CN105468932A - Heating efficiency online calculation method for boiler - Google Patents

Abstract

The invention discloses a heating efficiency online calculation method for a boiler. The method includes the specific steps that firstly, data is preprocessed; secondly, parameter models are built, wherein a soft measurement model with exhaust gas temperature as the target and a soft measurement model with fly ash carbon content as the target are built by means of a support vector machine (SVM); thirdly, heating efficiency is calculated, wherein the heating efficiency of the boiler is predicted and calculated online through a simplified heat efficiency formula and key parameters; fourthly, errors are analyzed. According to the method, the soft measurement model with exhaust gas temperature as the target and the soft measurement model with fly ash carbon content as the target are built through the support vector machine (SVM), and on the basis, the heating efficiency of the boiler is obtained online through the simplified heating efficiency computational formula. Modeling time is shortened and complexity is reduced while high fitting precision and prediction precision are obtained, data support is provided for monitoring and measuring of the performance of the boiler of a power plant, industrial needs are met, and the method has great guiding significance in operation and production.

Description

A kind of on-line calculation method of boiler thermal output

[technical field]

The present invention relates to boiler technology field, particularly the technical field of station boiler performance online purpose monitoring.

[background technology]

Boiler thermal output is the most important index weighing boiler output.At present, the computing method of boiler thermal output can be divided into two classes: 1, back balance computing method: national standard " GB10184-88 station boiler performance test code " discloses the computing method of the pulverized coal firing boiler thermal efficiency, wherein, Q _r---the boiler input heat of unit of fuel; Q ₂---the heat loss due to exhaust gas heat of unit of fuel; Q ₃---the loses heat that the inflammable gas of unit of fuel is imperfect combustion; Q ₄---the solid-unburning hot loss heat of unit of fuel; Q ₅---the boiler radiation loss heat of unit of fuel; Q ₆---the lime-ash physical sensible heat loses heat of unit of fuel; 2, positive balance computing method: wherein, Q ₁for the boiler of unit fuel effectively utilizes heat, generally increased by steam enthalpy and characterize with the product of steam flow.

Have the related application of boiler thermal output computing method at present, as CN200810053039, coal-burning boiler thermal efficiency on-line checkingi, obtains boiler thermal output by data acquisition and data processing, but does not show computing method; CN201010564627, the pulverized coal firing boiler thermal efficiency and coal data method of real-time, pass through data acquisition system (DAS), obtain each data that thermal efficiency back balance calculates, substitute into formulae discovery and obtain boiler efficiency, but required data variable is very many, generally more than 30, calculate comparatively complicated; CN201210193799, a kind of boiler thermal output on-line monitoring method based on coal quality database, by supposition boiler efficiency, pick out coal-fired calorific value, seek the mapping relations between coal burning caloricity and coal-fired composition, then carry out back balance calculating, then contrast with the boiler efficiency of supposition, carry out interative computation; CN201310335778, Thermal Efficiency of Circulating Fluidized Bed Boiler prognoses system and method, wherein, the predictor formula of boiler thermal output is: $\mathrm{\η}=100-(K_1\widehat{y{\left(t\right)}_1}+K_2)\frac{\hat{y}{\left(t\right)}_2}{100}-31223\frac{A_{ar}}{Q_{ar,net,p}}\×\frac{\hat{y}{\left(t\right)}_3}{100-\hat{y}{\left(t\right)}_3}-q_x,$ Wherein, K ₁, K ₂for the design factor relevant with coal; A _arfor the as received basis ash content percentage of fire coal; Q _{ar, net, p}for the net calorific value as received basis of fire coal; for the predicted value of t excess air coefficient; for the predicted value of the t smoke evacuation temperature difference; for the predicted value of t unburned carbon in flue dust percentage; q _xfor the empirical value of other a small amount of thermal loss content percentages of Circulating Fluidized Bed Boiler except heat loss due to exhaust gas and heat loss due to unburned carbon, get 1%.The present invention considers the requirement of real-time of boiler thermal output needed for line computation, the model of exhaust gas temperature and unburned carbon in flue dust is set up by support vector machine (SVM), then the Efficiency Calculation formula by simplifying obtains boiler thermal output, required variable is few, the modeling time is short, fitting precision and precision of prediction are all higher, meet industrial requirement.

[summary of the invention]

Object of the present invention solves the problems of the prior art exactly, a kind of on-line calculation method of boiler thermal output is proposed, not only reduce the variable number needed for Efficiency Calculation, reduce model complexity and modeling time, and there is higher fitting precision and precision of prediction, for the performance monitoring monitoring of boiler of power plant provides Data support.

For achieving the above object, the present invention proposes a kind of on-line calculation method of boiler thermal output, be based on based on support vector machine (SVM) on the basis of exhaust gas temperature and unburned carbon in flue dust modeling, utilize the Efficiency Calculation formula simplified to obtain boiler thermal output, concrete steps comprise:

(a) data prediction: utilize 3 σ criterions to reject the data with gross error gathered from Distributed Control System (DCS) (DCS);

B () parametric model is set up: the soft-sensing model that to utilize support vector machine (SVM) to set up with exhaust gas temperature, unburned carbon in flue dust be respectively target;

(c) Efficiency Calculation: utilize the thermal efficiency formula simplified to carry out on-line prediction in conjunction with key parameters and calculate boiler thermal output;

(d) error analysis: error analysis is carried out, calibration model parameter to the boiler thermal output that approximating and forecasting obtains.

As preferably, in described (b) step, the independent variable of parametric model comprises coal-fired calorific value, boiler load, total blast volume, coal-supplying amount, absorbing quantity, coal property, and coal property comprises application base ash content, carbon content, hydrogen richness, nitrogen content.

As optimization, in described (b) step, the parameter of support vector machine (SVM) is obtained by the way selection of cross validation.

As preferably, in described (c) step, the variable that the boiler thermal output simplified calculates has oxygen content of smoke gas, application base ash content, coal-fired calorific value, reference temperature, and the output of key parameters model, and key parameters comprises exhaust gas temperature, unburned carbon in flue dust predicted value.

Beneficial effect of the present invention:

The present invention is by support vector machine (SVM) soft-sensing model that to establish with exhaust gas temperature and unburned carbon in flue dust be target, and on this basis, the Efficiency Calculation formula simplified is utilized to obtain boiler thermal output online, while the higher fitting precision of acquisition and precision of prediction, reduce modeling time and complexity, for the performance monitoring monitoring of boiler of power plant provides Data support, meet industry needs, to operation production, there is directive significance.

[embodiment]

The on-line calculation method of a kind of boiler thermal output of the present invention, concrete steps comprise:

Step one, the data with gross error utilizing 3 σ criterions rejectings to gather from Distributed Control System (DCS) (DCS);

Step 2, the soft-sensing model that to utilize support vector machine (SVM) to set up with exhaust gas temperature, unburned carbon in flue dust be respectively target;

The thermal efficiency formula that step 3, utilization simplify carrys out on-line prediction in conjunction with key parameters (exhaust gas temperature, unburned carbon in flue dust predicted value) and calculates boiler thermal output;

Step 4, to approximating and forecasting obtain boiler thermal output carry out error analysis, calibration model parameter.

Support vector machine (SVM) uses nonlinear mapping function that sample set inseparable in lower dimensional space is mapped to higher dimensional space, in higher dimensional space, adopt linear learning method to solve classification and the regression problem of sample, small sample, non-linear and higher-dimension sample problem show good characteristic, makes it be used widely.

SVM regression problem is exactly under the condition of a given input amendment X, obtains corresponding output y by machine learning.Its mathematical description is: given input sample of data collection ((x ₁, y ₁), (x ₂, y ₂) ... (x _i, y _i)), if sample is P (X by select probability in sample set _i, y _i), the function output of input amendment being carried out to loss appraisal is c (x _i, y _i, f), wherein ε is unwise sensitivity, and f is the mapping function between input amendment and output sample.

$c(x_i,y_i,f)={|y_i-f\left(x_i\right)|}_{\mathrm{\ϵ}}=\left\{\begin{array}{c}0,|y_i-f\left(x_i\right)|\≤\mathrm{\ϵ}\\ |y_i-f\left(x_i\right)|-\mathrm{\ϵ},|y_i-f\left(x_i\right)|>\mathrm{\ϵ}\end{array}\right.$

SVM finds suitable mapping function f, makes formula (3.7) sample set loss function get minimum value:

R[f(·)]＝∫c(x _i,y _i,f)dP(x _i,y _i)

Wherein, Nonlinear Support Vector Machines regression algorithm is:

If the constrained input of sample meets nonlinear relationship, we need to find a nonlinear function ψ (X), be mapped in higher dimensional space by original sample, thus problem nonlinear in lower dimensional space is converted into the linear problem in higher dimensional space.Suppose that nonlinear solshing is:

f(x)＝w·Ψ(x)+b

Optimization problem corresponding to nonlinear case can be expressed as:

$\left\{\begin{array}{cc}\min & \frac{1}{2}\left|\right|w|{|}^2+C\underset{i=1}{\overset{l}{\Σ}}({\mathrm{\ξ}}_i+{{\mathrm{\ξ}}_i}^{*})\\ s.t.& \left\{\begin{array}{c}{y}_i-w\·\mathrm{\Ψ}\left(x_i\right)-b\≤\mathrm{\ϵ}+{\mathrm{\ξ}}_i\\ w\·\mathrm{\Ψ}\left(x_i\right)+b-y_i\≤\mathrm{\ϵ}+{{\mathrm{\ξ}}_i}^{*}\\ {\mathrm{\ξ}}_i,{{\mathrm{\ξ}}_i}^{*}\≥0\end{array}\right.\end{array}\right.$

Analogy linear SVM regression algorithm, solving above-mentioned optimization problem can draw

$w=\underset{i=1}{\overset{l}{\Σ}}(\mathrm{\α}-{\mathrm{\α}}^{*})\mathrm{\Ψ}\left(x_i\right)$

Then nonlinear estimation function is:

$f\left(x\right)=\underset{i=1}{\overset{l}{\Σ}}(\mathrm{\α}-{\mathrm{\α}}^{*})\mathrm{\Ψ}\left(x_i\right)\mathrm{\Ψ}\left(x\right)+b$

Definition Mercer kernel function K (X _i, X _j)=ψ (X _i) ^tψ (X _j), obtaining nonlinear regression function is:

$f\left(x\right)=\underset{i=1}{\overset{l}{\Σ}}(\mathrm{\α}-{\mathrm{\α}}^{*})K(x_i,x)+b$

In actual applications, the selection of kernel function is very large for the impact of soft-sensing model, Radial basis kernel function K (X _i, X _j)=exp (-| X _i-X _j| ²/ 2 σ ²) be current most popular kernel function.

The Efficiency Calculation formula simplified is mainly:

Dry gas loss: $L_G=W_G{C}_{PG}\frac{{\mathrm{\θ}}_A-{\mathrm{\θ}}_b}{1000H_u};$ Moisture thermal loss: $L_M=(W_{H_{2}O}+W_A{W}_{MA})\frac{H-H_b}{1000H_u};$ Unburned carbon thermal loss: $L_C=33730\frac{C_f}{100-C_f}{r}_f\frac{A}{1000H_u};$ Boiler radiation loss: $L_R=C_q{Q_N}^{-0.3}\frac{{10}^{-6}}{h}.$ Wherein, θ _bfor pressure fan outlet temperature; C _pGfor dry flue gas mean specific heat; W _mAfor the absolute humidity of air; H is water vapor enthalpy; H _bfor the enthalpy that saturated vapour is corresponding; r _ffor flying dust ratio; Q _nfor the maximum output of boiler; C is surface coefficient; Q is mean heat flux; H is loading coefficient.

The course of work of the present invention:

The present invention is by support vector machine (SVM) soft-sensing model that to establish with exhaust gas temperature and unburned carbon in flue dust be target, and on this basis, the Efficiency Calculation formula simplified is utilized to obtain boiler thermal output online, while the higher fitting precision of acquisition and precision of prediction, reduce modeling time and complexity, for the performance monitoring monitoring of boiler of power plant provides Data support, meet industry needs, to operation production, there is directive significance.

Above-described embodiment is to explanation of the present invention, is not limitation of the invention, anyly all belongs to protection scope of the present invention to the scheme after simple transformation of the present invention.

Claims (4)

1. the on-line calculation method of a boiler thermal output, it is characterized in that: described computing method be based on based on support vector machine (SVM) on the basis of exhaust gas temperature and unburned carbon in flue dust modeling, utilize the Efficiency Calculation formula simplified to obtain boiler thermal output, concrete steps comprise:

(a) data prediction: utilize 3 σ criterions to reject the data with gross error gathered from Distributed Control System (DCS) (DCS);

B () parametric model is set up: the soft-sensing model that to utilize support vector machine (SVM) to set up with exhaust gas temperature, unburned carbon in flue dust be respectively target;

(c) Efficiency Calculation: utilize the thermal efficiency formula simplified to carry out on-line prediction in conjunction with key parameters and calculate boiler thermal output;

(d) error analysis: error analysis is carried out, calibration model parameter to the boiler thermal output that approximating and forecasting obtains.

2. the on-line calculation method of a kind of boiler thermal output as claimed in claim 1, it is characterized in that: in described (b) step, the independent variable of parametric model comprises coal-fired calorific value, boiler load, total blast volume, coal-supplying amount, absorbing quantity, coal property, and coal property comprises application base ash content, carbon content, hydrogen richness, nitrogen content.

3. the on-line calculation method of a kind of boiler thermal output as claimed in claim 1, is characterized in that: in described (b) step, and the parameter of support vector machine (SVM) is obtained by the way selection of cross validation.

4. the on-line calculation method of a kind of boiler thermal output as claimed in claim 1, it is characterized in that: in described (c) step, the variable that the boiler thermal output simplified calculates has oxygen content of smoke gas, application base ash content, coal-fired calorific value, reference temperature, and the output of key parameters model, key parameters comprises exhaust gas temperature, unburned carbon in flue dust predicted value.