Method and device for designing model predictive controller for frequency modulation participated by IGV of gas-steam combined cycle unit, and medium

Abstract

The invention discloses a method and equipment for designing a model prediction controller for frequency modulation participated by an IGV of a gas-steam combined cycle unit, and a medium, and belongs to the technical field of automatic control, and the method comprises the steps: obtaining the operation data of the unit; unit design parameters are obtained; based on the unit operation data, utilizing a subspace identification method to identify and obtain a unit state space model for controller design; and on the basis of the unit state space model, a model prediction controller is designed based on the unit design parameters and used for controlling the output power of the unit. According to the method, the capacity of tracking load change of the whole system is improved at the cost of slightly weakening the control effect of the exhaust temperature of the gas turbine in a short time, and assistance is provided for continuous load change of the system in an AGC mode.

Description

Technical Field

[0001] The present invention relates to automatic control technology, and in particular to a model predictive controller design method, equipment and medium for IGV frequency modulation of a gas-steam combined cycle unit. Background Art

[0002] PID control is currently the most widely used control method for gas-steam combined cycle units, which can meet basic setpoint requirements. However, AGC frequency regulation is widely used in gas-steam combined cycle units. Introducing advanced control methods such as model predictive control can improve overall control effectiveness.

[0003] The core of AGC frequency regulation is to quickly track load requirements. Therefore, improving the response rate of unit power while ensuring safe and stable operation is a key factor in controller design.

[0004] Currently, some model predictive controllers used in gas-steam combined cycle units primarily utilize fuel quantity to control power, and the turbine exhaust temperature is controlled by the turbine inlet guide vane opening. To ensure turbine efficiency, the turbine exhaust temperature must track the setpoint. However, to balance this control, the controller, as a multi-input, multi-output controller, reduces its power control effectiveness. This limits the unit's power response rate and, in turn, reduces the unit's ability to participate in AGC frequency regulation. Summary of the Invention

[0005] In response to the deficiencies in the prior art, the present invention provides a method, device, and medium for designing a model predictive controller for the IGV (intake guide vane) participation in frequency modulation of a gas-steam combined cycle unit. By reducing the exhaust temperature deviation limit within an allowable range, the regulation of the power tracking deviation by the opening of the gas turbine inlet guide vane is strengthened to increase the power response rate, thereby improving the unit's ability to participate in AGC frequency modulation.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for designing a model predictive controller for frequency modulation of an IGV in a gas-steam combined cycle unit includes the following steps:

[0008] Obtain unit operation data;

[0009] Obtain unit design parameters;

[0010] Based on the unit operating data, the subspace identification method is used to identify the unit state space model for controller design;

[0011] Based on the unit state space model and the unit design parameters, a model predictive controller is designed to control the unit output power.

[0012] To optimize the above technical solutions, specific measures taken also include:

[0013] Furthermore, the unit operation data includes the amount of fuel input during the unit operation, the opening of the gas turbine inlet guide vanes during the unit operation, the total power output during the unit operation, and the temperature of the gas turbine exhaust during the unit operation.

[0014] Furthermore, the unit design parameters include the fuel quantity allowed by the unit, the upper and lower limits of the opening of the gas turbine inlet guide vanes, and the upper and lower limits of the change rate of the fuel quantity and the opening of the gas turbine inlet guide vanes.

[0015] Furthermore, the specific process of using the subspace identification method to identify the unit state space model for controller design is as follows:

[0016] The expression defining the state space model is:

[0017]

[0018] Where x k+1 is the state vector at time k+1, x k is the state vector at time k, A a represents the state transition matrix, B b represents the input matrix, u k is the input vector, K represents the Kalman filter gain, e k represents the observation noise, y k represents the output vector, C c Denotes the output matrix, D d represents the direct transfer matrix;

[0019] The input vector is represented as follows:

[0020]

[0021] In the formula, Mf k Indicates fuel quantity, IGV k Indicates the opening of the inlet guide vane, R 2 represents the two-dimensional real number field;

[0022] The output vector is represented as follows:

[0023]

[0024] Where Pall k Indicates output power, TX k Indicates exhaust temperature;

[0025] Identify the state transfer matrix A using the subspace identification method a , input matrix B b , output matrix C c and direct transfer matrix D d .

[0026] Furthermore, the subspace identification method is specifically as follows:

[0027] The input data include: the amount of fuel input during the operation of the unit, the opening of the gas turbine inlet guide vanes during the operation of the unit, and the output data used include the total power output during the operation of the unit, and the temperature of the gas turbine exhaust during the operation of the unit;

[0028] Construct the Hankel matrix U of the input data std :

[0029]

[0030] Where U p is the block matrix of the past moment, U f is the block matrix of the future moment;

[0031] The Hankel matrix Y of the output data std Also divided into the block matrix Y of the past moment p and the future block matrix Y f ;

[0032] Y f =Γ nn X f +H d U f +H s E f (1-3)

[0033] Y p =Γ nn X p +H d U p +H s E p (1-4)

[0034] Where, Γ nn Represents the output coefficient matrix, X f represents the future state sequence, H d Represents the dynamic influence matrix from input to output, U f is the block matrix of the future time, H s Represents the dynamic impact matrix of noise to output, E f represents the noise at the future time, Xp Represents the past state sequence, U p represents the past time block matrix, E p represents the noise at the past moment;

[0035] The state matrix X std Defined as:

[0036]

[0037] X p represents the past state sequence, X f Represents a sequence of future states;

[0038] X f =L nn W p (1-6)

[0039] Where, L nn represents the subspace matrix, W p Represents the data matrix of the past moment;

[0040] Substituting formula (1-6) into formula (1-3) yields:

[0041] Y f =L w X f +L u U f +L e E f (1-7)

[0042] Where, L w represents the state weight matrix, L u represents the input influence matrix, L e represents the noise dynamic matrix;

[0043] The Hankel matrix U of the input data std Do the QR decomposition as follows:

[0044]

[0045] Where R 11 Represents the upper triangular block of the input data matrix after orthogonal projection, R 21 Represents the sub-block of cross-projection of input and output data, R 31 represents the associated block between the input and residual subspace, R 22 represents the main projection block of the output data matrix, R 32 represents the coupling block between the output and residual subspace, R 33represents the upper triangular block of the residual subspace, Q1 represents the orthogonal basis of the input data, Q2 represents the orthogonal basis of the output data, and Q3 represents the orthogonal basis of the residual subspace;

[0046] Comparing Equation (1-8) with Equation (1-7), we can calculate the subspace joint description matrix L = (L w L u )as follows:

[0047]

[0048] The state transfer matrix A is obtained from the subspace joint description matrix a , input matrix B b , output matrix C c and direct transfer matrix D d .

[0049] Furthermore, the model predictive controller is designed based on the unit design parameters as follows:

[0050] The number of inputs, outputs, and states of the gas-steam combined cycle unit are r, m, and n respectively. The unit state space model is rewritten into a discrete form:

[0051]

[0052] Among them, x m (k), u(k), y(k) are the n-dimensional state variables, r-dimensional input variables and m-dimensional output variables of the unit at time k respectively. m represents the discrete state transfer matrix, B m represents the discrete input matrix, C m represents the discrete output matrix; x m (k+1) is the n-dimensional state variable of the unit at time k+1;

[0053] A model predictive controller is constructed based on the discrete state space model of the unit. The expression of the model predictive controller is as follows:

[0054] Y=Fx(k)+ΦΔU (2-7)

[0055] Where Y represents the output matrix at time k, F represents the terminal cost weight matrix, x(k) represents the n+m-dimensional augmented state variable, and x(k) = (Δx m (k) T y(k) T ) T , Δx m(k) represents the increment of the state variable at time k, y(k) represents the m-dimensional output variable of the unit at time k, Φ represents the dynamic response matrix that quantifies the cumulative impact, and ΔU represents the input quantity increment matrix under the premise of time k, which is used as the control sequence;

[0056] The output matrix Y at time k is as follows:

[0057] Y=(y(k+1|k)y(k+2|k)…y(k+N p |k)) T

[0058] y(k+N p |k) represents k+N at time k p The output of the unit at this moment, N p is the prediction step length;

[0059] The input increment matrix ΔU at time k is as follows:

[0060] ΔU=(Δu(k)Δu(k+1)…Δu(k+N c -1)) T

[0061] Δu(k+N c -1) means k+N c -1 time input increment, N c To control the step length;

[0062] Design the objective function, take the minimum value of the objective function as the goal, and solve the optimal control sequence;

[0063] The objective function is designed as:

[0064]

[0065] Where, J QP represents the objective function, ΔU represents the input increment matrix, Φ represents the dynamic response matrix that quantifies the cumulative impact, Q represents the weighted matrix of the output error, and L represents the weighted matrix of the control increment; R represents the input weight matrix, F represents the terminal cost weight matrix, and x(k) represents the n+m-dimensional augmented state variable;

[0066] Write the current input as the previous input plus the increment, and the input is:

[0067] U=A QP ΔU+u(k-1)B QP (2-12)

[0068] U represents the input quantity, ΔU represents the input quantity increment matrix, A QP represents the QP decomposition incremental coefficient matrix, BQP represents the QP decomposition input coefficient matrix; u(k-1) represents the input at the previous moment;

[0069] The constraints of the objective function are:

[0070]

[0071] U min Indicates the minimum value allowed for the input quantity, U max Indicates the maximum value allowed for the input;

[0072] The condition for the objective function to achieve the minimum value is Substitute this condition into the objective function (2-12) and solve it to obtain the optimal control sequence at time k:

[0073] ΔU=(Φ T QΦ+L) -1 Φ T Q(R-Fx(k)) (2-15)

[0074] Where R represents the input weight matrix, R = (r(k+1)r(k+2)…r(k+N p )), r(k+N p ) represents k+N p Input weight value at any time, N p is the prediction step length;

[0075] Only the first increment Δu(k) in the optimal control sequence will be applied, so the control input increment at time k is:

[0076] Δu(k)=(I m 0…0)(Φ T QΦ+L) -1 Φ T Q(R-Fx(k)) (2-16)

[0077] Where, I m Represents the m-dimensional identity matrix.

[0078] Furthermore, in the terminal cost weight matrix F of the model predictive controller, the weight of the total output power is set higher than the weight of the gas turbine exhaust temperature.

[0079] The present invention also proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the model predictive controller design method for the IGV participating in frequency modulation of the gas-steam combined cycle unit as described above is implemented.

[0080] The present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the above-mentioned model predictive controller design method for IGV-participated frequency modulation of a gas-steam combined cycle unit.

[0081] The beneficial effects of the present invention are:

[0082] When a unit operates in AGC mode, it receives variable load commands with short intervals and small amplitude changes, requiring the unit to have rapid response capabilities. This invention collects and analyzes unit design and operating data to design a unit predictive controller. By adjusting the predictive controller's weight matrix parameters, the unit's response rate to load commands is increased at the expense of reduced temperature control capability within a short period of time, thereby improving the unit's frequency regulation capability in AGC mode. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 : A flow chart of a method for designing a model predictive controller for frequency modulation of an IGV of a gas-steam combined cycle unit according to the present invention.

[0084] Figure 2 : Schematic diagram of gas-steam combined cycle unit model.

[0085] Figure 3: Input and output data diagram for recognition. Figure 3a Enter the data map for the fuel quantity, Figure 3b Input data graph for opening margin, Figure 3c is the total power output data diagram, Figure 3d Output data plot for gas turbine exhaust temperature.

[0086] Figure 4 : AGC frequency modulation instruction.

[0087] Figure 5: Comparison of controller control effects. Figure 5a This is a fuel quantity comparison chart. Figure 5b This is the IGV comparison chart. Figure 5c This is a power comparison chart. Figure 5d A comparison chart of gas turbine exhaust temperatures. DETAILED DESCRIPTION

[0088] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0089] Example 1

[0090] The present invention proposes a model predictive controller design method for IGV frequency modulation of a gas-steam combined cycle unit. The overall process of the method is as follows: Figure 1 As shown, the following steps are included:

[0091] The unit operation data includes the amount of fuel input during the unit operation, the opening of the gas turbine inlet guide vanes during the unit operation, the total power output during the unit operation, and the temperature of the gas turbine exhaust during the unit operation.

[0092] The unit design parameters are obtained; the unit design parameters include the allowable fuel quantity of the unit, the upper and lower limits of the opening of the gas turbine inlet guide vane, and the upper and lower limits of the change rate of the fuel quantity and the opening of the gas turbine inlet guide vane.

[0093] Based on the unit operation data, the subspace identification method is used to identify the unit state space model for controller design. The specific process of using the subspace identification method to identify the unit state space model for controller design is as follows:

[0094] The expression defining the state space model is:

[0095]

[0096] Where x k+1 is the state vector at time k+1, x k is the state vector at time k, A a represents the state transition matrix, B b represents the input matrix, u k is the input vector, K represents the Kalman filter gain, e k represents the observation noise, y k represents the output vector, C c Denotes the output matrix, D d represents the direct transfer matrix;

[0097] The input vector is represented as follows:

[0098]

[0099] In the formula, Mf k Indicates fuel quantity, IGV k Indicates the opening of the inlet guide vane, R 2 represents the two-dimensional real number field;

[0100] The output vector is represented as follows:

[0101]

[0102] Where Pall k Indicates output power, TXk Indicates exhaust temperature;

[0103] Identify the state transfer matrix A using the subspace identification method a , input matrix B b , output matrix C c and direct transfer matrix D d The subspace identification method is as follows:

[0104] The input data include: the amount of fuel input during the operation of the unit, the opening of the gas turbine inlet guide vanes during the operation of the unit, and the output data used include the total power output during the operation of the unit, and the temperature of the gas turbine exhaust during the operation of the unit;

[0105] Construct the Hankel matrix U of the input data std :

[0106]

[0107] Where U p is the block matrix of the past moment, U f is the block matrix of the future moment; nn and mm are matrices U p and U f The number of rows and columns.

[0108] The Hankel matrix Y of the output data std Also divided into the block matrix Y of the past moment p and the future block matrix Y f ;

[0109] Y f =Γ nn X f +H d U f +H s E f (1-3)

[0110] Y p =Γ nn X p +H d U p +H s E p (1-4)

[0111]

[0112] Where, Γ nn Represents the output coefficient matrix, X f represents the future state sequence, H d Represents the dynamic influence matrix from input to output, U fis the block matrix of the future time, H s Represents the dynamic impact matrix of noise to output, E f represents the noise at the future time, X p Represents the past state sequence, U p represents the past time block matrix, E p represents the noise at the past moment; Y represents the output constraint matrix, Ψ U represents the input constraint matrix, Represents the estimated value of the state transition matrix.

[0113] Output coefficient matrix Γ nn is defined as follows:

[0114]

[0115] C c represents the output matrix, A a State transition matrix.

[0116] Dynamic influence matrix H from input to output d is defined as follows:

[0117]

[0118] A a represents the state transition matrix, B b represents the input matrix, C c Denotes the output matrix, D d Denotes the direct transfer matrix. The dynamic influence matrix H of noise to output s is defined as follows:

[0119]

[0120] Output constraint matrix Ψ Y is defined as follows:

[0121]

[0122] Input constraint matrix Ψ U is defined as follows:

[0123] Represents the estimated value of the input matrix.

[0124] The state matrix X std Defined as:

[0125]

[0126] X p represents the past state sequence, X fRepresents a sequence of future states;

[0127] When nn is greater than 10, formula (1-4) converges approximately to formula (1-3), and we have:

[0128] X f =L nn W p (1-6)

[0129] Where, L nn represents the subspace matrix, W p Represents the past moment data matrix;

[0130] L nn =(Ψ Y Ψ U )

[0131] W p =((Y p ) T (U p ) T ) T

[0132] Substituting formula (1-6) into formula (1-3) yields:

[0133] Y f =L w X f +L u U f +L e E f (1-7)

[0134] Where, L w represents the state weight matrix, L u represents the input influence matrix, L e represents the noise dynamic matrix;

[0135] L w =Γ nn L nn ,L u =H d ,L e =H s

[0136] The Hankel matrix U of the input data std Do the QR decomposition as follows:

[0137]

[0138] Where R 11 Represents the upper triangular block of the input data matrix after orthogonal projection, R 21Represents the sub-block of cross-projection of input and output data, R 31 represents the associated block between the input and residual subspace, R 22 represents the main projection block of the output data matrix, R 32 represents the coupling block between the output and residual subspace, R 33 represents the upper triangular block of the residual subspace, Q1 represents the orthogonal basis of the input data, Q2 represents the orthogonal basis of the output data, and Q3 represents the orthogonal basis of the residual subspace;

[0139] Comparing Equation (1-8) with Equation (1-7), we can calculate the subspace joint description matrix L = (L w L u )as follows:

[0140]

[0141] The state transfer matrix A is obtained from the subspace joint description matrix a , input matrix B b , output matrix C c and direct transfer matrix D d The details are as follows:

[0142] For the state weight matrix L w Perform singular value decomposition and get:

[0143]

[0144] U1 represents the first l columns of the left singular vector matrix, U2 represents the remaining columns of the left singular vector matrix, S1 represents the diagonal matrix composed of l main singular values, S2 represents the diagonal matrix composed of the remaining singular values, V1 represents the first l rows of the right singular vector matrix, V2 represents the remaining rows of the right singular vector matrix, and l is the model order.

[0145] Combining formula (1-7), we can get:

[0146]

[0147] By Γ nn Get the state transfer matrix A a And the output matrix C c .

[0148] From formula (1-3) we get

[0149]

[0150] (·) ⊥ represents the orthogonal complementary projection of the object, Represents the pseudo-inverse of an object.

[0151] Combining equations (1-3), (1-4), and (1-12), we get the input matrix B b and direct transfer matrix D d .

[0152] Based on the unit state space model, a model predictive controller is designed based on the unit design parameters to control the unit output power. The specific steps of designing a model predictive controller based on the unit design parameters are as follows:

[0153] The number of inputs, outputs, and states of the gas-steam combined cycle unit are r, m, and n respectively. The unit state space model is rewritten into a discrete form:

[0154]

[0155] Among them, x m (k), u(k), y(k) are the n-dimensional state variables, r-dimensional input variables and m-dimensional output variables of the unit at time k respectively. m represents the discrete state transfer matrix, B m represents the discrete input matrix, C m represents the discrete output matrix; x m (k+1) is the n-dimensional state variable of the unit at time k+1.

[0156] Since D in the system model m are all zero matrices and can therefore be omitted.

[0157] In this embodiment, r=2, m=2, n=1, where 1 is the model order of subspace identification.

[0158] The input variables are:

[0159]

[0160] The output variables are:

[0161]

[0162] Mf k Indicates fuel quantity, IGV k Indicates the opening of the inlet guide vane, Pall k Indicates output power, TX k Indicates the exhaust temperature, R 2 represents the two-dimensional real number field;

[0163] Define Δx m (k+1)=x m (k+1)-x m (k) and Δu(k)=u(k)-u(k-1), we get:

[0164] Δxm (k+1)=A m Δx m (k)+B m Δu(k) (2-2)

[0165] Δx m (k+1) represents the increment of the state variable at time k+1, x m (k+1) represents the state variable of the system at time k+1, Δu(k) represents the increment of the input at time k+1, and u(k-1) represents the input of the system at time k-1.

[0166] The increment of the output variable can be expressed as:

[0167] Δy(k+1)=C m A m Δx m (k)+C m B m Δu(k) (2-3)

[0168] Δx m (k) represents the increment of the state variable at time k

[0169] In the formula, Δy(k+1)=y(k+1)-y(k)

[0170] Define n+m dimensional augmented state variable x(k)

[0171] x(k)=(Δx m (k) T y(k) T ) T

[0172] The discrete state space equation can be rewritten as:

[0173]

[0174] In the formula

[0175] C=(0I q ), I q is the identity matrix.

[0176] A represents the augmented discrete state transfer matrix, B represents the augmented discrete input matrix, and C represents the augmented discrete output matrix.

[0177] Define r(k) as the expected output value, N c To control the step size, N p To predict the step length, the input increment and state sequence at the future moment are recorded as

[0178] Δu(k),Δu(k+1),…,Δu(k+N c -1)

[0179] x(k+1|k),x(k+2|k),…,x(k+N p |k)

[0180] Δu(k+N c -1) means k+N c -1 time input increment, x(k+N p |k) represents k+N at time k p The state variables of the system at time t.

[0181] get:

[0182]

[0183] From formula (2-5), the future output can be obtained as

[0184]

[0185] y(k+N p |k) represents k+N at time k p The output of the system at that moment.

[0186] definition:

[0187] Y=(y(k+1|k)y(k+2|k)…y(k+N p |k)) T

[0188] ΔU=(Δu(k)Δu(k+1)…Δu(k+N c -1)) T

[0189] Y represents the output matrix under the premise of k time, and ΔU represents the input increment matrix under the premise of k time.

[0190] A model predictive controller is constructed based on the discrete state space model of the unit. The expression of the model predictive controller is as follows:

[0191] Y=Fx(k)+ΦΔU (2-7)

[0192] Where Y represents the output matrix at time k, and F represents the terminal cost weight matrix. In the terminal cost weight matrix F of the model predictive controller, the power weight is set to a high value, and the gas turbine exhaust temperature weight is set to a small value close to zero. At the cost of reducing the temperature control ability in a short period of time, the response rate of the unit to the load command is improved, thereby improving the frequency regulation ability of the unit in the AGC mode. x(k) represents the n+m-dimensional augmented state variable, x(k)=(Δx m (k) T y(k) T ) T , Δx m (k) represents the increment of the state variable at time k, y(k) represents the m-dimensional output variable of the unit at time k, Φ represents the dynamic response matrix that quantifies the cumulative impact, and ΔU represents the input quantity increment matrix under the premise of time k, which is used as the control sequence;

[0193] The output matrix Y at time k is as follows:

[0194] Y=(y(k+1|k)y(k+2|k)…y(k+N p |k)) T

[0195] y(k+N p |k) represents k+N at time k p The output of the unit at this moment, N p is the prediction step length;

[0196] The input increment matrix ΔU at time k is as follows:

[0197] ΔU=(Δu(k)Δu(k+1)…Δu(k+N c -1)) T

[0198] Δu(k+N c -1) means k+N c -1 time input increment, N c To control the step length;

[0199] Design the objective function, take the minimum value of the objective function as the goal, and solve the optimal control sequence;

[0200] Define the objective function J as:

[0201] J=Y T QY+ΔU T LΔU (2-8)

[0202] Where Q represents the weighted matrix of the output error, and L is the weighted matrix of the control increment.

[0203] It is necessary to consider the constraints of the input variables, which are:

[0204]

[0205] U min Indicates the minimum value allowed for the input quantity, U indicates the input quantity, U max Indicates the maximum value allowed for the input quantity, ΔU min Indicates the minimum value allowed for the input change, ΔU max Indicates the maximum value allowed for the input variable.

[0206] Then update the objective function expression to:

[0207]

[0208] Where, J QP represents the objective function, ΔU represents the input increment matrix, Φ represents the dynamic response matrix that quantifies the cumulative impact, Q represents the weighted matrix of the output error, and L represents the weighted matrix of the control increment; R represents the input weight matrix, F represents the terminal cost weight matrix, and x(k) represents the n+m-dimensional augmented state variable;

[0209] Write the current input as the previous input plus the increment, and the input is:

[0210]

[0211] The simplified form of formula (2-13) is:

[0212] U=A QP ΔU+u(k-1)B QP (2-12)

[0213] U represents the input quantity, ΔU represents the input quantity increment matrix, A QP represents the QP decomposition incremental coefficient matrix, B QP Represents the QP decomposition input coefficient matrix; u(k-1) represents the input amount at the previous moment.

[0214]

[0215] The constraints of the objective function are rewritten as:

[0216]

[0217] U min Indicates the minimum value allowed for the input quantity, U max Indicates the maximum value allowed for the input;

[0218] Formula (2-12) and Formula (2-15) can be written in simplified form as follows:

[0219]

[0220] Where H is the coefficient matrix of the quadratic term of the objective function, η is the coefficient vector of the linear term of the objective function, ψ is the coefficient matrix of the inequality constraint, and γ is the boundary vector of the inequality constraint.

[0221] H=Φ T QΦ+L,η=Φ T (R-Fx(k))

[0222]

[0223] The condition for the objective function to achieve the minimum value is Substitute this condition into the objective function (2-12) and solve it to obtain the optimal control sequence at time k:

[0224] ΔU=(Φ T QΦ+L) -1 Φ T Q(R-Fx(k)) (2-15)

[0225] Where R represents the input weight matrix, R = (r(k+1)r(k+2)…r(k+N p )), r(k+N p ) represents k+N p Input weight value at any time, N p is the prediction step length;

[0226] Only the first increment Δu(k) in the optimal control sequence will be applied, so the control input increment at time k is:

[0227] Δu(k)=(I m 0…0)(Φ T QΦ+L) -1 Φ T Q(R-Fx(k)) (2-16)

[0228] Where, I m Represents the m-dimensional identity matrix.

[0229] Example 2

[0230] The present invention proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for designing a model predictive controller for frequency modulation of an IGV of a gas-steam combined cycle unit as described in Example 1 is implemented.

[0231] Example 3

[0232] The present invention provides a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the model predictive controller design method for IGV-participated frequency modulation of a gas-steam combined cycle unit as described in the first embodiment.

[0233] In order to better illustrate the control effect of the present invention, a comparison is made between the control effects of a conventional controller and the controller designed by the present invention on a gas-steam combined cycle unit model, and the superiority of the controller is shown through the difference comparison.

[0234] The gas-steam combined cycle unit model used is as follows: Figure 2 shown.

[0235] From the analysis of the system, it can be found that the influence of the inlet guide vane opening IGV on the total power Pall and the gas turbine exhaust temperature TX is opposite to that of the fuel quantity Wf. For the convenience of identification, this input quantity is replaced by the opening margin 1-IGV.

[0236] The input fuel quantity Wf is taken as a random signal with a mean value of 0.7 and a sampling time of 1.3s, and the input quantity 1-IGV is taken as a random signal with a mean value of 0.3 and a sampling time of 0.5s. The input and output data used for identification are shown in Figure 3.

[0237] The state space model obtained by identification is shown in formula (3-1):

[0238]

[0239] The continuous state space parameters in formula (3-1) are converted into discrete state space parameters consistent with MPC control as shown in formula (3-2).

[0240]

[0241] The state space parameters are brought into the model predictive controller design process. The parameters of the conventional model predictive controller and the controller designed under the present invention are selected as shown in Table 1 and Table 2 respectively.

[0242] The controller design part is completed.

[0243] Table 1 Conventional predictive controller parameter selection table

[0244]

[0245] Table 2 Predictive controller parameter selection table under this design

[0246]

[0247] The following is a comparison of the performance of the designed controller with that of the PID controller and the conventional model predictive controller.

[0248] Figure 4 The figure shows the AGC command curve. Figure 4 The following figure shows the capability comparison of the original PID controller, the conventional predictive controller and the predictive controller of the present invention under the AGC instruction. As shown in FIG5 , the present invention improves the capability of the system to track the load instruction under the AGC mode.

[0249] In the embodiments disclosed herein, computer storage media can be tangible media that can contain or store programs for use by or in conjunction with an instruction execution system, device, or apparatus. Computer storage media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0250] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0251] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A model predictive controller design method for IGV frequency modulation of a gas-steam combined cycle unit, characterized in that: The following steps are involved: Obtain unit operation data; Obtain unit design parameters; Based on the unit operating data, the subspace identification method is used to identify the unit state space model for controller design; Based on the unit state space model and the unit design parameters, a model predictive controller is designed to control the unit output power.

2. The method for designing a model predictive controller for frequency modulation of an IGV in a gas-steam combined cycle unit according to claim 1, characterized in that: The unit operation data includes the amount of fuel input during the unit operation, the opening of the gas turbine inlet guide vanes during the unit operation, the total power output during the unit operation, and the temperature of the gas turbine exhaust during the unit operation.

3. The method for designing a model predictive controller for frequency modulation of an IGV in a gas-steam combined cycle unit according to claim 1, characterized in that: The unit design parameters include the fuel quantity allowed by the unit, the upper and lower limits of the opening of the gas turbine inlet guide vanes, and the upper and lower limits of the change rate of the fuel quantity and the opening of the gas turbine inlet guide vanes.

4. The method for designing a model predictive controller for frequency modulation of an IGV in a gas-steam combined cycle unit according to claim 1, wherein: The specific process of using the subspace identification method to identify the unit state space model for controller design is as follows: The expression defining the state space model is: Where x k+1 is the state vector at time k+1, x k is the state vector at time k, A a represents the state transition matrix, B b represents the input matrix, u k is the input vector, K represents the Kalman filter gain, e k represents the observation noise, y k represents the output vector, C c Denotes the output matrix, D d represents the direct transfer matrix; The input vector is represented as follows: In the formula, Mf k Indicates fuel quantity, IGV k Indicates the opening of the inlet guide vane, R 2 represents the two-dimensional real number field; The output vector is represented as follows: Where Pall k Indicates output power, TX k Indicates exhaust temperature; Identify the state transfer matrix A using the subspace identification method a , input matrix B b , output matrix C c and direct transfer matrix D d .

5. The method for designing a model predictive controller for frequency modulation of an IGV in a gas-steam combined cycle unit according to claim 4, characterized in that: The subspace identification method is specifically as follows: The input data include: the amount of fuel input during the operation of the unit, the opening of the gas turbine inlet guide vanes during the operation of the unit, and the output data used include the total power output during the operation of the unit, and the temperature of the gas turbine exhaust during the operation of the unit; Construct the Hankel matrix U of the input data std : Where U p is the block matrix of the past moment, U f is the block matrix of the future moment; The Hankel matrix Y of the output data std Also divided into the block matrix Y of the past moment p and the future block matrix Y f ; THE f =Γ nn X f +H d U f +H s ITS f (1-3) THE p =Γ nn X p +H d U p +H s ITS p (1-4) Where, Γ nn Represents the output coefficient matrix, X f represents the future state sequence, H d Represents the dynamic influence matrix from input to output, U f is the block matrix of the future time, H s Represents the dynamic impact matrix of noise to output, E f represents the noise at the future time, X p Represents the past state sequence, U p represents the past time block matrix, E p represents the noise at the past moment; The state matrix X std Defined as: X p represents the past state sequence, X f Represents a sequence of future states; X f =L nn W p (1-6) Where, L nn represents the subspace matrix, W p Represents the past moment data matrix; Substituting formula (1-6) into formula (1-3) yields: Y f =L w X f +L u U f +L e E f (1-7) Where, L w represents the state weight matrix, L u represents the input influence matrix, L e represents the noise dynamic matrix; The Hankel matrix U of the input data std Do the QR decomposition as follows: Where R 11 Represents the upper triangular block of the input data matrix after orthogonal projection, R 21 Represents the sub-block of cross-projection of input and output data, R 31 represents the associated block between the input and residual subspace, R 22 represents the main projection block of the output data matrix, R 32 represents the coupling block between the output and residual subspace, R 33 represents the upper triangular block of the residual subspace, Q1 represents the orthogonal basis of the input data, Q2 represents the orthogonal basis of the output data, and Q3 represents the orthogonal basis of the residual subspace; Comparing Equation (1-8) with Equation (1-7), we can calculate the subspace joint description matrix L = (L w L u )as follows: The state transfer matrix A is obtained from the subspace joint description matrix a , input matrix B b , output matrix C c and direct transfer matrix D d .

6. The method for designing a model predictive controller for frequency modulation of an IGV in a gas-steam combined cycle unit according to claim 1, wherein: The model predictive controller designed based on the unit design parameters is specifically: The number of inputs, outputs, and states of the gas-steam combined cycle unit are r, m, and n respectively. The unit state space model is rewritten into a discrete form: Among them, x m (k), u(k), y(k) are the n-dimensional state variables, r-dimensional input variables and m-dimensional output variables of the unit at time k respectively. m represents the discrete state transfer matrix, B m represents the discrete input matrix, C m represents the discrete output matrix; x m (k+1) is the n-dimensional state variable of the unit at time k+1; A model predictive controller is constructed based on the discrete state space model of the unit. The expression of the model predictive controller is as follows: Y=Fx(k)+ΦΔU (2-7) Where Y represents the output matrix at time k, F represents the terminal cost weight matrix, x(k) represents the n+m-dimensional augmented state variable, and x(k) = (Δx m (k) T y(k) T ) T , Δx m (k) represents the increment of the state variable at time k, y(k) represents the m-dimensional output variable of the unit at time k, Φ represents the dynamic response matrix that quantifies the cumulative impact, and ΔU represents the input quantity increment matrix under the premise of time k, which is used as the control sequence; The output matrix Y at time k is as follows: Y=(y(k+1|k) y(k+2|k)…y(k+N p |k)) T y(k+N p |k) represents k+N at time k p Output of the unit at this moment, N p is the prediction step length; The input increment matrix ΔU at time k is as follows: ΔU=(Δu(k) Δu(k+1)…Δu(k+N c -1)) T Δu(k+N c -1) means k+N c -1 time input increment, N c To control the step length; Design the objective function, take the minimum value of the objective function as the goal, and solve the optimal control sequence; The objective function is designed as: Where, J QP represents the objective function, ΔU represents the input increment matrix, Φ represents the dynamic response matrix that quantifies the cumulative impact, Q represents the weighted matrix of the output error, and L represents the weighted matrix of the control increment; R represents the input weight matrix, F represents the terminal cost weight matrix, and x(k) represents the n+m-dimensional augmented state variable; Write the current input as the previous input plus the increment, and the input is: U=A QP ΔU+u(k-1)B QP (2-12) U represents the input quantity, ΔU represents the input quantity increment matrix, A QP represents the QP decomposition incremental coefficient matrix, B QP represents the QP decomposition input coefficient matrix; u(k-1) represents the input at the previous moment; The constraints of the objective function are: U min Indicates the minimum value allowed for the input quantity, U max Indicates the maximum value allowed for the input; The condition for the objective function to achieve the minimum value is Substitute this condition into the objective function (2-12) and solve it to obtain the optimal control sequence at time k: ΔU=(Φ T QΦ+L) -1 F T Q(R-Fx(k)) (2-15) Where R represents the input weight matrix, R=(r(k+1) r(k+2)…r(k+N p )), r(k+N p ) represents k+N p Input weight value at any time, N p is the prediction step length; Only the first increment Δu(k) in the optimal control sequence will be applied, so the control input increment at time k is: Δu(k)=(I m 0…0)(Φ T QΦ+L) -1 F T Q(R-Fx(k)) (2-16) Where, I m Represents the m-dimensional identity matrix.

7. The method for designing a model predictive controller for frequency modulation of an IGV in a gas-steam combined cycle unit according to claim 6, characterized in that: In the terminal cost weight matrix F of the model predictive controller, the weight of the total output power is set higher than the weight of the gas turbine exhaust temperature.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for designing a model predictive controller for frequency modulation of an IGV of a gas-steam combined cycle unit as described in any one of claims 1 to 7 is implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: The computer program enables a computer to execute the model predictive controller design method for IGV participation in frequency regulation of a gas-steam combined cycle unit as described in any one of claims 1 to 7.

Images