工大学 | Garbage incinerator denitration control method based on feedforward model prediction error self-correction |
| CN113970937A (en) * | 2021-09-22 | 2022-01-25 | 同兴环保科技股份有限公司 | SCR denitration accurate ammonia injection control method |
| CN114159959A (en) * | 2021-10-29 | 2022-03-11 | 广东惠州天然气发电有限公司 | Denitration control method and device for gas-steam combined cycle unit |
| CN114225662A (en) * | 2021-12-07 | 2022-03-25 | 国网河北能源技术服务有限公司 | Flue gas desulfurization and denitrification optimization control method based on hysteresis model |
| CN114326387A (en) * | 2021-12-07 | 2022-04-12 | 江苏方天电力技术有限公司 | Denitration control device and method for thermal power generating unit |
| CN114326616A (en) * | 2021-12-22 | 2022-04-12 | 大唐东北电力试验研究院有限公司 | Industrial process control method based on multivariate load improved dynamic matrix prediction algorithm |
| CN114442486A (en) * | 2022-01-12 | 2022-05-06 | 华能罗源发电有限责任公司 | SCR denitration optimization control system and control method for thermal power generating unit |
| CN114609986A (en) * | 2022-03-16 | 2022-06-10 | 中国中材国际工程股份有限公司 | Cement decomposing furnace denitration regulation and control optimization system and method based on predictive control |
| CN114664389A (en) * | 2022-03-24 | 2022-06-24 | 南方电网电力科技股份有限公司 | Method and device for predicting reaction conditions for urea hydrolysis ammonia production |
| CN114878745A (en) * | 2022-04-25 | 2022-08-09 | 北京市生态环境保护科学研究院 | Gas boiler nitrogen oxide prediction monitoring system based on support vector machine |
| CN115155310A (en) * | 2022-07-13 | 2022-10-11 | 浙江大学 | Ammonia spraying accurate optimization method for SCR denitration system |
| CN115591378A (en) * | 2022-12-12 | 2023-01-13 | 清华大学（Cn） | Feedforward compensation and disturbance suppression control system and method for SCR denitration of thermal power generating unit |
| CN115993787A (en) * | 2021-10-18 | 2023-04-21 | 国能智深控制技术有限公司 | Dynamic feedforward method, device and storage medium of generator set coordination control system |
| CN116610170A (en) * | 2023-07-17 | 2023-08-18 | 北京中科润宇环保科技股份有限公司 | Efficient SNCR intelligent three-dimensional temperature partition control method |
| CN116954058A (en) * | 2023-07-13 | 2023-10-27 | 淮阴工学院 | Boiler NOx concentration prediction and intelligent control method and system |
| CN117406817A (en) * | 2023-12-13 | 2024-01-16 | 福建德尔科技股份有限公司 | Fluorination reaction control system and method for preparing monofluoromethane |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105786035A (en) * | 2016-03-29 | 2016-07-20 | 东南大学 | Thermal power set SCR denitration optimization control system based on heuristic prediction control technology |
| CN106681381A (en) * | 2017-01-03 | 2017-05-17 | 华北电力大学 | SCR denitration system ammonia spraying quantity optimal control system and method based on intelligent feedforward signals |
| WO2017133316A1 (en) * | 2016-02-02 | 2017-08-10 | 华北电力科学研究院有限责任公司 | Method, device, and automatic control system for determining air intake amount for opposed firing |
-
2020
- 2020-11-12 CN CN202011261792.XA patent/CN112580250A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017133316A1 (en) * | 2016-02-02 | 2017-08-10 | 华北电力科学研究院有限责任公司 | Method, device, and automatic control system for determining air intake amount for opposed firing |
| CN105786035A (en) * | 2016-03-29 | 2016-07-20 | 东南大学 | Thermal power set SCR denitration optimization control system based on heuristic prediction control technology |
| CN106681381A (en) * | 2017-01-03 | 2017-05-17 | 华北电力大学 | SCR denitration system ammonia spraying quantity optimal control system and method based on intelligent feedforward signals |
Non-Patent Citations (1)
| Title |
|---|
| 毛清华、马宏伟等: "《矿用带式输送机智能监测及防护技术》", 31 May 2020 * |
Cited By (31)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113082969B (en) * | 2021-04-07 | 2024-01-23 | 唐锦婷 | Exhaust gas treatment system |
| CN113082969A (en) * | 2021-04-07 | 2021-07-09 | 唐锦婷 | Exhaust gas treatment system |
| CN113433911B (en) * | 2021-06-30 | 2022-05-20 | 浙江大学 | Accurate control system and method for ammonia spraying of denitration device based on accurate concentration prediction |
| CN113433911A (en) * | 2021-06-30 | 2021-09-24 | 浙江大学 | Denitration device ammonia injection accurate control system and method based on concentration accurate prediction |
| CN113426264A (en) * | 2021-07-15 | 2021-09-24 | 国电环境保护研究院有限公司 | Intelligent operation control method and control platform for flue gas purification island |
| CN113592163A (en) * | 2021-07-22 | 2021-11-02 | 汉谷云智(武汉)科技有限公司 | Method and equipment for predicting concentration of nitrogen oxide at inlet of SCR (Selective catalytic reduction) denitration reactor |
| CN113970937A (en) * | 2021-09-22 | 2022-01-25 | 同兴环保科技股份有限公司 | SCR denitration accurate ammonia injection control method |
| CN113847609A (en) * | 2021-09-29 | 2021-12-28 | 华南理工大学 | Garbage incinerator denitration control method based on feedforward model prediction error self-correction |
| CN113847609B (en) * | 2021-09-29 | 2024-02-06 | 华南理工大学 | Garbage incinerator denitration control method based on feedforward model prediction error self-correction |
| CN115993787A (en) * | 2021-10-18 | 2023-04-21 | 国能智深控制技术有限公司 | Dynamic feedforward method, device and storage medium of generator set coordination control system |
| CN114159959B (en) * | 2021-10-29 | 2024-04-26 | 广东惠州天然气发电有限公司 | Denitration control method and device for gas-steam combined cycle unit |
| CN114159959A (en) * | 2021-10-29 | 2022-03-11 | 广东惠州天然气发电有限公司 | Denitration control method and device for gas-steam combined cycle unit |
| CN114326387A (en) * | 2021-12-07 | 2022-04-12 | 江苏方天电力技术有限公司 | Denitration control device and method for thermal power generating unit |
| CN114225662B (en) * | 2021-12-07 | 2024-02-13 | 国网河北能源技术服务有限公司 | Hysteresis model-based flue gas desulfurization and denitrification optimal control method |
| CN114225662A (en) * | 2021-12-07 | 2022-03-25 | 国网河北能源技术服务有限公司 | Flue gas desulfurization and denitrification optimization control method based on hysteresis model |
| CN114326616A (en) * | 2021-12-22 | 2022-04-12 | 大唐东北电力试验研究院有限公司 | Industrial process control method based on multivariate load improved dynamic matrix prediction algorithm |
| CN114326616B (en) * | 2021-12-22 | 2023-09-19 | 大唐东北电力试验研究院有限公司 | Industrial process control method based on multivariate load improved dynamic matrix prediction algorithm |
| CN114442486A (en) * | 2022-01-12 | 2022-05-06 | 华能罗源发电有限责任公司 | SCR denitration optimization control system and control method for thermal power generating unit |
| CN114609986A (en) * | 2022-03-16 | 2022-06-10 | 中国中材国际工程股份有限公司 | Cement decomposing furnace denitration regulation and control optimization system and method based on predictive control |
| CN114664389A (en) * | 2022-03-24 | 2022-06-24 | 南方电网电力科技股份有限公司 | Method and device for predicting reaction conditions for urea hydrolysis ammonia production |
| CN114664389B (en) * | 2022-03-24 | 2023-09-22 | 南方电网电力科技股份有限公司 | Prediction method and device for reaction conditions of urea hydrolysis ammonia production |
| CN114878745A (en) * | 2022-04-25 | 2022-08-09 | 北京市生态环境保护科学研究院 | Gas boiler nitrogen oxide prediction monitoring system based on support vector machine |
| CN115155310B (en) * | 2022-07-13 | 2024-04-26 | 浙江大学 | SCR denitration system ammonia spraying accurate optimization method |
| CN115155310A (en) * | 2022-07-13 | 2022-10-11 | 浙江大学 | Ammonia spraying accurate optimization method for SCR denitration system |
| CN115591378A (en) * | 2022-12-12 | 2023-01-13 | 清华大学（Cn） | Feedforward compensation and disturbance suppression control system and method for SCR denitration of thermal power generating unit |
| CN116954058A (en) * | 2023-07-13 | 2023-10-27 | 淮阴工学院 | Boiler NOx concentration prediction and intelligent control method and system |
| CN116954058B (en) * | 2023-07-13 | 2024-02-23 | 淮阴工学院 | Boiler NOx concentration prediction and intelligent control method and system |
| CN116610170B (en) * | 2023-07-17 | 2023-10-13 | 北京中科润宇环保科技股份有限公司 | Efficient SNCR intelligent three-dimensional temperature partition control method |
| CN116610170A (en) * | 2023-07-17 | 2023-08-18 | 北京中科润宇环保科技股份有限公司 | Efficient SNCR intelligent three-dimensional temperature partition control method |
| CN117406817A (en) * | 2023-12-13 | 2024-01-16 | 福建德尔科技股份有限公司 | Fluorination reaction control system and method for preparing monofluoromethane |
| CN117406817B (en) * | 2023-12-13 | 2024-03-01 | 福建德尔科技股份有限公司 | Fluorination reaction control system and method for preparing monofluoromethane |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112580250A (en) | Thermal power generating unit denitration system based on deep learning and optimization control method | |
| CN104826492B (en) | Improvement method for selective catalytic reduction flue gas denitrification and ammonia injection control system | |
| CN107243257B (en) | It is suitble to the intelligence spray ammonia control system of full load | |
| CN109343349B (en) | SCR flue gas denitration optimal control system and method based on ammonia injection amount compensator | |
| CN106681381A (en) | SCR denitration system ammonia spraying quantity optimal control system and method based on intelligent feedforward signals | |
| CN113433911B (en) | Accurate control system and method for ammonia spraying of denitration device based on accurate concentration prediction | |
| CN104826493A (en) | Control method for selective catalytic reduction flue gas denitrification system | |
| CN110263395A (en) | The power plant's denitration running optimizatin method and system analyzed based on numerical simulation and data | |
| CN108628177A (en) | A kind of SCR denitration intelligence spray ammonia optimization method and system based on model adaptation PID | |
| CN110263452B (en) | Flue gas time distribution characteristic analysis method and system in flue and denitration system | |
| CN105629738A (en) | SCR (Selective Catalytic Reduction) flue gas denitration system control method and apparatus | |
| CN105069185A (en) | Method for establishing air pre-heater clean factor calculation model by using smoke pressure difference, and application | |
| CN105786035A (en) | Thermal power set SCR denitration optimization control system based on heuristic prediction control technology | |
| CN111841276B (en) | SNCR denitration control method and device for circulating fluidized bed unit and storage medium | |
| CN111897373A (en) | Model prediction-based ammonia injection flow adjusting method for SCR denitration device | |
| CN104715142A (en) | NOx emission dynamic soft-sensing method for power station boiler | |
| CN115111601A (en) | Multi-target boiler combustion optimization control method with embedded algorithm fusion under variable loads | |
| CN115145152A (en) | Boiler combustion and denitration process collaborative optimization control method | |
| CN115685743A (en) | Intelligent control coal-fired boiler and intelligent prediction regulation and control flue gas emission method thereof | |
| CN107561944A (en) | A kind of denitrating system adaptive prediction control method based on Laguerre model | |
| CN109933884B (en) | Neural network inverse control method for SCR denitration system of coal-fired unit | |
| CN112783115A (en) | Online real-time optimization method and device for steam power system | |
| CN117270387A (en) | SCR denitration system low ammonia escape control method and system based on deep learning | |
| CN116036849A (en) | CFB boiler flue gas denitration automatic control method and system | |
| CN114397813B (en) | Power generation boiler combustion continuous sliding mode control method based on slow time-varying disturbance observer |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210330 |