CN109519957A - A kind of ultra-supercritical boiler closed loop optimized control method of combustion - Google Patents
A kind of ultra-supercritical boiler closed loop optimized control method of combustion Download PDFInfo
- Publication number
- CN109519957A CN109519957A CN201811149972.1A CN201811149972A CN109519957A CN 109519957 A CN109519957 A CN 109519957A CN 201811149972 A CN201811149972 A CN 201811149972A CN 109519957 A CN109519957 A CN 109519957A
- Authority
- CN
- China
- Prior art keywords
- moment
- indicate
- nox
- boieff
- formula
- 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.)
- Granted
Links
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F23—COMBUSTION APPARATUS; COMBUSTION PROCESSES
- F23N—REGULATING OR CONTROLLING COMBUSTION
- F23N1/00—Regulating fuel supply
- F23N1/02—Regulating fuel supply conjointly with air supply
- F23N1/022—Regulating fuel supply conjointly with air supply using electronic means
Landscapes
- Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Mechanical Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a kind of ultra-supercritical boiler closed loop optimized control method of combustion, it can accurately reflect boiler efficiency and NOx emission with the dynamic characteristic of load variations based on the combustion system dynamic model that Unscented kalman filtering least square method supporting vector machine is established, while introducings of update mechanism also ensures dynamic model under different working conditions still with good adaptive ability and predictive ability.1000MW coal-burning boiler is in varying duty and steady load, regulated quantity, controlled index and relevant parameter are in zone of reasonableness and the stable situation of variation, boiler efficiency can be set to maintain to stablize, while NOx concentration is substantially reduced before relatively putting into operation after SCR inlet conversion.
Description
Technical field
The present invention relates to thermal technics fields, more particularly to a kind of ultra-supercritical boiler closed loop combustion control
Method.
Background technique
The adjustable parameters such as the method for operation of the air distribution mode of boiler, oxygen amount and pulverized coal preparation system are to boiler efficiency and NOx emission
It has a major impact.DCS control system is generally only capable of realizing the control to total coal amount and total blast volume according to workload demand at present.For
The coal amount and allocation of the amount of air of each layer burner then use and evenly distribute mode, or are rule of thumb adjusted by operations staff.It is aobvious
The right this method of operation still compares " extensive ", can not also adapt to coal type change, therefore, it is necessary to pass through burning optimization into one
Step excavates the potentiality of boiler energy-saving emission-reduction.
Boiler combustion optimization mainly has at present: non-linear modeling method and intelligent optimization algorithm phase based on data
In conjunction with, boiler combustion characteristic model is first established, it is then optimal for performance indicator with boiler efficiency and discharge, operating parameter is carried out
Optimization, such methods are suitable for the burning optimization of steady state condition, lack the adaptability to changing factors such as coals；It proposes later
Boiler system dynamic modeling is carried out based on Unscented kalman filtering least square method supporting vector machine, when model accuracy is unsatisfactory for requiring
When, Sample Refreshment strategy is devised, and avoid line matrix and invert, to reduce on-line calculation, but above-mentioned online calculation
A maximum deficiency existing for method is have the nuclear parameter σ of great influence still can only be by trying to gather repeatedly model accuracy
Method determine offline, cannot achieve online updating.
Summary of the invention
Goal of the invention: the object of the present invention is to provide a kind of ultra-supercritical boiler closed loop optimized control method of combustion, have
To the adaptability of the changing factors such as coal, online updating can be realized.
Technical solution: to reach this purpose, the invention adopts the following technical scheme:
Ultra-supercritical boiler closed loop optimized control method of combustion of the present invention, comprising the following steps:
S1: the prediction model of boiler efficiency and the prediction model of NOx concentration are established, and is counted respectively according to two prediction models
Calculate the boiler efficiency predicted value and NOx concentration predicted value at current k moment；Boiler efficiency prediction model and NOx concentration prediction model
Constitute combustion system dynamic model；
S2: two predicted values at step S1 obtained current k moment are compared with respective measured value, judgment bias
Whether meet required precision: if conditions are not met, being then updated to model parameter and sample data, then carrying out step S3；Such as
Fruit precision meets, then continues step S3；
S3: by the control variable at the following M moment acquired in the model parameter and k-1 time step S4 at current k moment
Calculate the boiler efficiency predicted value and NOx concentration predicted value at the following P moment；
S4: the boiler efficiency predicted value at the step S3 obtained following P moment and NOx concentration predicted value are brought into target
In function, by line solver constrained nonlinear systems problem, obtains the control variable at the following M moment and export；
S5: enabling k=k+1, and returns to step S1.
Further, in the step S1, boiler efficiency prediction model and NOx concentration prediction model all have input variable,
Controlled variable and output variable；The input variable of boiler efficiency prediction model includes unit load xload, flue gas oxygen contentCombustion
Cinder throttle opening xofa, secondary air register aperture xseca, coal-supplying amount bias xcoalWith the boiler efficiency at preceding 2 moment, NOx concentration prediction
The input variable of model includes unit load xload, flue gas oxygen contentBurn throttle opening xofa, secondary air register aperture xseca、
Coal-supplying amount biases xcoalWith the NOx concentration at preceding 2 moment；The controlled variable of boiler efficiency prediction model includes ultra supercritical fire coal
Boiler efficiency yBoiEff, the controlled variable of NOx concentration prediction model includes the NOx of selective-catalytic-reduction denitrified system entry dense
Spend yNOx；
The input variable and output variable for taking the top n moment are as model training sample set TNOxAnd TBoiEff,
TNOx={ (xNOx1,yNOx1),…,(xNOxN,yNOxN), TBoiEff={ (xBoiEff1,yBoiEff1),…,(xBoiEffN,
yBoiEffN),
Wherein yNOxi=yNOx
(i), yBoiEffi=yBoiEff(i), i=1,2 ..., N,Indicate i-1 moment flue gas oxygen content, xofa(i-1) i-1 is indicated
Moment burns throttle opening, xseca(i-1) i-1 moment secondary air register aperture, x are indicatedcoal(i-1) indicate that i-1 moment coal-supplying amount is inclined
It sets, xload(i-1) i-1 moment unit load, y are indicatedNOx(i) i moment NOx concentration, y are indicatedNOx(i-1) i-1 moment NOx is indicated
Concentration, yNOx(i-2) i-2 moment NOx concentration, y are indicatedBoiEff(i) i moment boiler efficiency, y are indicatedBoiEff(i-1) when indicating i-1
Carve boiler efficiency, yBoiEff(i-2) i-2 moment boiler efficiency is indicated.
Further, the prediction model for the NOx concentration established in the step S1 is obtained according to formula (1):
In formula (1), yNOx(k) the NOx concentration predicted value at current k moment, x are indicatedNOxIndicate that current k moment prediction model is defeated
Enter variable；N≥1；αNOxiRepresenting matrix αNOxIn i-th of element, αNOxIndicate NOx concentration prediction model Lagrange multiplier,
Initial time αNOx=αNOx0, bNOxIndicate NOx concentration prediction model decision function parameter, initial time bNOx=bNOx0, αNOx0With
bNOx0It is obtained according to formula (3)；K(xNOx,xNOxi) it is kernel function, it is obtained according to formula (2), xNOxiIndicate training sample set TNOxI-th
Input variable；
In formula (2),
Indicate k-1 moment flue gas oxygen content, xofa(k-1) indicate that the k-1 moment burns throttle opening, xseca(k-1) k- is indicated
1 moment secondary air register aperture, xcoal(k-1) biasing of k-1 moment coal-supplying amount, x are indicatedload(k-1) k-1 moment unit load is indicated,
yNOx(k-1) k-1 moment NOx concentration, y are indicatedNOx(k-2) k-2 moment NOx concentration, σ are indicatedNOxIndicate NOx concentration prediction model
Nuclear parameter；
In formula (3),Y=[yNOx1 yNOx2 … yNOxN]T, yNOxi=yNOx(i), yNOx(i) when indicating i
Carve NOx concentration, αNOx0iFor αNOx0In i-th of element, Υ obtains according to formula (4)；
In formula (4), c representative function penalty factor,xNOxlTable
Show training sample set TNOxFirst of input variable.
Further, the prediction model for the boiler efficiency established in the step S1 is obtained according to formula (5):
In formula (5), yBoiEff(k) the boiler efficiency predicted value at current k moment, x are indicatedBoiEffIndicate that the current k moment is predicted
Mode input variable；N≥1；αBoiEffiRepresenting matrix αBoiEffIn i-th of element, αBoiEffIndicate boiler efficiency prediction model
Lagrange multiplier, initial time αBoiEff=αBoiEff0, bBoiEffIndicate boiler efficiency prediction model decision function parameter, initially
Moment bBoiEff=bBoiEff0, αBoiEff0And bBoiEff0It is obtained according to formula (7)；K(xBoiEff,xBoiEffi) it is kernel function, according to formula
(6) it obtains, xBoiEffiIndicate training sample set TBoiEffIn i-th of input variable；
In formula,
Indicate k-1 moment flue gas oxygen content, xofa(k-1) indicate that the k-1 moment burns throttle opening, xseca(k-1) k- is indicated
1 moment secondary air register aperture, xcoal(k-1) biasing of k-1 moment coal-supplying amount, x are indicatedload(k-1) k-1 moment unit load is indicated,
yBoiEff(k-1) k-1 moment NOx concentration, y are indicatedBoiEff(k-2) k-2 moment NOx concentration, σ are indicatedBoiEffIndicate that boiler efficiency is pre-
Survey model nuclear parameter；
In formula (7),Y=[yBoiEff1 yBoiEff2 … yBoiEffN]T, yBoiEffiFor training sample set
TBoiEffIn i-th of output variable, αBoiEff0iRepresenting matrix αBoiEff0In i-th of element, Υ obtains according to formula (8)；
In formula (8), c representative function penalty factor,
xBoiEfflIndicate training sample set TBoiEffIn first of input variable.
Further, in the step S2, the NOx concentration predicted value at current k moment is compared with measured value, judgement is inclined
Whether difference meets required precision, is updated if being unsatisfactory for model parameter and sample data, specifically includes the following steps:
S2.1: by the NOx concentration predicted value y at current k momentNOx(k) and measured valueIt is compared, it is dense to obtain NOx
Spend the deviation Error of predictionNOx(k):
If ErrorNOx(k) > MaxErrorNOx, MaxErrorNOxIndicate the maximum deviation that NOx concentration prediction allows, then
Carry out step S2.2；
S2.2: building Multidimensional Parametric Vectors pNOxSVM=[bNOx αNOx σNOx]T∈RN+2, αNOxIndicate that NOx concentration predicts mould
Type Lagrange multiplier, bNOxIndicate NOx concentration prediction model decision function parameter, σNOxIndicate NOx concentration prediction model core ginseng
Number carries out parameter Estimation using Unscented kalman filtering, specifically includes the following steps:
S2.2.1: initialization
In formula,Indicate the priori mean value of stochastic variable.
In formula, P0Indicate the covariance matrix of stochastic variable, N >=1；
S2.2.2: 2d+1 sigma sampled point is chosen, d=N+2:
Wherein, χk-1Indicate the 2d+1 sigma sampled point at k-1 moment,Indicate the estimates of parameters at k-1 moment,Representing matrixZ column,Representing matrix?
Z-d column, Pk-1Indicate that the covariance matrix d=N+2, ψ of the parameter Estimation at k-1 moment are proportionality coefficient, Wm1Indicate the 1st sampling
The desired weight of point, Wc1Indicate the weight of the 1st sampled point variance, WmzIndicate the desired weight of z-th of sampled point, WczIt indicates
The weight of z-th of sampled point variance；
S2.2.3: it is calculated by the following formula:
χk|k-1=F (χk-1)=χk-1 (15)
Wherein, χk|k-1Indicate that the k-1 moment carries out the sample information after nonlinear state functional transformation, F (χ to sampled pointk-1)
Indicate nonlinear state function, χk-1Indicate the 2d+1 sigma sampled point at k-1 moment,Indicate a step state at k-1 moment
Prediction,Indicate the error co-variance matrix at k-1 moment, PwIndicate the variance matrix of systematic procedure noise, χz,k|k-1It indicates
χk|k-1Z column；
S2.2.4: it is calculated by the following formula:
yk|k-1=G (χk|k-1) (18)
Wherein, yk|k-1Indicate the output valve that k-1 moment sigma point is obtained by non-linear measurement functional transformation, yz,k|k-1
For yk|k-1Z column, G (χk|k-1) indicate non-linear measurement functional transformation,Indicate that a step at k-1 moment exports prediction,Table
Show the auto-covariance matrix at k-1 moment,Indicate that the Cross-covariance at k-1 moment, K indicate filtering gain matrix,Table
The k moment parameter Estimation shown,Indicate a step status predication at k-1 moment, ykIndicate measuring value, PkIndicate that the k moment joins
The covariance matrix of number estimation,Indicate the error co-variance matrix at k-1 moment, PvIndicate the Positive Definite square of system measurements noise
Battle array, thus acquires the NOx concentration prediction model parameters of subsequent time: bNOx=pNOxSVM(1), αNOx=pNOxSVM(2:N+1), σNOx
=pNOxSVM(N+2)。
Further, in the step S2, the boiler efficiency predicted value at current k moment is compared with measured value, is judged
Whether deviation meets required precision, is updated if being unsatisfactory for model parameter and sample data, specifically includes following step
It is rapid:
S2.3: by the boiler efficiency predicted value y at current k momentBoiEff(k) and measured valueIt is compared, obtains
The deviation Error of NOx concentration predictionBoiEff(k):
If ErrorBoiEff(k) > MaxErrorBoiEff, MaxErrorBoiEffIndicate the maximum that boiler efficiency prediction allows
Deviation then carries out step S2.4；
S2.4: building Multidimensional Parametric Vectors pBoiEffSVM=[bBoiEff αBoiEff σBoiEff]T∈RN+2, αBoiEffIndicate boiler
EFFICIENCY PREDICTION model Lagrange multiplier, bBoiEffIndicate boiler efficiency prediction model decision function parameter, σBoiEffIndicate boiler
EFFICIENCY PREDICTION model nuclear parameter carries out parameter Estimation using Unscented kalman filtering, specifically includes the following steps:
S2.4.1: initialization
In formula,Indicate the priori mean value of stochastic variable；
In formula, P0Indicate the covariance matrix of initial time, N >=1；
S2.4.2: 2d+1 sigma sampled point is chosen, d=N+2:
In formula, χk-1Indicate the 2d+1 sigma sampled point at k-1 moment,Indicate the equal of the stochastic variable at k-1 moment
Value,Representing matrixZ column,Representing matrix
Z-d column, Pk-1Indicate the covariance matrix of the parameter Estimation at k-1 moment, d=N+2, ψ are proportionality coefficient, Wm1Indicate the 1st
A desired weight of sampled point, Wc1Indicate the weight of the 1st sampled point variance, WmzIndicate the desired weight of z-th of sampled point,
WczIndicate the weight of z-th of sampled point variance；
S2.4.3: it is calculated by the following formula:
χk|k-1=F (χk-1)=χk-1 (31)
In formula, χk|k-1Indicate that the k-1 moment carries out the sample information after nonlinear state functional transformation, F (χ to sampled pointk-1)
Indicate nonlinear state function, χk-1Indicate the 2d+1 sigma sampled point at k-1 moment,Indicate a step state at k-1 moment
Prediction,Indicate the error co-variance matrix at k-1 moment, PwIndicate the variance matrix of systematic procedure noise, χz,k|k-1It indicates
χk|k-1Z column；
S2.4.4: it is calculated by the following formula:
yk|k-1=G (χk|k-1) (34)
In formula, yk|k-1Indicate the output valve that the sigma point at k-1 moment is obtained by non-linear measurement functional transformation,
yz,k|k-1For yk|k-1Z column, G (χk|k-1) indicate non-linear measurement functional transformation,Indicate that a step at k-1 moment exports prediction,Indicate the auto-covariance matrix at k-1 moment,Indicate that the Cross-covariance at k-1 moment, K indicate filtering gain matrix,Indicate obtained k moment parameter Estimation,Indicate a step status predication at k-1 moment, ykIndicate measuring value, PkWhen indicating k
The covariance matrix of parameter Estimation is carved,Indicate the error co-variance matrix at k-1 moment, PvIndicate the positive definite side of system measurements noise
Thus poor matrix acquires the boiler efficiency prediction model parameters of subsequent time: bBoiEff=pBoiEffSVM(1), αBoiEff=
pBoiEffSVM(2:N+1), σBoiEff=pBoiEffSVM(N+2)。
Further, in the step S3, the NOx concentration predicted value at the following P moment are as follows:
As 1≤j1When≤M,
Wherein, M indicates control time domain, yNOx(k+j1) indicate future j1The NOx concentration predicted value at a moment, αNOxiIndicate square
Battle array αNOxIn i-th of element, αNOxIndicate current time NOx concentration prediction model Lagrange multiplier, bNOxIndicate current time
NOx concentration prediction model decision function parameter, K (xNOx(k+j1),xNOxi) it is kernel function, it is calculated by formula (41), N >=1,
In formula, xNOxiFor NOx concentration training sample set TNOxIn i-th of input variable, xNOx(k+j1) indicate k+j1Moment
Input variable is obtained according to formula (42):
In formula (42),Indicate the k+j acquired1The flue gas oxygen content at -1 moment, xofa(k+j1- 1) it indicates to ask
The k+j obtained1- 1 moment burnt throttle opening, xseca(k+j1- 1) k+j acquired is indicated1The secondary air register aperture at -1 moment,
xcoal(k+j1- 1) k+j acquired is indicated1The coal-supplying amount at -1 moment biases, xload(k) current k moment unit load, y are indicatedNOx
(k+j1- 1) k+j is indicated1The NOx emission concentration prediction value at -1 moment, yNOx(k+j1- 2) k+j is indicated1The NOx emission at -2 moment is dense
Spend predicted value；
As M≤j1When≤P,
Wherein, M indicates control time domain, yNOx(k+j1) indicate future j1The NOx concentration predicted value at a moment, αNOxiIndicate square
Battle array αNOxIn i-th of element, αNOxIndicate current time NOx concentration prediction model Lagrange multiplier, bNOxIndicate current time
NOx concentration prediction model decision function parameter, K (xNOx(k+j1),xNOxi) it is kernel function, it is calculated by formula (43), N >=1,
In formula, xNOxiFor NOx concentration training sample set TNOxIn i-th of input variable, xNOx(k+j1) indicate k+j1Moment
Input variable is obtained according to formula (44):
In formula,Indicate the flue gas oxygen content at the k+M-1 moment acquired, xofa(k+M-1) expression acquires
The k+M-1 moment burns throttle opening, xseca(k+M-1) the secondary air register aperture at the k+M-1 moment acquired, x are indicatedcoal(k+M-
1) the coal-supplying amount biasing at the k+M-1 moment acquired, x are indicatedload(k) current k moment unit load, y are indicatedNOx(k+j1- 1) it indicates
k+j1The NOx emission concentration prediction value at -1 moment, yNOx(k+j1- 2) k+j is indicated1The NOx emission concentration prediction value at -2 moment.
Further, in the step S3, the boiler efficiency predicted value at the following P moment are as follows:
As 1≤j2When≤M,
Wherein, M indicates control time domain, yBoiEff(k+j2) indicate future j2The boiler efficiency predicted value at a moment, αBoiEffi
Representing matrix αBoiEffiIn i-th of element, αBoiEffIndicate current time boiler efficiency prediction model Lagrange multiplier,
bBoiEffIndicate current time boiler efficiency prediction model decision function parameter, K (xBoiEff(k+j2),xBoiEffi) it is kernel function, by
Formula (45) is calculated, N >=1,
In formula, xBoiEffiFor boiler efficiency training sample set TBoiEfiIn i-th of input variable, xBoiEff(k+j2) indicate k+
j2The input variable at moment is obtained according to formula (46):
In formula,Indicate the k+j acquired2The flue gas oxygen content at -1 moment, xofa(k+j2- 1) it indicates to acquire
K+j2- 1 moment burnt throttle opening, xseca(k+j2- 1) k+j acquired is indicated2The secondary air register aperture at -1 moment, xcoal
(k+j2- 1) k+j acquired is indicated2The coal-supplying amount at -1 moment biases, xload(k) current k moment unit load, y are indicatedBoiEff(k+
j2- 1) k+j is indicated2The boiler efficiency predicted value at -1 moment, yBoiEff(k+j2- 2) k+j is indicated2The boiler efficiency at -2 moment is predicted
Value；
As M≤j2When≤P,
Wherein, M indicates control time domain, yBoiEff(k+j2) indicate future j2The boiler efficiency predicted value at a moment, αBoiEffi
Representing matrix αBoiEffIn i-th of element, αBoiEffIndicate current time boiler efficiency prediction model Lagrange multiplier,
bBoiEffIndicate current time boiler efficiency prediction model decision function parameter, K (xBoiEff(k+j2),xBoiEffi) it is kernel function, by
Formula (47) is calculated, N >=1,
In formula, xBoiEffiFor boiler efficiency training sample set TBoiEfiIn i-th of input variable, xBoiEff(k+j2) indicate k+
j2The input variable at moment is obtained according to formula (48):
In formula,Indicate the flue gas oxygen content at the k+M-1 moment acquired, xofa(k+M-1) k acquired is indicated
+ M-1 the moment burns throttle opening, xseca(k+M-1) the secondary air register aperture at the k+M-1 moment acquired, x are indicatedcoal(k+M-
1) the coal-supplying amount biasing at the k+M-1 moment acquired, x are indicatedload(k) current k moment unit load, y are indicatedBoiEff(k+j2- 1) table
Show k+j2The boiler efficiency predicted value at -1 moment, yBoiEff(k+j2- 2) k+j is indicated2The boiler efficiency predicted value at -2 moment.
Further, the step S4 specifically includes the following steps:
S4.1: objective function is determined:
s.t.MVmin< MV < MVmax (50)
ΔMVmin< Δ MV < Δ MVmax (51)
yNOx(k+j1)≤NOxmax, j=1,2 ..., N (52)
Wherein,For objective function, MV represent oxygen amount definite value, burn throttle opening, secondary air register aperture or
The biasing of person's coal-supplying amount,yBoiEff(k+j2) prediction of the expression to the following P moment boiler efficiency
Value, MVmaxIndicate the upper limit value of MV, MVminIndicate that the lower limit value of MV, Δ MV indicate the rate of change of MV, Δ MVmaxIndicate Δ MV
Upper limit value, Δ MVminIndicate the lower limit value of Δ MV, yNOx(k+j1) indicate the predicted value of the following P moment NOx emission concentration,
NOxmaxIndicate NOxThe upper limit value of discharge, N >=1；
S4.2: nonlinear optimal problem is solved, the control variable at the following M moment is obtained and exports；Wherein, the following M
The control variable at moment includesxofa(k+1)…xofa(k+M)、xseca(k+1)…xseca(k+M)
And xcoal(k+1)…xcoal(k+M),Indicate the flue gas oxygen content at the following j moment, xofa(k+j) future j are indicated
Moment burns throttle opening, xseca(k+j) the secondary air register aperture at the following j moment, x are indicatedcoal(k+j) when indicating j following
The coal-supplying amount at quarter biases, 1≤j≤M.
The utility model has the advantages that the invention discloses a kind of ultra-supercritical boiler closed loop optimized control method of combustion, with the prior art
Compare, have it is following the utility model has the advantages that
1) the combustion system dynamic model established based on Unscented kalman filtering least square method supporting vector machine can be accurate
Ground reflects boiler efficiency and NOx emission with the dynamic characteristic of load variations, while the introducing of update mechanism also ensures dynamic analog
Type still has good adaptive ability and predictive ability under different working conditions；
2) in varying duty and steady load, regulated quantity, controlled index and relevant parameter are in 1000MW coal-burning boiler
Under zone of reasonableness and the stable situation of variation, boiler efficiency can be made to maintain to stablize, while NOx concentration after SCR inlet conversion
It is substantially reduced before relatively putting into operation.
Detailed description of the invention
Fig. 1 is the schematic diagram of the control method of the specific embodiment of the invention；
Fig. 2 is the structure chart of combustion system dynamic model in the specific embodiment of the invention；
Fig. 3 is the test of the prediction model of the prediction model and NOx concentration of boiler efficiency in the specific embodiment of the invention
As a result；
Fig. 3 (a) is the test result of the prediction model of NOx concentration；
Fig. 3 (b) is the test result of the prediction model of boiler efficiency；
Fig. 4 is the Contrast on effect that present embodiment method and traditional control method are used in same load
Figure；
Fig. 4 (a) is that varying duty process combusts optimize the effect picture that puts into operation；
Burning optimization puts into operation effect picture when Fig. 4 (b) is steady load.
Specific embodiment
Technical solution of the present invention is further introduced with attached drawing With reference to embodiment.
Present embodiment discloses a kind of ultra-supercritical boiler closed loop optimized control method of combustion, and Fig. 1 is method
Schematic diagram, method the following steps are included:
S1: the prediction model of boiler efficiency and the prediction model of NOx concentration are established, and is counted respectively according to two prediction models
Calculate the boiler efficiency predicted value and NOx concentration predicted value at current k moment；
S2: two predicted values at step S1 obtained current k moment are compared with respective measured value, judgment bias
Whether meet required precision: if conditions are not met, being then updated to model parameter and sample data, then carrying out step S3；Such as
Fruit precision meets, then continues step S3；
S3: by the control variable at the following M moment acquired in the model parameter and k-1 time step S4 at current k moment
Calculate the boiler efficiency predicted value and NOx concentration predicted value at the following P moment；P=10, M=2；
S4: the boiler efficiency predicted value at the step S3 obtained following P moment and NOx concentration predicted value are brought into target
In function, by line solver constrained nonlinear systems problem, obtains the control variable at the following M moment and export；
S5: enabling k=k+1, and returns to step S1.
In step S1, boiler efficiency prediction model and NOx concentration prediction model all have input variable, controlled variable and defeated
Variable out；The input variable of boiler efficiency prediction model includes unit load xload, flue gas oxygen contentBurn throttle opening
xofa, secondary air register aperture xseca, coal-supplying amount bias xcoalWith the boiler efficiency at preceding 2 moment, the input of NOx concentration prediction model
Variable includes unit load xload, flue gas oxygen contentBurn throttle opening xofa, secondary air register aperture xseca, coal-supplying amount biasing
xcoalWith the NOx concentration at preceding 2 moment；The controlled variable of boiler efficiency prediction model includes ultra supercritical coal-burning boiler efficiency
yBoiEff, the controlled variable of NOx concentration prediction model includes the NOx concentration y of selective-catalytic-reduction denitrified system entryNOx；Pot
Efficiency of furnace prediction model and NOx concentration prediction model constitute combustion system dynamic model, as shown in Figure 2；Wherein, it is fired using 1
The instruction of cinder airdoor control indicates the influence of 4 layers of burnt wind (each 2 layers of front-back wall)；It is secondary using 36 layers of expression of secondary air register instruction
The influence (each 3 layers of front-back wall, identical height takes same instructions) of wind, xseca=[xseca1,xseca2,xseca3]；Coal is given using 3
Measuring offset instructions indicates influence (each 3 layers of front-back wall, identical height takes same instructions) x of 6 layers of coal-supplying amount biasingcoal=[xcoal1,
xcoal2,xcoal3]；
The input variable and output variable for taking the top n moment are as model training sample set TNOxAnd TBoiEff,
TNOx={ (xNOx1,yNOx1),…,(xNOxN,yNOxN), TBoiEff={ (xBoiEff1,yBoiEff1),…,(xBoiEffN,
yBoiEffN),
Wherein yNOxi=yNOx
(i), yBoiEffi=yBoiEff(i), i=1,2 ..., N,Indicate i-1 moment flue gas oxygen content, xofa(i-1) i-1 is indicated
Moment burns throttle opening, xseca(i-1) i-1 moment secondary air register aperture, x are indicatedcoal(i-1) indicate that i-1 moment coal-supplying amount is inclined
It sets, xload(i-1) i-1 moment unit load, y are indicatedNOx(i) i moment NOx concentration, y are indicatedNOx(i-1) i-1 moment NOx is indicated
Concentration, yNOx(i-2) i-2 moment NOx concentration, y are indicatedBoiEff(i) i moment boiler efficiency, y are indicatedBoiEff(i-1) when indicating i-1
Carve boiler efficiency, yBoiEff(i-2) i-2 moment boiler efficiency is indicated.
The prediction model for the NOx concentration established in step S1 is obtained according to formula (1):
In formula (1), yNOx(k) the NOx concentration predicted value at current k moment, x are indicatedNOxIndicate that current k moment prediction model is defeated
Enter variable；N≥1；αNOxiRepresenting matrix αNOxIn i-th of element, αNOxIndicate NOx concentration prediction model Lagrange multiplier,
Initial time αNOx=αNOx0, bNOxIndicate NOx concentration prediction model decision function parameter, initial time bNOx=bNOx0, αNOx0With
bNOx0It is obtained according to formula (3)；K(xNOx,xNOxi) it is kernel function, it is obtained according to formula (2), xNOxiIndicate training sample set TNOxI-th
Input variable；
In formula (2),
Indicate k-1 moment flue gas oxygen content, xofa(k-1) indicate that the k-1 moment burns throttle opening, xseca(k-1) k- is indicated
1 moment secondary air register aperture, xcoal(k-1) biasing of k-1 moment coal-supplying amount, x are indicatedload(k-1) k-1 moment unit load is indicated,
yNOx(k-1) k-1 moment NOx concentration, y are indicatedNOx(k-2) k-2 moment NOx concentration, σ are indicatedNOxIndicate NOx concentration prediction model
Nuclear parameter；
In formula (3),Y=[yNOx1 yNOx2 … yNOxN]T, yNOxi=yNOx(i), yNOx(i) when indicating i
Carve NOx concentration, αNOx0iFor αNOx0In i-th of element, Υ obtains according to formula (4)；
In formula (4), c representative function penalty factor,xNOxlTable
Show training sample set TNOxFirst of input variable.
The prediction model for the boiler efficiency established in step S1 is obtained according to formula (5):
In formula (5), yBoiEff(k) the boiler efficiency predicted value at current k moment, x are indicatedBoiEffIndicate that the current k moment is predicted
Mode input variable；N≥1；αBoiEffiRepresenting matrix αBoiEffIn i-th of element, αBoiEffIndicate boiler efficiency prediction model
Lagrange multiplier, initial time αBoiEff=αBoiEff0, bBoiEffIndicate boiler efficiency prediction model decision function parameter, initially
Moment bBoiEff=bBoiEff0, αBoiEff0And bBoiEff0It is obtained according to formula (7)；K(xBoiEff,xBoiEffi) it is kernel function, according to formula
(6) it obtains, xBoiEffiIndicate training sample set TBoiEffIn i-th of input variable；
In formula,
Indicate k-1 moment flue gas oxygen content, xofa(k-1) indicate that the k-1 moment burns throttle opening, xseca(k-1) k- is indicated
1 moment secondary air register aperture, xcoal(k-1) biasing of k-1 moment coal-supplying amount, x are indicatedload(k-1) k-1 moment unit load is indicated,
yBoiEff(k-1) k-1 moment NOx concentration, y are indicatedBoiEff(k-2) k-2 moment NOx concentration, σ are indicatedBoiEffIndicate that boiler efficiency is pre-
Survey model nuclear parameter；
In formula (7),Y=[yBoiEff1 yBoiEff2 … yBoiEffN]T, yBoiEffiFor training sample set
TBoiEffIn i-th of output variable, αBoiEff0iRepresenting matrix αBoiEff0In i-th of element, Υ obtains according to formula (8)；
In formula (8), c representative function penalty factor,
xBoiEfflIndicate training sample set TBoiEffIn first of input variable.
In step S2, the NOx concentration predicted value at current k moment is compared with measured value, whether judgment bias meets
Required precision is updated model parameter and sample data if being unsatisfactory for, specifically includes the following steps:
S2.1: by the NOx concentration predicted value y at current k momentNOx(k) and measured valueIt is compared, it is dense to obtain NOx
Spend the deviation Error of predictionNOx(k):
If ErrorNOx(k) > MaxErrorNOx, MaxErrorNOxIndicate the maximum deviation that NOx concentration prediction allows, then
Carry out step S2.2；
S2.2: building Multidimensional Parametric Vectors pNOxSVM=[bNOx αNOx σNOx]T∈RN+2, αNOxIndicate that NOx concentration predicts mould
Type Lagrange multiplier, bNOxIndicate NOx concentration prediction model decision function parameter, σNOxIndicate NOx concentration prediction model core ginseng
Number carries out parameter Estimation using Unscented kalman filtering, specifically includes the following steps:
S2.2.1: initialization
In formula,Indicate the priori mean value of stochastic variable.
In formula, P0Indicate the covariance matrix of stochastic variable, N >=1；
S2.2.2: 2d+1 sigma sampled point is chosen, d=N+2:
Wherein, χk-1Indicate the 2d+1 sigma sampled point at k-1 moment,Indicate the estimates of parameters at k-1 moment,Representing matrixZ column,Representing matrix?
Z-d column, Pk-1Indicate that the covariance matrix d=N+2, ψ of the parameter Estimation at k-1 moment are proportionality coefficient, Wm1Indicate the 1st sampling
The desired weight of point, Wc1Indicate the weight of the 1st sampled point variance, WmzIndicate the desired weight of z-th of sampled point, WczIt indicates
The weight of z-th of sampled point variance；
S2.2.3: it is calculated by the following formula:
χk|k-1=F (χk-1)=χk-1 (15)
Wherein, χk|k-1Indicate that the k-1 moment carries out the sample information after nonlinear state functional transformation, F (χ to sampled pointk-1)
Indicate nonlinear state function, χk-1Indicate the 2d+1 sigma sampled point at k-1 moment,Indicate a step state at k-1 moment
Prediction,Indicate the error co-variance matrix at k-1 moment, PwIndicate the variance matrix of systematic procedure noise, χz,k|k-1It indicates
χk|k-1Z column；
S2.2.4: it is calculated by the following formula:
yk|k-1=G (χk|k-1) (18)
Wherein, yk|k-1Indicate the output valve that k-1 moment sigma point is obtained by non-linear measurement functional transformation, yz,k|k-1
For yk|k-1Z column, G (χk|k-1) indicate non-linear measurement functional transformation,Indicate that a step at k-1 moment exports prediction,Table
Show the auto-covariance matrix at k-1 moment,Indicate that the Cross-covariance at k-1 moment, K indicate filtering gain matrix,Table
The k moment parameter Estimation shown,Indicate a step status predication at k-1 moment, ykIndicate measuring value, PkIndicate that the k moment joins
The covariance matrix of number estimation,Indicate the error co-variance matrix at k-1 moment, PvIndicate the Positive Definite square of system measurements noise
Battle array, thus acquires the NOx concentration prediction model parameters of subsequent time: bNOx=pNOxSVM(1), αNOx=pNOxSVM(2:N+1), σNOx
=pNOxSVM(N+2)。
In step S2, the boiler efficiency predicted value at current k moment is compared with measured value, whether judgment bias meets
Required precision is updated model parameter and sample data if being unsatisfactory for, specifically includes the following steps:
S2.3: by the boiler efficiency predicted value y at current k momentBoiEff(k) and measured valueIt is compared, obtains
The deviation Error of NOx concentration predictionBoiEff(k):
If ErrorBoiEff(k) > MaxErrorBoiEff, MaxErrorBoiEffIndicate the maximum that boiler efficiency prediction allows
Deviation then carries out step S2.4；
S2.4: building Multidimensional Parametric Vectors pBoiEffSVM=[bBoiEff αBoiEff σBoiEff]T∈RN+2, αBoiEffIndicate boiler
EFFICIENCY PREDICTION model Lagrange multiplier, bBoiEffIndicate boiler efficiency prediction model decision function parameter, σBoiEffIndicate boiler
EFFICIENCY PREDICTION model nuclear parameter carries out parameter Estimation using Unscented kalman filtering, specifically includes the following steps:
S2.4.1: initialization
In formula,Indicate the priori mean value of stochastic variable；
In formula, P0Indicate the covariance matrix of initial time, N >=1；
S2.4.2: 2d+1 sigma sampled point is chosen, d=N+2:
In formula, χk-1Indicate the 2d+1 sigma sampled point at k-1 moment,Indicate the equal of the stochastic variable at k-1 moment
Value,Representing matrixZ column,Representing matrix
Z-d column, Pk- 1 indicates the covariance matrix of the parameter Estimation at k-1 moment, and d=N+2, ψ are proportionality coefficient, Wm1Indicate the 1st
A desired weight of sampled point, Wc1Indicate the weight of the 1st sampled point variance, WmzIndicate the desired weight of z-th of sampled point,
WczIndicate the weight of z-th of sampled point variance；
S2.4.3: it is calculated by the following formula:
χk|k-1=F (χk-1)=χk-1 (31)
In formula, χk|k-1Indicate that the k-1 moment carries out the sample information after nonlinear state functional transformation, F (χ to sampled pointk-1)
Indicate nonlinear state function, χk-1Indicate the 2d+1 sigma sampled point at k-1 moment,Indicate a step state at k-1 moment
Prediction,Indicate the error co-variance matrix at k-1 moment, PwIndicate the variance matrix of systematic procedure noise, χz,k|k-1It indicates
χk|k-1Z column；
S2.4.4: it is calculated by the following formula:
yk|k-1=G (χk|k-1) (34)
In formula, yk|k-1Indicate the output valve that the sigma point at k-1 moment is obtained by non-linear measurement functional transformation,
yz,k|k-1For yk|k-1Z column, G (χk|k-1) indicate non-linear measurement functional transformation,Indicate that a step at k-1 moment exports prediction,Indicate the auto-covariance matrix at k-1 moment,Indicate that the Cross-covariance at k-1 moment, K indicate filtering gain matrix,Indicate obtained k moment parameter Estimation,Indicate a step status predication at k-1 moment, ykIndicate measuring value, PkWhen indicating k
The covariance matrix of parameter Estimation is carved,Indicate the error co-variance matrix at k-1 moment, PvIndicate the positive definite side of system measurements noise
Thus poor matrix acquires the boiler efficiency prediction model parameters of subsequent time: bBoiEff=pBoiEffSVM(1), αBoiEff=
pBoiEffSVM(2:N+1), σBoiEff=pBoiEffSVM(N+2)。
In step S3, the NOx concentration predicted value at the following P moment are as follows:
As 1≤j1When≤M,
Wherein, M indicates control time domain, yNOx(k+j1) indicate future j1The NOx concentration predicted value at a moment, αNOxiIndicate square
Battle array αNOxIn i-th of element, αNOxIndicate current time NOx concentration prediction model Lagrange multiplier, bNOxIndicate current time
NOx concentration prediction model decision function parameter, K (xNOx(k+j1),xNOxi) it is kernel function, it is calculated by formula (41), N >=1,
In formula, xNOxiFor NOx concentration training sample set TNOxIn i-th of input variable, xNOx(k+j1) indicate k+j1Moment
Input variable is obtained according to formula (42):
In formula (42),Indicate the k+j acquired1The flue gas oxygen content at -1 moment, xofa(k+j1- 1) it indicates to ask
The k+j obtained1- 1 moment burnt throttle opening, xseca(k+j1- 1) k+j acquired is indicated1The secondary air register aperture at -1 moment,
xcoal(k+j1- 1) k+j acquired is indicated1The coal-supplying amount at -1 moment biases, xload(k) current k moment unit load, y are indicatedNOx
(k+j1- 1) k+j is indicated1The NOx emission concentration prediction value at -1 moment, yNOx(k+j1- 2) k+j is indicated1The NOx emission at -2 moment is dense
Spend predicted value；
As M≤j1When≤P,
Wherein, M indicates control time domain, yNOx(k+j1) indicate future j1The NOx concentration predicted value at a moment, αNOxiIndicate square
Battle array αNOxIn i-th of element, αNOxIndicate current time NOx concentration prediction model Lagrange multiplier, bNOxIndicate current time
NOx concentration prediction model decision function parameter, K (xNOx(k+j1),xNOxi) it is kernel function, it is calculated by formula (43), N >=1,
In formula, xNOxiFor NOx concentration training sample set TNOxIn i-th of input variable, xNOx(k+j1) indicate k+j1Moment
Input variable is obtained according to formula (44):
In formula,Indicate the flue gas oxygen content at the k+M-1 moment acquired, xofa(k+M-1) k acquired is indicated
+ M-1 the moment burns throttle opening, xseca(k+M-1) the secondary air register aperture at the k+M-1 moment acquired, x are indicatedcoal(k+M-
1) the coal-supplying amount biasing at the k+M-1 moment acquired, x are indicatedload(k) current k moment unit load, y are indicatedNOx(k+j1- 1) it indicates
k+j1The NOx emission concentration prediction value at -1 moment, yNOx(k+j1- 2) k+j is indicated1The NOx emission concentration prediction value at -2 moment.
In step S3, the boiler efficiency predicted value at the following P moment are as follows:
As 1≤j2When≤M,
Wherein, M indicates control time domain, yBoiEff(k+j2) indicate future j2The boiler efficiency predicted value at a moment, αBoiEffi
Representing matrix αBoiEffiIn i-th of element, αBoiEffIndicate current time boiler efficiency prediction model Lagrange multiplier,
bBoiEffIndicate current time boiler efficiency prediction model decision function parameter, K (xBoiEff(k+j2),xBoiEffi) it is kernel function, by
Formula (45) is calculated, N >=1,
In formula, xBoiEffiFor boiler efficiency training sample set TBoiEfiIn i-th of input variable, xBoiEff(k+j2) indicate k+
j2The input variable at moment is obtained according to formula (46):
In formula,Indicate the k+j acquired2The flue gas oxygen content at -1 moment, xofa(k+j2- 1) it indicates to acquire
K+j2- 1 moment burnt throttle opening, xseca(k+j2- 1) k+j acquired is indicated2The secondary air register aperture at -1 moment, xcoal
(k+j2- 1) k+j acquired is indicated2The coal-supplying amount at -1 moment biases, xload(k) current k moment unit load, y are indicatedBoiEff(k+
j2- 1) k+j is indicated2The boiler efficiency predicted value at -1 moment, yBoiEff(k+j2- 2) k+j is indicated2The boiler efficiency at -2 moment is predicted
Value；
As M≤j2When≤P,
Wherein, M indicates control time domain, yBoiEff(k+j2) indicate future j2The boiler efficiency predicted value at a moment, αBoiEffi
Representing matrix αBoiEffIn i-th of element, αBoiEffIndicate current time boiler efficiency prediction model Lagrange multiplier,
bBoiEffIndicate current time boiler efficiency prediction model decision function parameter, K (xBoiEff(k+j2),xBoiEffi) it is kernel function, by
Formula (47) is calculated, N >=1,
In formula, xBoiEffiFor boiler efficiency training sample set TBoiEfiIn i-th of input variable, xBoiEff(k+j2) indicate k+
j2The input variable at moment is obtained according to formula (48):
In formula,Indicate the flue gas oxygen content at the k+M-1 moment acquired, xofa(k+M-1) expression acquires
The k+M-1 moment burns throttle opening, xseca(k+M-1) the secondary air register aperture at the k+M-1 moment acquired, x are indicatedcoal(k+M-
1) the coal-supplying amount biasing at the k+M-1 moment acquired, x are indicatedload(k) current k moment unit load, y are indicatedBoiEff(k+j2- 1) table
Show k+j2The boiler efficiency predicted value at -1 moment, yBoiEff(k+j2- 2) k+j is indicated2The boiler efficiency predicted value at -2 moment.
Step S4 specifically includes the following steps:
S4.1: objective function is determined:
s.t.MVmin< MV < MVmax (50)
ΔMVmin< Δ MV < Δ MVmax (51)
yNOx(k+j1)≤NOxmax, j=1,2 ..., N (52)
Wherein,For objective function, MV represent oxygen amount definite value, burn throttle opening, secondary air register aperture or
The biasing of person's coal-supplying amount,yBoiEff(k+j2) prediction of the expression to the following P moment boiler efficiency
Value, MVmaxIndicate the upper limit value of MV, MVminIndicate that the lower limit value of MV, Δ MV indicate the rate of change of MV, Δ MVmaxIndicate Δ MV
Upper limit value, Δ MVminIndicate the lower limit value of Δ MV, yNOx(k+j1) indicate the predicted value of the following P moment NOx emission concentration,
NOxmaxIndicate NOxThe upper limit value of discharge, N >=1；
S4.2: nonlinear optimal problem is solved, the control variable at the following M moment is obtained and exports；Wherein, the following M
The control variable at moment includesxofa(k+1)…xofa(k+M)、xseca(k+1)…xseca(k+M)
And xcoal(k+1)…xcoal(k+M),Indicate the flue gas oxygen content at the following j moment, xofa(k+j) future j are indicated
Moment burns throttle opening, xseca(k+j) the secondary air register aperture at the following j moment, x are indicatedcoal(k+j) when indicating j following
The coal-supplying amount at quarter biases, 1≤j≤M.
Using data collected from the DCS of certain power plant 1000MW unit progress model measurement, the selection sampling time is
10s chooses and is wherein used as training sample for 500 groups, in addition chooses the data of 10000 groups of consecutive variations as test sample.For just
In the forecast result of observing and nursing, the test result for wherein 3000 groups of test samples is only provided herein, such as Fig. 3 institute
Show.Fig. 3 (a) is the test result of NOx model, and Fig. 3 (b) is the test result of boiler efficiency model.Test result shows, NOx
The update times of model and boiler efficiency model are respectively 40 times and 17 times.From test result as can be seen that based on no mark card
The combustion system dynamic model that Kalman Filtering least square method supporting vector machine is established can accurately reflect boiler efficiency and NOx row
Put the dynamic characteristic with load variations, at the same the introducing of update mechanism also ensure dynamic model under different working conditions according to
So there is good adaptive ability and predictive ability.
Fig. 4 is in zone of reasonableness and variation in regulated quantity, controlled index and relevant parameter for certain 1000MW coal-burning boiler
In stable situation, burning optimization puts into operation the efficiency of non-period of putting into operation and the comparison diagram of NOx concentration under period and similar operating condition.From
It can be seen that boiler efficiency maintains to stablize in figure, NOx concentration is substantially reduced before relatively putting into operation after SCR inlet conversion.Middle and high load
In the case of, NOx concentration averagely reduces 30mg/m3Or so, in the case of underload, NOx concentration is averagely reduced more than 50mg/m3.Fig. 4
(a) under unit varying load condition, burning optimization puts into operation the efficiency and NOx concentration of non-period of putting into operation under period and similar operating condition
Comparison.The two approaches in time, and the conditions such as load variations situation, coal quality and environment temperature are almost the same.Comparing from figure can
To find out, boiler efficiency slightly rises after burning optimization puts into operation, and former flue gas conversion NOx concentration reduces 30mg/m3Left and right.Fig. 4 (b)
When stablizing near 900MW for unit load, the comparison that puts into operation with the non-situation that puts into operation.After visible combustion optimization puts into operation, boiler effect
Rate is almost the same, and former flue gas conversion NOx concentration reduces about 45mg/m3。
Claims (9)
1. a kind of ultra-supercritical boiler closed loop optimized control method of combustion, it is characterised in that: the following steps are included:
S1: the prediction model of boiler efficiency and the prediction model of NOx concentration are established, and is calculated separately and is worked as according to two prediction models
The boiler efficiency predicted value and NOx concentration predicted value at preceding k moment；Boiler efficiency prediction model and NOx concentration prediction model are constituted
Combustion system dynamic model；
S2: two predicted values at step S1 obtained current k moment are compared with respective measured value, whether judgment bias
Meet required precision: if conditions are not met, being then updated to model parameter and sample data, then carrying out step S3；If smart
Degree meets, then continues step S3；
S3: it is calculated by the control variable at the following M moment acquired in the model parameter and k-1 time step S4 at current k moment
The boiler efficiency predicted value and NOx concentration predicted value at the following P moment；
S4: the boiler efficiency predicted value at the step S3 obtained following P moment and NOx concentration predicted value are brought into objective function
In, by line solver constrained nonlinear systems problem, obtains the control variable at the following M moment and export；
S5: enabling k=k+1, and returns to step S1.
2. ultra-supercritical boiler closed loop optimized control method of combustion according to claim 1, it is characterised in that: the step
In S1, boiler efficiency prediction model and NOx concentration prediction model all have input variable, controlled variable and output variable；Boiler
The input variable of EFFICIENCY PREDICTION model includes unit load xload, flue gas oxygen contentBurn throttle opening xofa, secondary air register
Aperture xseca, coal-supplying amount bias xcoalWith the boiler efficiency at preceding 2 moment, the input variable of NOx concentration prediction model includes unit
Load xload, flue gas oxygen contentBurn throttle opening xofa, secondary air register aperture xseca, coal-supplying amount bias xcoalWith first 2
The NOx concentration at moment；The controlled variable of boiler efficiency prediction model includes ultra supercritical coal-burning boiler efficiency yBoiEff, NOx concentration
The controlled variable of prediction model includes the NOx concentration y of selective-catalytic-reduction denitrified system entryNOx；
The input variable and output variable for taking the top n moment are as model training sample set TNOxAnd TBoiEff,
TNOx={ (xNOx1,yNOx1),…,(xNOxN,yNOxN), TBoiEff={ (xBoiEff1,yBoiEff1),…,(xBoiEffN,
yBoiEffN),
Wherein yNOxi=yNOx
(i), yBoiEffi=yBoiEff(i), i=1,2 ..., N,Indicate i-1 moment flue gas oxygen content, xofa(i-1) i-1 is indicated
Moment burns throttle opening, xseca(i-1) i-1 moment secondary air register aperture, x are indicatedcoal(i-1) indicate that i-1 moment coal-supplying amount is inclined
It sets, xload(i-1) i-1 moment unit load, y are indicatedNOx(i) i moment NOx concentration, y are indicatedNOx(i-1) i-1 moment NOx is indicated
Concentration, yNOx(i-2) i-2 moment NOx concentration, y are indicatedBoiEff(i) i moment boiler efficiency, y are indicatedBoiEff(i-1) when indicating i-1
Carve boiler efficiency, yBoiEff(i-2) i-2 moment boiler efficiency is indicated.
3. ultra-supercritical boiler closed loop optimized control method of combustion according to claim 1, it is characterised in that: the step
The prediction model for the NOx concentration established in S1 is obtained according to formula (1):
In formula (1), yNOx(k) the NOx concentration predicted value at current k moment, x are indicatedNOxIndicate that current k moment prediction model input becomes
Amount；N≥1；αNOxiRepresenting matrix αNOxIn i-th of element, αNOxIndicate NOx concentration prediction model Lagrange multiplier, initially
Moment αNOx=αNOx0, bNOxIndicate NOx concentration prediction model decision function parameter, initial time bNOx=bNOx0, αNOx0And bNOx0
It is obtained according to formula (3)；K(xNOx,xNOxi) it is kernel function, it is obtained according to formula (2), xNOxiIndicate training sample set TNOxI-th of input
Variable；
In formula (2),
Indicate k-1 moment flue gas oxygen content, xofa(k-1) indicate that the k-1 moment burns throttle opening, xseca(k-1) table
Show k-1 moment secondary air register aperture, xcoal(k-1) biasing of k-1 moment coal-supplying amount, x are indicatedload(k-1) k-1 moment unit is indicated
Load, yNOx(k-1) k-1 moment NOx concentration, y are indicatedNOx(k-2) k-2 moment NOx concentration, σ are indicatedNOxIndicate NOx concentration prediction
Model nuclear parameter；
In formula (3),Y=[yNOx1 yNOx2 … yNOxN]T, yNOxi=yNOx(i), yNOx(i) the i moment is indicated
NOx concentration, αNOx0iFor αNOx0In i-th of element, Υ obtains according to formula (4)；
In formula (4), c representative function penalty factor,1≤l≤N, xNOxlIndicate instruction
Practice sample set TNOxFirst of input variable.
4. ultra-supercritical boiler closed loop optimized control method of combustion according to claim 1, it is characterised in that: the step
The prediction model for the boiler efficiency established in S1 is obtained according to formula (5):
In formula (5), yBoiEff(k) the boiler efficiency predicted value at current k moment, x are indicatedBoiEffIndicate current k moment prediction model
Input variable；N≥1；αBoiEffiRepresenting matrix αBoiEffIn i-th of element, αBoiEffIndicate boiler efficiency prediction model glug
Bright day multiplier, initial time αBoiEff=αBoiEff0, bBoiEffIndicate boiler efficiency prediction model decision function parameter, initial time
bBoiEff=bBoiEff0, αBoiEff0And bBoiEff0It is obtained according to formula (7)；K(xBoiEff,xBoiEffi) it is kernel function, it is obtained according to formula (6)
It arrives, xBoiEffiIndicate training sample set TBoiEffIn i-th of input variable；
In formula,
Indicate k-1 moment flue gas oxygen content, xofa(k-1) indicate that the k-1 moment burns throttle opening, xseca(k-1) table
Show k-1 moment secondary air register aperture, xcoal(k-1) biasing of k-1 moment coal-supplying amount, x are indicatedload(k-1) k-1 moment unit is indicated
Load, yBoiEff(k-1) k-1 moment NOx concentration, y are indicatedBoiEff(k-2) k-2 moment NOx concentration, σ are indicatedBoiEffIndicate boiler
EFFICIENCY PREDICTION model nuclear parameter；
In formula (7),Y=[yBoiEff1 yBoiEff2 … yBoiEffN]T, yBoiEffiFor training sample set TBoiEff
In i-th of output variable, αBoiEff0iRepresenting matrix αBoiEff0In i-th of element, Υ obtains according to formula (8)；
In formula (8), c representative function penalty factor,1≤l, i≤N, xBoiEffl
Indicate training sample set TBoiEffIn first of input variable.
5. ultra-supercritical boiler closed loop optimized control method of combustion according to claim 1, it is characterised in that: the step
In S2, the NOx concentration predicted value at current k moment is compared with measured value, whether judgment bias meets required precision, if
It is unsatisfactory for, model parameter and sample data is updated, specifically includes the following steps:
S2.1: by the NOx concentration predicted value y at current k momentNOx(k) and measured valueIt is compared, it is pre- to obtain NOx concentration
The deviation Error of surveyNOx(k):
If ErrorNOx(k) > MaxErrorNOx, MaxErrorNOxIt indicates the maximum deviation that NOx concentration prediction allows, then carries out
Step S2.2；
S2.2: building Multidimensional Parametric Vectors pNOxSVM=[bNOx αNOx σNOx]T∈RN+2, αNOxIndicate NOx concentration prediction model glug
Bright day multiplier, bNOxIndicate NOx concentration prediction model decision function parameter, σNOxIt indicates NOx concentration prediction model nuclear parameter, utilizes
Unscented kalman filtering carries out parameter Estimation, specifically includes the following steps:
S2.2.1: initialization
In formula,Indicate the priori mean value of stochastic variable.
In formula, P0Indicate the covariance matrix of stochastic variable, N >=1；
S2.2.2: 2d+1 sigma sampled point is chosen, d=N+2:
Wherein, χk-1Indicate the 2d+1 sigma sampled point at k-1 moment,Indicate the estimates of parameters at k-1 moment,Representing matrixZ column,Representing matrix?
Z-d column, Pk-1Indicate that the covariance matrix d=N+2, ψ of the parameter Estimation at k-1 moment are proportionality coefficient, Wm1Indicate the 1st sampling
The desired weight of point, Wc1Indicate the weight of the 1st sampled point variance, WmzIndicate the desired weight of z-th of sampled point, WczIt indicates
The weight of z-th of sampled point variance；
S2.2.3: it is calculated by the following formula:
χk|k-1=F (χk-1)=χk-1 (15)
Wherein, χk|k-1Indicate that the k-1 moment carries out the sample information after nonlinear state functional transformation, F (χ to sampled pointk-1) indicate
Nonlinear state function, χk-1Indicate the 2d+1 sigma sampled point at k-1 moment,Indicate that a step state at k-1 moment is pre-
It surveys,Indicate the error co-variance matrix at k-1 moment, PwIndicate the variance matrix of systematic procedure noise, χz,k|k-1Indicate χk|k-1
Z column；
S2.2.4: it is calculated by the following formula:
yk|k-1=G (χk|k-1) (18)
Wherein, yk|k-1Indicate the output valve that k-1 moment sigma point is obtained by non-linear measurement functional transformation, yz,k|k-1For
yk|k-1Z column, G (χk|k-1) indicate non-linear measurement functional transformation,Indicate that a step at k-1 moment exports prediction,It indicates
The auto-covariance matrix at k-1 moment,Indicate that the Cross-covariance at k-1 moment, K indicate filtering gain matrix,It indicates
The k moment parameter Estimation arrived,Indicate a step status predication at k-1 moment, ykIndicate measuring value, PkIndicate that k moment parameter is estimated
The covariance matrix of meter,Indicate the error co-variance matrix at k-1 moment, PvIndicate the Positive Definite matrix of system measurements noise,
Thus the NOx concentration prediction model parameters of subsequent time: b are acquiredNOx=pNOxSVM(1), αNOx=pNOxSVM(2:N+1), σNOx=
pNOxSVM(N+2)。
6. ultra-supercritical boiler closed loop optimized control method of combustion according to claim 1, it is characterised in that: the step
In S2, the boiler efficiency predicted value at current k moment is compared with measured value, whether judgment bias meets required precision, such as
Fruit is unsatisfactory for, and is updated to model parameter and sample data, specifically includes the following steps:
S2.3: by the boiler efficiency predicted value y at current k momentBoiEff(k) and measured valueIt is compared, it is dense to obtain NOx
Spend the deviation Error of predictionBoiEff(k):
If ErrorBoiEff(k) > MaxErrorBoiEff, MaxErrorBoiEffThe maximum for indicating that boiler efficiency prediction allows is inclined
Difference then carries out step S2.4；
S2.4: building Multidimensional Parametric Vectors pBoiEffSVM=[bBoiEff αBoiEff σBoiEff]T∈RN+2, αBoiEffIndicate boiler efficiency
Prediction model Lagrange multiplier, bBoiEffIndicate boiler efficiency prediction model decision function parameter, σBoiEffIndicate boiler efficiency
Prediction model nuclear parameter carries out parameter Estimation using Unscented kalman filtering, specifically includes the following steps:
S2.4.1: initialization
In formula,Indicate the priori mean value of stochastic variable；
In formula, P0Indicate the covariance matrix of initial time, N >=1；
S2.4.2: 2d+1 sigma sampled point is chosen, d=N+2:
In formula, χk-1Indicate the 2d+1 sigma sampled point at k-1 moment,Indicate the mean value of the stochastic variable at k-1 moment,Representing matrixZ column,Representing matrix?
Z-d column, Pk-1Indicate the covariance matrix of the parameter Estimation at k-1 moment, d=N+2, ψ are proportionality coefficient, Wm1Indicate that the 1st is adopted
The desired weight of sampling point, Wc1Indicate the weight of the 1st sampled point variance, WmzIndicate the desired weight of z-th of sampled point, WczTable
Show the weight of z-th of sampled point variance；
S2.4.3: it is calculated by the following formula:
χk|k-1=F (χk-1)=χk-1 (31)
In formula, χk|k-1Indicate that the k-1 moment carries out the sample information after nonlinear state functional transformation, F (χ to sampled pointk-1) indicate
Nonlinear state function, χk-1Indicate the 2d+1 sigma sampled point at k-1 moment,Indicate a step status predication at k-1 moment,Indicate the error co-variance matrix at k-1 moment, PwIndicate the variance matrix of systematic procedure noise, χz,k|k-1Indicate χk|k-1?
Z column；
S2.4.4: it is calculated by the following formula:
yk|k-1=G (χk|k-1) (34)
In formula, yk|k-1Indicate the output valve that the sigma point at k-1 moment is obtained by non-linear measurement functional transformation, yz,k|k-1For
yk|k-1Z column, G (χk|k-1) indicate non-linear measurement functional transformation,Indicate that a step at k-1 moment exports prediction,It indicates
The auto-covariance matrix at k-1 moment,Indicate that the Cross-covariance at k-1 moment, K indicate filtering gain matrix,It indicates
The k moment parameter Estimation arrived,Indicate a step status predication at k-1 moment, ykIndicate measuring value, PkIndicate that k moment parameter is estimated
The covariance matrix of meter,Indicate the error co-variance matrix at k-1 moment, PvIndicate the Positive Definite matrix of system measurements noise,
Thus the boiler efficiency prediction model parameters of subsequent time: b are acquiredBoiEff=pBoiEffSVM(1), αBoiEff=pBoiEffSVM(2:N+
1), σBoiEff=pBoiEffSVM(N+2)。
7. ultra-supercritical boiler closed loop optimized control method of combustion according to claim 1, it is characterised in that: the step
In S3, the NOx concentration predicted value at the following P moment are as follows:
As 1≤j1When≤M,
Wherein, M indicates control time domain, yNOx(k+j1) indicate future j1The NOx concentration predicted value at a moment, αNOxiRepresenting matrix
αNOxIn i-th of element, αNOxIndicate current time NOx concentration prediction model Lagrange multiplier, bNOxIndicate current time
NOx concentration prediction model decision function parameter, K (xNOx(k+j1),xNOxi) it is kernel function, it is calculated by formula (41), N >=1,
In formula, xNOxiFor NOx concentration training sample set TNOxIn i-th of input variable, xNOx(k+j1) indicate k+j1The input at moment
Variable is obtained according to formula (42):
In formula (42),Indicate the k+j acquired1The flue gas oxygen content at -1 moment, xofa(k+j1- 1) expression acquires
k+j1- 1 moment burnt throttle opening, xseca(k+j1- 1) k+j acquired is indicated1The secondary air register aperture at -1 moment, xcoal(k+
j1- 1) k+j acquired is indicated1The coal-supplying amount at -1 moment biases, xload(k) current k moment unit load, y are indicatedNOx(k+j1-1)
Indicate k+j1The NOx emission concentration prediction value at -1 moment, yNOx(k+j1- 2) k+j is indicated1The NOx emission concentration prediction at -2 moment
Value；
As M≤j1When≤P,
Wherein, M indicates control time domain, yNOx(k+j1) indicate future j1The NOx concentration predicted value at a moment, αNOxiRepresenting matrix
αNOxIn i-th of element, αNOxIndicate current time NOx concentration prediction model Lagrange multiplier, bNOxIndicate current time
NOx concentration prediction model decision function parameter, K (xNOx(k+j1),xNOxi) it is kernel function, it is calculated by formula (43), N >=1,
In formula, xNOxiFor NOx concentration training sample set TNOxIn i-th of input variable, xNOx(k+j1) indicate k+j1The input at moment
Variable is obtained according to formula (44):
In formula,Indicate the flue gas oxygen content at the k+M-1 moment acquired, xofa(k+M-1) k+M-1 acquired is indicated
Moment burns throttle opening, xseca(k+M-1) the secondary air register aperture at the k+M-1 moment acquired, x are indicatedcoal(k+M-1) table
Show the coal-supplying amount biasing at the k+M-1 moment acquired, xload(k) current k moment unit load, y are indicatedNOx(k+j1- 1) k+ is indicated
j1The NOx emission concentration prediction value at -1 moment, yNOx(k+j1- 2) k+j is indicated1The NOx emission concentration prediction value at -2 moment.
8. ultra-supercritical boiler closed loop optimized control method of combustion according to claim 1, it is characterised in that: the step
In S3, the boiler efficiency predicted value at the following P moment are as follows:
As 1≤j2When≤M,
Wherein, M indicates control time domain, yBoiEff(k+j2) indicate future j2The boiler efficiency predicted value at a moment, αBoiEffiIt indicates
Matrix αBoiEffiIn i-th of element, αBoiEffIndicate current time boiler efficiency prediction model Lagrange multiplier, bBoiEffTable
Show current time boiler efficiency prediction model decision function parameter, K (xBoiEff(k+j2),xBoiEffi) it is kernel function, by formula (45)
It is calculated, N >=1,
In formula, xBoiEffiFor boiler efficiency training sample set TBoiEfiIn i-th of input variable, xBoiEff(k+j2) indicate k+j2When
The input variable at quarter is obtained according to formula (46):
In formula,Indicate the k+j acquired2The flue gas oxygen content at -1 moment, xofa(k+j2- 1) k+ acquired is indicated
j2- 1 moment burnt throttle opening, xseca(k+j2- 1) k+j acquired is indicated2The secondary air register aperture at -1 moment, xcoal(k+
j2- 1) k+j acquired is indicated2The coal-supplying amount at -1 moment biases, xload(k) current k moment unit load, y are indicatedBoiEff(k+j2-
1) k+j is indicated2The boiler efficiency predicted value at -1 moment, yBoiEff(k+j2- 2) k+j is indicated2The boiler efficiency predicted value at -2 moment；
As M≤j2When≤P,
Wherein, M indicates control time domain, yBoiEff(k+j2) indicate future j2The boiler efficiency predicted value at a moment, αBoiEffiIt indicates
Matrix αBoiEffIn i-th of element, αBoiEffIndicate current time boiler efficiency prediction model Lagrange multiplier, bBoiEffTable
Show current time boiler efficiency prediction model decision function parameter, K (xBoiEff(k+j2),xBoiEffi) it is kernel function, by formula (47)
It is calculated, N >=1,
In formula, xBoiEffiFor boiler efficiency training sample set TBoiEfiIn i-th of input variable, xBoiEff(k+j2) indicate k+j2When
The input variable at quarter is obtained according to formula (48):
In formula,Indicate the flue gas oxygen content at the k+M-1 moment acquired, xofa(k+M-1) k+M-1 acquired is indicated
Moment burns throttle opening, xseca(k+M-1) the secondary air register aperture at the k+M-1 moment acquired, x are indicatedcoal(k+M-1) table
Show the coal-supplying amount biasing at the k+M-1 moment acquired, xload(k) current k moment unit load, y are indicatedBoiEff(k+j2- 1) k+ is indicated
j2The boiler efficiency predicted value at -1 moment, yBoiEff(k+j2- 2) k+j is indicated2The boiler efficiency predicted value at -2 moment.
9. ultra-supercritical boiler closed loop optimized control method of combustion according to claim 1, it is characterised in that: the step
S4 specifically includes the following steps:
S4.1: objective function is determined:
s.t.MVmin< MV < MVmax (50)
ΔMVmin< Δ MV < Δ MVmax (51)
yNOx(k+j1)≤NOxmax, j=1,2 ..., N (52)
Wherein,For objective function, MV represents oxygen amount definite value, burns throttle opening, secondary air register aperture or to coal
Amount biasing,yBoiEff(k+j2) indicate to the predicted value of the following P moment boiler efficiency,
MVmaxIndicate the upper limit value of MV, MVminIndicate that the lower limit value of MV, Δ MV indicate the rate of change of MV, Δ MVmaxIndicate that Δ MV's is upper
Limit value, Δ MVminIndicate the lower limit value of Δ MV, yNOx(k+j1) indicate the predicted value of the following P moment NOx emission concentration, NOxmax
Indicate NOxThe upper limit value of discharge, N >=1；
S4.2: nonlinear optimal problem is solved, the control variable at the following M moment is obtained and exports；Wherein, the following M moment
Control variable includexofa(k+1)…xofa(k+M)、xseca(k+1)…xseca(k+M) and
xcoal(k+1)…xcoal(k+M),Indicate the flue gas oxygen content at the following j moment, xofa(k+j) when indicating j following
Burn throttle opening, x quarterseca(k+j) the secondary air register aperture at the following j moment, x are indicatedcoal(k+j) the following j moment is indicated
Coal-supplying amount biasing, 1≤j≤M.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811149972.1A CN109519957B (en) | 2018-09-29 | 2018-09-29 | A kind of ultra-supercritical boiler closed loop optimized control method of combustion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811149972.1A CN109519957B (en) | 2018-09-29 | 2018-09-29 | A kind of ultra-supercritical boiler closed loop optimized control method of combustion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109519957A true CN109519957A (en) | 2019-03-26 |
| CN109519957B CN109519957B (en) | 2019-11-12 |
Family
ID=65771563
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811149972.1A Active CN109519957B (en) | 2018-09-29 | 2018-09-29 | A kind of ultra-supercritical boiler closed loop optimized control method of combustion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109519957B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110486749A (en) * | 2019-08-29 | 2019-11-22 | 国网河南省电力公司电力科学研究院 | A kind of thermal power unit boiler optimized control method of combustion and system |
| CN113341713A (en) * | 2021-05-31 | 2021-09-03 | 苏州西热节能环保技术有限公司 | Coal-air synchronous dynamic coordination control method for coal-fired unit |
| CN113671837A (en) * | 2021-08-24 | 2021-11-19 | 华能南京燃机发电有限公司 | Method and system for controlling NOx emission during cold start of combined cycle unit |
| CN115598974A (en) * | 2022-09-02 | 2023-01-13 | 南京天洑软件有限公司（Cn） | Denitration system model prediction control method and device based on linear system identification |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040181300A1 (en) * | 2003-03-11 | 2004-09-16 | Clark Robert L. | Methods, apparatus and computer program products for adaptively controlling a system by combining recursive system identification with generalized predictive control |
| EP1941328B1 (en) * | 2005-10-27 | 2010-06-16 | Honeywell International Inc. | Automated tuning of large-scale multivariable model predictive controllers for spatially-distributed processes |
| CN103324862A (en) * | 2013-07-11 | 2013-09-25 | 中国石油大学(华东) | Coal-fired boiler optimization method based on improved neural network and genetic algorithm |
| CN104613468A (en) * | 2015-01-27 | 2015-05-13 | 华北电力大学 | Circulating fluidized bedboiler combustion optimizing control method based on fuzzy adaptive inference |
| CN106500128A (en) * | 2017-01-12 | 2017-03-15 | 东北电力大学 | Power plant boiler NOx emission control method under a kind of varying load condition |
-
2018
- 2018-09-29 CN CN201811149972.1A patent/CN109519957B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040181300A1 (en) * | 2003-03-11 | 2004-09-16 | Clark Robert L. | Methods, apparatus and computer program products for adaptively controlling a system by combining recursive system identification with generalized predictive control |
| EP1941328B1 (en) * | 2005-10-27 | 2010-06-16 | Honeywell International Inc. | Automated tuning of large-scale multivariable model predictive controllers for spatially-distributed processes |
| CN103324862A (en) * | 2013-07-11 | 2013-09-25 | 中国石油大学(华东) | Coal-fired boiler optimization method based on improved neural network and genetic algorithm |
| CN104613468A (en) * | 2015-01-27 | 2015-05-13 | 华北电力大学 | Circulating fluidized bedboiler combustion optimizing control method based on fuzzy adaptive inference |
| CN106500128A (en) * | 2017-01-12 | 2017-03-15 | 东北电力大学 | Power plant boiler NOx emission control method under a kind of varying load condition |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110486749A (en) * | 2019-08-29 | 2019-11-22 | 国网河南省电力公司电力科学研究院 | A kind of thermal power unit boiler optimized control method of combustion and system |
| CN113341713A (en) * | 2021-05-31 | 2021-09-03 | 苏州西热节能环保技术有限公司 | Coal-air synchronous dynamic coordination control method for coal-fired unit |
| US11822316B2 (en) | 2021-05-31 | 2023-11-21 | Suzhou Tpri Ener & Enviro Tech Co., Ltd. | Coal-air synchronous dynamic coordinated control method for coal-fired unit |
| CN113671837A (en) * | 2021-08-24 | 2021-11-19 | 华能南京燃机发电有限公司 | Method and system for controlling NOx emission during cold start of combined cycle unit |
| CN113671837B (en) * | 2021-08-24 | 2024-05-24 | 华能南京燃机发电有限公司 | Cold start NOx emission control method and system for combined cycle unit |
| CN115598974A (en) * | 2022-09-02 | 2023-01-13 | 南京天洑软件有限公司（Cn） | Denitration system model prediction control method and device based on linear system identification |
| CN115598974B (en) * | 2022-09-02 | 2024-02-20 | 南京天洑软件有限公司 | Denitration system model predictive control method and device based on linear system identification |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109519957B (en) | 2019-11-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109519957B (en) | A kind of ultra-supercritical boiler closed loop optimized control method of combustion | |
| CN107526927B (en) | Blast furnace molten iron quality online robust soft measurement method | |
| CN109147878B (en) | Soft measurement method for free calcium of cement clinker | |
| CN109001979A (en) | Blast-melted quality adaptation Robust Predictive Control method based on Lazy learning | |
| CN104750131B (en) | A kind of fluidized-bed temperature control method based on parameter identification | |
| CN111829003B (en) | Power plant combustion control system and control method | |
| CN101498554B (en) | Serial automatic coal injection control system and method for blast furnace | |
| CN103048058B (en) | Online detecting method of coke-oven flue temperatures | |
| CN107016176A (en) | A kind of hybrid intelligent overall boiler burning optimization method | |
| CN102176221B (en) | Coke furnace temperature predicting method based on dynamic working conditions in coke furnace heating and burning process | |
| CN109190848A (en) | A kind of SCR system NO based on Time-delay PredictionxConcentration of emission prediction technique | |
| CN104715142B (en) | A kind of station boiler NOxDischarge dynamic soft-measuring method | |
| CN109583585A (en) | A kind of station boiler wall temperature prediction neural network model | |
| CN109342703B (en) | Method and system for measuring content of free calcium in cement clinker | |
| CN106594794B (en) | A kind of hybrid intelligent boiler efficiency burning optimization model update method | |
| CN110334840A (en) | A kind of thermal power plant's coal yard field damage prediction technique based on neural network algorithm | |
| CN108845501A (en) | A kind of blast-melted quality adaptation optimal control method based on Lazy learning | |
| CN102842066A (en) | Modeling method for combustion optimization of biomass furnace | |
| CN111915080A (en) | Raw fuel cost optimal proportioning method based on molten iron quality constraint | |
| CN102277468B (en) | Real-time forecasting method of LF refining furnace molten steel temperature | |
| CN101968832A (en) | Coal ash fusion temperature forecasting method based on construction-pruning mixed optimizing RBF (Radial Basis Function) network | |
| CN105808945B (en) | A kind of hybrid intelligent boiler efficiency burning optimization method | |
| CN109992844A (en) | A kind of boiler flyash carbon content prediction technique based on ADQPSO-SVR model | |
| CN104408317A (en) | Metallurgy enterprise gas flow interval predicting method based on Bootstrap echo state network integration | |
| CN110533247B (en) | Monthly power consumption prediction method adopting temperature data anomaly point compensation |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |