CN107450325A

CN107450325A - CO after one kind burning2The Multi model Predictive Controllers of trapping system

Info

Publication number: CN107450325A
Application number: CN201710795146.3A
Authority: CN
Inventors: 吴啸; 梁修凡; 李益国; 沈炯
Original assignee: Southeast University
Current assignee: Southeast University
Priority date: 2017-09-06
Filing date: 2017-09-06
Publication date: 2017-12-08
Anticipated expiration: 2037-09-06
Also published as: CN107450325B

Abstract

CO after being burnt the invention discloses one kind₂The Multi model Predictive Controllers of trapping system, the forecast Control Algorithm is with CO after the burning based on chemisorbed₂Trapping system is controlled device, and lean solution valve opening and turbine low pressure cylinder valve opening of drawing gas are system control input amount, CO₂Capture rate and reboiler temperature are system output quantity；Subspace state space system identification is primarily based on, using data caused by system operation, the local state spatial model of system is established at different operating points；Then the nonlinear Distribution of controlled device is investigated using the method for gap metric；And then predictive controller is established at suitable local operating point, and membership function is designed by its weighted array, CO after foundation burning₂Trapping system multiple model predictive control system.The method of the present invention has good global nonlinear Control ability, is capable of the demand of the effective a wide range of variable working condition of adaptive system, fast track CO₂Capture rate setting value, improve CO₂The level of trapping system depth fast and flexible operation.

Description

CO after one kind burning2The Multi model Predictive Controllers of trapping system

Technical field

The present invention relates to forecast Control Algorithm technical field, CO after especially a kind of burning₂The multi-model of trapping system is pre- Survey control method.

Background technology

With greenhouse effects and increasingly serious, the emission reduction CO of relevant climate ecological problem₂International community's reply gas is turned into Wait the crucial behave of change.As the capital equipment of supply of electric power, fired power generating unit is CO₂Emission source that is most stable, most concentrating, generation Boundary 30%-40%, the CO in China 40%~50%₂Discharge comes from fired power generating unit.In actively development new energy technology, make great efforts to carry While high fired power generating unit generating efficiency, fired power generating unit CO₂Trapping is known as realizing greatly in Future 30 Years by numerous authoritative institutions Scale CO₂The most direct effective technological means of emission reduction.

In existing fired power generating unit CO₂In trapping technique, CO after the burning based on chemical absorption method₂Trapping technique is directly from electricity CO is separated in flue gas after factory's burning₂, there is the inheritance outstanding to existing unit and the preferable adaptability of technology, be current CO₂ Trap the mainstream technology that power station uses.Due to CO₂Trapping needs to extract a large amount of steam from fired power generating unit, in low pressure (LP) cylinder for poor Liquid regenerates, and it implements very big for the influence of fired power generating unit generating efficiency.Therefore, CO₂Trapping system must realize large-scale spirit Operation living, such as, it is urgent or electricity price higher period reduces CO in power demands₂Capture rate, and environmental protection pressure is larger or carbon valency Higher period improves capture rate.However, with CO₂The a wide range of variable parameter operation of catching apparatus, its system show stronger non- Linear characteristic, predictive controller control performance of the tradition based on linear model design is caused to reduce, stability decline.Therefore it is a kind of CO after combustion₂The exploitation that the predictive control algorithm utilized to smoke signal is added in trapping system is necessary.

The content of the invention

The technical problems to be solved by the invention are, there is provided CO after one kind burning₂The pre- observing and controlling of multi-model of trapping system Method processed, it is possible to increase CO₂The regulation quality of a wide range of variable working condition of trapping system, improve the energy that its fast deep is flexibly run Power.

In order to solve the above technical problems, the present invention provides CO after a kind of burning₂The multiple model predictive control side of trapping system Method, comprise the following steps：

(1) CO after burning₂Trapping system is switched to manual mode, near different capture rate operating points, with poor liquid stream Measure valve opening u_aDrawn gas valve opening amount signal u with turbine low pressure cylinder_bTo input, to CO₂Trapping system enters row energization, obtains CO₂Capture rate y_aWith reboiler temperature y_bOpen-loop response data；

(2) sampling period Ts is selected, withTo input,For output, subspace is utilized Discrimination method, build CO at different capture rate operating points₂Trapping system local state spatial model；

(3) method for using gap metric, analyzes the difference between each adjacent partial model, investigates the non-of CO2 trapping systems Linear distribution；

(4) model predictive controller is established at suitable partial duty point；In each sampling instant, each son is utilized respectively The CO of model pre-estimating system within following certain time₂Capture rateAnd reboiler temperaturePart is calculated most by optimization Excellent lean solution flow valve aperture uⁱ _a-opDrawn gas valve opening amount signal u with turbine low pressure cylinderⁱ _b-op；

(5) the final suitable membership function of design, each controller is exported into weighted array, obtains final poor flow quantity Valve opening u_a-opDrawn gas valve opening amount signal u with turbine low pressure cylinder_b-opAnd act on CO2 trapping systems；Whereinω_iFor membership function corresponding to i-th of controller, It is scheduling variable, capture rate CR function, uⁱ _a-opAnd uⁱ _b-opThe optimum control signal calculated for i-th of controller；

(6) the prediction matrix ψ exported in fixed each local control_x、ψ_u、ψ_y, repeat step (3)-(4) are realized continuous Control.

Preferably, in step (2), T95/T_s=5~15, wherein T95 are the regulating time that transient process rises to 95%.

Preferably, in step (2), CO at different capture rate operating points is built₂Trapping system local state spatial model, tool Body step is：

(21) output data from the 0th moment to 2N+j-2 moment that will continuously be obtained near a certain given operating point Y and input data u are arranged as Hankel matrix forms respectively：

Wherein, N is matrix line number, and N is more than CO₂Trapping system order, j are matrix columns, Y and U represent respectively output with The Hankel matrixes of input data composition, Y^fAnd Y^pThe Future Data and past data of output data, U are represented respectively^fAnd U^pRespectively Represent the Future Data and past data of input data, y_jRepresent j-th of output data, u_jRepresent j-th of input data；

(22) W is made^p=[(Y^p)^T (U^p)^T]^T, QR decomposition is carried out to following matrix：

Obtain matrix L：

(23) so as to obtaining matrix L_w=L (:,1:N (m+l)), L_u=L (:,N(m+l)+1:End), m ties up for input variable Number, l are output variable dimension, L (:,1:N (m+l)) representing matrix L preceding N (m+l) row, L (:,N(m+l)+1:End square) is represented Battle array L from N (m+l)+1 row after all row；

(24) to L_wMatrix does singular value decomposition：

Obtain matrix Γ_N=U₁(S₁)^1/2, and then obtain：Model parameterWhereinL rows before expression is removed Γ_N,Γ _NThe Γ of rear l rows is removed in expression_N,Represent Moore-Penrose pseudoinverses；Model parameter C=Γ_N(1:l,:) can be from Γ_NPreceding l rows in directly obtain；

(25) system of linear equations is solved：

Obtain model parameter B and D.

Preferably, in step (3), the clearence degree value between adjacent partial model is calculated, is concretely comprised the following steps：

(31) partial model P adjacent to two₁、P₂Orthogonal fight coprime factorization is done, is obtained：

(32) computation model P₁、P₂Between distance：

Wherein, H_∞For a special Hardy normed spaces：σ_max(G (j ω)) represents G The maximum singular value of (j ω)；Q is H_∞Arbitrary function in space；

Gap metric between (33) two systems

Preferably, in step (4), the CO of the system within following a period of time is estimated using formula (1)₂Capture rate and boil again Device temperature

Wherein, prediction matrix ψ_x、ψ_u、ψ_yRespectively：

Respectively k moment CO₂Estimated state, input and the output of trapping system, F are observer gain, u_fFor The input data at following Nu moment, It is following N_yWhen etching system estimate it is defeated Go out,

Using equation below calculation of performance indicators function J：

Wherein, Q_fAnd R_fIt is the weight matrix for adjusting input and output Control platform,r_fIt is Following N₁When etching system CO₂Capture rate and reboiler temperature setting value sequence,

Represent the k+1 moment to k+N respectively₁When etching system CO₂Capture rate r_aWith reboiler temperature r_bSetting value, It is following N_yWhen etching system CO₂Capture rate and reboiler temperature estimate value sequence,

Represent the k+1 moment to k+N respectively_yWhen etching system CO₂Capture rate y_aWith reboiler temperature y_bEstimate Value,

Δu_fIt is following N_uThe lean solution flow valve opening amount signal u at moment_aDrawn gas valve opening amount signal u with low pressure (LP) cylinder_bSequenceIncrement；

Wherein

CO2 trapping system lean solution flow valves and low pressure (LP) cylinder draw gas valve opening amount signal u Filters with Magnitude Constraints (u_min,u_max) and Increment restriction (Δ u_min,Δu_max) be：

Wherein, u_min,u_maxRepresent respectively lean solution flow valve and low pressure (LP) cylinder draw gas valve opening amount signal u minimum value with most Big value, Δ u_min,Δu_maxRepresent respectively lean solution flow valve and low pressure (LP) cylinder draw gas valve opening amount signal u smallest incremental with it is maximum Increment；

Each sampling instant, formula (1) is substituted into formula (2), and minimized in the case where meeting formula (3) and (4) Performance index function J, obtain optimal controlling increment sequence u_f：

Extract optimum control increment sequence u_fIn first piece of u_k+1, taken out as optimal lean solution flow valve and low pressure (LP) cylinder Steam valve door opening amount signal

Beneficial effects of the present invention are：The Multi model Predictive Controllers of the present invention have good global nonlinear Control Ability, CO after being burnt applied to thermal power station₂Trapping system is capable of the demand of the effective a wide range of variable working condition of adaptive system, fast track CO₂Capture rate setting value, improve CO₂The level of trapping system depth fast and flexible operation.

Brief description of the drawings

Fig. 1 is the Method And Principle schematic diagram of the present invention.

Fig. 2 is the non-linear finding schematic diagram of the present invention.

Fig. 3 is the membership function schematic diagram designed by the present invention.

Fig. 4 is that multiple model predictive control of the present invention (solid line) controls (dotted line) in CO with conventional proportional integral differential₂Trapping Control effect contrast schematic diagram under the change of rate setting value small range (dotted line is setting value).

Fig. 5 is multiple model predictive control of the present invention (solid line) and conventional linear PREDICTIVE CONTROL (dotted line) in CO₂Capture rate is set Control effect contrast schematic diagram under definite value wide variation (dotted line is setting value).

Embodiment

The Multi model Predictive Controllers of the present invention are applied to CO after certain 1MW fired power generating unit is burnt₂Trapping system system In simulation model, the target of control is under conditions of input constraint is met, makes CO₂Capture rate and reboiler temperature tracking setting Value, realizes CO₂The a wide range of variable parameter operation of trapping system.

CO after the burning of the present invention₂Trapping system Multi model Predictive Controllers, the forecast Control Algorithm is with based on chemistry CO after the burning of absorption₂Trapping system is controlled device, and lean solution valve opening and turbine low pressure cylinder draw gas valve opening to be System control input amount, CO₂Capture rate and reboiler temperature are system output quantity, are primarily based on subspace state space system identification, utilize system Data caused by system operation, the local state spatial model of system is established at different operating points；Then using gap metric Method investigates the nonlinear Distribution of controlled device, and then establishes predictive controller at suitable local operating point, and designs person in servitude Category degree function is by its weighted array, CO after foundation burning₂Trapping system multiple model predictive control system.Phase is controlled with Classical forecast Than the present invention improves CO₂The Control platform of a wide range of variable parameter operation of trapping system, enhance its ability flexibly run.

As shown in figure 1, CO after the burning of the present invention₂Trapping system Multi model Predictive Controllers, specifically include following step Suddenly：

Step 1, near a certain given capture rate operating point, design change in 30 seconds once, continue the poor liquid stream of 30000 seconds Measure valve opening signal u_aDrawn gas valve opening amount signal u with steam turbine low pressure (LP) cylinder_b, row energization is entered to system, obtains a series of CO₂Catch Collection rate y_aWith reboiler temperature y_bOpen-loop response data；

Step 2, sampling period T is selected_s=30s, withTo input,For output, profit With subspace state space system identification, CO at different capture rate operating points is built₂Trapping system local state spatial model, specific steps For：

A：Continuously obtain 1000 groups of output data Y and amplification input data U are arranged as Hankel matrix forms respectively (2N+j-2=1000)：

Wherein, N is matrix line number, take N=10；N is more than CO₂Trapping system order, j is matrix columns, in hardware condition It is the bigger the better in the case of permission, Y and U represent output and the Hankel matrixes of input data composition, Y respectively^fAnd Y^pRepresent respectively The Future Data and past data of output data, U^fAnd U^pThe Future Data and past data of input data, y are represented respectively_jRepresent J-th of output data, u_jRepresent j-th of input data；

B：Make W^p=[(Y^p)^T (U^p)^T]^T, QR decomposition is carried out to following matrix：

Obtain matrix L：

C：Obtain matrix L_w=L (:,1:N (m+l)), L_u=L (:,N(m+l)+1:End), m=2, m tie up for input variable Number, l=2, l are input/output variable dimension, L (:,1:N (m+l)) represent that L preceding N (m+l) is arranged, L (:,N(m+l)+1:end) Represent L from all row after the row of N (m+l)+1；

D：To L_wMatrix does singular value decomposition：

Obtain matrix Γ_N=U₁(S₁)^1/2, and then obtain：Model parameterWhereinL rows before expression is removed Γ_N,Γ _NThe Γ of rear l rows is removed in expression_N,Represent Moore-Penrose pseudoinverses；Model parameter C=Γ_N(1:l,:) can be from Γ_N Preceding l rows in directly obtain.

Subspace matrices l_w=L_w(1:l,:),l_u=L_u(1:l,1:m)；

E：Solve system of linear equations：

Obtain model parameter B and D.

Step 3, using the method for gap metric, the difference between each adjacent partial model is analyzed, investigation CO2 trapping systems Nonlinear Distribution, concretely comprise the following steps：

A：Partial model P adjacent to two₁、P₂Orthogonal fight coprime factorization is done, is obtained：

B：Computation model P₁、P₂Between distance：

Wherein, H_∞For a special Hardy normed spaces：σ_max(G (j ω)) represents G The maximum singular value of (j ω)；Q is H_∞Arbitrary function in space, takes Q=1.

C：Gap metric between two systems

D：Local operating point is selected according to the result of gap metric and designs membership function.In this example, its clearence degree is adjusted It is as shown in Figure 2 to grind result, it is seen that, in low capture rate and high capture rate section, mission nonlinear is stronger, and degree of membership letter is designed for this Number is as shown in Figure 3.

Step 4：Model predictive controller is established at suitable partial duty point.Each sampling instant, using formula (1) Estimate the CO of the system within following a period of time₂Capture rate and reboiler temperature

Wherein, prediction matrix ψ_x、ψ_u、ψ_yRespectively：

Respectively k moment CO₂Estimated state, input and the output of trapping system, F are observer gain.u_fFor Following N_uThe input data at individual moment, It is following N_yWhen etching system estimate it is defeated Go out,In this example, N is taken_u=2, N_y=100.

Modus ponens (2) is used as performance index function formula：

Wherein,It is the weight matrix for adjusting input and output Control platform,

r_fIt is following N_yWhen etching system CO2 capture rates and reboiler temperature setting value Sequence, Represent the k+1 moment to k+N respectively_yWhen etching system CO₂Capture rate r_aWith reboiler temperature r_bSetting value,

Etching system CO when being following Ny₂Capture rate and reboiler temperature estimate value sequence,

Represent the k+1 moment to k+N respectively_yWhen etching system CO₂Capture rate y_aWith reboiler temperature y_bEstimate Value,It can be described by formula (1), take N_y=10；Δu_fIt is following N_uThe lean solution flow valve opening amount signal at quarter and Low pressure (LP) cylinder draws gas valve opening amount signal sequenceIncrement, wherein N_u=2.

Consider CO₂Filters with Magnitude Constraints (the u of trapping system valve opening signal_min=[0.4 0.02]^T,u_max=[1 0.075 ]^T) and increment restriction (Δ u_min=[- 0.007-0.001]^T,Δu_max=[0.007 0.001]^T)：

Each sampling instant, (1) is substituted into performance indications formula (2), and minimized in the case where meeting to constrain (3) and (4) (2) local controlling increment sequence u, is obtained_f：Extract Partial controll increment sequence u_fIn First piece of u_k+1, lean solution flow valve and low pressure (LP) cylinder as part draw gas valve opening amount signal

Step 5, using membership function, each controller is exported into weighted array, final lean solution flow valve is obtained and opens Spend u_a-opDrawn gas valve opening amount signal u with turbine low pressure cylinder_b-opAnd act on CO2 trapping systems.Whereinω_iFor membership function corresponding to i-th of controller, It is scheduling variable, capture rate CR function, uⁱ _a-opAnd uⁱ _b-opThe optimum control signal calculated for i-th of controller.

Step 6, the prediction matrix ψ exported in fixed each local control_x、ψ_u、ψ_y, repeat step 3~4 is with the company of realization Continuous control.

The present embodiment is for CO after the burning in more of the invention₂Trapping system Multi model Predictive Controllers, conventional ratio The control effect of example integral differential control method and general forecast control method, has done two groups of l-G simulation tests：Emulation experiment 1, CO₂ The initial capture rate of trapping system is stable at 80%, in t=15min and 115min, CO₂Capture rate setting value is slow respectively from 80% 70% and 75% are changed to, it is constant that reboiler temperature setting value is maintained at 383K；Emulation experiment 2, CO₂Trapping system initially traps Rate is stable at 80%, t=15min and 115min, CO₂Capture rate setting value is slowly varying to 90% and 55% respectively from 80%, It is constant that reboiler temperature setting value is maintained at 383K.

As shown in figure 4, in CO₂In the case of capture rate setting value increaseds or decreases, the present invention is to CO after burning₂Trapping system Optimal control effect curve be substantially better than conventional proportional plus integral controller, there is satisfied setting value tracking and regulation energy Power.As shown in figure 5, in CO₂In the case of capture rate setting value wide variation, optimal control method of the invention can be more preferable Ground is coordinated to draw gas and poor flow quantity using reboiler, realizes the fast track control to capture rate, while it is possible to prevente effectively from big The controller concussion that scope variable parameter operation Linear Model with Side mismatch is brought, has more stably control effect, improves CO₂Catch The riding quality of collecting system.

CO after present invention burning₂Trapping system Multi model Predictive Controllers, CO is established using subspace state space system identification₂Catch Model of the collecting system under different operating points, on the basis of to the non-linear investigation of trapping system, select partial model, establish in advance Survey controller and design membership function, it is big to greatly improve system on the premise of all advantages of conventional linear PREDICTIVE CONTROL are possessed Scope variable parameter operation is horizontal, so as to further improve CO₂The ability that trapping system fast deep is flexibly run.

Although the present invention is illustrated and described with regard to preferred embodiment, it is understood by those skilled in the art that Without departing from scope defined by the claims of the present invention, variations and modifications can be carried out to the present invention.

Claims

1. CO after one kind burning₂The Multi model Predictive Controllers of trapping system, it is characterised in that comprise the following steps：

(1) CO after burning₂Trapping system is switched to manual mode, near different capture rate operating points, with lean solution flow valve Aperture u_aDrawn gas valve opening amount signal u with turbine low pressure cylinder_bTo input, to CO₂Trapping system enters row energization, obtains CO₂Trapping Rate y_aWith reboiler temperature y_bOpen-loop response data；

(2) sampling period Ts is selected, withTo input,For output, Subspace Identification is utilized Method, build CO at different capture rate operating points₂Trapping system local state spatial model；

(3) method for using gap metric, analyzes the difference between each adjacent partial model, investigates CO₂Non-linear point of trapping system Cloth；

(4) model predictive controller is established at suitable partial duty point；In each sampling instant, each submodel is utilized respectively Estimate the CO of the system within following certain time₂Capture rateAnd reboiler temperatureLocal optimum is calculated by optimization Lean solution flow valve apertureDrawn gas valve opening amount signal with turbine low pressure cylinder

(5) the final suitable membership function of design, each controller is exported into weighted array, obtains final lean solution flow valve Aperture u_a-opDrawn gas valve opening amount signal u with turbine low pressure cylinder_b-opAnd act on CO₂Trapping system；

Whereinω_iFor person in servitude corresponding to i-th of controller Category degree function, it is scheduling variable, capture rate CR function,WithThe optimum control calculated for i-th of controller Signal；

(6) the prediction matrix ψ exported in fixed each local control_x、ψ_u、ψ_y, continuous control is realized in repeat step (3)-(4).

2. CO after burning as claimed in claim 1₂The Multi model Predictive Controllers of trapping system, it is characterised in that step (2) in, T95/T_s=5~15, wherein T95 are the regulating time that transient process rises to 95%.

3. CO after burning as claimed in claim 1₂The Multi model Predictive Controllers of trapping system, it is characterised in that step (2) in, CO at different capture rate operating points is built₂Trapping system local state spatial model, is concretely comprised the following steps：(21) will be at certain The output data y and input data u from the 0th moment to 2N+j-2 moment that one given operating point nearby continuously obtains are arranged respectively It is classified as Hankel matrix forms：

<mrow> <mi>Y</mi> <mo>=</mo> <mo>&lsqb;</mo> <mfrac> <msup> <mi>Y</mi> <mi>p</mi> </msup> <msup> <mi>Y</mi> <mi>f</mi> </msup> </mfrac> <mo>&rsqb;</mo> <mo>=</mo> <mrow> <mo>&lsqb;</mo> <mfrac> <mtable> <mtr> <mtd> <msub> <mi>y</mi> <mn>0</mn> </msub> </mtd> <mtd> <msub> <mi>y</mi> <mn>1</mn> </msub> </mtd> <mtd> <mn>...</mn> </mtd> <mtd> <msub> <mi>y</mi> <mrow> <mi>j</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> </mtd> </mtr> <mtr> <mtd> <msub> <mi>y</mi> <mn>1</mn> </msub> </mtd> <mtd> <msub> <mi>y</mi> <mn>2</mn> </msub> </mtd> <mtd> <mn>...</mn> </mtd> <mtd> <msub> <mi>y</mi> <mi>j</mi> </msub> </mtd> </mtr> <mtr> <mtd> <mn>...</mn> </mtd> <mtd> <mn>...</mn> </mtd> <mtd> <mn>...</mn> </mtd> <mtd> <mn>...</mn> </mtd> </mtr> <mtr> <mtd> <msub> <mi>y</mi> <mrow> <mi>N</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> </mtd> <mtd> <msub> <mi>y</mi> <mi>N</mi> </msub> </mtd> <mtd> <mn>...</mn> </mtd> <mtd> <msub> <mi>y</mi> <mrow> <mi>N</mi> <mo>+</mo> <mi>j</mi> <mo>-</mo> <mn>2</mn> </mrow> </msub> </mtd> </mtr> </mtable> <mtable> <mtr> <mtd> <msub> <mi>y</mi> <mi>N</mi> </msub> </mtd> <mtd> <msub> <mi>y</mi> <mrow> <mi>N</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> </mtd> <mtd> <mn>...</mn> </mtd> <mtd> <msub> <mi>y</mi> <mrow> <mi>N</mi> <mo>+</mo> <mi>j</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> </mtd> </mtr> <mtr> <mtd> <msub> <mi>y</mi> <mrow> <mi>N</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> </mtd> <mtd> <msub> <mi>y</mi> <mrow> <mi>N</mi> <mo>+</mo> <mn>2</mn> </mrow> </msub> </mtd> <mtd> <mn>...</mn> </mtd> <mtd> <msub> <mi>y</mi> <mrow> <mi>N</mi> <mo>+</mo> <mi>j</mi> </mrow> </msub> </mtd> </mtr> <mtr> <mtd> <mn>...</mn> </mtd> <mtd> <mn>...</mn> </mtd> <mtd> <mn>...</mn> </mtd> <mtd> <mn>...</mn> </mtd> </mtr> <mtr> <mtd> <msub> <mi>y</mi> <mrow> <mn>2</mn> <mi>N</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> </mtd> <mtd> <msub> <mi>y</mi> <mrow> <mn>2</mn> <mi>N</mi> </mrow> </msub> </mtd> <mtd> <mn>...</mn> </mtd> <mtd> <msub> <mi>y</mi> <mrow> <mn>2</mn> <mi>N</mi> <mo>+</mo> <mi>j</mi> <mo>-</mo> <mn>2</mn> </mrow> </msub> </mtd> </mtr> </mtable> </mfrac> <mo>&rsqb;</mo> </mrow> <mo>,</mo> </mrow> 1

Wherein, N is matrix line number, and N is more than CO₂Trapping system order, j are matrix columns, and Y and U represent output and input number respectively According to the Hankel matrixes of composition, Y^fAnd Y^pThe Future Data and past data of output data, U are represented respectively^fAnd U^pRepresent respectively defeated Enter the Future Data and past data of data, y_jRepresent j-th of output data, u_jRepresent j-th of input data；

Obtain matrix L：

(23) so as to obtaining matrix L_w=L (:,1:N (m+l)), L_u=L (:,N(m+l)+1:End), m is input variable dimension, l For output variable dimension, L (:,1:N (m+l)) representing matrix L preceding N (m+l) row, L (:,N(m+l)+1:End) representing matrix L All row from after the row of N (m+l)+1；

(24) to L_wMatrix does singular value decomposition：

Obtain matrix Γ_N=U₁(S₁)^1/2, and then obtain：Model parameterWhereinRepresent l rows before removing Γ_N,Γ _NThe Γ of rear l rows is removed in expression_N,Represent Moore-Penrose pseudoinverses；Model parameter C=Γ_N(1:l,:) can be from Γ_N Preceding l rows in directly obtain；

(25) system of linear equations is solved：

Obtain model parameter B and D.

4. CO after burning as claimed in claim 1₂The Multi model Predictive Controllers of trapping system, it is characterised in that step (3) in, the clearence degree value between adjacent partial model is calculated, is concretely comprised the following steps：

(32) computation model P₁、P₂Between distance：

<mrow> <mover> <mi>&delta;</mi> <mo>&RightArrow;</mo> </mover> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mn>1</mn> </msub> <mo>,</mo> <msub> <mi>P</mi> <mn>2</mn> </msub> <mo>)</mo> </mrow> <mo>=</mo> <munder> <mrow> <mi>i</mi> <mi>n</mi> <mi>f</mi> </mrow> <mrow> <mi>Q</mi> <mo>&Element;</mo> <msub> <mi>H</mi> <mi>&infin;</mi> </msub> </mrow> </munder> <mo>|</mo> <mo>|</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <msub> <mi>N</mi> <mn>1</mn> </msub> </mtd> </mtr> <mtr> <mtd> <msub> <mi>D</mi> <mn>1</mn> </msub> </mtd> </mtr> </mtable> </mfenced> <mo>-</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <msub> <mi>N</mi> <mn>2</mn> </msub> </mtd> </mtr> <mtr> <mtd> <msub> <mi>D</mi> <mn>2</mn> </msub> </mtd> </mtr> </mtable> </mfenced> <mi>Q</mi> <mo>|</mo> <msub> <mo>|</mo> <mi>&infin;</mi> </msub> <mo>,</mo> </mrow>

Wherein, H_∞For a special Hardy normed spaces：σ_max(G (j ω)) represents G (j Maximum singular value ω)；Q is H_∞Arbitrary function in space；

Gap metric between (33) two systems

5. CO after burning as claimed in claim 1₂The Multi model Predictive Controllers of trapping system, it is characterised in that step (4) in, the CO of the system within following a period of time is estimated using formula (1)₂Capture rate and reboiler temperature

Wherein, prediction matrix ψ_x、ψ_u、ψ_yRespectively：

Respectively k moment CO₂Estimated state, input and the output of trapping system, F are observer gain, u_fFor future The input data at Nu moment, It is following N_yWhen etching system estimate output,

Using equation below calculation of performance indicators function J：

<mrow> <mi>J</mi> <mo>=</mo> <msup> <mrow> <mo>(</mo> <msub> <mover> <mi>y</mi> <mo>^</mo> </mover> <mi>f</mi> </msub> <mo>-</mo> <msub> <mi>r</mi> <mi>f</mi> </msub> <mo>)</mo> </mrow> <mi>T</mi> </msup> <msub> <mi>Q</mi> <mi>f</mi> </msub> <mrow> <mo>(</mo> <msub> <mover> <mi>y</mi> <mo>^</mo> </mover> <mi>f</mi> </msub> <mo>-</mo> <msub> <mi>r</mi> <mi>f</mi> </msub> <mo>)</mo> </mrow> <mo>+</mo> <msubsup> <mi>&Delta;u</mi> <mi>f</mi> <mi>T</mi> </msubsup> <msub> <mi>R</mi> <mi>f</mi> </msub> <msub> <mi>&Delta;u</mi> <mi>f</mi> </msub> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> <mo>;</mo> </mrow>

Wherein, Q_fAnd R_fIt is the weight matrix for adjusting input and output Control platform,r_fIt is future N_yWhen etching system CO₂Capture rate and reboiler temperature setting value sequence,

<mrow> <msub> <mi>r</mi> <mi>f</mi> </msub> <mo>=</mo> <msup> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <msubsup> <mi>r</mi> <mrow> <mi>k</mi> <mo>+</mo> <mn>1</mn> </mrow> <mi>T</mi> </msubsup> </mtd> <mtd> <msubsup> <mi>r</mi> <mrow> <mi>k</mi> <mo>+</mo> <mn>2</mn> </mrow> <mi>T</mi> </msubsup> </mtd> <mtd> <mo>...</mo> </mtd> <mtd> <msubsup> <mi>r</mi> <mrow> <mi>k</mi> <mo>+</mo> <mi>N</mi> <mi>y</mi> </mrow> <mi>T</mi> </msubsup> </mtd> </mtr> </mtable> </mfenced> <mi>T</mi> </msup> <mo>,</mo> </mrow> 3

Represent the k+1 moment to k+N respectively_yWhen etching system CO₂Capture rate r_aWith reboiler temperature r_bSetting value, It is following N_yWhen etching system CO₂Capture rate and reboiler temperature estimate value sequence,

Represent the k+1 moment to k+N respectively_yWhen etching system CO₂Capture rate y_aWith reboiler temperature y_bDiscreet value,

<mrow> <msub> <mi>&Delta;u</mi> <mi>f</mi> </msub> <mo>=</mo> <mi>&psi;</mi> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <msub> <mi>u</mi> <mi>k</mi> </msub> </mtd> </mtr> <mtr> <mtd> <msub> <mi>u</mi> <mi>f</mi> </msub> </mtd> </mtr> </mtable> </mfenced> <mo>,</mo> <mi>&psi;</mi> <mo>=</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <mrow> <mo>-</mo> <msub> <mi>I</mi> <mi>m</mi> </msub> </mrow> </mtd> <mtd> <msub> <mi>I</mi> <mi>m</mi> </msub> </mtd> <mtd> <mn>0</mn> </mtd> <mtd> <mo>...</mo> </mtd> <mtd> <mn>0</mn> </mtd> </mtr> <mtr> <mtd> <mo>...</mo> </mtd> <mtd> <mrow> <mo>-</mo> <msub> <mi>I</mi> <mi>m</mi> </msub> </mrow> </mtd> <mtd> <msub> <mi>I</mi> <mi>m</mi> </msub> </mtd> <mtd> <mn>0</mn> </mtd> <mtd> <mo>...</mo> </mtd> </mtr> <mtr> <mtd> <mo>...</mo> </mtd> <mtd> <mo>...</mo> </mtd> <mtd> <mo>...</mo> </mtd> <mtd> <mo>...</mo> </mtd> <mtd> <mo>...</mo> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mo>...</mo> </mtd> <mtd> <mo>...</mo> </mtd> <mtd> <mrow> <mo>-</mo> <msub> <mi>I</mi> <mi>m</mi> </msub> </mrow> </mtd> <mtd> <msub> <mi>I</mi> <mi>m</mi> </msub> </mtd> </mtr> </mtable> </mfenced> <mo>,</mo> </mrow>

Wherein

CO2 trapping system lean solution flow valves and low pressure (LP) cylinder draw gas valve opening amount signal u Filters with Magnitude Constraints (u_min,u_max) and increment Constrain (Δ u_min,Δu_max) be：

<mrow> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <msub> <mi>I</mi> <mi>m</mi> </msub> </mtd> </mtr> <mtr> <mtd> <msub> <mi>I</mi> <mi>m</mi> </msub> </mtd> </mtr> <mtr> <mtd> <mo>.</mo> </mtd> </mtr> <mtr> <mtd> <mo>.</mo> </mtd> </mtr> <mtr> <mtd> <mo>.</mo> </mtd> </mtr> <mtr> <mtd> <msub> <mi>I</mi> <mi>m</mi> </msub> </mtd> </mtr> </mtable> </mfenced> <msub> <mi>&Delta;u</mi> <mi>min</mi> </msub> <mo>&le;</mo> <mi>&psi;</mi> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <msub> <mi>u</mi> <mi>k</mi> </msub> </mtd> </mtr> <mtr> <mtd> <msub> <mi>u</mi> <mi>f</mi> </msub> </mtd> </mtr> </mtable> </mfenced> <mo>&le;</mo> <mfenced open = "[" close = "]"> <mtable> <mtr> <mtd> <msub> <mi>I</mi> <mi>m</mi> </msub> </mtd> </mtr> <mtr> <mtd> <msub> <mi>I</mi> <mi>m</mi> </msub> </mtd> </mtr> <mtr> <mtd> <mo>.</mo> </mtd> </mtr> <mtr> <mtd> <mo>.</mo> </mtd> </mtr> <mtr> <mtd> <mo>.</mo> </mtd> </mtr> <mtr> <mtd> <msub> <mi>I</mi> <mi>m</mi> </msub> </mtd> </mtr> </mtable> </mfenced> <msub> <mi>&Delta;u</mi> <mrow> <mi>m</mi> <mi>a</mi> <mi>x</mi> </mrow> </msub> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>4</mn> <mo>)</mo> </mrow> <mo>,</mo> </mrow>

Wherein, u_min,u_maxRepresent respectively lean solution flow valve and low pressure (LP) cylinder draw gas valve opening amount signal u minimum value with it is maximum Value, Δ u_min,Δu_maxRepresent respectively lean solution flow valve and low pressure (LP) cylinder draw gas valve opening amount signal u smallest incremental with most increasing Amount；

Each sampling instant, formula (1) is substituted into formula (2), and performance is minimized in the case where meeting formula (3) and (4) Target function J, obtain optimal controlling increment sequence u_f：

Extract optimum control increment sequence u_fIn first piece of u_k+1, drawn gas valve as optimal lean solution flow valve and low pressure (LP) cylinder Opening amount signal