CN100440081C - Generalized predictable control system and method of air separating tower - Google Patents

Abstract

A generalized prediction-control system of air separation tower consists of on-site intelligent instrument connected directly with air separation tower, data storage unit for storing historical data and up-level computer. It is featured as connecting said on-site instruments and data storage unit as well as up-level computer in sequence, using said up-level computer as generalized prediction-control formed by composition judgment-control unit and generalized prediction-control unit. The method of utilizing said system to realize control of air separation tower is also disclosed.

Description

The generalized predictable control system of air separation column and method

(1) technical field

The present invention relates to the control system and the method design field of air separation column, especially, relate to a kind of generalized predictable control system and method for air separation column.

(2) background technology

Air separation unit separates the sky branch exactly, and obtains the device of high-purity industrial gasses such as oxygen, nitrogen, argon.It is the supportive unit operations of numerous industries that concern the life-blood of the national economy, as chemical industry, metallurgy, electronics, the energy, Aero-Space, food and drink etc., belong to national substantial equipment, its development scale and technology status are to weigh the industry of a country and an important symbol of development in science and technology level.

Empty branch operation is one and relates to low temperature, many equipment, long flow process, complicated operation, the exigent complex process of safety in production.In the production, the purity of oxygen, nitrogen, argon product often requires up to more than 99%, belong to high-purity distillation control problem, stationarity to the air separation column operation requires very high, and the high-purity distillation process is because the coupling between dynamic perfromance, strong nonlinear and the loop of the complexity that it showed, and traditional is difficult to it is controlled effect preferably as Linear Control schemes such as PID.

(3) summary of the invention

In order to overcome coupling between empty dynamic perfromance, strong nonlinear and the loop of dividing operation of can not adapting to of existing air separation column controlling schemes, can not to obtain the deficiency of good control effect, the invention provides a kind of coupling problem that can solve between empty dynamic perfromance, strong nonlinear and the loop of dividing operation, and obtain good control effect air separation column based on generalized predictable control system and method.

The technical solution adopted for the present invention to solve the technical problems is:

A kind of generalized predictable control system of air separation column, comprise and the direct-connected field intelligent instrument of air separation column, the data storage device that is used for storing history data and host computer, intelligence instrument, data storage device and host computer link to each other successively, described host computer is the universal model controller, described universal model controller comprises component deduction control section and universal model control section, described component infers that control section comprises: the measuring instrument module, comprise detector unit and pressure detecting element, be used to detect the temperature and pressure of the last tower of air separation column; The I/O component module is used for the transmission between controller inside and controller and data storage device of electric signal and data-signal, and the component inference module is used for inferring component according to detecting the temperature and the pressure data that obtain, and its formula is (1), (2):

${\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">Y</figure-callout>}}_1=\frac{\mathrm{\&alpha;}}{\mathrm{\&alpha;}-1}-\frac{{10}^{(a-\frac{\mathrm{<figure-callout\; id="1"\; label="b\; T"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">b</figure-callout>}}{T_{}})}}{(\mathrm{\&alpha;}-1)P}---\left(1\right)$

$\mathrm{Xn}=\frac{\mathrm{P\&alpha;}}{{10}^{(a-\frac{T_n+c}{b})}(\mathrm{\&alpha;}-1)}-\frac{1}{\mathrm{\&alpha;}-1}---\left(2\right)$

Wherein, Y ₁Be the component of nitrogen in the nitrogen product in the air separation column, Xn is the component of nitrogen in the liquid oxygen product, and P is last tower pressure, T ₁, T _nBe respectively cat head, column bottom temperature, α is a relative volatility, and a, b, c are the Peter Antonie constant; Described generalized predictive control partly comprises: the I/O component module is used for the inside of generalized predictive controller and the electric signal between controller and the data storage device, the transmission of data-signal; The forecast model module is used for that DCS historical data base or test gained data are carried out least square and fits the system of obtaining and have the discrete differential equation model of step disturbance nonstationary noise at random, and its formula is (3);

A(z ^-1)y(k)＝B(z ^-1)u(k-1)+C(z ^-1)ε(k)/Δ (3)

Wherein $A\left(z^{-1}\right)=1+\underset{i=1}{\overset{n_a}{\mathrm{\&Sigma;}}}{a}_i{z}^{-i},$ $B\left(z^{-1}\right)=\underset{i=0}{\overset{n_b}{\mathrm{\&Sigma;}}}{b}_i{z}^{-i},$ Δ-difference operator, Δ=1-z ^-1

Model self-correcting module is used to adopt the Recursive Least Squares with forgetting factor, on-line correction system model parameter, and concrete steps are as follows:

(1) upgrades vectorial X (k-1) ^T=[Δ y (k-1) ... ,-Δ y (k-n _a), Δ u (k-1) ..., Δ u (k-n _b-1)];

(2) read y (k) and calculate ε (k)=Δ y (k)-X (k-1) ^Tθ (k-1);

(3) use the recursive least squares with forgetting factor to obtain θ (k), its formula is (4):

$\mathrm{\&theta;}\left(k\right)=\mathrm{\&theta;}(k-1)+\frac{P(k-2)X(k-1)\mathrm{\&epsiv;}\left(k\right)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}---\left(4\right)$

Wherein, $P(k-1)=\frac{1}{\mathrm{\&rho;}}[P(k-2)-\frac{P(k-2)X(k-1)X{(k-1)}^{T}P(k-2)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}],$ P (1) is enough big positive definite matrix, $\mathrm{\&theta;}=[a_1,...,a_{n_a},b_0,...,b_{n_b}];$

(4) from θ (k), extract master mould parameter A (z ^-1), B (z ^-1), and calculate A (z ^-1)=A (z ^-1) Δ, B (z ^-1)=B (z ^-1) Δ;

The rolling optimization module is used for finding the solution controlled quentity controlled variable u (k) based on forecast model module, feedback compensation module and GPC (Generalized Predictive Control) algorithm computing, concrete formula following (5):

Δu(k)＝d ₁ ^T[Y _r(k+1)-F ₀ΔU(k-1)-SY(k)]

，u(k)＝u(k-1)+Δu(k) (5)

d ₁ ^TBe (G ^TQG+ λ) ^-1G ^TThe i row vector of Q

Wherein, Y (k)=[y (k) y (k-1) L y (k-n _a)] ^T,

ΔU(k-1)＝[Δu(k-n _b)Δu(k-n _b+1)L?Δu(k-1)] ^T；

The control output module, the data-signal that is used for the u (k) that will calculate outputs to air separation column.

As preferred a kind of scheme: described general model control system also comprises the DCS system, and described DCS system is made of data-interface, control station and historical data base, and described data storage device is the historical data base of DCS system.

As preferred another kind of scheme: described field intelligent instrument, DCS system, universal model controller connect successively by fieldbus

As preferred another scheme: described generalized predictive controller also comprises human-computer interface module, is used for the value of the control variable u (k) that will calculate, and with it and detect the Y that obtains ₁, Xn value on the man-machine interface of controller, show.

A kind of generalized forecast control method of air separation column may further comprise the steps:

(1) determines the bi-component setting value Y of air separation column _1set, X _Nset, and sampling period T;

(2) each sampling instant KT infers component according to detecting the temperature and the pressure data that obtain, and its formula is (1), (2):

${\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">Y</figure-callout>}}_1=\frac{\mathrm{\&alpha;}}{\mathrm{\&alpha;}-1}-\frac{{10}^{(a-\frac{\mathrm{<figure-callout\; id="1"\; label="b\; T"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">b</figure-callout>}}{T_{}})}}{(\mathrm{\&alpha;}-1)P}---\left(1\right)$

$\mathrm{Xn}=\frac{\mathrm{P\&alpha;}}{{10}^{(a-\frac{T_n+c}{b})}(\mathrm{\&alpha;}-1)}-\frac{1}{\mathrm{\&alpha;}-1}---\left(2\right)$

Wherein, Y ₁Be the component of nitrogen in the nitrogen product in the air separation column, Xn is the component of nitrogen in the liquid oxygen product, and P is last tower pressure, T ₁～T _nRespectively be last tower temperature, α is a relative volatility, and a, b, c are the Peter Antonie constant;

(3) by the DCS historical data base or do the data fitting system that obtains of test and have the discrete differential equation model of step disturbance nonstationary noise at random, its formula is (3):

A(z ^-1)y(k)＝B(z ^-1)u(k-1)+C(z ^-1)ε(k)/Δ (3)

Wherein $A\left(z^{-1}\right)=1+\underset{i=1}{\overset{n_a}{\mathrm{\&Sigma;}}}{a}_i{z}^{-i},$ $B\left(z^{-1}\right)=\underset{i=0}{\overset{n_b}{\mathrm{\&Sigma;}}}{b}_i{z}^{-i},$ Δ-difference operator, Δ=1-z ^-1

(4) put initial value P, M, Q, λ, n _a, n _b, ρ, α, θ (0), length of field, Q were that error weighting matrix, λ are that forgetting factor, α are the softening factor for control weighting matrix, ρ when length of field, M were for control when wherein P was for prediction, $\mathrm{\&theta;}=[a_1,...,a_{n_a},b_0,...,b_{n_b}]$

(5) self-correcting recursion update system model, concrete steps are as follows:

(5.1) upgrade vector

X(k-1) ^T＝[-Δy(k-1)，...，-Δy(k-n _a)，Δu(k-1)，...，Δu(k-n _b-1)]

(5.2) read y (k) and calculate ε (k)=Δ y (k)-X (k-1) ^Tθ (k-1)

(5.3) use the recursive least squares with forgetting factor to obtain θ (k), its calculating formula is (4): $\mathrm{\&theta;}\left(k\right)=\mathrm{\&theta;}(k-1)+\frac{P(k-2)X(k-1)\mathrm{\&epsiv;}\left(k\right)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}---\left(4\right)$

Wherein, $P(k-1)=\frac{1}{\mathrm{\&rho;}}[P(k-2)-\frac{P(k-2)X(k-1)X{(k-1)}^{T}P(k-2)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}],$ P (1) is enough big positive definite matrix;

(5.4) from θ (k), extract master mould parameter A (z ^-1), B (z ^-1), and calculate A (z ^-1)=A (z ^-1) Δ, B (z ^-1)=B (z ^-1) Δ;

(6) find the solution controlled quentity controlled variable u (k), concrete formula following (4):

Δu(k)＝d ₁ ^T[Y _r(k+1)-F ₀ΔU(k-1)-SY(k)]

，u(k)＝u(k-1)+Δu(k) (4)

d ₁ ^TBe (G ^TQG+ λ) ^-1G ^TThe i row vector of Q

Y (k)=[y (k) y (k-1) L y (k-n wherein _a)] ^T,

ΔU(k-1)＝[Δu(k-n _b)Δu(k-n _b+1)L?Δu(k-1)] ^T

(7) data-signal with u (k) returns to air separation column.

As preferred a kind of scheme: described control method also comprises: (8), fall into a trap in described (6) and to have calculated the value of control variable u (k), and with it and detect the Y that obtains ₁, Xn value on the man-machine interface of controller, show.

As preferred another kind of scheme: described data storage device is the historical data base of DCS system, described DCS system is made of data-interface, control station and historical data base, in described (8), data are passed to the DCS system, and at the control station procedure for displaying state of DCS.

Technical conceive of the present invention is: the composition X that adopts the product light constituent at high-purity air separation column two ends _D, X _BAs controlled variable, the reflux ratio R/ (R+D) of corresponding overhead product and the reboil ratio V/B of bottom product are as control variable.

Infer control section, be used to solve the difficult problem that the industry spot product component can not directly be measured,, can eliminate greatly and measure hysteresis and have stronger reliability relatively with respect to the chromatographic way of online applicable industry.The generalized predictive control part is used to use GPC (Generalized Predictive Control) algorithm to obtain the value of real-time control variable.

Generalized predictive control and traditional PID control algorithm difference are that algorithm itself is based on matrix operation, the coupled problem of solution multivariable Control that can be essential, algorithm has comprised controlling object parameter self-correcting module, effectively solved the process nonlinear problem, the fast development of the widespread use of DCS and computer technology in recent years also makes its arithmetic speed increase greatly, and commercial Application becomes possibility.

Beneficial effect of the present invention mainly shows: 1, can adapt to the coupling between dynamic perfromance, strong nonlinear and the loop of high-purity air separation column operation, realize the quiet run to bi-component control at the bottom of the high-purity distillation process cat head tower; 2, can access good control effect; 3, simple to operate, applicability is strong.

(4) description of drawings

Fig. 1 is the hardware connection layout of air separation column high-purity distillation control system based on generalized predictive control proposed by the invention.

Fig. 2 is the theory diagram of generalized predictive control of the present invention.

Fig. 3 is the on-the-spot connection layout of air separation column high-purity distillation control system based on generalized predictive control proposed by the invention.

(5) embodiment

Below in conjunction with accompanying drawing the present invention is further described.The embodiment of the invention is used for the present invention that explains, rather than limits the invention, and in the protection domain of spirit of the present invention and claim, any modification and change to the present invention makes all fall into protection scope of the present invention.

Embodiment 1

With reference to Fig. 1, Fig. 2, Fig. 3, a kind of generalized predictable control system of air separation column, comprise and air separation column 1 direct-connected field intelligent instrument 2, the data storage device and the host computer 6 that are used for storing history data, intelligence instrument 2, data storage device and host computer 6 link to each other successively, described host computer 6 is the universal model controller, described universal model controller comprises component deduction control section and universal model control section, described component infers that control section comprises: measuring instrument module 7, comprise detector unit and pressure detecting element, be used to detect the temperature and pressure of the last tower of air separation column; I/O component module 9 is used for the transmission between controller inside and controller and DCS of electric signal and data-signal, and component inference module 10 is used for inferring component according to detecting the temperature and the pressure data that obtain, and its formula is (1), (2):

${\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">Y</figure-callout>}}_1=\frac{\mathrm{\&alpha;}}{\mathrm{\&alpha;}-1}-\frac{{10}^{(a-\frac{\mathrm{<figure-callout\; id="1"\; label="b\; T"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">b</figure-callout>}}{T_{}})}}{(\mathrm{\&alpha;}-1)P}---\left(1\right)$

$\mathrm{Xn}=\frac{\mathrm{P\&alpha;}}{{10}^{(a-\frac{T_n+c}{b})}(\mathrm{\&alpha;}-1)}-\frac{1}{\mathrm{\&alpha;}-1}---\left(2\right)$

Wherein, Y ₁Be the component of nitrogen in the nitrogen product in the air separation column, Xn is the component of nitrogen in the liquid oxygen product, and P is last tower pressure, T ₁, T _nBe respectively cat head, column bottom temperature, α is a relative volatility, and a, b, c are the Peter Antonie constant;

Described generalized predictive control part 11 comprises: the I/O component module is used for the inside of universal model controller and the electric signal between controller and the data storage device, the transmission of data-signal; Forecast model module 11 is used for that DCS historical data base or test gained data are carried out least square and fits the system of obtaining and have the discrete differential equation model of step disturbance nonstationary noise at random, and its formula is (3);

A(z ^-1)y(k)＝B(z ^-1)u(k-1)+C(z ^-1)ε(k)/Δ (3)

Wherein $A\left(z^{-1}\right)=1+\underset{i=1}{\overset{n_a}{\mathrm{\&Sigma;}}}{a}_i{z}^{-i},$ $B\left(z^{-1}\right)=\underset{i=0}{\overset{n_b}{\mathrm{\&Sigma;}}}{b}_i{z}^{-i},$ Δ-difference operator, Δ=1-z ^-1

Model self-correcting module 12 is used to adopt the Recursive Least Squares with forgetting factor, on-line correction system model parameter, and concrete steps are as follows:

(1) upgrades vectorial X (k-1) ^T=[Δ y (k-1) ... ,-Δ y (k-n _a), Δ u (k-1) ..., Δ u (k-n _b-1)];

(2) read y (k) and calculate ε (k)=Δ y (k)-X (k-1) ^Tθ (k-1);

(3) use the recursive least squares with forgetting factor to obtain θ (k), its formula is (4):

$\mathrm{\&theta;}\left(k\right)=\mathrm{\&theta;}(k-1)+\frac{P(k-2)X(k-1)\mathrm{\&epsiv;}\left(k\right)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}---\left(4\right)$

Wherein, $P(k-1)=\frac{1}{\mathrm{\&rho;}}[P(k-2)-\frac{P(k-2)X(k-1)X{(k-1)}^{T}P(k-2)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}],$ P (1) is enough big positive definite matrix, $\mathrm{\&theta;}=[a_1,...,a_{n_a},b_0,...,b_{n_b}];$

(4) from θ (k), extract master mould parameter A (z ^-1), B (z ^-1), and calculate A (z ^-1)=A (z ^-1) Δ, B (z ^-1)=B (z ^-1) Δ;

Rolling optimization module 13 is used for finding the solution controlled quentity controlled variable u (k) based on forecast model module, feedback compensation module and GPC (Generalized Predictive Control) algorithm computing, concrete formula following (5):

Δu(k)＝d ₁ ^T[Y _r(k+1)-F ₀ΔU(k-1)-SY(k)]

，u(k)＝u(k-1)+Δu(k) (5)

d ₁ ^TBe (G ^TQG+ λ) ^-1G ^TThe i row vector of Q

Wherein, Y (k)=[y (k) y (k-1) L y (k-n _a)] ^T,

ΔU(k-1)＝[Δu(k-n _b)Δu(k-n _b+1)L?Δu(k-1)] ^T

The control output module, the data-signal that is used for the u (k) that will calculate outputs to air separation column.

Described generalized predictable control system also comprises DCS system 12, and described DCS system 12 is made of data-interface 3, control station 4 and historical data base 5, and described data storage device is the historical data base 5 of DCS system.Described field intelligent instrument 2, DCS system, generalized predictive controller 6 connect successively by fieldbus.

With reference to Fig. 1, the generalized predictable control system of the air separation column of present embodiment, comprise the field intelligent instrument 2, DCS system and the generalized predictive controller 6 that link to each other with on-the-spot air separation column 1, described DCS system is made of data-interface 3, control station 4 and historical data base 5; On-the-spot air separation column object 1, intelligence instrument 2, DCS system, generalized predictive controller 6 connect successively by fieldbus.

The generalized predictable control system hardware structure diagram of the air separation column of present embodiment as shown in Figure 1, the core of described generalized predictable control system is a generalized predictive controller 6, comprises in addition: field intelligent instrument 2, DCS system and fieldbus.On-the-spot air separation column 1, intelligence instrument 2, DCS system, generalized predictive controller 6 link to each other successively by fieldbus, and uploading of information of realization assigned.Generalized predictable control system in time obtains the value of the control variable of current time by industry spot data detected and that extract from historical data base 5, and returns to first floor system, in time system dynamics is made a response.

The theory diagram of the generalized predictive controller of the air separation column of present embodiment as shown in Figure 2, the generalized predictive controller of described air separation column comprises:

Infer control section, be used to solve the difficult problem that the industry spot product component can not directly be measured,, can eliminate greatly and measure hysteresis and have stronger reliability relatively with respect to the chromatographic way of online applicable industry.

1) the measuring instrument module 7: comprise detector unit, can adopt the thermojunction type temperature transmitter, and pressure detecting element, can adopt the pressure resistance type transmitter.

2) the I/O component module 9: be used for the transmission between controller inside and controller and DCS of electric signal and data-signal.

3) the component inference module 10: be used for inferring component according to detecting the temperature and the pressure data that obtain.Its formula is (1), (2):

${\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">Y</figure-callout>}}_1=\frac{\mathrm{\&alpha;}}{\mathrm{\&alpha;}-1}-\frac{{10}^{(a-\frac{\mathrm{<figure-callout\; id="1"\; label="b\; T"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">b</figure-callout>}}{T_{}})}}{(\mathrm{\&alpha;}-1)P}---\left(1\right)$

$\mathrm{Xn}=\frac{\mathrm{P\&alpha;}}{{10}^{(a-\frac{T_n+c}{b})}(\mathrm{\&alpha;}-1)}-\frac{1}{\mathrm{\&alpha;}-1}---\left(2\right)$

Y wherein ₁Be the component of nitrogen in the nitrogen product in the air separation column, Xn is the component of nitrogen in the liquid oxygen product, and P is last tower pressure, T ₁, T _nBe respectively cat head, column bottom temperature, α is a relative volatility, and a, b, c are the Peter Antonie constant.

The generalized predictive control part is used to use GPC (Generalized Predictive Control) algorithm to obtain the value of real-time control variable, comprising:

1) I/O element: be used for the inside of generalized predictive control and the electric signal between controller and the DCS, the transmission of data-signal.

2) the forecast model module 11, are used for that DCS historical data base or test gained data are carried out least square and fit the system of obtaining and have the discrete differential equation model of step disturbance nonstationary noise at random, and its formula is (3):

A(z ^-1)y(k)＝B(z ^-1)u(k-1)+C(z ^-1)ε(k)/Δ (3)

Wherein $A\left(z^{-1}\right)=1+\underset{i=1}{\overset{n_a}{\mathrm{\&Sigma;}}}{a}_i{z}^{-i},$ $B\left(z^{-1}\right)=\underset{i=0}{\overset{n_b}{\mathrm{\&Sigma;}}}{b}_i{z}^{-i},$ Δ-difference operator, Δ=1-z ^-1

3) model self-correcting module 12 is used to adopt the Recursive Least Squares with forgetting factor, on-line correction system prediction model 11 parameters.Concrete steps are as follows:

(1) upgrades vectorial X (k-1) ^T=[Δ y (k-1) ... ,-Δ y (k-n _a), Δ u (k-1) ..., Δ u (k-n _b-1)]

(2) read y (k) and calculate ε (k)=Δ y (k)-X (k-1) ^Tθ (k-1)

(3) use the recursive least squares with forgetting factor to obtain θ (k), its formula is (4):

$\mathrm{\&theta;}\left(k\right)=\mathrm{\&theta;}(k-1)+\frac{P(k-2)X(k-1)\mathrm{\&epsiv;}\left(k\right)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}---\left(4\right)$

$P(k-1)=\frac{1}{\mathrm{\&rho;}}[P(k-2)-\frac{P(k-2)X(k-1)X{(k-1)}^{T}P(k-2)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}],$ P (1) is enough big positive definite matrix, $\mathrm{\&theta;}=[a_1,...,a_{n_a},b_0,...,b_{n_b}].$

(4) from θ (k), extract master mould parameter A (z ^-1), B (z ^-1), and calculate A (z ^-1)=A (z ^-1) Δ, B (z ^-1)=B (z ^-1) Δ.

4) the rolling optimization module 13, are used for obtaining based on forecast model module, feedback compensation module and GPC (Generalized Predictive Control) algorithm computing the controller output of current time.Concrete operation formula following (5):

Δu(k)＝d ₁ ^T[Y _r(k+1)-F ₀ΔU(k-1)-SY(k)]

，u(k)＝u(k-1)+Δu(k) (5)

d ₁ ^TBe (G ^TQG+ λ) ^-1G ^TThe i row vector of Q

Y (k)=[y (k) y (k-1) L y (k-n wherein _a)] ^T,

ΔU(k-1)＝[Δu(k-n _b)Δu(k-n _b+1)L?Δu(k-1)] ^T

5) human-computer interface module 8, are used for the selection of controller parameter and the demonstration of historical data and system's current state, and the choosing, set of controller parameter.

The on-the-spot connection layout of the generalized predictable control system of the air separation column of present embodiment as shown in Figure 3, system adopts the component Y that goes up nitrogen in the tower 15 top nitrogen products ₁, the component Xn that goes up nitrogen in the tower 1 bottom liquid oxygen product is controlled variable, the flow of the liquid oxygen of following tower 16 supreme towers 14 and the capacity of returns of liquid air, liquid oxygen product is the control corresponding variable.Connect a detector unit TT and pressure detecting element PT respectively and be delivered to upper system at the bottom of the last tower 14 cat head towers, generalized predictive controller by the data computation current time of on-the-spot and historical data base the control variable value and pass to down layer system, the scene changes the value of control variable by the change valve opening by flow controller FC.

The generalized forecast control method of described air separation column is realized according to following steps:

1, system initialization

(1) in generalized predictive controller 6, sets the bi-component setting value Y of air separation column _1set, X _Nset, and sampling period among the DCS is set.

(2) by the DCS historical data base or do the data that obtain of test and carry out least square and fit the system of obtaining and have the discrete differential equation initial model of step disturbance nonstationary noise at random, its formula is (3):

A(z ^-1)y(k)＝B(z ^-1)u(k-1)+C(z ^-1)ε(k)/Δ (3)

Wherein $A\left(z^{-1}\right)=1+\underset{i=1}{\overset{n_a}{\mathrm{\&Sigma;}}}{a}_i{z}^{-i},$ $B\left(z^{-1}\right)=\underset{i=0}{\overset{n_b}{\mathrm{\&Sigma;}}}{b}_i{z}^{-i},$ Δ-difference operator, Δ=1-z ^-1, and in forecast model module 11, establish initial value n _a, n _b

(3) put initial value P, M, Q, λ, ρ, α, θ (0) in feedback compensation module 12, the rolling optimization module 13.Wherein P length of field, Q when when prediction, length of field, M were for control is the error weighting matrix, and generally the unit's of getting diagonal matrix, λ are the control weighting matrix, generally gets enough little number, ρ is a forgetting factor, generally gets 0.95～1, α is the softening factor, $\mathrm{\&theta;}=[a_1,...,a_{n_a},b_0,...,b_{n_b}].$

(4) put rolling optimization module 13 initial value R ₁(z ^-1)=1, S ₁(z ^-1)=z[1-A (z ^-1)], G ₁(z ^-1)=B (z ^-1)

2, system puts into operation.

1) each DCS sampling instant, intelligence instrument 2 detects temperature, the pressure data of on-the-spot air separation column 1 and is sent in the DCS historical data base 5;

2) each controller sampling instant, generalized predictive controller 6 reads the temperature and pressure data from DCS historical data base 5, calculate current time controlled variable Y by component inference module 10 ₁, the value of Xn, its formula is (1), (2):

${\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">Y</figure-callout>}}_1=\frac{\mathrm{\&alpha;}}{\mathrm{\&alpha;}-1}-\frac{{10}^{(a-\frac{\mathrm{<figure-callout\; id="1"\; label="b\; T"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">b</figure-callout>}}{T_{}})}}{(\mathrm{\&alpha;}-1)P}---\left(1\right)$

$\mathrm{Xn}=\frac{\mathrm{P\&alpha;}}{{10}^{(a-\frac{T_n+c}{b})}(\mathrm{\&alpha;}-1)}-\frac{1}{\mathrm{\&alpha;}-1}---\left(2\right)$

Y wherein ₁Be the component of nitrogen in the nitrogen product in the air separation column, Xn is the component of nitrogen in the liquid oxygen product, and P is last tower pressure, T ₁, T _nBe respectively cat head, column bottom temperature, α is a relative volatility, and a, b, c are the Peter Antonie constant;

3) Y by obtaining from component inference module 10 ₁, Xn value, obtain the system model parameter of current time by the computing of model self-correcting module 12, its concrete steps are as follows:

(1) upgrades vectorial X (k-1) ^T=[Δ y (k-1) ... ,-Δ y (k-n _a), Δ u (k-1) ..., Δ u (k-n _b-1)]

(2) read y (k) and calculate ε (k)=Δ y (k)-X (k-1) ^Tθ (k-1)

(3) use the recursive least squares with forgetting factor to obtain θ (k), its formula is (4):

$\mathrm{\&theta;}\left(k\right)=\mathrm{\&theta;}(k-1)+\frac{P(k-2)X(k-1)\mathrm{\&epsiv;}\left(k\right)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}---\left(4\right)$

$P(k-1)=\frac{1}{\mathrm{\&rho;}}[P(k-2)-\frac{P(k-2)X(k-1)X{(k-1)}^{T}P(k-2)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}],$ P (1) is enough big positive definite matrix.

(4) from θ (k), extract master mould parameter A (z ^-1), B (z ^-1), and calculate A (z ^-1)=A (z ^-1) Δ, B (z ^-1)=B (z ^-1) Δ.

4) calculate R _j(z ^-1), S _j(z ^-1), G _j(z ^-1), the concrete operation formula is as follows:

R _j+1(z ^-1)＝R _j(z ^-1)+r _j+1，jz ^-j，S _j+1(z ^-1)＝z[S _j(z ^-1)-r _j+1，jA(z ^-1)]，

G _j+1(z ^-1)＝G _j(z ^-1)+z ^-js _j，0B(z ^-1)

In the formula $R_j\left(z^{-1}\right)=1+\underset{i=1}{\overset{j-1}{\mathrm{\&Sigma;}}}{r}_{j,i}{z}^{-i},$ $S_j\left(z^{-1}\right)=\underset{i=0}{\overset{n_a}{\mathrm{\&Sigma;}}}{s}_{j,i}{z}^{-i},$

${\stackrel{\&OverBar;}{G}}_j\left(z^{-1}\right)=B\left(z^{-1}\right)R_j\left(z^{-1}\right)=g_{j,0}+g_{j,1}{z}^{-1}+...+g_{j,n_b+j-1}{z}^{-(n_b+j-1)}$

r _j+1，j＝s _j，0＝s _j(0)

s _j+1，0＝s _j，1-a ₁s _j，0

s _j+1，1＝s _j，2-a ₂s _j，0

.

.

.

s _j+1，i＝s _j，i+1-a _i+1s _j，0(0≤i＜n _a)

.

.

.

$s_{j+1,n_a}=-{\stackrel{\&OverBar;}{a}}_{n_a+1}{s}_{j,0}$

5) compute matrix $G,F_0,S,G=\left[\begin{array}{cccc}{g}_0& 0& 0& 0\\ g_1& g_0& 0& 0\\ \&CenterDot;& & & \\ \&CenterDot;& & g_0& 0\\ \&CenterDot;& & & \\ g_{M-1}& g_{M-2}& \&CenterDot;\&CenterDot;\&CenterDot;& g_0\\ \&CenterDot;& & & \\ \&CenterDot;& & & \\ \&CenterDot;& & & \\ g_{P-1}& g_{P-2}& \&CenterDot;\&CenterDot;\&CenterDot;& g_{P-M}\end{array}\right],$

$F_0=\left[\begin{array}{ccccc}{\mathrm{<figure-callout\; id="1"\; label="g"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">g</figure-callout>}}_{1,\mathrm{<figure-callout\; id="1"\; label="nb\; g"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">nb</figure-callout>}}& g_{}{}_{1,\mathrm{nb}-1}& \&CenterDot;\&CenterDot;\&CenterDot;& {\mathrm{<figure-callout\; id="1"\; label="g"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">g</figure-callout>}}_{\mathrm{1,2}}& {\mathrm{<figure-callout\; id="1"\; label="g"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">g</figure-callout>}}_{\mathrm{1,1}}\\ {\mathrm{<figure-callout\; id="2"\; label="g"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">g</figure-callout>}}_{2,\mathrm{nb}+1}& {\mathrm{<figure-callout\; id="2"\; label="g"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">g</figure-callout>}}_{2,\mathrm{nb}}& \&CenterDot;\&CenterDot;\&CenterDot;& {\mathrm{<figure-callout\; id="2"\; label="g"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">g</figure-callout>}}_{\mathrm{2,3}}& {\mathrm{<figure-callout\; id="2"\; label="g"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">\; <figure-callout\; id="2"\; label="g"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">g</figure-callout>\; </figure-callout>}}_{\mathrm{2,2}}\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;& \&CenterDot;\\ g_{P,\mathrm{nb}+P-1}& g_{P,\mathrm{nb}+P-2}& \&CenterDot;\&CenterDot;\&CenterDot;& g_{P,P+1}& g_{P,P}\end{array}\right],$ $S=\left[\begin{array}{ccc}{\mathrm{<figure-callout\; id="1"\; label="s"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">s</figure-callout>}}_{\mathrm{1,0}}& \&CenterDot;\&CenterDot;\&CenterDot;& {\mathrm{<figure-callout\; id="1"\; label="s"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">s</figure-callout>}}_{1,n_a}\\ \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& & \&CenterDot;\\ s_{P,0}& \&CenterDot;\&CenterDot;\&CenterDot;& S_{P,n_a}\end{array}\right]$

6) according to the output valve of setting value, current system, determine the interior output reference locus of future anticipation time domain that current time rises, choose and adopt following single order exponential form:

y _r(k+i)＝α _r ⁱy(k)+(1-α _r ⁱ)y _set

(i＝1，2，...)

y _r(k)＝y(k)

Y wherein _SetBe the setting value of y, α is the flexible factor, should choose suitable value in the practical operation, and the value of α is big more, and systematically flexibility is good more, and robustness is strong more, but control ground rapidity is poor.

7) find the solution controlled quentity controlled variable u (k), concrete formula following (5):

Δu(k)＝d ₁ ^T[Y _r(k+1)-F ₀ΔU(k-1)-SY(k)]

，u(k)＝u(k-1)+Δu(k) (5)

d ₁ ^TBe (G ^TQG+ λ) ^-1G ^TThe i row vector of Q

Y (k)=[y (k) y (k-1) wherein ... y (k-n _a)] ^T,

ΔU(k-1)＝[Δu(k-n _b)Δu(k-n _b+1)…Δu(k-1)] ^T

8) result is delivered on the display module of each level system and show, make things convenient for the slip-stick artist in time process dynamically to be made a response and operate, comprise human-computer interface module 8, DCS system operation station 4 and the operator station of generalized predictive controller.

Embodiment 2

With reference to Fig. 1, Fig. 2 and Fig. 3, the control method that the generalized predictable control system of the described air separation column of a kind of usefulness is realized, described control method may further comprise the steps:

(1) determines the bi-component setting value Y of air separation column _1set, X _Nset, and sampling period T;

(2) each sampling instant KT infers component according to detecting the temperature and the pressure data that obtain, and its formula is (1), (2):

${\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">Y</figure-callout>}}_1=\frac{\mathrm{\&alpha;}}{\mathrm{\&alpha;}-1}-\frac{{10}^{(a-\frac{\mathrm{<figure-callout\; id="1"\; label="b\; T"\; filenames="C20061015548500211.png"\; state="\{\{state\}\}">b</figure-callout>}}{T_{}})}}{(\mathrm{\&alpha;}-1)P}---\left(1\right)$

$\mathrm{Xn}=\frac{\mathrm{P\&alpha;}}{{10}^{(a-\frac{T_n+c}{b})}(\mathrm{\&alpha;}-1)}-\frac{1}{\mathrm{\&alpha;}-1}---\left(2\right)$

Wherein, Y ₁Be the component of nitrogen in the nitrogen product in the air separation column, Xn is the component of nitrogen in the liquid oxygen product, and P is last tower pressure, T ₁, T _nBe respectively cat head, column bottom temperature, α is a relative volatility, and a, b, c are the Peter Antonie constant;

(3) by the DCS historical data base or do the data fitting system that obtains of test and have the discrete differential equation model of step disturbance nonstationary noise at random, its formula is (3):

A(z ^-1)y(k)＝B(z ^-1)u(k-1)+C(z ^-1)ε(k)/Δ (3)

Wherein $A\left(z^{-1}\right)=1+\underset{i=1}{\overset{n_a}{\mathrm{\&Sigma;}}}{a}_i{z}^{-i},$ $B\left(z^{-1}\right)=\underset{i=0}{\overset{n_b}{\mathrm{\&Sigma;}}}{b}_i{z}^{-i},$ Δ-difference operator, Δ=1-z ^-1

(4) put initial value P, M, Q, λ, n _a, n _b, ρ, α, θ (0), length of field, Q were that error weighting matrix, λ are that forgetting factor, α are the softening factor for control weighting matrix, ρ when length of field, M were for control when wherein P was for prediction, $\mathrm{\&theta;}=[a_1,...,a_{n_a},b_0,...,b_{n_b}]$

(5) self-correcting recursion update system model, concrete steps are as follows:

(5.1) upgrade vector

X(k-1) ^T＝[-Δy(k-1)，...，-Δy(k-n _a)，Δu(k-1)，...，Δu(k-n _b-1)]

(5.2) read y (k) and calculate ε (k)=Δ y (k)-X (k-1) ^Tθ (k-1)

(5.3) use the recursive least squares with forgetting factor to obtain θ (k), its calculating formula is (4):

$\mathrm{\&theta;}\left(k\right)=\mathrm{\&theta;}(k-1)+\frac{P(k-2)X(k-1)\mathrm{\&epsiv;}\left(k\right)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}---\left(4\right)$

Wherein, $P(k-1)=\frac{1}{\mathrm{\&rho;}}[P(k-2)-\frac{P(k-2)X(k-1)X{(k-1)}^{T}P(k-2)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}],$ P (1) is enough big positive definite matrix;

(5.4) from θ (k), extract master mould parameter A (z ^-1), B (z ^-1), and calculate A (z ^-1)=A (z ^-1) Δ, B (z ^-1)=B (z ^-1) Δ;

(6) find the solution controlled quentity controlled variable u (k), concrete formula following (4):

Δu(k)＝d ₁ ^T[Y _r(k+1)-F ₀ΔU(k-1)-SY(k)]

，u(k)＝u(k-1)+Δu(k) (4)

d ₁ ^TBe (G ^TQG+ λ) ^-1G ^TThe i row vector of Q

Y (k)=[y (k) y (k-1) wherein ... y (k-n _a)] ^T,

ΔU(k-1)＝[Δu(k-n _b)Δu(k-n _b+1)…Δu(k-1)] ^T

(7) data-signal with u (k) returns to air separation column.

Control method also comprises: (8), fall into a trap in described (6) and to have calculated the value of control variable u (k), and with it and detect the Y that obtains ₁, Xn value on the man-machine interface of controller, show.Described data storage device is the historical data base of DCS system, and described DCS system is made of data-interface, control station and historical data base, in described (8), data is passed to the DCS system, and at the control station procedure for displaying state of DCS.

Claims (7)

1, a kind of generalized predictable control system of air separation column, comprise and the direct-connected field intelligent instrument of air separation column, the data storage device that is used for storing history data and host computer, intelligence instrument, data storage device and host computer link to each other successively, it is characterized in that: described host computer is a generalized predictive controller, described generalized predictive controller comprises component deduction control section and generalized predictive control part

Described component infers that control section comprises:

The measuring instrument module comprises detector unit and pressure detecting element, is used to detect the temperature and pressure of the last tower of air separation column;

The I/O component module is used for the transmission between controller inside and controller and data storage device of electric signal and data-signal,

The component inference module is used for inferring component according to detecting the temperature and the pressure data that obtain, and its formula is (1), (2):

$Y_1=\frac{\mathrm{\&alpha;}}{\mathrm{\&alpha;}-1}-\frac{{10}^{(a-\frac{b}{T_1+c})}}{(\mathrm{\&alpha;}-1)P}---\left(1\right)$

$\mathrm{Xn}=\frac{\mathrm{P\&alpha;}}{{10}^{(a-\frac{T_n+c}{b})}(\mathrm{\&alpha;}-1)}-\frac{1}{\mathrm{\&alpha;}-1}---\left(2\right)$

Wherein, Y ₁Be the component of nitrogen in the nitrogen product in the air separation column, Xn is the component of nitrogen in the liquid oxygen product, and P is last tower pressure, T ₁, T _nBe respectively cat head, column bottom temperature, α is a relative volatility, and a, b, c are the Peter Antonie constant; Described generalized predictive control partly comprises:

The I/O component module is used for the inside of generalized predictive controller and the electric signal between controller and the data storage device, the transmission of data-signal;

The forecast model module is used for that DCS historical data base or test gained data are carried out least square and fits the system of obtaining and have the discrete differential equation model of step disturbance nonstationary noise at random, and its formula is (3);

A(z ^-1)y(k)＝B(z ^-1)u(k-1)+C(z ^-1)ε(k)/Δ (3)

Wherein $A\left(z^{-1}\right)=1+\underset{i=1}{\overset{n_a}{\mathrm{\&Sigma;}}}{a}_i{z}^{-i},$ $B\left(z^{-1}\right)=\underset{i=0}{\overset{n_b}{\mathrm{\&Sigma;}}}{b}_i{z}^{-i},$ Δ-difference operator, Δ=1-z ^-1

Model self-correcting module is used to adopt the Recursive Least Squares with forgetting factor, on-line correction system model parameter, and concrete steps are as follows:

1) upgrades vectorial X (k-1) ^T=[Δ y (k-1) ... ,-Δ y (k-n _a), Δ u (k-1) ..., Δ u (k-n _b-1)];

2) read y (k) and calculate ε (k)=Δ y (k)-X (k-1) ^Tθ (k-1);

3) use the recursive least squares with forgetting factor to obtain θ (k), its formula is (4):

$\mathrm{\&theta;}\left(k\right)=\mathrm{\&theta;}(k-1)+\frac{P(k-2)X(k-1)\mathrm{\&epsiv;}\left(k\right)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}---\left(4\right)$

Wherein, $P(k-1)=\frac{1}{\mathrm{\&rho;}}[P(k-2)-\frac{P(k-2)X(k-1)X{(k-1)}^{T}P(k-2)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}],$ P (k-1) is enough big positive definite matrix, $\mathrm{\&theta;}=[a_1,...,a_{n_a},b_0,...,b_{n_b}];$

4) from θ (k), extract master mould parameter A (z ^-1), B (z ^-1), and calculate A (z ^-1)=A (z ^-1) Δ, B (z ^-1)=B (z ^-1) Δ;

The rolling optimization module is used for finding the solution controlled quentity controlled variable u (k) based on forecast model module, model self-correcting module and GPC (Generalized Predictive Control) algorithm computing, concrete formula following (5):

Δu(k)＝d ₁ ^T[Y _r(k+1)-F ₀ΔU(k-1)-SY(k)]，u(k)＝u(k-1)+Δu(k) (5)

d ₁ ^TBe (G ^TQG+ λ) ^-1G ^TThe i row vector of Q

Wherein, Y (k)=[y (k) y (k-1) ... y (k-n _a)] ^T,

ΔU(k-1)＝[Δu(k-n _b)Δu(k-n _b+1)…Δu(k-1)] ^T

The control output module, the data-signal that is used for the u (k) that will calculate outputs to air separation column.

2, the generalized predictable control system of air separation column as claimed in claim 1, it is characterized in that: described generalized predictable control system also comprises the DCS system, described DCS system is made of data-interface, control station and historical data base, and described data storage device is the historical data base of DCS system.

3, the generalized predictable control system of air separation column as claimed in claim 2 is characterized in that: described field intelligent instrument, DCS system, generalized predictive controller connect successively by fieldbus.

4, as the generalized predictable control system of the described air separation column of one of claim 1～3, it is characterized in that: described generalized predictive controller also comprises human-computer interface module, is used for the value of the controlled quentity controlled variable u (k) that will calculate, and with it and detect the Y that obtains ₁, Xn value on the man-machine interface of controller, show.

5, the control method of the generalized predictable control system of a kind of usefulness air separation column as claimed in claim 1 realization, it is characterized in that: described control method may further comprise the steps:

1) determines the bi-component setting value Y of air separation column _1set, X _Nset, and sampling period T;

Each sampling instant KT infers component according to detecting the temperature and the pressure data that obtain, and its formula is (1), (2):

$Y_1=\frac{\mathrm{\&alpha;}}{\mathrm{\&alpha;}-1}-\frac{{10}^{(a-\frac{b}{T_1+c})}}{(\mathrm{\&alpha;}-1)P}---\left(1\right)$

$\mathrm{Xn}=\frac{\mathrm{P\&alpha;}}{{10}^{(a-\frac{T_n+c}{b})}(\mathrm{\&alpha;}-1)}-\frac{1}{\mathrm{\&alpha;}-1}---\left(2\right)$

Wherein, Y ₁Be the component of nitrogen in the nitrogen product in the air separation column, Xn is the component of nitrogen in the liquid oxygen product, and P is last tower pressure, T ₁, T _nBe respectively cat head, column bottom temperature, α is a relative volatility, and a, b, c are the Peter Antonie constant;

2) by the DCS historical data base or do the data fitting system that obtains of test and have step disturbance at random

The discrete differential equation model of nonstationary noise, its formula are (3):

A(z ^-1)y(k)＝B(z ^-1)u(k-1)+C(z ^-1)ε(k)/Δ (3)

Wherein $A\left(z^{-1}\right)=1+\underset{i=1}{\overset{n_a}{\mathrm{\&Sigma;}}}{a}_i{z}^{-i},$ $B\left(z^{-1}\right)=\underset{i=0}{\overset{n_b}{\mathrm{\&Sigma;}}}{b}_i{z}^{-i},$ Δ-difference operator, Δ=1-z ^-1

3) put initial value P, M, Q, λ, n _a, n _b, ρ, α, θ (0), length of field when wherein P is for prediction,

Length of field, Q were that error weighting matrix, λ are that forgetting factor, α are the softening factor for control weighting matrix, ρ when M was control, $\mathrm{\&theta;}=[a_1,...,a_{n_a},b_0,...,b_{n_b}]$

4) self-correcting recursion update system model, concrete steps are as follows:

(4.1) upgrade vector

X(k-1) ^T＝[-Δy(k-1)，...，-Δy(k-n _a)，Δu(k-1)，...，Δu(k-n _b-1)]

(4.2) read y (k) and calculate ε (k)=Δ y (k)-X (k-1) ^Tθ (k-1)

(4.3) use the recursive least squares with forgetting factor to obtain θ (k), its calculating formula is (4): $\mathrm{\&theta;}\left(k\right)=\mathrm{\&theta;}(k-1)+\frac{P(k-2)X(k-1)\mathrm{\&epsiv;}\left(k\right)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}---\left(4\right)$

Wherein, $P(k-1)=\frac{1}{\mathrm{\&rho;}}[P(k-2)-\frac{P(k-2)X(k-1)X{(k-1)}^{T}P(k-2)}{\mathrm{\&rho;}+X{(k-1)}^{T}P(k-2)X(k-1)}],$ P (k-1) is enough big positive definite matrix;

(4.4) from θ (k), extract master mould parameter A (z ^-1), B (z ^-1), and calculate A (z ^-1)=A (z ^-1) Δ, B (z ^-1)=B (z ^-1) Δ;

5) find the solution controlled quentity controlled variable u (k), concrete formula following (5):

Δu(k)＝d ₁ ^T[Y _r(k+1)-F ₀ΔU(k-1)-SY(k)]，u(k)＝u(k-1)+Δu(k) (5)

d ₁ ^TBe (G ^TQG+ λ) ^-1G ^TThe i row vector of Q

Y (k)=[y (k) y (k-1) wherein ... y (k-n _a)] ^T,

ΔU(k-1)＝[Δu(k-n _b)Δu(k-n _b+1)…Δu(k-1)] ^T

6) data-signal with u (k) returns to air separation column.

6, the generalized forecast control method of air separation column as claimed in claim 5 is characterized in that: described control method also comprises:

7), fall into a trap in described step 6) and to have calculated the value of controlled quentity controlled variable u (k), and with it and detect the Y that obtains ₁, Xn value on the man-machine interface of controller, show.

7, the generalized forecast control method of air separation column as claimed in claim 6, it is characterized in that: described data storage device is the historical data base of DCS system, described DCS system is made of data-interface, control station and historical data base, in described step 7), data are passed to the DCS system, and at the control station procedure for displaying state of DCS.

Images