CN106647240A - Subcritical unit coordination prediction function control algorithm based on leading disturbance model - Google Patents

Abstract

The invention discloses a subcritical unit coordination prediction function control algorithm based on a leading disturbance model. The conventional PID control system can not consider both the load quickness and the pressure stability, and the quality of a coordination control system needs to be improved. The prediction function control is applied to a pressure closed loop of the subcritical unit coordination control system, a prediction function control system is formed, and the pressure closed loop has a delay inertial link; the load closed loop of the subcritical unit coordination control system adopts PID control, and a load instruction adopts a datum line plus differential mode for a coal quantity feedforward system; an actual turbine control valve instruction serves as a disturbance signal source, and the optimal control law is acquired according to a prediction model of the prediction function control system; and simplification is carried out on the basis of the optimal control law, and a simplified optimal control law is obtained. The control quality of the subcritical unit coordination control system is improved, and both the load quickness and the pressure stability can be considered.

Description

Subcritical Units coordinate forecast function control algolithm based on leading Disturbance Model

Technical field

The present invention relates to the control of subcritical fired power generating unit coordinated control system, specifically a kind of to be disturbed based on leading The Subcritical Units coordinate forecast function control algolithm of model.

Background technology

Large-scale subcritical fired power generating unit coordinated control system is the BIBO system with close coupling, boiler, vapour Expander system can be reduced to a BIBO system with close coupling, pressure-controlled object under given operating point With large delay, big inertia and time variation, response is relatively slow, and spatial load forecasting object has rapidity, has between them very strong Coupling, constitutes a conflicting unification.

In the coordination control of fired power generating unit coordinated control system, the Mathematical Modeling of control object has close coupling characteristic, Regulatory PID control system is difficult the stability of the rapidity and pressure for taking into account load, and coordinated control system quality has much room for improvement.

The content of the invention

The technical problem to be solved is the defect for overcoming above-mentioned prior art to exist, there is provided one kind is based on leading The Subcritical Units coordinate forecast function control algolithm of Disturbance Model, effectively to alleviate the work of the coupling between boiler and steam turbine With, it is ensured that Predictive Control System has stronger robustness；The Control platform of Subcritical Units coordinated control system is improved, is taken into account The rapidity of load and the stability of pressure.

For this purpose, the present invention is adopted the following technical scheme that：Subcritical Units coordinate forecast letter based on leading Disturbance Model Number control algolithm, comprises the steps：

1) Predictive function control is applied to the pressure closed loop loop of subcritical unit Coordinated Control Systems, forms pre- Function control system is surveyed, described pressure closed loop loop carries delays inertial element；Subcritical unit Coordinated Control Systems Load closed loop adopt PID control, load instruction to add by the way of differential coal amount feedforward system using datum line；

2) using actual steam turbine pitch instruct as disturbing signal source, anticipation in Predictive function control system, according to pre- The forecast model for surveying function control system obtains optimal control law；

3) simplified on the basis of optimal control law, index coefficient is reduced into multiplier coefficients, will be predicted that time domain is excellent Change length to be reduced to predict regulation coefficient, the optimal control law being simplified.

The present invention realizes the optimal control to Subcritical Units coordination target by adjustment prediction regulation coefficient.By the present invention Algorithm be applied to Subcritical Units coordinated control system, control effect demonstrates the premium properties of the algorithm.

The Predictive function control system architecture designed based on the algorithm is simple, and parameter adjustment is convenient, can preferably fit Duty requirements are answered, Control platform is improved.

Further, described Predictive function control system includes coal amount to the forecast model and pitch of pressure to pressure Forecast model, Predictive function control system is acted on by pitch to the forecast model of pressure, while load instruction passes through coal amount Feedforward system produces signal and instructs superposition in Stress control main channel with PREDICTIVE CONTROL；

Forecast model of the coal amount to pressure, for Mathematical Modeling of the approximate coal amount to pressure characteristic, it is：

In formula, K_m1For model gain, T_m1For model inertia time, T_dm1For model pure delay time (s is integrating factor)；

Forecast model of the pitch to pressure, for Mathematical Modeling of the approximate pitch to pressure characteristic, it is：

In formula, K_m2For model gain, T_m2For model inertia time, T_dm2For the model pure delay time.

Further, when using a basic function, have：

U (k+i)=u (k), i=1,2 ..., H-1,

Bu (k+i)=ff+u (k), i=1,2 ..., H-1,

Tu (k+i)=Tu (k), i=1,2 ..., H-1,

In above formula, k represents sampling instant, and H represents prediction time-domain step size, and u is optimum control amount, and Bu is control of steam turbine governing valve Instruction, ff is load to coal amount feed-forward signal, and Tu is the instruction of steam coal amount；

The prediction of each forecast model is output as：

In above formula, α_m1For forecast model G of the coal amount to pressure_m1(s) difference equation coefficient, α_m2It is pitch to the pre- of pressure Survey model G_m2(s) difference equation coefficient；

Forecast model is always output as：

y_m(k+H)=y_m1(k+H)+y_m2(k+H)

Obtaining optimal control law according to the extreme value of optimizing index is：

In above formula, c is setting value, and β is that system expects closed-loop dynamic characteristic, and y (k) is controlled volume, y_m1K () is coal amount to pressure The forecast model output of power, y_m2K () is that pitch is exported to the forecast model of pressure, y_mK () is y_m1(k) and y_m2It is defeated after (k) superposition Go out.

Further, described prediction regulation coefficient includes that reference locus predict regulation coefficient b and control forecasting adjustment system Number a, simplified optimal control law is as follows：

In above formula,T_sFor sampling period, T_RFor reference locus time constant；

The invention has the advantages that：Effectively alleviate the coupling between boiler and steam turbine；Ensure that prediction Control system has stronger robustness；The Control platform that Subcritical Units coordinate system is improve, the quick of load has been taken into account The stability of property and pressure.

Description of the drawings

Fig. 1 is that (in figure, c is setting value to conventional belt disturbance Predictive function control systematic schematic diagram, y_rFor reference locus, u is Optimum control amount, D be disturbing signal source, y_m2For Disturbance Model output, y_m1For object model output, y_mAlways export for model, y is Controlled device is exported.)

Fig. 2 is that (in figure, c is setting value, y for the schematic diagram of the present invention_rFor reference locus, u is optimum control amount, and ff is negative To coal amount feed-forward signal, Tu is control of steam turbine governing valve instruction to lotus, and Bu is the instruction of steam coal amount, y_m2It is pitch to pressure disturbance model Output, y_m1Pressure model is exported for coal amount, y_mAlways export for pressure model, y is actual pressure signal.)

Fig. 3 is load responding curve of the present invention (A is load setting value in figure, and B is load responding curve).

Fig. 4 is pressure response curve of the present invention (A is pressure set points in figure, and B is pressure response curve).

Specific embodiment

With reference to specification drawings and specific embodiments, the invention will be further described.

First, conventional belt disturbance algorithm of predictive functional control

According to Fig. 1, when the external disturbance signal of control system can be surveyed, the disturbance mould of test determination system can be passed through Type, now system is with two models, i.e. object model G_m1(s) and Disturbance Model G_m2(s), it is assumed that be single order and add and delay mould Type, i.e.,：

When using a basic function, have：

U (k+i)=u (k) i=1,2 ..., H-1

D (k+i)=D (k) i=1,2 ..., H-1 (3)

The prediction of each model is output as：

Forecast model is output as：

y_m(k+H)=y_m1(k+H)+y_m2(k+H) (5)

Can obtain optimal control law according to the extreme value of optimizing index is：

2nd, Subcritical Units coordinate forecast function Control System Design

For Subcritical Units, the premise of, feed water normal and generator excited system normal work normal in boiler combustion Under, boiler, turbine system can be reduced to a BIBO system with close coupling under given operating point：

Wherein：Δ N is unit electrical power variable quantity；Δ P is pressure variety before machine；Δ μ is the change of steam turbine pitch aperture Amount；Δ B is coal amount variable quantity.The form of each function in formula (7), wherein G can determine according to test₁₁Differential ring can be approximately Section, G₁₂、G₂₁、G₂₂Inertia can be approximately add and delay link.

According to Fig. 2, G_m1It is forecast model of the coal amount to pressure, for approximate coal amount to pressure characteristic；G_m2It is pitch to pressure The forecast model of power, for approximate pitch to pressure characteristic.

The present invention propose it is a kind of based on Predictive function control system coordination control algolithm, will pitch instruct as whole The disturbing signal source of individual pressure system, Predictive Control System is acted on by pitch to the Mathematical Modeling of pressure, while load refers to Order produces signal and instructs superposition in Stress control main channel with PREDICTIVE CONTROL by feedforward system, can derive from Fig. 2 The control law of the Predictive function control system.

When the external disturbance signal of control system can be surveyed, as steam turbine pitch instruction Tu, now system is with two moulds The forecast model G of type, i.e. coal amount to pressure_m1(s) and forecast model G of the pitch to pressure_m2(s), it is assumed that be single order and add and delay Model, i.e.,：

When using a basic function, have：

U (k+i)=u (k) i=1,2 ..., H-1

Bu (k+i)=ff+u (k) i=1,2 ..., H-1

Tu (k+i)=Tu (k) i=1,2 ..., H-1 (10)

The prediction of each model is output as：

Forecast model is always output as：

y_m(k+H)=y_m1(k+H)+y_m2(k+H) (12)

Can obtain optimal control law according to the extreme value of optimizing index is：

The present invention carries out simplifying design on the basis of above-mentioned optimal control law, and index coefficient is reduced into multiplier coefficients, will Prediction Optimization of Time Domain length is reduced to predict regulation coefficient, respectively reference locus prediction regulation coefficient b and control forecasting adjustment Coefficient a, it is simplified after the optimal control law designed it is as follows：

In above formula,T_sFor sampling period, T_RFor reference locus time constant, 4～12 are traditionally arranged to be Second；In adjustment control system quality, prediction regulation coefficient a and b need to be only adjusted, Just good Control platform can be obtained.

3rd, inventive algorithm control effect

Inventive control algorithm is applied into certain 300MW Subcritical Units coordinated control system, simulation study is carried out.According to Each transfer function characteristics of formula (7) coordination target are

Using PI controls, proportionality coefficient is 1 to turbine main control, and integral coefficient is 0.008；Feed-forward signal is rolled over using load instruction The coal amount reference signal of calculation and differential sum.

Using the Predictive function control system shown in Fig. 2, the first spy to pitch to the characteristic and coal amount of pressure to pressure Property is fitted simplification, obtains equivalent single order plus pure object model of delaying is：

Coal amount is to the forecast model of pressure：

Pitch is to the forecast model of pressure：

Two forecast models of PFC are taken as：K_m1=0.18, T_m1=300, T_dm1=170；K_m2=-0.09, T_m2=380, T_dm2=30；PFC adopts a basic function, sampling period to be set to 4, and reference locus time constant 12, reference locus prediction is adjusted Coefficient 4, control forecasting regulation coefficient is 2.2, carries out varying duty response experiment, and load instruction is changed into the load that 190MW is obtained With Stress control response curve as shown in Figure 3, Figure 4.Load deviation is within ± 2MW in Fig. 3, in Fig. 4 pressure divergence ± Within 0.3MPa.

Claims (4)

1. the Subcritical Units coordinate forecast function control algolithm based on leading Disturbance Model, comprises the steps：

1) Predictive function control is applied to the pressure closed loop loop of subcritical unit Coordinated Control Systems, prediction letter is formed Number control system, described pressure closed loop loop carries delays inertial element；Subcritical unit Coordinated Control Systems it is negative Lotus closed loop adopts PID control, load instruction to add by the way of differential coal amount feedforward system using datum line；

2) using the instruction of actual steam turbine pitch as disturbing signal source, anticipation in Predictive function control system, according to prediction letter The forecast model of number control system obtains optimal control law；

3) simplified on the basis of optimal control law, index coefficient is reduced into multiplier coefficients, prediction Optimization of Time Domain is long Degree is reduced to predict regulation coefficient, the optimal control law being simplified.

2. Subcritical Units coordinate forecast function control algolithm according to claim 1, it is characterised in that described prediction Function control system includes forecast model of the coal amount to the forecast model and pitch of pressure to pressure, by pitch to the pre- of pressure Survey model and act on Predictive function control system, while load instruction produces signal by coal amount feedforward system referring to PREDICTIVE CONTROL Superposition is made in Stress control main channel；

Forecast model of the coal amount to pressure, for Mathematical Modeling of the approximate coal amount to pressure characteristic, it is：

$G_{m1}\left(s\right)=\frac{K_{m1}}{T_{m1}s+1}{e}^{-T_{dm1}s},$

In formula, K_m1For model gain, T_m1For model inertia time, T_dm1For the model pure delay time；

Forecast model of the pitch to pressure, for Mathematical Modeling of the approximate pitch to pressure characteristic, it is：

$G_{m2}\left(s\right)=\frac{K_{m2}}{T_{m2}s+1}{e}^{-T_{dm2}s},$

In formula, K_m2For model gain, T_m2For model inertia time, T_dm2For the model pure delay time.

3. Subcritical Units coordinate forecast function control algolithm according to claim 2, it is characterised in that

When using a basic function, have：

U (k+i)=u (k), i=1,2 ..., H-1,

Bu (k+i)=ff+u (k), i=1,2 ..., H-1,

Tu (k+i)=Tu (k), i=1,2 ..., H-1,

In above formula, k represents sampling instant, and H represents prediction time-domain step size, and u is optimum control amount, and Bu refers to for control of steam turbine governing valve Order, ff is load to coal amount feed-forward signal, and Tu is the instruction of steam coal amount；

The prediction of each forecast model is output as：

$\left\{\begin{array}{c}{y}_{m1}(k+H)={\mathrm{\α}}_{m1}^H{y}_{m1}\left(k\right)+K_{m1}(1-{\mathrm{\α}}_{m1}^H)u\left(k\right)\\ y_{m2}(k+H)={\mathrm{\α}}_{m2}^H{y}_{m2}\left(k\right)+K_{m2}(1-{\mathrm{\α}}_{m2}^H)Tu\left(k\right)\end{array}\right.$

In above formula, α_m1For forecast model G of the coal amount to pressure_m1(s) difference equation coefficient, α_m2For prediction mould of the pitch to pressure Type G_m2(s) difference equation coefficient；

Forecast model is always output as：

y_m(k+H)=y_m1(k+H)+y_m2(k+H)

Obtaining optimal control law according to the extreme value of optimizing index is：

$\begin{array}{c}u\left(k\right)=\frac{c(k+H)-{\mathrm{\β}}^{H}c\left(k\right)-(1-{\mathrm{\β}}^H)y\left(k\right)-{\mathrm{\α}}_{m1}^H{y}_{m1}\left(k\right)}{K_{m1}(1-{\mathrm{\α}}_{m1}^H)}\\ -\frac{{\mathrm{\α}}_{m2}^H{y}_{m2}\left(k\right)+K_{m2}(1-{\mathrm{\α}}_{m2}^H)Tu\left(k\right)-y_m\left(k\right)}{K_{m1}(1-{\mathrm{\α}}_{m1}^H)}\end{array}$

In above formula, c is setting value, and β is that system expects closed-loop dynamic characteristic, and y (k) is controlled volume, y_m1K () is coal amount to pressure Forecast model is exported, y_m2K () is that pitch is exported to the forecast model of pressure, y_mK () is y_m1(k) and y_m2Export after (k) superposition.

4. fired power generating unit steam temperature algorithm of predictive functional control according to claim 3, it is characterised in that described prediction is adjusted Integral coefficient includes that reference locus predict regulation coefficient b and control forecasting regulation coefficient a, and simplified optimal control law is as follows：

$\begin{array}{c}u\left(k\right)=\frac{c(k+H)-\mathrm{\β}*b*c\left(k\right)-(1-\mathrm{\β}*b)*y\left(k\right)-{\mathrm{\α}}_{m1}*a*y_{m1}\left(k\right)}{K_{m1}*(1-{\mathrm{\α}}_{m1}*a)}\\ -\frac{{\mathrm{\α}}_{m2}*a*y_{m2}\left(k\right)+K_{m2}*(1-{\mathrm{\α}}_{m2}*a)*Tu\left(k\right)-y_m\left(k\right)}{K_{m1}*(1-{\mathrm{\α}}_{m1}*a)}\end{array}$

In above formula,T_sFor sampling period, T_RFor reference locus time constant；