CN103413039B - Circulating fluid bed boiler secondary air dynamo-electric stream prognoses system and method - Google Patents

Abstract

The invention discloses a kind of circulating fluid bed boiler secondary air dynamo-electric stream prognoses system and method, system comprises the field intelligent instrument, database, data-interface, control station and the host computer that are connected with CFBB; Field intelligent instrument is connected with control station, database and host computer, and host computer comprises: standardization module, for gathering the training sample of key variables from database, the column criterion of going forward side by side processing; Forecasting mechanism forms module, for setting up forecast model; Prediction Executive Module, for real-time estimate overfire air fan electric current; Model modification module; Signal acquisition module; Result display module. The present invention predicts overfire air fan electric current according to the operating condition of CFBB and performance variable, so that suggestion and guides operation, by the overfire air fan Current Control of CFBB at ideal range, effectively improve boiler operating efficiency, extension device service life, and for further operational efficiency being optimized and being laid the foundation.

Description

Circulating fluid bed boiler secondary air dynamo-electric stream prognoses system and method

Technical field

The present invention relates to energy project field, especially, relate to a kind of circulating fluid bed boiler secondary air dynamo-electric stream prognoses system andMethod.

Background technology

CFBB has the advantages such as pollutant emission is few, fuel tolerance wide, Load Regulation ability is strong, exists in recent yearsIn the industry such as electric power, heat supply, obtain applying more and more widely. Along with growing tension and people's energy-conserving and environment-protective of the energy are realized notDisconnected enhancing, user excavates in the urgent need to the operation potentiality to boiler unit, improves the operational efficiency of unit. But at presentMost of CFBB all exists automaticity low, and operation relies on the feature of artificial experience, makes the energy-saving potential of boilerBe difficult to be taped the latent power fully, a major reason that causes this situation is to lack rational prognoses system and method. Secondary AirMachine is one of three large blower fans of CFBB, is also one of main current consuming apparatus, and overfire air fan electric current directly reflects twoThe size of inferior fan energy consumption. Based on the consideration of energy-conservation object, set up the prognoses system of the dynamo-electric stream of circulating fluid bed boiler secondary air,Energy efficient operation, operating analysis and operation optimization to CFBB are significant.

Summary of the invention

The object of the invention is to for the deficiencies in the prior art, a kind of CFBB heat loss due to exhaust gas prognoses system is providedAnd method.

The technical solution adopted for the present invention to solve the technical problems is: a kind of circulating fluid bed boiler secondary air machine current forecasting isSystem, comprises the field intelligent instrument, database, data-interface, control station and the host computer that are connected with CFBB;Field intelligent instrument is connected with control station, database and host computer, and described host computer comprises:

Standardization module, for gather the historical record of operating condition variable and performance variable from database, composition is from becomingThe training sample matrix X of amount, gathers the historical record of corresponding overfire air fan current signal, composition dependent variable training sample vectorY, carries out standardization to training sample X, Y, and the average that makes each variable is 0, and variance is 1, obtain after standardization fromVariable training sample matrix X^*(dependent variable training sample vector Y after n × p), standardization^*(n × 1), adopts following process to complete:

1.1) average:

${\stackrel{\‾}{x}}_j=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{x}_{\mathrm{ij}},(i=\mathrm{1,2},\·\·\·,n;j=\mathrm{1,2},\·\·\·,p)---\left(1\right)$

$\stackrel{\‾}{y}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{y}_i,(i=1,2,\·\·\·,n)---\left(2\right)$

1.2) ask standard deviation

$s_{x,j}=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-{\stackrel{\‾}{x}}_j)}^2},(i=\mathrm{1,2},\·\·\·,n;j=\mathrm{1,2},\·\·\·,p)---\left(3\right)$

$s_y=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(y_i-\stackrel{\‾}{y})}^2},(i=\mathrm{1,2},\·\·\·,n)---\left(4\right)$

1.3) standardization

$x_{\mathrm{ij}}^{*}=\frac{x_{\mathrm{ij}}-{\stackrel{\‾}{x}}_j}{s_{x,j}},(i=\mathrm{1,2},\·\·\·,n;j=\mathrm{1,2},\·\·\·,p)---\left(5\right)$

$y_i^{*}=\frac{y_i-\stackrel{\‾}{y}}{s_y},(i=\mathrm{1,2},\·\·\·,n)---\left(6\right)$

Wherein, x_ij、y_iFor the initial value of training sample point, n is training sample number, and p is independent variable number,For training sampleThis average, s_x,j、s_yFor the standard deviation of training sample, For the standardized value of training sample point, wherein subscript i, jRepresent respectively i training sample point, a j independent variable.

Forecasting mechanism forms module, and for setting up forecast model, implementation step is as follows:

2.1) make initial argument's residual error battle array E₀=X^*；

2.2) make the initial residual vector F of dependent variable₀=Y^*；

2.3) make Initial Composition count h=1;

2.4) successively by following various solving:

$w_h=\frac{E_{h-1}^T{F}_{h-1}}{\left|\right|E_{h-1}^T{F}_{h-1}\left|\right|}---\left(7\right)$

t_h=E_h-1w_h（8）

$r_h=\frac{F_{h-1}^T{t}_h}{{\left|\right|t_h\left|\right|}^2}---\left(9\right)$

$p_h=\frac{E_{h-1}^T{t}_h}{{\left|\right|t_h\left|\right|}^2}---\left(10\right)$

$w_h^{*}=\underset{j=1}{\overset{h-1}{\mathrm{\Π}}}(I-w_j{p}_j^T)w_h---\left(11\right)$

$E_h=E_{h-1}-t_h{p}_h^T---\left(12\right)$

F_h=F_h-1-t_hr_h（13）

$\mathrm{\β}=r_1{w}_1^{*}+r_2{w}_2^{*}+\·\·\·+r_h{w}_h^{*}---\left(14\right)$

Wherein, w_hThe axial vector of h composition, t_hH composition, r_hPilot process coefficient, p_h、Mistake in the middle of beingCheng Xiangliang, E_hH residual error battle array of independent variable, F_hBe h residual error battle array of dependent variable, β is the predictive coefficient of dependent variableVector, the transposition of subscript T representing matrix, subscript h, h-1 represent respectively containing the corresponding composition sequence number of lower target physical quantity be h,h-1；

2.5) read test sample from database, carries out nondimensionalization processing by (15), (16) formula:

${x_{\mathrm{ij}}^{*}}^{\′}=\frac{{x_{\mathrm{ij}}}^{\′}-{\stackrel{\‾}{x}}_j}{s_{x,j}},(i=\mathrm{1,2},\·\·\·,m;j=\mathrm{1,2},\·\·\·,p)---\left(15\right)$

${y_i^{*}}^{\′}=\frac{{y_i}^{\′}-\stackrel{\‾}{y}}{s_y},(i=\mathrm{1,2},\·\·\·,m)---\left(16\right)$

Wherein, x_ij'、y_i' be the initial value of test sample book point,Be the nondimensionalization value of test sample book point, m is test sample bookNumber;

2.6) ask the predicted value of dependent variable:

${{\hat{y}}_i^{*}}^{\′}=({x_{i1}^{*}}^{\′},{x_{i2}^{*}}^{\′},...,{x_{\mathrm{ip}}^{*}}^{\′})\mathrm{\β}---\left(17\right)$

Wherein,It is the nondimensionalization predicted value of dependent variable test sample book point;

2.7) ask current predicated error:

${S_{\mathrm{SS},h}}^{\′}=\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{({y_i}^{\′}-{{\hat{y}}_i}^{\′})}^2---\left(18\right)$

Wherein, S_SS,h' be residual sum of squares (RSS);

2.8) if h=1 makes h=2, return to 2.4), otherwise turn 2.9);

2.9) ask discriminant coefficientWork as P_h, think that introducing h composition can obviously improve prediction at >=0.02 o'clockAbility, makes h=h+1, returns to 2.4), otherwise make h=h-1, turn 2.10);

2.10) predictive coefficient vector β is transmitted and stores into prediction Executive Module.

Prediction Executive Module, for predicting Secondary Air electromechanics according to the performance variable of the operating condition of CFBB and settingStream, implementation step is as follows:

3.1) the independent variable signal of input is pressed to (19) formula processing:

${x\left(t\right)}_j^{*}=\frac{x{\left(t\right)}_j-{\stackrel{\‾}{x}}_j}{s_{x,j}},(j=\mathrm{1,2},\·\·\·,p)---\left(19\right)$

Wherein, x (t)_jFor t moment j independent variable initial value,Be the average of j independent variable training sample, s_x,jBe j independent variable instructionPractice the standard deviation of sample,For t moment j independent variable nondimensionalization value, t represents that time, unit are second;

3.2) ask the nondimensionalization predicted value of overfire air fan electric current by following formula:

$\hat{y}{\left(t\right)}^{*}=\left(\begin{array}{cccc}{x\left(t\right)}_1^{*}& {x\left(t\right)}_2^{*}& \·\·\·& {x\left(t\right)}_p^{*}\end{array}\right)\mathrm{\β}---\left(20\right)$

Wherein,For the nondimensionalization predicted value of t moment overfire air fan electric current;

3.3) ask the former dimension predicted value of overfire air fan electric current by following formula:

$\hat{y}\left(t\right)=\hat{y}{\left(t\right)}^{*}\·s_y+\stackrel{\‾}{y}---\left(21\right)$

Wherein,For the former dimension predicted value of t moment overfire air fan electric current.

As preferred a kind of scheme: described host computer also comprises: model modification module, for the time interval by setting willActual overfire air fan electric current and predicted value comparison, if relative error is greater than 10%, add new data training sample data,Re-execute standardization module and forecasting mechanism and form module.

Further, described host computer also comprises:

Signal acquisition module, for the sampling time interval by setting, gathers real time data from field intelligent instrument, and from numberAccording to gathering historical data in storehouse.

Result display module, for reading parameters from control station, and passes to control station by overfire air fan current forecasting value and carries outShow, and provide suggestion for operation: under current operating mode, how performance variable is adjusted is conducive to overfire air fan Current Control mostAt ideal range, so that control station staff, according to overfire air fan current forecasting value and suggestion for operation, adjusts operation bar in timePart, at ideal range, improves boiler operating efficiency by overfire air fan Current Control, simultaneously extension device service life. Wherein,How performance variable is adjusted is conducive to overfire air fan Current Control at ideal range most, and a short-cut method is by performance variableMultiple combination value, substitution overfire air fan current forecasting system, obtains corresponding overfire air fan current forecasting value, thereby very directly perceivedGround obtains by big or small.

As preferred another kind of scheme: described independent variable comprises: operating condition variable: main steam flow, environment temperature, toCoolant-temperature gage, combustion chamber draft, bed pressure, coal-fired moisture, coal-fired volatile matter, coal-fired ash content, coal-fired sulphur content; Performance variable: onceWind total blast volume, Secondary Air total blast volume.

The dynamo-electric stream of a kind of circulating fluid bed boiler secondary air Forecasting Methodology, described Forecasting Methodology comprises the following steps:

1) from database, gather the historical record of operating condition variable and performance variable, the training sample matrix of composition independent variableX, gathers the historical record of corresponding overfire air fan current signal, composition dependent variable training sample vector Y, to training sample X,Y carries out standardization, and the average that makes each variable is 0, and variance is 1, obtains independent variable training sample matrix after standardizationX^*(dependent variable training sample vector Y after n × p), standardization^*(n × 1), adopts following process to complete:

1.1) average:

${\stackrel{\‾}{x}}_j=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{x}_{\mathrm{ij}},(i=\mathrm{1,2},\·\·\·,n;j=\mathrm{1,2},\·\·\·,p)---\left(1\right)$

$\stackrel{\‾}{y}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{y}_i,(i=1,2,\·\·\·,n)---\left(2\right)$

1.2) ask standard deviation

$s_{x,j}=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-{\stackrel{\‾}{x}}_j)}^2},(i=\mathrm{1,2},\·\·\·,n;j=\mathrm{1,2},\·\·\·,p)---\left(3\right)$

$s_y=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(y_i-\stackrel{\‾}{y})}^2},(i=\mathrm{1,2},\·\·\·,n)---\left(4\right)$

1.3) standardization

$x_{\mathrm{ij}}^{*}=\frac{x_{\mathrm{ij}}-{\stackrel{\‾}{x}}_j}{s_{x,j}},(i=\mathrm{1,2},\·\·\·,n;j=\mathrm{1,2},\·\·\·,p)---\left(5\right)$

$y_i^{*}=\frac{y_i-\stackrel{\‾}{y}}{s_y},(i=\mathrm{1,2},\·\·\·,n)---\left(6\right)$

Wherein, x_ij、y_iFor the initial value of training sample point, n is training sample number, and p is independent variable number,For training sampleThis average, s_x,j、s_yFor the standard deviation of training sample,For the standardized value of training sample point, wherein subscript i, jRepresent respectively i training sample point, a j independent variable.

2) the standardized training sample obtaining is set up to forecast model by following process:

2.1) make initial argument's residual error battle array E₀=X^*；

2.2) make the initial residual vector F of dependent variable₀=Y^*；

2.3) make Initial Composition count h=1;

2.4) successively by following various solving:

$w_h=\frac{E_{h-1}^T{F}_{h-1}}{\left|\right|E_{h-1}^T{F}_{h-1}\left|\right|}---\left(7\right)$

t_h=E_h-1w_h（8）

$r_h=\frac{F_{h-1}^T{t}_h}{{\left|\right|t_h\left|\right|}^2}---\left(9\right)$

$p_h=\frac{E_{h-1}^T{t}_h}{{\left|\right|t_h\left|\right|}^2}---\left(10\right)$

$w_h^{*}=\underset{j=1}{\overset{h-1}{\mathrm{\Π}}}(I-w_j{p}_j^T)w_h---\left(11\right)$

$E_h=E_{h-1}-t_h{p}_h^T---\left(12\right)$

F_h=F_h-1-t_hr_h（13）

$\mathrm{\β}=r_1{w}_1^{*}+r_2{w}_2^{*}+\·\·\·+r_h{w}_h^{*}---\left(14\right)$

Wherein, w_hThe axial vector of h composition, t_hH composition, r_hPilot process coefficient, p_h、Mistake in the middle of beingCheng Xiangliang, E_hH residual error battle array of independent variable, F_hBe h residual error battle array of dependent variable, β is the predictive coefficient of dependent variableVector, the transposition of subscript T representing matrix, subscript h, h-1 represent respectively containing the corresponding composition sequence number of lower target physical quantity be h,h-1；

2.5) read test sample from database, carries out nondimensionalization processing by (15), (16) formula:

${x_{\mathrm{ij}}^{*}}^{\′}=\frac{{x_{\mathrm{ij}}}^{\′}-{\stackrel{\‾}{x}}_j}{s_{x,j}},(i=\mathrm{1,2},\·\·\·,m;j=\mathrm{1,2},\·\·\·,p)---\left(15\right)$

${y_i^{*}}^{\′}=\frac{{y_i}^{\′}-\stackrel{\‾}{y}}{s_y},(i=\mathrm{1,2},\·\·\·,m)---\left(16\right)$

Wherein, x_ij'、y_i' be the initial value of test sample book point,Be the nondimensionalization value of test sample book point, m is test sample bookNumber;

2.6) ask the predicted value of dependent variable:

${{\hat{y}}_i^{*}}^{\′}=({x_{i1}^{*}}^{\′},{x_{i2}^{*}}^{\′},...,{x_{\mathrm{ip}}^{*}}^{\′})\mathrm{\β}---\left(17\right)$

Wherein,It is the nondimensionalization predicted value of dependent variable test sample book point;

2.7) ask current predicated error:

${S_{\mathrm{SS},h}}^{\′}=\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{({y_i}^{\′}-{{\hat{y}}_i}^{\′})}^2---\left(18\right)$

Wherein, S_SS,h' be residual sum of squares (RSS);

2.8) if h=1 makes h=2, return to 2.4), otherwise turn 2.9);

2.9) ask discriminant coefficientWork as P_h, think that introducing h composition can obviously improve prediction at >=0.02 o'clockAbility, makes h=h+1, returns to 2.4), otherwise make h=h-1, turn 2.10);

2.10) preserve the predictive coefficient vector β finally obtaining.

3) using the performance variable of the operating condition variable of CFBB and setting as input signal, according to predictive coefficient toAmount, predicts overfire air fan electric current, and implementation step is as follows:

3.1) the independent variable signal of input is pressed to (19) formula processing:

${x\left(t\right)}_j^{*}=\frac{x{\left(t\right)}_j-{\stackrel{\‾}{x}}_j}{s_{x,j}},(j=\mathrm{1,2},\·\·\·,p)---\left(19\right)$

Wherein, x (t)_jFor t moment j independent variable initial value,Be the average of j independent variable training sample, s_x,jBe j independent variable instructionPractice the standard deviation of sample,For t moment j independent variable nondimensionalization value, t represents that time, unit are second;

3.2) ask the nondimensionalization predicted value of overfire air fan electric current by following formula:

$\hat{y}{\left(t\right)}^{*}=\left(\begin{array}{cccc}{x\left(t\right)}_1^{*}& {x\left(t\right)}_2^{*}& \·\·\·& {x\left(t\right)}_p^{*}\end{array}\right)\mathrm{\β}---\left(20\right)$

Wherein,For the nondimensionalization predicted value of t moment overfire air fan electric current;

3.3) ask the former dimension predicted value of overfire air fan electric current by following formula:

$\hat{y}\left(t\right)=\hat{y}{\left(t\right)}^{*}\·s_y+\stackrel{\‾}{y}---\left(21\right)$

Wherein,For the former dimension predicted value of t moment overfire air fan electric current.

As preferred a kind of scheme: described method also comprises: 4) by the sampling time interval of setting, collection site Intelligent InstrumentTable signal, by the actual overfire air fan electric current and the predicted value comparison that obtain, if relative error is greater than 10%, adds new dataEnter training sample data, re-execute step 1), 2), so that forecast model is upgraded.

Further, in described step 3), read parameters from control station, and overfire air fan current forecasting value is passed to controlSystem station shows, and provides suggestion for operation: under current operating mode, how performance variable is adjusted is conducive to overfire air fan mostCurrent Control is at ideal range, so that control station staff, according to overfire air fan current forecasting value and suggestion for operation, adjusts in timeWhole operating condition, at ideal range, improves boiler operating efficiency by overfire air fan Current Control, simultaneously extension device service life.Wherein, how performance variable is adjusted is conducive to overfire air fan Current Control at ideal range most, and a short-cut method is by operationThe multiple combination value of variable, substitution overfire air fan current forecasting system, obtains corresponding overfire air fan current forecasting value, from but notOften obtain by big or small intuitively.

As preferred another kind of scheme: described independent variable comprises: operating condition variable: main steam flow, environment temperature, toCoolant-temperature gage, combustion chamber draft, bed pressure, coal-fired moisture, coal-fired volatile matter, coal-fired ash content, coal-fired sulphur content; Performance variable: onceWind total blast volume, Secondary Air total blast volume.

Beneficial effect of the present invention is mainly manifested in: the overfire air fan electric current to CFBB is predicted, advises and refers toLead production operation, overfire air fan Current Control, at ideal range, is excavated to device energy-saving potential, extension fixture service life.

Brief description of the drawings

Fig. 1 is the hardware structure diagram of system proposed by the invention.

Fig. 2 is the functional block diagram of host computer of the present invention.

Detailed description of the invention

Below in conjunction with drawings and Examples, the invention will be further described.

Embodiment 1

With reference to Fig. 1, Fig. 2, the dynamo-electric stream of a kind of circulating fluid bed boiler secondary air prognoses system, comprises and CFBB 1The field intelligent instrument 2, data-interface 3, database 4, control station 5 and the host computer 6 that connect, field intelligent instrument 2 withFieldbus connects, and data/address bus is connected with data-interface 3, data-interface 3 and database 4, control station 5 and host computer 6Connect, described host computer 6 comprises:

Standardization module 7, for gather the historical record of operating condition variable and performance variable from database, composition certainlyThe training sample matrix X of variable, gathers the historical record of corresponding overfire air fan current signal, composition dependent variable training sample toAmount Y, carries out standardization to training sample X, Y, and the average that makes each variable is 0, and variance is 1, obtains after standardizationIndependent variable training sample matrix X^*(dependent variable training sample vector Y after n × p), standardization^*(n × 1), adopts following processBecome:

1.1) average:

${\stackrel{\‾}{x}}_j=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{x}_{\mathrm{ij}},(i=\mathrm{1,2},\·\·\·,n;j=\mathrm{1,2},\·\·\·,p)---\left(1\right)$

$\stackrel{\‾}{y}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{y}_i,(i=1,2,\·\·\·,n)---\left(2\right)$

1.2) ask standard deviation

$s_{x,j}=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-{\stackrel{\‾}{x}}_j)}^2},(i=\mathrm{1,2},\·\·\·,n;j=\mathrm{1,2},\·\·\·,p)---\left(3\right)$

$s_y=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(y_i-\stackrel{\‾}{y})}^2},(i=\mathrm{1,2},\·\·\·,n)---\left(4\right)$

1.3) standardization

$x_{\mathrm{ij}}^{*}=\frac{x_{\mathrm{ij}}-{\stackrel{\‾}{x}}_j}{s_{x,j}},(i=\mathrm{1,2},\·\·\·,n;j=\mathrm{1,2},\·\·\·,p)---\left(5\right)$

$y_i^{*}=\frac{y_i-\stackrel{\‾}{y}}{s_y},(i=\mathrm{1,2},\·\·\·,n)---\left(6\right)$

Wherein, x_ij、y_iFor the initial value of training sample point, n is training sample number, and p is independent variable number,For training sampleThis average, s_x,j、s_yFor the standard deviation of training sample,For the standardized value of training sample point, wherein subscript i, jRepresent respectively i training sample point, a j independent variable.

Forecasting mechanism forms module 8, and for setting up forecast model, implementation step is as follows:

2.1) make initial argument's residual error battle array E₀=X^*；

2.2) make the initial residual vector F of dependent variable₀=Y^*；

2.3) make Initial Composition count h=1;

2.4) successively by following various solving:

$w_h=\frac{E_{h-1}^T{F}_{h-1}}{\left|\right|E_{h-1}^T{F}_{h-1}\left|\right|}---\left(7\right)$

t_h=E_h-1w_h（8）

$r_h=\frac{F_{h-1}^T{t}_h}{{\left|\right|t_h\left|\right|}^2}---\left(9\right)$

$p_h=\frac{E_{h-1}^T{t}_h}{{\left|\right|t_h\left|\right|}^2}---\left(10\right)$

$w_h^{*}=\underset{j=1}{\overset{h-1}{\mathrm{\Π}}}(I-w_j{p}_j^T)w_h---\left(11\right)$

$E_h=E_{h-1}-t_h{p}_h^T---\left(12\right)$

F_h=F_h-1-t_hr_h（13）

$\mathrm{\β}=r_1{w}_1^{*}+r_2{w}_2^{*}+\·\·\·+r_h{w}_h^{*}---\left(14\right)$

Wherein, w_hThe axial vector of h composition, t_hH composition, r_hPilot process coefficient, p_h、Mistake in the middle of beingCheng Xiangliang, E_hH residual error battle array of independent variable, F_hBe h residual error battle array of dependent variable, β is the predictive coefficient of dependent variableVector, the transposition of subscript T representing matrix, subscript h, h-1 represent respectively containing the corresponding composition sequence number of lower target physical quantity be h,h-1；

2.5) read test sample from database, carries out nondimensionalization processing by (15), (16) formula:

${x_{\mathrm{ij}}^{*}}^{\′}=\frac{{x_{\mathrm{ij}}}^{\′}-{\stackrel{\‾}{x}}_j}{s_{x,j}},(i=\mathrm{1,2},\·\·\·,m;j=\mathrm{1,2},\·\·\·,p)---\left(15\right)$

${y_i^{*}}^{\′}=\frac{{y_i}^{\′}-\stackrel{\‾}{y}}{s_y},(i=\mathrm{1,2},\·\·\·,m)---\left(16\right)$

Wherein, x_ij'、y_i' be the initial value of test sample book point,Be the nondimensionalization value of test sample book point, m is test sample bookNumber;

2.6) ask the predicted value of dependent variable:

${{\hat{y}}_i^{*}}^{\′}=({x_{i1}^{*}}^{\′},{x_{i2}^{*}}^{\′},...,{x_{\mathrm{ip}}^{*}}^{\′})\mathrm{\β}---\left(17\right)$

Wherein,It is the nondimensionalization predicted value of dependent variable test sample book point;

2.7) ask current predicated error:

${S_{\mathrm{SS},h}}^{\′}=\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{({y_i}^{\′}-{{\hat{y}}_i}^{\′})}^2---\left(18\right)$

Wherein, S_SS,h' be residual sum of squares (RSS);

2.8) if h=1 makes h=2, return to 2.4), otherwise turn 2.9);

2.9) ask discriminant coefficientWork as P_h, think that introducing h composition can obviously improve prediction at >=0.02 o'clockAbility, makes h=h+1, returns to 2.4), otherwise make h=h-1, turn 2.10);

2.10) predictive coefficient vector β is transmitted and stores into prediction Executive Module.

Prediction Executive Module 9, for predicting overfire air fan according to the performance variable of the operating condition of CFBB and settingElectric current, implementation step is as follows:

3.1) the independent variable signal of input is pressed to (19) formula processing:

${x\left(t\right)}_j^{*}=\frac{x{\left(t\right)}_j-{\stackrel{\‾}{x}}_j}{s_{x,j}},(j=\mathrm{1,2},\·\·\·,p)---\left(19\right)$

Wherein, x (t)_jFor t moment j independent variable initial value,Be the average of j independent variable training sample, s_x,jBe j independent variable training sampleThis standard deviation,For t moment j independent variable nondimensionalization value, t represents that time, unit are second;

3.2) ask the nondimensionalization predicted value of overfire air fan electric current by following formula:

$\hat{y}{\left(t\right)}^{*}=\left(\begin{array}{cccc}{x\left(t\right)}_1^{*}& {x\left(t\right)}_2^{*}& \·\·\·& {x\left(t\right)}_p^{*}\end{array}\right)\mathrm{\β}---\left(20\right)$

Wherein,For the nondimensionalization predicted value of t moment overfire air fan electric current;

3.3) ask the former dimension predicted value of overfire air fan electric current by following formula:

$\hat{y}\left(t\right)=\hat{y}{\left(t\right)}^{*}\·s_y+\stackrel{\‾}{y}---\left(21\right)$

Wherein,For the former dimension predicted value of t moment overfire air fan electric current.

Described host computer 6 also comprises: signal acquisition module 11, and for the sampling time interval by setting, from site intelligent instrumentTable gathers real time data, and from database, gathers historical data.

Described host computer 6 also comprises: model modification module 12, and for pressing the time interval of setting by actual Secondary Air electromechanicsStream and predicted value comparison, if relative error is greater than 10%, add new data training sample data, re-executes standardizationProcessing module and forecasting mechanism form module.

Described host computer 6 also comprises: result display module 10, and for reading parameters from control station, and by overfire air fanCurrent forecasting value is passed to control station and is shown, and provides suggestion for operation: under current operating mode, how performance variable is adjustedBe conducive to overfire air fan Current Control at ideal range so that control station staff, according to overfire air fan current forecasting value andSuggestion for operation, adjusts operating condition in time, and overfire air fan Current Control, at ideal range, is improved to boiler operating efficiency, simultaneouslyExtension device service life. Wherein, how performance variable is adjusted is conducive to overfire air fan Current Control at ideal range most, and oneIndividual short-cut method is by the multiple combination value of performance variable, and substitution overfire air fan current forecasting system, obtains corresponding overfire air fanCurrent forecasting value, thus obtain by big or small very intuitively.

The hardware components of described host computer 6 comprises: I/O element, for the collection of data and the transmission of information; Data storage,Data sample and operational factor etc. that storage running is required; Program storage, the software program of storage practical function module; ComputingDevice, performing a programme, realizes the function of specifying; Display module, shows the parameter, the operation result that arrange, and provides suggestion for operation.

Embodiment 2

With reference to Fig. 1, Fig. 2, the dynamo-electric stream of a kind of circulating fluid bed boiler secondary air Forecasting Methodology, described Forecasting Methodology comprises followingStep:

1) from database, gather the historical record of operating condition variable and performance variable, the training sample matrix of composition independent variableX, gathers the historical record of corresponding overfire air fan current signal, composition dependent variable training sample vector Y, to training sample X,Y carries out standardization, and the average that makes each variable is 0, and variance is 1, obtains independent variable training sample matrix after standardizationX^*(dependent variable training sample vector Y after n × p), standardization^*(n × 1), adopts following process to complete:

1.1) average:

${\stackrel{\‾}{x}}_j=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{x}_{\mathrm{ij}},(i=\mathrm{1,2},\·\·\·,n;j=\mathrm{1,2},\·\·\·,p)---\left(1\right)$

$\stackrel{\‾}{y}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{y}_i,(i=1,2,\·\·\·,n)---\left(2\right)$

1.2) ask standard deviation

$s_{x,j}=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(x_{\mathrm{ij}}-{\stackrel{\‾}{x}}_j)}^2},(i=\mathrm{1,2},\·\·\·,n;j=\mathrm{1,2},\·\·\·,p)---\left(3\right)$

$s_y=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(y_i-\stackrel{\‾}{y})}^2},(i=\mathrm{1,2},\·\·\·,n)---\left(4\right)$

1.3) standardization

$x_{\mathrm{ij}}^{*}=\frac{x_{\mathrm{ij}}-{\stackrel{\‾}{x}}_j}{s_{x,j}},(i=\mathrm{1,2},\·\·\·,n;j=\mathrm{1,2},\·\·\·,p)---\left(5\right)$

$y_i^{*}=\frac{y_i-\stackrel{\‾}{y}}{s_y},(i=\mathrm{1,2},\·\·\·,n)---\left(6\right)$

Wherein, x_ij、y_iFor the initial value of training sample point, n is training sample number, and p is independent variable number,For training sampleThis average, s_x,j、s_yFor the standard deviation of training sample,For the standardized value of training sample point, wherein subscript i, jRepresent respectively i training sample point, a j independent variable.

2) the standardized training sample obtaining is set up to forecast model by following process:

2.1) make initial argument's residual error battle array E₀=X^*；

2.2) make the initial residual vector F of dependent variable₀=Y^*；

2.3) make Initial Composition count h=1;

2.4) successively by following various solving:

$w_h=\frac{E_{h-1}^T{F}_{h-1}}{\left|\right|E_{h-1}^T{F}_{h-1}\left|\right|}---\left(7\right)$

t_h=E_h-1w_h（8）

$r_h=\frac{F_{h-1}^T{t}_h}{{\left|\right|t_h\left|\right|}^2}---\left(9\right)$

$p_h=\frac{E_{h-1}^T{t}_h}{{\left|\right|t_h\left|\right|}^2}---\left(10\right)$

$w_h^{*}=\underset{j=1}{\overset{h-1}{\mathrm{\Π}}}(I-w_j{p}_j^T)w_h---\left(11\right)$

$E_h=E_{h-1}-t_h{p}_h^T---\left(12\right)$

F_h=F_h-1-t_hr_h（13）

$\mathrm{\β}=r_1{w}_1^{*}+r_2{w}_2^{*}+\·\·\·+r_h{w}_h^{*}---\left(14\right)$

Wherein, w_hThe axial vector of h composition, t_hH composition, r_hPilot process coefficient, p_h、Mistake in the middle of beingCheng Xiangliang, E_hH residual error battle array of independent variable, F_hBe h residual error battle array of dependent variable, β is the predictive coefficient of dependent variableVector, the transposition of subscript T representing matrix, subscript h, h-1 represent respectively containing the corresponding composition sequence number of lower target physical quantity be h,h-1；

2.5) read test sample from database, carries out nondimensionalization processing by (15), (16) formula:

${x_{\mathrm{ij}}^{*}}^{\′}=\frac{{x_{\mathrm{ij}}}^{\′}-{\stackrel{\‾}{x}}_j}{s_{x,j}},(i=\mathrm{1,2},\·\·\·,m;j=\mathrm{1,2},\·\·\·,p)---\left(15\right)$

${y_i^{*}}^{\′}=\frac{{y_i}^{\′}-\stackrel{\‾}{y}}{s_y},(i=\mathrm{1,2},\·\·\·,m)---\left(16\right)$

Wherein, x_ij'、y_i' be the initial value of test sample book point,Be the nondimensionalization value of test sample book point, m is test sample bookNumber;

2.6) ask the predicted value of dependent variable:

${{\hat{y}}_i^{*}}^{\′}=({x_{i1}^{*}}^{\′},{x_{i2}^{*}}^{\′},...,{x_{\mathrm{ip}}^{*}}^{\′})\mathrm{\β}---\left(17\right)$

Wherein,It is the nondimensionalization predicted value of dependent variable test sample book point;

2.7) ask current predicated error:

${S_{\mathrm{SS},h}}^{\′}=\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{({y_i}^{\′}-{{\hat{y}}_i}^{\′})}^2---\left(18\right)$

Wherein, S_SS,h' be residual sum of squares (RSS);

2.8) if h=1 makes h=2, return to 2.4), otherwise turn 2.9);

2.9) ask discriminant coefficientWork as P_h, think that introducing h composition can obviously improve prediction at >=0.02 o'clockAbility, makes h=h+1, returns to 2.4), otherwise make h=h-1, turn 2.10);

2.10) preserve the predictive coefficient vector β finally obtaining.

3) using the performance variable of the operating condition variable of CFBB and setting as input signal, according to predictive coefficient toAmount, predicts overfire air fan electric current, and implementation step is as follows:

3.1) the independent variable signal of input is pressed to (19) formula processing:

${x\left(t\right)}_j^{*}=\frac{x{\left(t\right)}_j-{\stackrel{\‾}{x}}_j}{s_{x,j}},(j=\mathrm{1,2},\·\·\·,p)---\left(19\right)$

Wherein, x (t)_jFor t moment j independent variable initial value,Be the average of j independent variable training sample, s_x,jBe j independent variable instructionPractice the standard deviation of sample,For t moment j independent variable nondimensionalization value, t represents that time, unit are second;

3.2) ask the nondimensionalization predicted value of overfire air fan electric current by following formula:

$\hat{y}{\left(t\right)}^{*}=\left(\begin{array}{cccc}{x\left(t\right)}_1^{*}& {x\left(t\right)}_2^{*}& \·\·\·& {x\left(t\right)}_p^{*}\end{array}\right)\mathrm{\β}---\left(20\right)$

Wherein,For the nondimensionalization predicted value of t moment overfire air fan electric current;

3.3) ask the former dimension predicted value of overfire air fan electric current by following formula:

$\hat{y}\left(t\right)=\hat{y}{\left(t\right)}^{*}\·s_y+\stackrel{\‾}{y}---\left(21\right)$

Wherein,For the former dimension predicted value of t moment overfire air fan electric current.

Described method also comprises: 4), by the sampling time interval of setting, collection site intelligence instrument signal, by the reality obtainingOverfire air fan electric current and predicted value comparison, if relative error is greater than 10%, add new data training sample data, againExecution step 1), 2), so that forecast model is upgraded.

In described step 3), read parameters from control station, and overfire air fan current forecasting value is passed to control station and carry outShow, and provide suggestion for operation: under current operating mode, how performance variable is adjusted is conducive to overfire air fan Current Control mostAt ideal range, so that control station staff, according to overfire air fan current forecasting value and suggestion for operation, adjusts operation bar in timePart, at ideal range, improves boiler operating efficiency by overfire air fan Current Control, simultaneously extension device service life. Wherein,How performance variable is adjusted is conducive to overfire air fan Current Control at ideal range most, and a short-cut method is by performance variableMultiple combination value, substitution overfire air fan current forecasting system, obtains corresponding overfire air fan current forecasting value, thereby very directly perceivedGround obtains by big or small.

Described independent variable comprises: operating condition variable: main steam flow, environment temperature, feed temperature, combustion chamber draft, bed press,Coal-fired moisture, coal-fired volatile matter, coal-fired ash content, coal-fired sulphur content; Performance variable: wind total blast volume, Secondary Air total blast volume.

Circulating fluid bed boiler secondary air dynamo-electric stream prognoses system and method proposed by the invention, by above-mentioned concrete enforcement stepSuddenly be described, person skilled obviously can be not departing from content of the present invention, spirit and scope device as herein describedChange with method of operating or suitably change and combination, realize the technology of the present invention. Special needs to be pointed out is all phase classesLike replace and change apparent to one skilled in the artly, they all can be deemed to be included in spirit of the present invention, modelEnclose with content in.

Claims (1)

1. the dynamo-electric stream of a circulating fluid bed boiler secondary air Forecasting Methodology, is characterized in that, described Forecasting Methodology comprises the following steps:

1) from database, gather the historical record of operating condition variable and performance variable, the training sample matrix of composition independent variableX, gathers the historical record of corresponding overfire air fan current signal, composition dependent variable training sample vector Y, to training sample X,Y carries out standardization, and the average that makes each variable is 0, and variance is 1, obtains independent variable training sample matrix after standardizationX^*(dependent variable training sample vector Y after n × p), standardization^*(n × 1), adopts following process to complete:

1.1) average:

${\stackrel{\‾}{x}}_j=\frac{1}{n}\underset{i=1}{\overset{n}{\Σ}}{x}_{ij},(i=1,2,...,n;j=1,2,...,p)---\left(1\right)$

$\stackrel{\‾}{y}=\frac{1}{n}\underset{i=1}{\overset{n}{\Σ}}{y}_i,(i=1,2,...,n)---\left(2\right)$

1.2) ask standard deviation

$s_{x,j}=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\Σ}}{(x_{ij}-{\stackrel{\‾}{x}}_j)}^2},(i=1,2,...,n;j=1,2,...,p)---\left(3\right)$

$s_y=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\Σ}}{(y_i-\stackrel{\‾}{y})}^2},(i=1,2,...,n)---\left(4\right)$

1.3) standardization

$x_{ij}^{*}=\frac{x_{ij}-{\stackrel{\‾}{x}}_j}{s_{x,j}},(i=1,2,...,n;j=1,2,...,p)---\left(5\right)$

$y_i^{*}=\frac{y_i-\stackrel{\‾}{y}}{s_y},(i=1,2,...,n)---\left(6\right)$

Wherein, x_ij、y_iFor the initial value of training sample point, n is training sample number, and p is independent variable number,For training sampleThis average, s_x,j、s_yFor the standard deviation of training sample,For the standardized value of training sample point, wherein subscript i, jRepresent respectively i training sample point, a j independent variable;

2) the standardized training sample obtaining is set up to forecast model by following process:

2.1) make initial argument's residual error battle array E₀＝X^*；

2.2) make the initial residual vector F of dependent variable₀＝Y^*；

2.3) make Initial Composition count h=1;

2.4) successively by following various solving:

$w_h=\frac{E_{h-1}^T{F}_{h-1}}{\left|\right|E_{h-1}^T{F}_{h-1}\left|\right|}---\left(7\right)$

t_h＝E_h-1w_h(8)

$r_h=\frac{F_{h-1}^T{t}_h}{\left|\right|t_h|{|}^2}---\left(9\right)$

$p_h=\frac{E_{h-1}^T{t}_h}{\left|\right|t_h|{|}^2}---\left(10\right)$

$w_h^{*}=\underset{j=1}{\overset{h-1}{\Π}}(I-w_j{p}_j^T)w_h---\left(11\right)$

$E_h=E_{h-1}-t_h{p}_h^T---\left(12\right)$

F_h＝F_h-1-t_hr_h(13)

$\mathrm{\β}=r_1{w}_1^{*}+r_2{w}_2^{*}+...+r_h{w}_h^{*}---\left(14\right)$

Wherein, w_hThe axial vector of h composition, t_hH composition, r_hPilot process coefficient, p_h、Mistake in the middle of beingCheng Xiangliang, E_hH residual error battle array of independent variable, F_hBe h residual error battle array of dependent variable, β is the predictive coefficient of dependent variableVector, the transposition of subscript T representing matrix, subscript h, h-1 represent respectively containing the corresponding composition sequence number of lower target physical quantity be h,h-1；

2.5) read test sample from database, carries out nondimensionalization processing by (15), (16) formula:

${x_{ij}^{*}}^{\′}=\frac{{x_{ij}}^{\′}-{\stackrel{\‾}{x}}_j}{s_{x,j}},(i=1,2,...,m;j=1,2,...,p)---\left(15\right)$

${y_i^{*}}^{\′}=\frac{{y_i}^{\′}-\stackrel{\‾}{y}}{s_y},(i=1,2,...,m)---\left(16\right)$

Wherein, x_ij'、y_i' be the initial value of test sample book point,Be the nondimensionalization value of test sample book point, m is test sample bookNumber;

2.6) ask the predicted value of dependent variable:

${{\hat{y}}_i^{*}}^{\′}=({x_{i1}^{*}}^{\′},{x_{i2}^{*}}^{\′},...,{x_{ip}^{*}}^{\′})\mathrm{\β}---\left(17\right)$

Wherein,It is the nondimensionalization predicted value of dependent variable test sample book point;

2.7) ask current predicated error:

${S_{SS,h}}^{\′}=\underset{i=1}{\overset{m}{\Σ}}{({y_i}^{\′}-{{\hat{y}}_i}^{\′})}^2---\left(18\right)$

Wherein, S_SS,h' be residual sum of squares (RSS);

2.8) if h=1 makes h=2, return to 2.4), otherwise turn 2.9);

2.9) ask discriminant coefficientWork as P_h, think that introducing h composition can obviously improve prediction at >=0.02 o'clockAbility, makes h=h+1, returns to 2.4), otherwise make h=h-1, turn 2.10);

2.10) preserve the predictive coefficient vector β finally obtaining;

3) using the performance variable of the operating condition variable of CFBB and setting as input signal, according to predictive coefficient toAmount, predicts overfire air fan electric current, and implementation step is as follows:

3.1) the independent variable signal of input is pressed to (19) formula processing:

$x{\left(t\right)}_j^{*}=\frac{x{\left(t\right)}_j-{\stackrel{\‾}{x}}_j}{s_{x,j}},(j=1,2,...,p)---\left(19\right)$

Wherein, x (t)_jFor t moment j independent variable initial value,Be the average of j independent variable training sample, s_x,jBe j independent variable training sampleThis standard deviation,For t moment j independent variable nondimensionalization value, t represents that time, unit are second;

3.2) ask the nondimensionalization predicted value of overfire air fan electric current by following formula:

$\hat{y}{\left(t\right)}^{*}=\left(\begin{array}{cccc}x{\left(t\right)}_1^{*}& x{\left(t\right)}_2^{*}& ...& x{\left(t\right)}_p^{*}\end{array}\right)\mathrm{\β}---\left(20\right)$

Wherein,For the nondimensionalization predicted value of t moment overfire air fan electric current;

3.3) ask the former dimension predicted value of overfire air fan electric current by following formula:

$\hat{y}\left(t\right)=\hat{y}{\left(t\right)}^{*}\·s_y+\stackrel{\‾}{y}---\left(21\right)$

Wherein,For the former dimension predicted value of t moment overfire air fan electric current;

4) by the sampling time interval of setting, collection site intelligence instrument signal, by the actual overfire air fan electric current and the prediction that obtainValue relatively, if relative error is greater than 10%, adds new data training sample data, re-executes step 1), 2), withForecast model is upgraded;

In described step 3) in, read parameters from control station, and overfire air fan current forecasting value is passed to control station and enterRow shows, and provides suggestion for operation: under current operating mode, how performance variable is adjusted is conducive to the control of overfire air fan electric current mostBuilt in ideal range, so that control station staff, according to overfire air fan current forecasting value and suggestion for operation, adjusts operation in timeCondition, at ideal range, improves boiler operating efficiency by overfire air fan Current Control, simultaneously extension device service life; Wherein,How described performance variable is adjusted is conducive to overfire air fan Current Control in the implementation method of ideal range performance variable mostMultiple combination value, substitution overfire air fan current forecasting system, obtains corresponding overfire air fan current forecasting value, thereby very directly perceivedGround obtains by big or small;

Described independent variable comprises: operating condition variable: main steam flow, environment temperature, feed temperature, combustion chamber draft, bed press,Coal-fired moisture, coal-fired volatile matter, coal-fired ash content, coal-fired sulphur content; Performance variable: wind total blast volume, Secondary Air total blast volume.