CN104750131A - Fluidized bed temperature control method based on parameter identification - Google Patents

Abstract

The invention discloses a fluidized bed temperature control method based on parameter identification. The fluidized bed temperature control method based on parameter identification includes establishing a parameter identification function with a stable point according to a coal feeding quantity and a coal quantity feedback value so as to obtain a difference value between a constant parameter and a coal feeding quantity expectation value; identifying a model by a least square method according to the coal feeding quantity and fluidized bed temperature, and performing self-adaptive PID (proportion-integration-differentiation) control so as to obtain an increment of the coal feeding quantity; identifying a constant parameter by the least square method so as to well compensate deviation caused by an execution mechanism. Fuel can be utilized maximally according to the difference value and the increment computed in the first two steps. The fluidized bed temperature control method based on parameter identification has the advantages that the coal feeding quantity and the fluidized bed temperature can be in the stable states only by means of self-adaptive PID control without manual change of operators; the method is high in precision and is easily used for a fluidized bed boiler; energy waste is reduced, and pollutant discharge is reduced, and service life of the fluidized bed boiler is prolonged.

Description

A kind of fluidized-bed temperature control method based on parameter identification

Technical field

The invention belongs to fluidized-bed temperature control technology field, particularly relate to a kind of fluidized-bed temperature control method based on parameter identification.

Background technology

In recent years, China's rate of industrial development greatly exceed energy growth rate, and the energy is relatively in short supply.China is do one of many countries in the world producing and using in boiler; Boiler is important power-equipment in heat energy production, is again simultaneously take raw coal as the equipment that the power consumption of fuel is larger; But the boiler efficiency of present stage is not high, the waste of energy aspect is more serious; The pollutants such as the discharge of boiler combustion simultaneously oxides of nitrogen, sulphuric dioxide, flue dust, severe contamination air, has become one of China's air primary pollution source; Along with the improvement of people's living standard, environmental protection problem is taken seriously gradually, especially occurs in the numerous metropolitan haze phenomenon in the whole nation at the end of 2012 so far and has beaten alarm bell to the environmental protection of China.

Developed country's average specific in boiler operating efficiency wants high ten percentage points, but the low gap reason of China's boiler thermal output is mainly: China's boiler ubiquity operating load is low, calory burning loss is large, the problems such as excess air coefficient is large, so research boiler combustion control technology, improve the Control platform of boiler and the efficiency of boiler, larger economic benefit can not only be brought, but also the fume amount that can reduce in flue gas, reduce air pollution, the key benefit carrying out controlling in boiler combustion is that risk is little, successful, and raising operational efficiency can be reached, reduce the object of pollutant emission.

All much be realize feeding coal to system by coal dust rotor weigher at home, coal dust rotor weigher is by rotor, installation frame, gearing, the composition such as check weighing system and connecting hose, rotor and seal pad are arranged in an explosion-proof casing together, the gearing of whole rotor weigher housing and rotor is integrally suspended in a framework, framework there is the rolling bearing pedestal hanger bearing rotor weigher that two fixing, 3rd suspender is then connected with the weighing device with load sensor, the feed pipe of rotor weigher, discharge nozzle and compressed air hose all have connecting hose to connect with corresponding component.Coal dust is directly discharged from Pulverized Coal Bin by sliding gate, rotor portion is entered through entrance connecting hose, coal dust is taken away by the compartment of rotor, rotates 225 °, arrives discharge region, by the air of bottom compressed air hose, coal dust is blown into discharging opening, deliver to burner, although coal-supplying amount is stable, there is a lot of external factors and environmental factor in plant equipment in hello coal process, can change to the coal-supplying amount that rotor claims, thus affect the change of fluidized-bed temperature.

Claim to have the following disadvantages in fluidized-bed temperature at coal at present:

1, manually carry out that detection efficiency is low, accuracy of detection aspect does not reach desirable;

If 2 coal-supplying amounts can make fluidized-bed temperature not reach desirable very little, coal-supplying amount can cause the burning of fuel insufficient too much, causes unnecessary waste;

3, in the technique of reality, topworks also can affect to technique, actual coal-supplying amount is exceeded or lower than setting coal-supplying amount.

Summary of the invention

The object of the present invention is to provide a kind of fluidized-bed temperature control method based on parameter identification, be intended to solve that coal claims to exist fluidized-bed temperature aspect manually carries out that detection efficiency is low, accuracy of detection aspect does not reach desirable, coal-supplying amount can cause the burning of fuel insufficient too much, cause unnecessary waste, actual coal-supplying amount exceed or lower than setting coal-supplying amount problem.

The present invention realizes like this, a kind of the method that normal for least squares identification parameter is combined with the increment obtained by Adaptive PID Control to be controlled fluidized-bed temperature, should utilize based on the fluidized-bed temperature control method of parameter identification and feed coal amount and coal amount value of feedback, set up the parameter identification function with stable point, obtain normal parameter and the difference of feeding coal amount expectation value; Utilize hello coal amount and fluidized-bed temperature to go out model by least squares identification, recycling Adaptive PID Control obtains the increment of feeding coal amount; The difference calculated by first two steps and increment, namely maximizedly utilize fuel.

Further, should be as follows based on the concrete steps of the fluidized-bed temperature control method of parameter identification:

Step one, using detect in real time feed coal amount as input, the feedback quantity of the coal detected is as output;

Step 2, the coal amount of feeding collected in optional step one and feedback quantity, as the input and output of parameter identification, pick out one-component ξ by least-squares parameter discrimination method;

Step 3, the expectation value of feeding coal amount of required component ξ out and setting gets difference β, is first influence factor to feeding coal amount;

The fluidized-bed temperature value detected in real time is u by step 4, and the instantaneous value of feeding coal amount is designated as y;

Step 5, the Real-time streaming bed temperature angle value obtained in optional step four and real-time coal value of feeding, as the input and output of parameter identification, pick out one-component α by least-squares parameter discrimination method ₀, and draw the model now picked out;

Step 6, by the model picked out, utilizes self-regulated PID control method, weighs control the coal of system;

Step 7, the coal-supplying amount modified value that difference β step 3 calculated and step 6 are obtained is added, and feeds back to coal-supplying amount.

Further, in step 2, the value of feedback of coal-supplying amount and coal is as the input and output of linear least squares method, and pick out normal parameter ξ by linear least squares method method, concrete steps comprise:

The first step, provides that single-input single-output is linear, permanent, the mathematical model of stochastic system:

$y\left(k\right)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_{i}y(k-i)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_{i}u(k-i)+e\left(k\right)---\left(1\right)$

Then final output is:

$y\left(k\right)=-\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_{i}y(k-i)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_{i}u(k-i)+e\left(k\right)---\left(2\right)$

U (k) and y (k) they are the value of feedback data sequence { u (k) } of feeding coal amount and coal, { y (k) }, and e is model error, wherein k=1,2,, n, n are natural number, and in (2) formula, the value of exponent number n is 10, wherein i=1,2 ..., n, a _i, b _ibe all constant, pass through a _iand b _ivalue can finally try to achieve stable point ξ;

Second step, specializes model, thus obtains objective function:

Make θ=[a ₁, a ₂..., a _n, b ₁, b ₂..., b _n] ^t;

Then have:

Models fitting residual epsilon (k) is:

For n group data, obtain from (3) formula:

$\mathrm{\ϵ}\left(n\right)=y\left(n\right)-X\left(n\right)\widehat{\mathrm{\θ}}---\left(4\right)$

ε (n) is the models fitting residual error to n group data, the n group data of the value of feedback that y (n) is coal;

3rd step, obtains least-squares estimation: least square is the estimated value of a searching θ make the measured value of each time with by estimating the measurement determined estimates that only poor quadratic sum is minimum, from criterion of least squares derivation canonical equation, according to asking extremum principle, and least-squares estimation meet:

Obtain least-squares estimation

${\widehat{\mathrm{\θ}}}_{\mathrm{LS}}={\left(X^{T}X\right)}^{-1}{X}^{T}y---\left(6\right)$

4th step, derive and pick out normal parameter: the process picking out normal parameter ξ is: choose the model of impulse response model as system, and model is as shown in (7) formula:

$y\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}h\left(i\right)u(k-i)---\left(7\right)$

U (k-i) and y (k) are the value of feedback data sequence of feeding coal amount and coal, and h (i) is constant;

Adding normal parameter ξ, a ξ is in a model a parameter index that accurately can reflect that coal weighs, and model becomes:

$y\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}h\left(i\right)u(k-i)+\mathrm{\ξ}---\left(8\right)$

In formula, ξ is constant term, and by the experiment at scene, parameter ξ and real system contact closely, using ξ with to feeding the difference β of expectation value of coal amount as the controlling factor of first in Systematical control problem;

(8) are write as the form of vector:

Y(k)＝U(k)H (9)

Wherein:

$Y\left(k\right)=\left[\begin{array}{c}y\left(k\right)\\ y(k+1)\\ \·\\ \·\\ \·\\ y(k+N)\end{array}\right],H=\left[\begin{array}{c}{h}_1\\ h_2\\ \·\\ \·\\ \·\\ h_M\\ \mathrm{\ξ}\end{array}\right];$

$U\left(k\right)=\left[\begin{array}{ccccc}u(k-1)& u(k-2)& \·\·\·& u(k-M)& 1\\ u\left(k\right)& u(k-1)& \·\·\·& u(k-M+1)& 1\\ \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·\\ u(k+N-1)& u(k+N-2)& \·\·\·& u(k+N-M)& 1\end{array}\right];$

The matrix of H for being formed with the normal parameter ξ picked out by constant h (i), the matrix that U (k), Y (k) form for coal-supplying amount and feedback quantity;

Inputoutput data is consistent with the input and output of linear least squares method, tries to achieve and obtain optimum value when step size M=10 time, and through type (8) tries to achieve h ₁, h ₂... h _m, ξ;

Obtain normal parameter ξ, and record, and get difference β with the expectation value of feeding coal amount.

Further, in step 4, fluidized-bed temperature and coal-supplying amount are as the input and output of linear least squares method, pick out normal parameter alpha by linear least squares method method ₀, concrete steps are roughly the same with the method asking for normal parameter in step 2, and just input and output converted, concrete steps are:

The first step, provides that single-input single-output is linear, permanent, the mathematical model of stochastic system:

$y\left(k\right)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_{i}y(k-i)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_{i}u(k-i)+e\left(k\right)---\left(10\right)$

Then final output is:

$y\left(k\right)=-\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_{i}y(k-i)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_{i}u(k-i)+e\left(k\right)---\left(11\right)$

U (k) and y (k) are by the fluidized-bed temperature value detected in real time and the instantaneous value data sequence { u (k) } of feeding coal amount, { y (k) }, and e is model error, wherein k=1,2 ... n, n is natural number, and in (11) formula, the value of exponent number n is 10, wherein i=1,2,, n, a _i, b _ibe all constant, pass through a _iand b _ivalue can finally try to achieve stable point ξ;

Second step, specializes model, thus obtains objective function:

Make θ=[a ₁, a ₂..., a _n, b ₁, b ₂..., b _n] ^t;

Then have:

Models fitting residual epsilon (k) is:

For n group data, obtain from (3) formula:

$\mathrm{\ϵ}\left(n\right)=y\left(n\right)-X\left(n\right)\widehat{\mathrm{\θ}}---\left(13\right)$

ε (n) is the models fitting residual error to n group data, the n group data of the value of feedback that y (n) is coal;

3rd step, obtains least-squares estimation: least square is the estimated value of a searching θ make the measured value of each time with by estimating the measurement determined estimates that only poor quadratic sum is minimum, from criterion of least squares derivation canonical equation, according to asking extremum principle to know, and least-squares estimation meet:

By (10) formula to (14) formula, calculate least-squares estimation now

${\widehat{\mathrm{\θ}}}_{\mathrm{LS}}={\left(X^{T}X\right)}^{-1}{X}^{T}y---\left(15\right)$

Choose the model of impulse response model as system, model, such as formula shown in (7), increases a normal parameter alpha on this basis ₀, α ₀for accurately reflecting the parameter index that coal weighs, model becomes:

$y\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}h\left(i\right)u(k-i)+{\mathrm{\α}}_0---\left(16\right)$

The data sequence that u (k-i) and y (k) are fluidized-bed temperature and coal-supplying amount, h (i), α ₀for constant term.

Further, then normal parameter alpha is calculated ₀methods and steps two in the method for calculating parameter consistent, concrete steps are:

U (k-i) is fluidized-bed temperature value and the instantaneous value data sequence of feeding coal amount with y (k), and h (i) is constant;

Add a normal parameter alpha in a model ₀, α ₀for accurately reflecting the parameter index that coal weighs, model becomes:

$y\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}h\left(i\right)u(k-i)+{\mathrm{\α}}_0---\left(17\right)$

α in formula ₀for constant term;

(8) are write as the form of vector:

Y(k)＝U(k)H (18)

Wherein:

$Y\left(k\right)=\left[\begin{array}{c}y\left(k\right)\\ y(k+1)\\ \·\\ \·\\ \·\\ y(k+N)\end{array}\right],H=\left[\begin{array}{c}{h}_1\\ h_2\\ \·\\ \·\\ \·\\ h_M\\ \mathrm{\ξ}\end{array}\right];$

$U\left(k\right)=\left[\begin{array}{ccccc}u(k-1)& u(k-2)& \·\·\·& u(k-M)& 1\\ u\left(k\right)& u(k-1)& \·\·\·& u(k-M+1)& 1\\ \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·\\ u(k+N-1)& u(k+N-2)& \·\·\·& u(k+N-M)& 1\end{array}\right];$

H is by constant h (i) and the normal parameter alpha picked out ₀the matrix of composition, the matrix that U (k), Y (k) form for coal-supplying amount and feedback quantity;

Inputoutput data is consistent with the input and output of linear least squares method, tries to achieve and obtain optimum value when step size M=10 time, and through type (17) tries to achieve h ₁, h ₂... h _m, α ₀;

Calculate normal parameter alpha now ₀, and set up out with fluidized-bed temperature be input, specified rate be export model, for the self-regulated PID control of step 6 provides Controlling model.

Further, the Controlling model in step 6 is identification model out in step 4, and concrete steps comprise:

The first step, traditional PID control algorithm is made up of controller and controlled device, and belt controller is made up of ratio, integration, differential three links, and mathematical description is:

u(k)＝K _px(1)+K _dx(2)+K _ix(3) (19)

In formula, K _pfor scale-up factor; K _ifor integration time constant; K _dfor derivative time constant; The corrected value that the increase reduced value x (1) that u (k) weighs for the coal by obtaining after PID controller is ratio; The corrected value that x (2) is differential; The corrected value that x (3) is integration;

Second step, obtained x (1), x (2), the x (3) in the first step by the error of the measured value of temperature input quantity and the expectation value of temperature and sampling time, computing formula is:

x(1)＝error(k)；

x(2)＝[error(k)-error_1]/t _s；

x(3)＝x(3)+error(k)*t _s；

In formula, error (k) is the error calculated by measured value and expectation value in the k moment; t _sfor the sampling time;

3rd step, after upper two steps being programmed, the value u (k) of output is the modified value of coal-supplying amount, and records.

Further, be added with coal-supplying amount modified value u (k) that step 6 is obtained by the component ξ of step 3 in step 7, computing formula is:

η(k)＝u(k)+β (20)

In formula, the coal-supplying amount modified value of u (k) for being calculated by the control method of step 6, β is that η (k) is the final modified value to system coal-supplying amount by step 2 by least squares identification normal parameter out and the difference of hello coal amount expectation value.

Fluidized-bed temperature control method based on parameter identification provided by the invention, first, utilizes and feeds coal amount and coal amount value of feedback, set up the parameter identification function with stable point, obtains normal parameter and the difference of feeding coal amount expectation value; Secondly, utilize hello coal amount and fluidized-bed temperature to go out model by least squares identification, recycling Adaptive PID Control obtains the increment of feeding coal amount; Finally, the difference calculated by first two steps and increment, namely maximizedly can utilize fuel.

Beneficial effect of the present invention is:

1, in general fluidized-bed combustion boiler temperature controls, only can carry out manual adjustment coal-supplying amount by the height observing boiler temperature, reaching the stable of fluidized-bed temperature, in the present invention, making operating personnel temperature just can be made to reach stable without the need to manually controlling again by self-regulated PID control;

2, in some fluidized-bed combustion boiler temperature control, only can be controlled fluidized-bed combustion boiler temperature by general control method, and in the present invention, not only have employed self-regulated PID control, also go out a normal parameter by least squares identification, by the difference between normal parameter and expectation value, coal-supplying amount is changed, make temperature reach stable;

3, again coal-supplying amount is controlled after two kinds of methods being combined, than being used alone control method or manual operation more can make coal-supplying amount accurate;

4, precision is high, is easy to the use of fluidized-bed combustion boiler;

5, decrease the waste of the energy, reduce the discharge of pollutant;

6, the tenure of use of fluidized-bed combustion boiler is added.

Accompanying drawing explanation

Fig. 1 is the fluidized-bed temperature control method process flow diagram based on parameter identification that the embodiment of the present invention provides;

Fig. 2 is the process flow diagram of the embodiment 1 that the embodiment of the present invention provides;

Fig. 3 is the PID Control system architecture figure that the embodiment of the present invention provides;

Fig. 4 is the coal-supplying amount data plot of the reality that the embodiment of the present invention provides;

Fig. 5 be the embodiment of the present invention provide by the normal parameter after linear least squares method with feed the differential chart of coal amount expectation value;

Fig. 6 is the increment graph by obtaining after Adaptive PID Control that the embodiment of the present invention provides;

Fig. 7 is the original value of the coal-supplying amount that the embodiment of the present invention provides and the figure after being controlled by parameter and PID controlling increment.

Embodiment

In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with embodiment, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.

Below in conjunction with drawings and the specific embodiments, application principle of the present invention is further described.

As shown in Figure 1, the fluidized-bed temperature control method based on parameter identification of the embodiment of the present invention comprises the following steps:

S101: utilize and feed coal amount and coal amount value of feedback, set up the parameter identification function with stable point, obtains normal parameter and the difference of feeding coal amount expectation value;

S102: utilize hello coal amount and fluidized-bed temperature to go out model by least squares identification, recycling Adaptive PID Control obtains the increment of feeding coal amount;

S103: the difference calculated by first two steps and increment, namely maximizedly can utilize fuel.

The concrete steps of the embodiment of the present invention are as follows:

Step one, using detect in real time feed coal amount as input, the feedback quantity of the coal detected is as output;

Step 2, the coal amount of feeding collected in optional step one and feedback quantity, as the input and output of parameter identification, pick out one-component ξ by least-squares parameter discrimination method;

Step 3, the expectation value of feeding coal amount of required component ξ out and setting gets difference β, is first influence factor to feeding coal amount;

The fluidized-bed temperature value detected in real time is u by step 4, and the instantaneous value of feeding coal amount is designated as y;

Step 5, the Real-time streaming bed temperature angle value obtained in optional step four and real-time coal value of feeding, as the input and output of parameter identification, pick out one-component α by least-squares parameter discrimination method ₀, and draw the model now picked out;

Step 6, by the model picked out, utilizes self-regulated PID control method, weighs control the coal of system;

Step 7, the coal-supplying amount modified value that difference β step 3 calculated and step 6 are obtained is added, and feeds back to coal-supplying amount, the coal of system can be made to weigh and be issued to stable in the impact of the problems such as topworks's deviation, thus make fluidized-bed temperature reach stable.

In described step 2, the value of feedback of coal-supplying amount and coal is as the input and output of linear least squares method, and pick out normal parameter ξ by linear least squares method method, its concrete steps comprise:

The first step, provides that single-input single-output is linear, permanent, the mathematical model of stochastic system:

$y\left(k\right)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_{i}y(k-i)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_{i}u(k-i)+e\left(k\right)---\left(1\right)$

Then final output is:

$y\left(k\right)=-\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_{i}y(k-i)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_{i}u(k-i)+e\left(k\right)---\left(2\right)$

U (k) and y (k) are the value of feedback data sequence { u (k) } of feeding coal amount and coal, { y (k) }, and e is model error, wherein k=1,2 ... n, n is natural number, is shown that the value of the exponent number n of model is 10, wherein i=1 in (2) formula by contrast experiment, 2,, n, a _i, b _ibe all constant, pass through a _iand b _ivalue can finally try to achieve stable point ξ;

Second step, specializes model, thus obtains objective function:

Make θ=[a _i, a ₂..., a _n, b ₁, b ₂..., b _n] ^t;

Then have:

Models fitting residual epsilon (k) is:

For n group data, can obtain from (3) formula:

$\mathrm{\ϵ}\left(n\right)=y\left(n\right)-X\left(n\right)\widehat{\mathrm{\θ}}---\left(4\right)$

ε (n) is the models fitting residual error to n group data, the n group data of the value of feedback that y (n) is coal;

3rd step, obtains least-squares estimation: the thought of least square is exactly the estimated value of a searching θ make the measured value of each time with by estimating the measurement determined estimates that only poor quadratic sum is minimum, from criterion of least squares derivation canonical equation, according to asking extremum principle, and least-squares estimation meet:

Least-squares estimation can be obtained

${\widehat{\mathrm{\θ}}}_{\mathrm{LS}}={\left(X^{T}X\right)}^{-1}{X}^{T}y---\left(6\right)$

4th step, derive and pick out normal parameter: the process picking out normal parameter ξ is: choose the model of impulse response model as system, and model is as shown in (7) formula:

$y\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}h\left(i\right)u(k-i)---\left(7\right)$

U (k-i) and y (k) are the value of feedback data sequence of feeding coal amount and coal, and h (i) is constant;

Adding normal parameter ξ, a ξ is in a model a parameter index that accurately can reflect that coal weighs, and model becomes:

$y\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}h\left(i\right)u(k-i)+\mathrm{\ξ}---\left(8\right)$

In formula, ξ is constant term, and by the experiment at scene, parameter ξ and real system contact closely, and can using ξ with to feeding the difference β of expectation value of coal amount as the controlling factor of first in Systematical control problem;

(8) are write as the form of vector:

Y(k)＝U(k)H (9)

Wherein:

$Y\left(k\right)=\left[\begin{array}{c}y\left(k\right)\\ y(k+1)\\ \·\\ \·\\ \·\\ y(k+N)\end{array}\right],H=\left[\begin{array}{c}{h}_1\\ h_2\\ \·\\ \·\\ \·\\ h_M\\ \mathrm{\ξ}\end{array}\right];$

$U\left(k\right)=\left[\begin{array}{ccccc}u(k-1)& u(k-2)& \·\·\·& u(k-M)& 1\\ u\left(k\right)& u(k-1)& \·\·\·& u(k-M+1)& 1\\ \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·\\ u(k+N-1)& u(k+N-2)& \·\·\·& u(k+N-M)& 1\end{array}\right];$

The matrix of H for being formed with the normal parameter ξ picked out by constant h (i), the matrix that U (k), Y (k) form for coal-supplying amount and feedback quantity;

Inputoutput data is consistent with the input and output of linear least squares method, tries to achieve by experiment and can obtain optimum value when step size M=10 time, and through type (8) tries to achieve h ₁, h ₂... h _m, ξ;

In step 2, obtain normal parameter ξ by said method, and record, and get difference β with the expectation value of feeding coal amount.

In described step 4, fluidized-bed temperature and coal-supplying amount are as the input and output of linear least squares method, pick out normal parameter alpha by linear least squares method method ₀, its concrete steps are roughly the same with the method asking for normal parameter in step 2, and just input and output converted, concrete steps are:

The first step, provides that single-input single-output is linear, permanent, the mathematical model of stochastic system:

$y\left(k\right)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_{i}y(k-i)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_{i}u(k-i)+e\left(k\right)---\left(10\right)$

Then final output is:

$y\left(k\right)=-\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_{i}y(k-i)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_{i}u(k-i)+e\left(k\right)---\left(11\right)$

U (k) and y (k) are by the fluidized-bed temperature value detected in real time and the instantaneous value data sequence { u (k) } of feeding coal amount, { y (k) }, and e is model error, wherein k=1,2 ... n, n is natural number, and in (11) formula, the value of exponent number n is 10, wherein i=1,2,, n, a _i, b _ibe all constant, pass through a _iand b _ivalue can finally try to achieve stable point ξ;

Second step, specializes model, thus obtains objective function:

Make θ=[a ₁, a ₂..., a _n, b ₁, b ₂..., b _n] ^t;

Then have:

Models fitting residual epsilon (k) is:

For n group data, obtain from (3) formula:

$\mathrm{\ϵ}\left(n\right)=y\left(n\right)-X\left(n\right)\widehat{\mathrm{\θ}}---\left(13\right)$

ε (n) is the models fitting residual error to n group data, the n group data of the value of feedback that y (n) is coal;

3rd step, obtains least-squares estimation: least square is the estimated value of a searching θ make the measured value of each time with by estimating the measurement determined estimates that only poor quadratic sum is minimum, from criterion of least squares derivation canonical equation, according to asking extremum principle to know, and least-squares estimation meet:

By (10) formula to (14) formula, least-squares estimation now can be calculated

${\widehat{\mathrm{\θ}}}_{\mathrm{LS}}={\left(X^{T}X\right)}^{-1}{X}^{T}y---\left(15\right)$

Choose the model of impulse response model as system, model, such as formula shown in (7), increases a normal parameter alpha on this basis ₀, α ₀for accurately reflecting the parameter index that coal weighs, model becomes:

$y\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}h\left(i\right)u(k-i)+{\mathrm{\α}}_0---\left(16\right)$

The data sequence that u (k-i) and y (k) are fluidized-bed temperature and coal-supplying amount, h (i), α ₀for constant term.

Then calculate normal parameter alpha ₀methods and steps two in the method for calculating parameter consistent, concrete steps are:

U (k-i) is fluidized-bed temperature value and the instantaneous value data sequence of feeding coal amount with y (k), and h (i) is constant;

Add a normal parameter alpha in a model ₀, α ₀for accurately reflecting the parameter index that coal weighs, model becomes:

$y\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}h\left(i\right)u(k-i)+{\mathrm{\α}}_0---\left(17\right)$

α in formula ₀for constant term;

(8) are write as the form of vector:

Y(k)＝U(k)H (18)

Wherein:

$Y\left(k\right)=\left[\begin{array}{c}y\left(k\right)\\ y(k+1)\\ \·\\ \·\\ \·\\ y(k+N)\end{array}\right],H=\left[\begin{array}{c}{h}_1\\ h_2\\ \·\\ \·\\ \·\\ h_M\\ \mathrm{\ξ}\end{array}\right];$

$U\left(k\right)=\left[\begin{array}{ccccc}u(k-1)& u(k-2)& \·\·\·& u(k-M)& 1\\ u\left(k\right)& u(k-1)& \·\·\·& u(k-M+1)& 1\\ \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·\\ u(k+N-1)& u(k+N-2)& \·\·\·& u(k+N-M)& 1\end{array}\right];$

H is by constant h (i) and the normal parameter alpha picked out ₀the matrix of composition, the matrix that U (k), Y (k) form for coal-supplying amount and feedback quantity;

Inputoutput data is consistent with the input and output of linear least squares method, tries to achieve and obtain optimum value when step size M=10 time, and through type (17) tries to achieve h ₁, h ₂... h _m, α ₀;

Calculate normal parameter alpha now ₀, and set up out with fluidized-bed temperature be input, specified rate be export model, for the self-regulated PID control of step 6 provides Controlling model;

Controlling model in described step 6 is identification model out in step 4, and its concrete steps comprise:

The first step, traditional PID control algorithm is a kind of control method of classics, and it formed primarily of controller and controlled device, and belt controller is made up of ratio, integration, differential three links, and its mathematical description is:

u(k)＝K _px(1)+K _dx(2)+K _ix(3) (19)

In formula, K _pfor scale-up factor; K _ifor integration time constant; K _dfor derivative time constant; The corrected value that the increase reduced value x (1) that u (k) weighs for the coal by obtaining after PID controller is ratio; The corrected value that x (2) is differential; The corrected value that x (3) is integration;

Second step, can be obtained x (1), x (2), the x (3) in the first step by the error of the measured value of temperature input quantity and the expectation value of temperature and sampling time, its computing formula is:

x(1)＝error(k)；

x(2)＝[error(k)-error_1]/t _s；

x(3)＝x(3)+error(k)*t _s；

In formula, error (k) is the error calculated by measured value and expectation value in the k moment; t _sfor the sampling time;

3rd step, after upper two steps being programmed, the value u (k) of output is the modified value of coal-supplying amount, and records;

Be added with coal-supplying amount modified value u (k) that step 6 is obtained by the component ξ of step 3 in described step 7, its computing formula is:

η(k)＝u(k)+β (20)

In formula, the coal-supplying amount modified value of u (k) for being calculated by the control method of step 6, β is that η (k) is the final modified value to system coal-supplying amount by step 2 by least squares identification normal parameter out and the difference of hello coal amount expectation value.

Specific embodiments of the invention:

Embodiment 1

As shown in figs 2-4, a kind of fluidized-bed temperature control method based on parameter identification of the embodiment of the present invention, specifically comprises the following steps:

Step one, the value of feedback of the fluidized-bed temperature at scene, hello coal amount and coal amount is chosen respectively 10000 groups of data of same time period, the time interval is 1s, 10000 groups of data is divided into 100 groups, and calculates its mean value.Every 100 hello coal amounts and coal amount value of feedback pick out a supplemental characteristic ξ, and the expectation value calculating parameter and hello the coal amount picked out does difference, records the principal element that difference now affects hello coal amount as topworks.Just can obtain affecting first data of feeding coal amount by identification by step one.

Step 2, what collect in optional step one feeds the input and output as parameter identification of coal amount and coal amount value of feedback, such as wherein one group feed coal amount data be=[25.038, 27.895, 25.306, 26.788, ..., 26.79, 25.573, 27.804], the value of feedback data of coal amount are=[25.923, 26.386, 27.489, 25.9, ..., 27.642, 27.44, 27.29], the exponent number setting out model is set to 10, by least square method, identification is carried out to the given model with parameter, by normal parameter ξ=26.252 obtained after identification, and calculate the difference β with expectation value.Fig. 5 is 100 differences calculated by 100 times.This normal parameter and difference are exactly core of the present invention.

Step 3, what collect in optional step one feeds the input and output as identification of coal amount and fluidized-bed temperature, choose feed coal amount data be=[25.038,27.895,25.306,26.788 ..., 26.79,25.573,27.804], fluidized-bed temperature data be=[919.72,917,71,908.1,912.54,907.71, ..., 918.97,913.82,912.31,917.9], arranging model order is 10, by least square method Modling model, for next step self-adaptive PID is prepared.

Step 4, establishes model in step 3, utilizes self-adaptive PID to control system, and calculate u (k)=1.912 now, Fig. 6 is the system increment size by obtaining after Adaptive PID Control.The increment calculated is added with the difference β calculated in step 2, obtains final increment of adjustment η (k)=u (k)+β=2.16.Now obtaining the increment of adjustment come also is core of the present invention.Fig. 7 be feed coal amount with through controls and identified parameters adjust after hello coal amount comparison diagram.

The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.

Claims (7)

1. the fluidized-bed temperature control method based on parameter identification, it is characterized in that, should utilize based on the fluidized-bed temperature control method of parameter identification and feed coal amount and coal amount value of feedback, and set up the parameter identification function with stable point, obtain normal parameter and the difference of feeding coal amount expectation value; Utilize and feed coal amount and fluidized-bed temperature and go out by least squares identification the increment that model and Adaptive PID Control obtain hello coal amount; By the difference that calculates and increment, namely maximizedly utilize fuel.

2., as claimed in claim 1 based on the fluidized-bed temperature control method of parameter identification, it is characterized in that, should be as follows based on the concrete steps of the fluidized-bed temperature control method of parameter identification:

Step one, using detect in real time feed coal amount as input, the feedback quantity of the coal detected is as output;

Step 2, the coal amount of feeding collected in optional step one and feedback quantity, as the input and output of parameter identification, pick out one-component ξ by least-squares parameter discrimination method;

Step 3, the expectation value of feeding coal amount of required component ξ out and setting gets difference β, is first influence factor to feeding coal amount;

The fluidized-bed temperature value detected in real time is u by step 4, and the instantaneous value of feeding coal amount is designated as y;

Step 5, the Real-time streaming bed temperature angle value obtained in optional step four and real-time coal value of feeding, as the input and output of parameter identification, pick out one-component α by least-squares parameter discrimination method ₀, and draw the model now picked out;

Step 6, by the model picked out, utilizes self-regulated PID control method, weighs control the coal of system;

Step 7, the coal-supplying amount modified value that difference β step 3 calculated and step 6 are obtained is added, and feeds back to coal-supplying amount.

3. as claimed in claim 2 based on the fluidized-bed temperature control method of parameter identification, it is characterized in that, in step 2, the value of feedback of coal-supplying amount and coal is as the input and output of linear least squares method, and pick out normal parameter ξ by linear least squares method method, concrete steps comprise:

The first step, provides that single-input single-output is linear, permanent, the mathematical model of stochastic system:

$y\left(k\right)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_{i}y(k-i)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_{i}u(k-i)+e\left(k\right)---\left(1\right)$

Then final output is:

$y\left(k\right)=-\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_{i}y(k-i)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_{i}u(k-i)+e\left(k\right)---\left(2\right)$

U (k) and y (k) they are the value of feedback data sequence { u (k) } of feeding coal amount and coal, { y (k) }, and e is model error, wherein k=1,2,, n, n are natural number, and in (2) formula, the value of exponent number n is 10, wherein i=1,2 ..., n, a _i, b _ibe all constant, pass through a _iand b _ivalue can finally try to achieve stable point ξ;

Second step, specializes model, thus obtains objective function:

Make θ=[a ₁, a ₂..., a _n, b ₁, b ₂..., b _n] ^t;

Then have:

Models fitting residual epsilon (k) is:

For n group data, obtain from (3) formula:

$\mathrm{\ϵ}\left(n\right)=y\left(n\right)-X\left(n\right)\widehat{\mathrm{\θ}}---\left(4\right)$

ε (n) is the models fitting residual error to n group data, the n group data of the value of feedback that y (n) is coal;

3rd step, obtains least-squares estimation: least square is the estimated value of a searching θ make the measured value of each time with by estimating the measurement determined estimates that only poor quadratic sum is minimum, from criterion of least squares derivation canonical equation, according to asking extremum principle to know, and least-squares estimation meet:

Obtain least-squares estimation

${\widehat{\mathrm{\θ}}}_{\mathrm{LS}}={\left(X^{T}X\right)}^{-1}{X}^{T}y---\left(6\right)$

4th step, derive and pick out normal parameter: the process picking out normal parameter ξ is: choose the model of impulse response model as system, and model is as shown in (7) formula:

$y\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}h\left(i\right)u(k-i)---\left(7\right)$

U (k-i) and y (k) are the value of feedback data sequence of feeding coal amount and coal, and h (i) is constant;

Adding normal parameter ξ, a ξ is in a model a parameter index that accurately can reflect that coal weighs, and model becomes:

$y\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}h\left(i\right)u(k-i)+\mathrm{\ξ}---\left(8\right)$

In formula, ξ is constant term, and by the experiment at scene, parameter ξ and real system contact closely, using ξ with to feeding the difference β of expectation value of coal amount as the controlling factor of first in Systematical control problem;

(8) are write as the form of vector:

Y(k)＝U(k)H (9)

Wherein:

$Y\left(k\right)=\left[\begin{array}{c}y\left(k\right)\\ y(k+1)\\ .\\ .\\ .\\ y(k+N)\end{array}\right],H=\begin{array}{c}\left[\begin{array}{c}{h}_1\\ h_2\\ .\\ .\\ .\\ h_M\\ \mathrm{\ξ}\end{array}\right]\end{array};$

$U\left(k\right)=\left[\begin{array}{ccccc}u(k-1)& u(k-2)& ...& u(k-M)& 1\\ u\left(k\right)& u(k-1)& ...& u(k-M+1)& 1\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ u(k+N-1)& u(k+N-2)& ...& u(k+N-M)& 1\end{array}\right];$

The matrix of H for being formed with the normal parameter ξ picked out by constant h (i), the matrix that U (k), Y (k) form for coal-supplying amount and feedback quantity;

Inputoutput data is consistent with the input and output of linear least squares method, tries to achieve and obtain optimum value when step size M=10 time, and through type (8) tries to achieve h ₁, h ₂... h _m, ξ;

Obtain normal parameter ξ, and record, and get difference β with the expectation value of feeding coal amount.

4., as claimed in claim 2 based on the fluidized-bed temperature control method of parameter identification, it is characterized in that, in step 4, fluidized-bed temperature and coal-supplying amount are as the input and output of linear least squares method, pick out normal parameter alpha by linear least squares method method ₀, concrete steps are roughly the same with the method asking for normal parameter in step 2, and just input and output converted, concrete steps are:

The first step, provides that single-input single-output is linear, permanent, the mathematical model of stochastic system:

$y\left(k\right)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_{i}y(k-i)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_{i}u(k-i)+e\left(k\right)---\left(10\right)$

Then final output is:

$y\left(k\right)=-\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_{i}y(k-i)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{b}_{i}u(k-i)+e\left(k\right)---\left(11\right)$

U (k) and y (k) are by the fluidized-bed temperature value detected in real time and the instantaneous value data sequence { u (k) } of feeding coal amount, { y (k) }, and e is model error, wherein k=1,2 ... .., n, n is natural number, and in (11) formula, the value of exponent number n is 10, wherein i=1,2,, n, a _i, b _ibe all constant, pass through a _iand b _ivalue can finally try to achieve stable point ξ;

Second step, specializes model, thus obtains objective function:

Make θ=[a ₁, a ₂..., a _n, b ₁, b ₂..., b _n] ^t;

Then have:

Models fitting residual epsilon (k) is:

For n group data, obtain from (3) formula:

$\mathrm{\ϵ}\left(n\right)=y\left(n\right)-X\left(n\right)\widehat{\mathrm{\θ}}---\left(13\right)$

ε (n) is the models fitting residual error to n group data, the n group data of the value of feedback that y (n) is coal;

3rd step, obtains least-squares estimation: least square is the estimated value of a searching θ make the measured value of each time with by estimating the measurement determined estimates that only poor quadratic sum is minimum, from criterion of least squares derivation canonical equation, according to asking extremum principle to know, and least-squares estimation meet:

By (10) formula to (14) formula, calculate least-squares estimation now

${\widehat{\mathrm{\θ}}}_{\mathrm{LS}}={\left(X^{T}X\right)}^{-1}{X}^{T}y---\left(15\right)$

Choose the model of impulse response model as system, model, such as formula shown in (7), increases a normal parameter alpha on this basis ₀, α ₀for accurately reflecting the parameter index that coal weighs, model becomes:

$y\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}h\left(i\right)u(k-i)+{\mathrm{\α}}_0---\left(16\right)$

The data sequence that u (k-i) and y (k) are fluidized-bed temperature and coal-supplying amount, h (i), α ₀for constant term.

5., as claimed in claim 4 based on the fluidized-bed temperature control method of parameter identification, it is characterized in that, then calculate normal parameter alpha ₀methods and steps two in the method for calculating parameter consistent, concrete steps are:

U (k-i) is fluidized-bed temperature value and the instantaneous value data sequence of feeding coal amount with y (k), and h (i) is constant;

Add a normal parameter alpha in a model ₀, α ₀for accurately reflecting the parameter index that coal weighs, model becomes:

$y\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}h\left(i\right)u(k-i)+{\mathrm{\α}}_0---\left(17\right)$

α in formula ₀for constant term;

(8) are write as the form of vector:

Y(k)＝U(k)H (18)

Wherein:

$Y\left(k\right)=\left[\begin{array}{c}y\left(k\right)\\ y(k+1)\\ .\\ .\\ .\\ y(k+N)\end{array}\right],H=\begin{array}{c}\left[\begin{array}{c}{h}_1\\ h_2\\ .\\ .\\ .\\ h_M\\ \mathrm{\ξ}\end{array}\right]\end{array};$

$U\left(k\right)=\left[\begin{array}{ccccc}u(k-1)& u(k-2)& ...& u(k-M)& 1\\ u\left(k\right)& u(k-1)& ...& u(k-M+1)& 1\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ u(k+N-1)& u(k+N-2)& ...& u(k+N-M)& 1\end{array}\right];$

H is by constant h (i) and the normal parameter alpha picked out ₀the matrix of composition, the matrix that U (k), Y (k) form for coal-supplying amount and feedback quantity;

Inputoutput data is consistent with the input and output of linear least squares method, tries to achieve and obtain optimum value when step size M=10 time, and through type (17) tries to achieve h ₁, h ₂... h _m, α ₀;

Calculate normal parameter alpha now ₀, and set up out with fluidized-bed temperature be input, specified rate be export model, for the self-regulated PID control of step 6 provides Controlling model.

6., as claimed in claim 2 based on the fluidized-bed temperature control method of parameter identification, it is characterized in that, the Controlling model in step 6 is identification model out in step 4, and concrete steps comprise:

The first step, traditional PID control algorithm is made up of controller and controlled device, and belt controller is made up of ratio, integration, differential three links, and mathematical description is:

u(k)＝K _px(1)+K _dx(2)+K _ix(3) (19)

In formula, K _pfor scale-up factor; K _ifor integration time constant; K _dfor derivative time constant; The corrected value that the increase reduced value x (1) that u (k) weighs for the coal by obtaining after PID controller is ratio; The corrected value that x (2) is differential; The corrected value that x (3) is integration;

Second step, obtained x (1), x (2), the x (3) in the first step by the error of the measured value of temperature input quantity and the expectation value of temperature and sampling time, computing formula is:

x(1)＝error(k)；

x(2)＝[error(k)-error_1]/t _s；

x(3)＝x(3)+error(k)*t _s；

In formula, error (k) is the error calculated by measured value and expectation value in the k moment; t _sfor the sampling time;

3rd step, after upper two steps being programmed, the value u (k) of output is the modified value of coal-supplying amount, and records.

7., as claimed in claim 2 based on the fluidized-bed temperature control method of parameter identification, it is characterized in that, be added with coal-supplying amount modified value u (k) that step 6 is obtained by the component ξ of step 3 in step 7, computing formula is:

η(k)＝u(k)+β (20)

In formula, the coal-supplying amount modified value of u (k) for being calculated by the control method of step 6, β is that η (k) is the final modified value to system coal-supplying amount by step 2 by least squares identification normal parameter out and the difference of hello coal amount expectation value.