A method for identifying closed-loop optimization control systems of deep peak-shaving thermal power units

Abstract

This invention primarily relates to the field of thermal power unit system technology, specifically a method for identifying a closed-loop optimized control system for a deep peak-shaving thermal power unit. The method includes: S100 establishing a spatial state model of the thermal power unit system; S200 designing linear quadratic controller parameters for the spatial state model of the thermal power unit system, calculating the optimal control law through a state feedback matrix, and forming a closed-loop system of the thermal power unit; and obtaining the control input, control output, and system setpoint of the closed-loop system; S300 decomposing the closed-loop system into two open-loop subsystems; S400 obtaining the sensitivity function of the closed-loop system by performing subspace identification on the control input and system setpoint, and constructing a noise-free intermediate signal using the sensitivity function; and S500 obtaining a consistent and unbiased estimate of the parameters of the thermal power unit system by performing subspace identification on the intermediate signal and the control output of the closed-loop system.

Description

Technical Field

[0001] This invention mainly relates to the field of thermal power unit system technology, specifically a method for identifying a closed-loop optimization control system for deep peak-shaving thermal power units. Background Technology

[0002] With the transformation of the energy structure, thermal power units will gradually become the main type of regulating power source, characterized by deep peak shaving and frequent start-stop peak shaving. Deep peak shaving of thermal power units refers to the fine regulation of the power system frequency by adjusting their own load when the power system load fluctuates significantly. Under deep peak shaving conditions, thermal power generating units often need to rapidly change the load to meet grid demand. During load increases and decreases, the output power and heat load of the thermal power unit change, causing changes in the temperature and pressure inside the combustion chamber, preventing the thermal power unit from operating continuously at a stable point. To achieve stable, safe, and efficient operation of thermal power units under deep peak shaving and rapid load changes, accurate modeling of the thermal power unit control system is a prerequisite for all control and optimization.

[0003] Traditional modeling methods for thermal power units mainly include mechanistic modeling and system identification modeling. Mechanistic models can comprehensively reflect the intrinsic relationships between various parameters of the unit, making them suitable for simulation studies of the characteristics between various inputs and outputs and for evaluating the performance of control algorithms. However, for thermal power units with complex physical structures and significant subsystem coupling characteristics, mechanistic modeling can lead to numerous steps, complex modeling of nonlinear components, and poor model transferability between units. Furthermore, mechanistic modeling often fails to adequately consider environmental factors, while actual processes are affected by coal quality variations and environmental disturbances, resulting in a poor fit between the simulation output of the mechanistic model and the actual operating output data. Therefore, system parameter identification is necessary in the modeling process of thermal power unit control systems.

[0004] For safety and economic reasons, the control of industrial processes in thermal power units is all closed-loop control. This poses a challenge to the accurate identification of the system. The main difficulty lies in the fact that system noise, after being processed by feedback control, will affect the system input at future moments; that is, there is a correlation between noise data at past moments and input data at future moments. Due to the correlation between noise and measurement data, it is difficult to eliminate the effect of noise through simple regression calculations, leading to identification biases in traditional open-loop identification methods. In recent years, with the increase in actual closed-loop control industrial processes, system identification in a closed-loop environment has received increasing attention.

[0005] Subspace identification, as an effective tool for modeling multivariable systems, uses a simple error minimization criterion and linear algebra tools to accurately identify the state-space model. The only required "parameter" is the system order, effectively avoiding problems such as numerical ill-conditioning, parameter overlap, and minimum system realization. Especially for data generated from closed-loop industrial processes, subspace identification can yield a state-space model that effectively represents the actual industrial production process. Therefore, research on subspace identification methods for closed-loop systems has significant theoretical implications and important engineering application value. Summary of the Invention

[0006] To address the problem of high-reliability optimal control and parameter identification for thermal power units under rapid load changes across all operating conditions during online operation, this invention provides a method for identifying the closed-loop optimization control system of a deep peak-shaving thermal power unit.

[0007] This invention adopts the following technical solution: a method for identifying a closed-loop optimization control system for deep peak-shaving thermal power units, comprising:

[0008] S100: Establish a spatial state model of the thermal power unit system, select the coal consumption command and the turbine control valve opening command as the input of the thermal power unit system, and the unit power and the pressure in front of the turbine as the output of the system.

[0009] S200: Design linear quadratic controller parameters for the spatial state model of the thermal power unit system, calculate the optimal control law through the state feedback matrix, form a closed-loop system of the thermal power unit, and obtain the control input, control output and system setpoint of the closed-loop system of the thermal power unit;

[0010] S300: Decomposes the closed-loop system of a thermal power unit into two open-loop subsystems;

[0011] S400: The sensitivity function of the closed-loop system of the thermal power unit is obtained by subspace identification of the control input and system setpoint of the closed-loop system of the thermal power unit, and an intermediate signal without noise pollution is constructed by the sensitivity function of the closed-loop system of the thermal power unit.

[0012] S500: By performing subspace identification on intermediate signals and the control output of the closed-loop system of the thermal power unit, a consistent and unbiased estimate of the system parameters of the thermal power unit is finally obtained.

[0013] In some embodiments, step S100 includes:

[0014] S101: The thermal power unit system is linearized with small deviation at a certain operating point to transform it into a linear model of the unit load-pressure object;

[0015] S102: Based on the transfer function form of the unit load-pressure object shown in equation (1), establish the state-space expression of the system.

[0016] In some embodiments, in step S101, the linear model of the unit load-pressure object is:

[0017]

[0018] in, The transfer function representing the change in inlet pressure caused by fuel variation. The transfer function representing the change in inlet pressure caused by the change in the turbine control valve opening command. The transfer function represents the change in unit power caused by changes in fuel. The transfer function representing the change in unit power caused by the change in the turbine control valve opening command; For fuel quantity instructions, This is the turbine control valve opening command;

[0019]

[0020]

[0021]

[0022]

[0023] For unit power, Pre-machine pressure; For the dynamic time of flour production, The boiler heat storage coefficient, For the dynamic time of the steam turbine. For fuel command gain, The superheater drag coefficient is dimensionless. For turbine gain; in the transfer function and It is a constant, denoted as and .

[0024] In some embodiments, in step S102, the state-space expression is:

[0025]

[0026] in, For system state variables, Indicates input variables, This indicates the output variable.

[0027] In some embodiments, step S200 includes:

[0028] S201: Design linear quadratic control for the state-space model of unit load-pressure object, and introduce a quadratic objective function;

[0029]

[0030] Where X is the state vector, u is the system input, Q and R are given real symmetric weight matrices, and Q is positive semi-definite and R is positive definite;

[0031] Assume the linear feedback control rate of the thermal power unit system Substituting the control law into the system objective function, we get

[0032]

[0033] Given Q and R, the system's performance metrics J The minimum value is equivalent to the Riccati equation having a positive definite matrix solution P: After obtaining the solution P, the control gain of the system Represented as ;

[0034] S202: For a thermal power unit system, to find the control action u that minimizes the quadratic objective function J, the solution to the steady-state LQR problem can be derived, and the optimal control law is:

[0035]

[0036] u is the system input, R is a given real symmetric matrix, B is the parameter in the state-space expression, P is the positive definite matrix solution of the Riccati equation, and X is the system state vector.

[0037] In step S300,

[0038] The first open-loop system is:

[0039]

[0040] The second open-loop system is:

[0041]

[0042] In the formula, This refers to the fuel quantity and turbine control valve opening command data for the thermal power unit, i.e., the output data of the first open-loop system. This includes data on unit power and upstream pressure. The setpoints for the fuel quantity and turbine control valve opening commands of the thermal power unit are the input data for the first open-loop system. G represents the thermal power unit model. It is the sensitivity function of the closed-loop system of the thermal power unit. Feedback control L is designed LQR Controller.

[0043] In some embodiments, step S400 includes:

[0044] S401: For the first open-loop subsystem model Converting it to a state-space expression yields the following result:

[0045]

[0046] in, , Let be the control input and system setpoint data vectors of the closed-loop system at time t, respectively. It is the state variable of the first open-loop subsystem at time t. Zero-mean white noise; Sensitivity function of closed-loop system of thermal power unit Parameters;

[0047] S402: The sensitivity function of the closed-loop system of the thermal power unit is obtained by performing subspace identification on the control input and system setpoint of the closed-loop system. ;

[0048] S403: For the identification result of the first open-loop subsystem, provide the same input quantity. If the obtained output value is the same as If the trend is consistent and the error is small, it indicates that the sensitivity function of the closed-loop system of the thermal power unit is consistent. To meet accuracy requirements, the sensitivity function Multiplying this by the system setting value yields an intermediate signal free of noise pollution.

[0049] Step S500 includes:

[0050] S501: For the second open-loop subsystem Convert it into a state-space expression:

[0051]

[0052] in, , These are the intermediate signal at time t and the control output data vector of the thermal power unit's closed-loop system, respectively. These are the state variables of the thermal power unit model at time t. Zero-mean white noise; The matrix represents the parameters of the thermal power unit system G;

[0053] S502: Perform subspace identification on the intermediate signals and the control output of the thermal power unit closed-loop system to obtain the thermal power unit model and thermal power unit closed-loop system parameters. , These are the uniform unbiased parameters of the thermal power unit model G.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] This invention provides an identification method for a closed-loop optimization control system of a deep peak-shaving thermal power unit. First, at any operating point, a linear model of the thermal power unit is obtained through small-deviation linearization, and the transfer function model of the unit's load-pressure object is converted into a state-space model of the system. Second, a linear quadratic regulator is used to obtain the optimal control quantity, improving the dynamic characteristics of the thermal power unit and obtaining the system's control input, control output data, and system setpoint data. Then, a two-stage method is used to decompose the closed-loop system of the thermal power unit into two open-loop subsystems, and a noise-free intermediate signal is constructed to replace the actual control input, thereby eliminating the correlation between noise and control input. Finally, a subspace identification method is used to identify the two open-loop subsystems respectively, ultimately obtaining a consistent and unbiased estimate of the thermal power unit system parameters, achieving accurate system modeling. Attached Figure Description

[0056] Figure 1 This is a flowchart of an identification method for a closed-loop optimization control system of a deep peak-shaving thermal power unit proposed in this invention.

[0057] Figure 2 This is a schematic diagram of the closed-loop system structure in this invention;

[0058] Figure 3 This is a schematic diagram of the two-stage method structure in this invention;

[0059] Figure 4 This is a flowchart of the subspace identification method in this invention;

[0060] Figure 5 This is a diagram showing the input-output relationship of a thermal power unit in this invention. Detailed Implementation

[0061] To facilitate understanding of the present invention, the present invention will be described more fully and in detail below with reference to the accompanying drawings and preferred embodiments, but the scope of protection of the present invention is not limited to the following specific embodiments.

[0062] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0063] A method for identifying a closed-loop optimization control system for a deep peak-shaving thermal power unit includes:

[0064] S100: Establish a spatial state model of the thermal power unit system, select the coal consumption command and the turbine control valve opening command as the input of the thermal power unit system, and the unit power and the pressure in front of the turbine as the output of the system.

[0065] Step S100 includes:

[0066] S101: A common approach to handling nonlinearity is to linearize the nonlinear model with small deviations at multiple operating points and then analyze the characteristics of the object using the fundamental theory of linear systems. Therefore, we choose to transform the thermal power unit system into a linear model by linearizing it with small deviations at a certain operating point. The linear model of the unit's load-pressure object can be described as follows:

[0067]

[0068] in, The transfer function representing the change in inlet pressure caused by fuel variation. The transfer function representing the change in inlet pressure caused by the change in the turbine control valve opening command. The transfer function represents the change in unit power caused by changes in fuel. The transfer function representing the change in unit power caused by the change in the turbine control valve opening command; For fuel quantity instructions, This is the turbine control valve opening command;

[0069]

[0070]

[0071]

[0072]

[0073] For unit power, Pre-machine pressure; For the dynamic time of flour production, The boiler heat storage coefficient, For the dynamic time of the steam turbine. For fuel command gain, The superheater drag coefficient is dimensionless. For turbine gain; in the transfer function and It is a constant, denoted as and .

[0074] S102: Based on the transfer function form of the unit load-pressure object shown in equation (1), establish the state-space expression of the system:

[0075]

[0076] in, For system state variables, Indicates input variables, This indicates the output variable.

[0077] The specific forms of the coefficient matrices in its state-space model are as follows:

[0078]

[0079]

[0080] .

[0081] S200: Designs linear quadratic controller parameters for the spatial state model of the thermal power unit system, calculates the optimal control law through the state feedback matrix, and forms a closed-loop system for the thermal power unit; and obtains the control input, control output, and system setpoints of the closed-loop system. The control input, control output, and system setpoint data of the thermal power unit system are actual data from the power plant and can be exported from the power plant's DCS system.

[0082] Specifically, step S200 includes:

[0083] S201: Design linear quadratic control for the state-space model of unit load-pressure object, and introduce a quadratic objective function;

[0084]

[0085] Where X is the state vector, u is the system input, Q and R are given real symmetric weight matrices, and Q is positive semi-definite and R is positive definite;

[0086] Assume the linear feedback control rate of the thermal power unit system Substituting the control law into the system objective function, we get

[0087]

[0088] Given Q and R, the system's performance metrics J The minimum value is equivalent to the Riccati equation having a positive definite matrix solution P: After obtaining the solution P, the control gain of the system Represented as ;

[0089] S202: For a thermal power unit system, to find the control action u that minimizes the quadratic objective function J, the solution to the steady-state LQR problem can be derived, and the optimal control law is:

[0090]

[0091] u is the system input, R is a given real symmetric matrix, B is the parameter in the state-space expression, P is the positive definite matrix solution of the Riccati equation, and X is the system state vector.

[0092] S300: Decompose the closed-loop system of the thermal power unit into two open-loop subsystems and construct an intermediate signal without noise pollution to replace the actual control input, thereby eliminating the correlation between noise and control input;

[0093] refer to Figure 2 The closed-loop system expression for a thermal power unit is as follows:

[0094]

[0095] Bundle Substitution We can obtain:

[0096]

[0097] refer to Figure 3 The first stage targets the above formula. and Irrelevant, and and All can be measured, therefore we can... Consider it as the first open-loop system, starting from the setpoint data and process input data Sensitivity function is identified ;

[0098] Furthermore, the second phase can include Treating it as a second open-loop system, firstly, based on the sensitivity function obtained from the first stage identification... Multiplying the data by the setpoint yields a noise-free intermediate signal. and process output data The parameters of the thermal power unit were identified.

[0099] In the formula, This contains data on fuel quantity for thermal power units and turbine control valve opening commands. This includes data on unit power and upstream pressure. G represents the setpoints for the fuel quantity and turbine control valve opening commands of the thermal power unit, where G is the thermal power unit model. It is the sensitivity function of the closed-loop system of the thermal power unit. Feedback control L is designed LQR Controller.

[0100] LQR control is state feedback control, which is a type of closed-loop optimal control. When LQR state feedback control is added to a thermal power unit system, it becomes a closed-loop system for thermal power units.

[0101] S400: The sensitivity function of the closed-loop system of the thermal power unit is obtained by subspace identification of the control input and system setpoint of the closed-loop system of the thermal power unit, and an intermediate signal without noise pollution is constructed by the sensitivity function of the closed-loop system of the thermal power unit.

[0102] Step S400 includes:

[0103] S401: For the first open-loop subsystem model Converting it to a state-space expression yields the following result:

[0104]

[0105] in, , Let be the control input and system setpoint data vectors of the closed-loop system at time t, respectively. It is the state variable of the first open-loop subsystem at time t. Zero-mean white noise; Sensitivity function of closed-loop system of thermal power unit Parameters;

[0106] S402: The sensitivity function of the closed-loop system of the thermal power unit is obtained by performing subspace identification on the control input and system setpoint of the closed-loop system. ;

[0107] First, construct a Hankel matrix using past and future data of the control inputs and system setpoints of the closed-loop system of the thermal power unit. Then, perform row projection on the Hankel matrix to obtain... Then, the projected singular value decomposition is performed to obtain the generalized observable matrix. Kalman filter state Finally, by and To determine the state matrix .

[0108] S403: For the identification result of the first open-loop subsystem, provide the same input quantity. If the obtained output value is the same as If the trend is consistent and the error is small, it indicates that the sensitivity function of the closed-loop system of the thermal power unit is consistent. To meet accuracy requirements, the sensitivity function Multiplying this by the system setting value yields an intermediate signal free of noise pollution.

[0109] Step S300 converts the closed-loop system of the thermal power unit into two open-loop subsystems using a two-stage method, and obtains the system model of the thermal power unit through open-loop identification. Step S400 performs open-loop subspace identification on the two open-loop subsystems of the thermal power unit, and obtains the state-space equations.

[0110] S500: By performing subspace identification on intermediate signals and thermal power unit system outputs, a consistent and unbiased estimate of the thermal power unit system parameters is finally obtained.

[0111] S501: For the second open-loop subsystem Convert it into a state-space expression:

[0112]

[0113] in, , These are the intermediate signal at time t and the control output data vector of the thermal power unit's closed-loop system, respectively. These are the state variables of the thermal power unit model at time t. Zero-mean white noise; The matrix represents the parameters of the thermal power unit system G;

[0114] S502: Perform subspace identification on the intermediate signals and the control output of the thermal power unit closed-loop system to obtain the thermal power unit model and thermal power unit closed-loop system parameters. , These are the uniform unbiased parameters of the thermal power unit model G.

[0115] A Hankel matrix is ​​constructed using the intermediate signals and past and future data from the control output of the thermal power unit's closed-loop system. The remaining identification process is the same as that of the first open-loop subsystem, ultimately yielding the parameters of the thermal power unit's closed-loop system. , These are the uniform unbiased parameters of the thermal power unit model G.

[0116] Furthermore, the subspace identification steps are as follows: Figure 4 As shown, the following past and future data vectors and input / output Hankel matrices are defined, each containing i rows and j columns:

[0117]

[0118]

[0119]

[0120]

[0121] Each element

[0122] ,

[0123] These are column vectors representing the input and output, respectively, where the subscripts... Indicates the past. Representing the future, matrix Depend on and constitute

[0124]

[0125] State matrix and A similar definition is also made:

[0126]

[0127]

[0128] The generalized data input-output matrix equations can then be written in the following form, and these equations play an important role in subspace identification algorithms:

[0129]

[0130] The meanings of each part of the equation are as follows:

[0131] Generalized observability matrix :

[0132]

[0133] Determine the low-dimensional block triangular Toeplitz matrix :

[0134]

[0135] Low-dimensional block triangular Toeplitz matrix :

[0136]

[0137] The inverse generalized observability matrix of {A,B} :

[0138]

[0139] The inverse generalized observability matrix of {A,K} :

[0140]

[0141] Solving the state matrix using the subspace method All of these require starting from the generalized observable matrix. Kalman filter state The process involves two main steps: First, calculating the row space projection of a specific Hankel matrix, which is typically obtained mathematically using QR decomposition. Second, calculating the singular value decomposition of this projection to obtain the generalized observable matrix. Kalman filter state The third step, by and To determine the state matrix .

[0142] First, we need to take the formula... Item, through Future input Find out the future output Optimal prediction :

[0143]

[0144] Depend on The line space in and Zhang Cheng's orthogonal projection of the line space

[0145]

[0146] By using oblique projection, we can obtain the following respectively. and :

[0147]

[0148]

[0149] Define projection :

[0150]

[0151] It is equal to the generalized observable matrix. and state sequence Non-stationary Kalman filter state estimation The product of.

[0152] available

[0153]

[0154] The second step of the subspace algorithm is to decompose the above expression into components. and This is entirely possible because if the number of columns j of the block Hankel matrix tends to infinity, i.e. , It is a rank n matrix, where n is the order of the system. However, in practice, due to noise and the influence of finite-length data, It may not be a full-rank matrix. A matrix with rank of can be used. n The singular value decomposition of the matrix is ​​used for approximation. Before performing singular value decomposition, it is necessary to... Use respectively and By performing left-weighted and right-weighted operations, we have:

[0155]

[0156] In order to maintain The rank remains unchanged. and It should be full rank. Observational principal singular value. The order of the system can be estimated. n The generalized observable matrix can be directly derived from the above SVD decomposition. and the system's state estimation sequence :

[0157]

[0158] Matrices A and C can be derived from the generalized observability matrix. From the above equation, we can see that matrices A and C can be directly derived from the matrix. get

[0159]

[0160] Once matrices A and C are obtained, matrices B and D can be solved:

[0161]

[0162] in For a full-rank row matrix to satisfy For simplicity, we will denote the left side of the equation as M. Let L represent the expression, the above formula can be written as:

[0163]

[0164] Clearly, the above equation is a system of linear equations with B and D as variables, which can be solved using the least squares method. From this, we can obtain the equations in the thermal power unit model. Matrices enable accurate modeling of the system.

Claims

1. A method for identifying a closed-loop optimization control system for a deep peak-shaving thermal power unit, characterized in that, include: S100: Establish a spatial state model of the thermal power unit system, select the coal consumption command and the turbine control valve opening command as the input of the thermal power unit system, and the unit power and the pressure in front of the turbine as the output of the system. S200: Design linear quadratic controller parameters for the spatial state model of thermal power unit system, calculate the optimal control law through state feedback matrix, and form a closed-loop system of thermal power unit. It also obtains the control input, control output, and system setpoints of the closed-loop system of the thermal power unit; S300: Decomposes the closed-loop system of a thermal power unit into two open-loop subsystems; In step S300, The first open-loop system is: The second open-loop system is: In the formula, This refers to the fuel quantity and turbine control valve opening command data for the thermal power unit, i.e., the output data of the first open-loop system. This includes data on unit power and upstream pressure. The setpoints for the fuel quantity and turbine control valve opening commands of the thermal power unit are the input data for the first open-loop system. G represents the thermal power unit model. It is the sensitivity function of the closed-loop system of the thermal power unit, and the feedback control L is the designed sensitivity function. LQR Controller; S400: The sensitivity function of the closed-loop system of the thermal power unit is obtained by subspace identification of the control input and system setpoint of the closed-loop system of the thermal power unit, and an intermediate signal without noise pollution is constructed by the sensitivity function of the closed-loop system of the thermal power unit. Step S400 includes: S401: For the first open-loop subsystem model Converting it to a state-space expression yields the following result: in, , Let be the control input and system setpoint data vectors of the closed-loop system at time t, respectively. It is the state variable of the first open-loop subsystem at time t. Zero-mean white noise; Sensitivity function of closed-loop system of thermal power unit Parameters; S402: The sensitivity function of the closed-loop system of the thermal power unit is obtained by performing subspace identification on the control input and system setpoint of the closed-loop system. ; S403: For the identification result of the first open-loop subsystem, provide the same input quantity. If the obtained output value is the same as If the trend is consistent and the error is small, it indicates that the sensitivity function of the closed-loop system of the thermal power unit is consistent. To meet accuracy requirements, the sensitivity function Multiply by the system setpoint to obtain an intermediate signal free of noise pollution; S500: By performing subspace identification on intermediate signals and the control output of the closed-loop system of the thermal power unit, a consistent and unbiased estimate of the system parameters of the thermal power unit is finally obtained.

2. The method for identifying the closed-loop optimization control system of a deep peak-shaving thermal power unit according to claim 1, characterized in that, Step S100 includes: S101: The thermal power unit system is linearized with small deviation at a certain operating point to transform it into a linear model of the unit load-pressure object; S102: Based on the transfer function form of the unit load-pressure object, establish the state-space expression of the system.

3. The method for identifying the closed-loop optimization control system of a deep peak-shaving thermal power unit according to claim 2, characterized in that, In step S101, the linear model of the unit load-pressure object is: in, The transfer function representing the change in inlet pressure caused by fuel variation. The transfer function representing the change in inlet pressure caused by the change in the turbine control valve opening command. The transfer function represents the change in unit power caused by changes in fuel. The transfer function representing the change in unit power caused by the change in the turbine control valve opening command; For fuel quantity instructions, This is the turbine control valve opening command; For unit power, Pre-machine pressure; For the dynamic time of flour production, The boiler heat storage coefficient, For the dynamic time of the steam turbine. For fuel command gain, The superheater drag coefficient is dimensionless. For turbine gain; in the transfer function and It is a constant, denoted as and .

4. The method for identifying the closed-loop optimization control system of a deep peak-shaving thermal power unit according to claim 3, characterized in that, In step S102, the state-space expression is: in, For system state variables, Indicates input variables, This indicates the output variable.

5. The method for identifying the closed-loop optimization control system of a deep peak-shaving thermal power unit according to claim 4, characterized in that, Step S200 includes: S201: Design linear quadratic control for the state-space model of unit load-pressure object, and introduce a quadratic objective function; Where X is the state vector, u is the system input, Q and R are given real symmetric weight matrices, and Q is positive semi-definite and R is positive definite; Assume the linear feedback control rate of the thermal power unit system Substituting the control law into the system objective function, we get Given Q and R, the system's performance metrics J The minimum value is equivalent to the Riccati equation having a positive definite matrix solution P: After obtaining the solution P, the control gain of the system Represented as ; S202: For a thermal power unit system, to find the control action u that minimizes the quadratic objective function J, the solution to the steady-state LQR problem can be derived, and the optimal control law is: u is the system input, R is a given real symmetric matrix, B is the parameter in the state-space expression, P is the positive definite matrix solution of the Riccati equation, and X is the state vector.

6. The identification method for the closed-loop optimization control system of deep peak-shaving thermal power units according to claim 1, characterized in that, Step S500 includes: S501: For the second open-loop subsystem Convert it into a state-space expression: in, , These are the intermediate signal at time t and the control output data vector of the thermal power unit's closed-loop system, respectively. These are the state variables of the thermal power unit model at time t. Zero-mean white noise; The matrix represents the parameters of the thermal power unit system G; S502: Perform subspace identification on the intermediate signals and the control output of the thermal power unit closed-loop system to obtain the thermal power unit model and thermal power unit closed-loop system parameters. , These are the uniform unbiased parameters of the thermal power unit model G.

Images