Control method and system for reactor starting separator

Abstract

The invention relates to a control method and system for a reactor starting separator. The control method for the reactor starting separator comprises the following steps: acquiring input and output data; performing space identification based on the input and output data to obtain a state space model; solving based on the state space model to obtain a prediction control instruction; performing feedforward compensation on the prediction control instruction based on the disturbance estimation value to obtain a target control instruction; and performing driving control on the execution mechanism according to the target control instruction. According to the method, the state space model is obtained through space identification, solution prediction is carried out based on the state space model, meanwhile, feedforward compensation is carried out in combination with disturbance estimation, the problems of multivariable coupling and mode switching delay can be solved, the problem of insufficient anti-interference performance can also be solved, and the robustness of the system is improved. And the function of efficiently and reliably controlling the pressure and the liquid level of the starting separator is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of reactor control, more particularly, to a reactor startup separator control method and system. BACKGROUND

[0002] During the start-up and shut-down process of a nuclear power plant, the startup separator is located downstream of the steam generator, and its main function is to absorb steam that does not meet the quality requirements of the steam, and to control the pressure of the steam generator. The upper part of the separator is provided with a steam discharge valve, and the lower part is provided with a drain valve. During the start-up of the unit, the drain valve is used to control the outlet pressure of the steam generator, and the steam discharge valve is closed. When the liquid level in the separator drops to a set value, the function of the drain valve is switched, the steam discharge valve is opened to control the pressure, and the drain valve is used to control the liquid level.

[0003] The current startup separator control has the following defects: 1. Multivariable coupling: independent control of pressure and liquid level leads to oscillation; 2. Mode switching delay: manual intervention is required; 3. Insufficient disturbance rejection: when the steam flow fluctuates by ±10%, the liquid level deviation is ±15%. The traditional control method may have the problems of response lag and insufficient precision, and a more accurate and efficient control strategy is needed. However, there is little research on the start-up system at present, and automatic control function is not considered. In addition, the control strategies of different reactor types are the same, and the control schemes are not the same. SUMMARY

[0004] The technical problem solved by the present application is to provide a reactor startup separator control method and system to solve the problems in the prior art.

[0005] The technical solution adopted by the present application to solve the technical problem is: a reactor startup separator control method is constructed, comprising the following steps:

[0006] Obtain input and output data; the input and output data includes: historical data and online data; the historical data and the online data both include: pressure data, liquid level data and running data of the actuator;

[0007] Based on the input and output data, spatial identification is performed to obtain a state space model;

[0008] Based on the state space model, a solution is obtained to obtain a predictive control instruction;

[0009] Based on the disturbance estimation value, the predictive control instruction is fed forward compensated to obtain a target control instruction;

[0010] According to the target control instruction, the actuator is driven and controlled.

[0011] In the control method of the reactor startup separator, the space identification based on the input and output data comprises:

[0012] The space identification is performed based on the input and output data by using an N4-SID subspace identification method to obtain the state space model.

[0013] In the control method of the reactor startup separator, the space identification based on the input and output data by using an N4-SID subspace identification method to obtain the state space model comprises:

[0014] The initial model is obtained by model construction based on the historical data and the online data.

[0015] The historical data and the online data in the initial model are balanced by using a preset method in combination with a forgetting factor.

[0016] The state space model is obtained by order reduction processing on the model after the balancing processing.

[0017] In the control method of the reactor startup separator, the initial model is obtained by model construction based on the historical data and the online data.

[0018] The initial model is obtained by projection calculation based on the historical data and the online data by using a Hankel matrix.

[0019] In the control method of the reactor startup separator, the balancing processing of the historical data and the online data in the initial model by using a preset method in combination with a forgetting factor comprises:

[0020] The balancing processing of the historical data and the online data in the initial model is performed by using a sliding window algorithm in combination with the forgetting factor.

[0021] In the control method of the reactor startup separator, the prediction control instruction is obtained by solving based on the state space model.

[0022] The prediction control instruction is obtained by solving the state space model by using an MPC model predictive control algorithm.

[0023] In the control method of the reactor startup separator, the prediction control instruction is obtained by solving the state space model by using an MPC model predictive control algorithm.

[0024] The prediction model is obtained by calculation on the state space model.

[0025] determine a model predictive control function;

[0026] dynamically solve the prediction model based on the model predictive control function to obtain the prediction control instruction.

[0027] In the control method of the reactor startup separator, the method further comprises:

[0028] In the process of dynamically solving the prediction model, the output prediction value of the intermediate process is corrected.

[0029] In the control method of the reactor startup separator, the disturbance estimation value is obtained by the following steps:

[0030] Obtain observation data;

[0031] Based on the observation data and combined with the disturbance estimation expression, the disturbance estimation value is obtained.

[0032] The present application also provides a control system for a reactor startup separator, comprising: an N4-SID subspace identification system and an MPC model predictive control system;

[0033] The N4-SID subspace identification system comprises:

[0034] A data acquisition module is configured to obtain input and output data; the input and output data includes historical data and online data; the historical data and the online data both include pressure data, liquid level data and operating data of an actuator;

[0035] A subspace identification module is configured to perform subspace identification based on the input and output data to obtain a state space model;

[0036] The MPC model predictive control system comprises:

[0037] A model predictive control module is configured to solve based on the state space model to obtain a prediction control instruction;

[0038] A robust compensation module is configured to perform feedforward compensation on the prediction control instruction based on a disturbance estimation value to obtain a target control instruction;

[0039] A drive control module is configured to drive and control the actuator according to the target control instruction.

[0040] In the control system of the reactor startup separator, the subspace identification module comprises:

[0041] The model construction submodule is configured to construct a model based on the historical data and the online data to obtain an initial model.

[0042] The data balancing processing submodule is configured to balance the historical data and the online data in the initial model by using a preset method and in combination with a forgetting factor.

[0043] The order reduction processing submodule is configured to perform order reduction processing on the model after the balancing processing to obtain the state space model.

[0044] In the control system of the reactor startup separator, the model construction submodule is configured to perform projection calculation by using a Hankel matrix to obtain the initial model.

[0045] In the control system of the reactor startup separator, the data balancing processing submodule is configured to balance the historical data and the online data in the initial model by using a sliding window algorithm and in combination with the forgetting factor.

[0046] In the control system of the reactor startup separator, the model predictive control module is configured to solve the state space model by using an MPC model predictive control algorithm to obtain the predictive control instruction.

[0047] In the control system of the reactor startup separator, the model predictive control module comprises:

[0048] The prediction model calculation submodule is configured to calculate the state space model to obtain a prediction model.

[0049] The control function determination submodule is configured to determine a model predictive control function.

[0050] The dynamic solving submodule is configured to dynamically solve the prediction model based on the model predictive control function to obtain the predictive control instruction.

[0051] The control method and system of the reactor startup separator according to the present application have the following beneficial effects: the control method of the reactor startup separator comprises the steps of: obtaining input and output data; performing spatial identification based on the input and output data to obtain a state space model; performing solving based on the state space model to obtain a predictive control instruction; performing feedforward compensation on the predictive control instruction based on a disturbance estimation value to obtain a target control instruction; and driving the actuator according to the target control instruction. The present application obtains a state space model by performing spatial identification, and performs solving prediction based on the state space model, and performs feedforward compensation in combination with disturbance estimation, so as to not only solve the problems of multivariable coupling and mode switching hysteresis, but also solve the problem of insufficient disturbance resistance, and realize efficient and reliable control of the pressure and liquid level of the startup separator. BRIEF DESCRIPTION OF DRAWINGS

[0052] The present application will be further described below in combination with the drawings and embodiments, wherein:

[0053] Figure 1 is a flowchart of the control method of the reactor startup separator provided by the present application;

[0054] Figure 2 is a logic block diagram of the control system of the reactor startup separator provided by the present application;

[0055] Figure 3 is a flowchart based on N4-SID-MPC control provided by the present application;

[0056] Figure 4 is a system diagram of the startup separator provided by the present application;

[0057] Figure 5 is a control scheme comparison effect diagram. DETAILED DESCRIPTION

[0058] The technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0059] In order to solve the problems of automatic control of the pressure and liquid level of the reactor startup separator, coupling of the pressure and liquid level, etc., the present application provides a control method of a reactor startup separator.

[0060] Figure 1 shows one preferred embodiment of the control method of the reactor startup separator provided by the present application.

[0061] Specifically, as shown in Figure 1 The control method of the reactor startup separator in this embodiment includes the following steps:

[0062] Step S101: Obtain input and output data; the input and output data includes historical data and online data; the historical data and the online data both include pressure data, liquid level data and operating data of the actuator.

[0063] Step S102: Perform spatial identification based on the input and output data to obtain a state space model.

[0064] Step S103: Solve based on the state space model to obtain a predictive control instruction.

[0065] Step S104: Perform feedforward compensation on the predictive control instruction based on the disturbance estimation value to obtain a target control instruction.

[0066] Step S105: Drive and control the actuator according to the target control instruction.

[0067] The present application can dynamically capture system characteristics by updating the predictive model by performing spatial identification on the historical data and the online data to obtain a state space model and simultaneously updating the state space model online based on the subspace identification. Meanwhile, the optimal predictive control instruction can be generated by performing rolling solution of the multi-objective optimization problem on the state space model. The oscillation problem caused by pressure and liquid level control and the mode switching delay problem are effectively solved. In addition, the robust compensation can be realized by adding the disturbance estimation value to perform feedforward compensation on the predictive control instruction, which not only solves the problem of insufficient anti-interference performance but also guarantees the reliability of the system.

[0068] Optionally, in some embodiments, in step S101, the state space model is obtained by performing spatial identification based on the input and output data, including: performing spatial identification based on the input and output data by using the N4-SID subspace identification method to obtain the state space model. The state space model can be updated online by using the N4-SID subspace identification method to perform spatial identification.

[0069] In some embodiments, spatial identification is performed using the N4-SID subspace identification method based on input and output data to obtain a state-space model, including: constructing a model based on historical data and online data to obtain an initial model; balancing the historical data and online data in the initial model using a preset method in combination with a forgetting factor; and reducing the balanced model to obtain a state-space model. Constructing a model based on historical data and online data to obtain an initial model includes: performing a projection calculation using a Hankel matrix based on the historical data and online data to obtain the initial model. Balancing the historical data and online data in the initial model using a preset method in combination with a forgetting factor includes: balancing the historical data and online data in the initial model using a sliding window algorithm in combination with a forgetting factor.

[0070] Specifically, considering that the separator model is linear time-invariant, its discrete state space equation is as follows:

[0071]

[0072] (1) In the formula, x k ∈R n , is the state vector of the system at time k, R n is x k dimension, n is an integer greater than 0; u k ∈R m , is the input vector of the system at time k; y k ∈R m , is the output vector of the system at time k, R m for u k 、y k The dimension of m is an integer greater than 0. Where [A, B, C, D] is the system state space, that is, the coefficient matrix, and is the required identification parameter.

[0073] The model is constructed using the MOESP algorithm through formula (1). Specifically, based on the discrete state space equation of formula (1), the initial model is constructed through Hankel matrix projection calculation; then the sliding window algorithm (wherein the length of the sliding window = the dominant time constant of the system × 2) and the forgetting factor (λ = 0.95 ~ 0.99) are used to balance the historical data and online data in the initial model; then the model is reduced in order to obtain the final state space model. By constructing the initial model through Hankel matrix projection, the complexity of mechanism modeling and incremental update can be avoided. By using a sliding window for balancing, data explosion can be prevented and the identification time can be avoided from being too long. Through order reduction processing, the small eigenvalue modes of the state matrix A can be truncated and the dominant dynamics can be retained. The final state space model is as follows:

[0074] y(k+1) = Ax + Bu + Ke (2)

[0075] (2) where x represents a state vector of the system, u represents a control input vector, y represents a system output vector, A represents a state transition matrix, B represents an input control matrix, K represents a Kalman gain matrix, e represents an output prediction error (a noise term), and k represents a k-th time point.

[0076] In some embodiments, in step S102, the solving based on the state space model to obtain the prediction control instruction includes: solving the state space model by using an MPC model prediction control algorithm to obtain the prediction control instruction. In this case, the solving of the state space model by using the MPC model prediction control algorithm to obtain the prediction control instruction includes: calculating the state space model to obtain a prediction model; determining a model prediction control function; and dynamically solving the prediction model based on the model prediction control function to obtain the prediction control instruction. Further, in the process of dynamically solving the prediction model, the output prediction value of an intermediate process is corrected.

[0077] Specifically, the state space equation identified based on the N4-SID subspace identification method is used to solve the prediction model. In this case, the expression of the prediction model (prediction equation) is as follows:

[0078]

[0079] (3) where y is the output of the prediction equation, x and u are the inputs of the prediction equation.

[0080] Determination of the model prediction control function: a model prediction controller is designed, an optimization objective function (i.e., the model prediction control function) is defined, and thus the rolling optimization solving is realized to obtain the prediction control instruction. The model prediction control function is as follows:

[0081]

[0082] (4) where Np is a prediction step length, Nc is a control step length, Wp is a liquid level weight, Wp is a pressure weight, and Wv is a valve weight. p(k) represents the pressure at the k-th time point, p h p represents a pressure set value, h(k) represents the height at the k-th time point, h p p represents a pressure set value, h(k) represents the height at the k-th time point, h u p represents a pressure set value, h(k) represents the height at the k-th time point, h sp p represents a pressure set value, h(k) represents the height at the k-th time point, h sp p represents a pressure set value, h(k) represents the height at the k-th time point, h

[0083] Based on equation (4), the prediction control instruction is obtained by introducing a constraint condition and a transition trajectory design and using a weight dynamic adjustment continuous rolling optimization.

[0084] Due to model error, weak nonlinear characteristics and other uncertain factors existing in actual process, the open-loop optimal control law formula obtained according to the prediction model cannot necessarily lead the system output to closely follow the expected value, and it also cannot take into account the disturbance suffered by the object. In order to correct the inconsistency between model prediction and actual, the error information of the process must be used to correct the output prediction value of the intermediate process in time, so as to ensure the accuracy and reliability of the prediction control instruction. Among them, the feedback correction output is:

[0085] y p (k+1)=y(k+1)+[y(k)-y(k)](5);

[0086] At t=(k+1)T, the system predicted after error correction is:

[0087]

[0088] Further, in order to improve the anti-interference, the disturbance estimation value is added to adjust the prediction control instruction to obtain the target control instruction.

[0089] In the embodiment of the application, the disturbance estimation value is obtained by the following steps: obtaining observation data; based on the observation data and combining the disturbance estimation expression, the disturbance estimation value is obtained. Specifically, the observation data is obtained by a disturbance observer (DOB), the observation data is substituted into the disturbance estimation expression for calculation to obtain the disturbance estimation value, and the disturbance estimation value is injected into the prediction control instruction as a feedforward compensation to adjust (i.e. the disturbance estimation value is injected into the model predictive control function) to obtain the target control instruction.

[0090] Wherein, the disturbance estimation expression is as follows:

[0091]

[0092] (6) In the formula, represents the disturbance estimation value, y k represents the actual output value; represents the estimated output value; C k , D k represents the state space model parameter; u k represents the control output value.

[0093] The application will be described below with a specific embodiment.

[0094] As Figure 4As shown, in this example, based on the mechanism analysis of the startup separator working process of the nuclear power plant, a nonlinear model is established, and the state variables of the nonlinear model include: OTSG inlet flow, startup separator outlet flow, OTSG outlet pressure, startup separator liquid level, exhaust valve opening degree and drain valve opening degree. Among them, the output variables include: OTSG outlet pressure and startup separator liquid level. The input variables include: exhaust valve opening degree and drain valve opening degree. The pressure dynamic equation is as follows:

[0095]

[0096] (7) In the formula, represents the pressure change rate, which is determined by the steam input mass (m in ), exhaust valve outflow mass and condensation loss mass. Wherein, R is the gas constant of the working medium; T is the working medium temperature; V is the volume of the separator; K P (u2) is the flow coefficient of the exhaust valve; K cond is the condensation heat transfer coefficient; T sat is the saturation temperature (unit: K); T wall is the temperature of the separator (unit: K); p represents the pressure measurement value, and p0 represents the pressure set value.

[0097] The liquid level dynamic equation is:

[0098]

[0099] (8) In the formula, is the liquid level change rate, which is determined by the difference between the inflow and outflow. Wherein, A is the cross-sectional area of the separator; p is the working medium density; Q in is the mass flow rate of the steam generator entering the separator; C v (u1) is the flow coefficient of the drain valve; h0 is the reference liquid level of the drain valve outlet (unit: m). The steam side flow and the water side flow are selected as the input, and the pressure and the liquid level are selected as the output to establish a two-input two-output model of the separator.

[0100] According to the collected data u k and y k , k = 1…n. The number of rows of the Hankel matrix is 2*k, and the number of columns of the overall Hankel matrix is: N = 2*k+1. The overall Hankel matrix is split in half according to the number of rows, and the Hankel matrices U p , U f , Y p and Y f are obtained as follows:

[0101] Past input / output Hankel matrix:

[0102]

[0103] Future input / output Hankel matrix:

[0104]

[0105] Further, through matrix QR decomposition and SVD decomposition, the following relationship can be constructed:

[0106]

[0107] (9) In the formula, indicates the predicted next state (at time step k+1); Y k∣k indicates the estimated output based on the current information (at time step k); indicates the current state estimate (at time step k); U k∣k indicates the current input vector (at time step k); all indicate coefficient vector matrices.

[0108] Wherein, the state prediction equation is: The state at the next time is predicted by the state prediction equation, which is calculated based on the current state and input.

[0109] The output estimation equation is: The current output can be estimated by the output estimation equation, which is calculated based on the current state and input.

[0110] (9) The formula is a linear matrix equation of a system, and the matrix parameters of the system are estimated by least square method:

[0111]

[0112] Further, the prediction model is obtained by the state space model, and the calculation expression of the prediction model is as follows:

[0113]

[0114] The model predictive controller is designed, and the optimization objective function is defined:

[0115]

[0116] Wherein, N p = 20 is the prediction step, N c = 5 is the control step, and the constraint conditions are introduced: valve opening and liquid level limit value: 0≤u1,u2<100%, |Δu1|≤2% / s, |Δu2|≤5% / s, 20%<h<80%.

[0117] Weight dynamic adjustment: liquid level weight: Wh =0.3h≥45%0.3+0.01(50-h)h<45%, pressure weight: W p =0.7, valve weight: W u =0.05.

[0118] Comparison and verification of the schemes are carried out. Compared with the traditional PID control, it can be seen that the control method of the present invention is superior to the traditional PID control in terms of response speed and overshoot. Figure 5 shown.

[0119] refer to Figure 2 The present invention also provides a control system for a reactor startup separator. The control process is as follows: Figure 3 In a preferred embodiment, as shown Figure 2 As shown, the control system of the reactor startup separator includes: an N4-SID subspace identification system 10 and an MPC model predictive control system 20.

[0120] The N4-SID subspace identification system 10 includes a data acquisition module 11 for acquiring input and output data, including historical and online data, both of which include pressure data, liquid level data, and actuator operation data. A space identification module 12 is used to perform space identification based on the input and output data to obtain a state-space model. The data acquisition module 11 acquires pressure and liquid level data using sensors such as pressure transmitters and liquid level gauges. Actuators may include, but are not limited to, regulating valves such as steam traps and exhaust valves, and their operation data can be acquired through feedback monitoring.

[0121] In some embodiments, the spatial recognition module 12 includes: a model construction submodule for constructing a model based on historical data and online data to obtain an initial model; a data balancing submodule for balancing the historical data and online data in the initial model using a preset method and incorporating a forgetting factor; and a state-space reduction submodule for performing order reduction on the balanced model to obtain a state-space model. The model construction submodule performs projection calculations using a Hankel matrix to obtain the initial model. The data balancing submodule uses a sliding window algorithm and incorporating a forgetting factor to balance the historical data and online data in the initial model.

[0122] The MPC model predictive control system 20 comprises a model predictive control module 21 configured to solve a state space model to obtain a predictive control instruction; a robust compensation module 22 configured to feed forward compensate the predictive control instruction based on a disturbance estimation value to obtain a target control instruction; and a drive control module 23 configured to drive control the actuator according to the target control instruction. Optionally, the model predictive control module 21 is configured to solve the state space model by using an MPC model predictive control algorithm to obtain the predictive control instruction.

[0123] In some embodiments, the model predictive control module 21 comprises a prediction model calculation sub-module configured to calculate the state space model to obtain a prediction model; a control function determination sub-module configured to determine a model predictive control function; and a dynamic solving sub-module configured to dynamically solve the prediction model based on the model predictive control function to obtain the predictive control instruction.

[0124] Specifically, the cooperation and operation process between the units / modules in the control system of the reactor startup separator can refer to the control method of the reactor startup separator, which will not be described here.

[0125] The present application can realize efficient and reliable control of the pressure and liquid level of the startup separator by designing N4-SID, MPC control, combining disturbance estimation and dynamic switching optimization coefficient function, and controlling the valve action. By using MPC model predictive control, the influence of pressure and liquid level control can be effectively eliminated, and the coupling problem between the two can be solved.

[0126] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed by the embodiments, since it corresponds to the method disclosed by the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0127] The skilled person can further realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware, computer software or a combination of the two. In order to clearly show the interchangeability of hardware and software, the components and steps of each example have been described in the above description. Whether the functions are realized by hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0128] The steps of a method or algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module can reside in random access memory (RAM), flash memory, read-only memory (ROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0129] The above embodiments are only to illustrate the technical concept and characteristics of the present application, and the purpose is to enable those skilled in the art to understand the content of the present application and to implement it accordingly, and cannot limit the protection scope of the present application. Any equivalent changes and modifications made within the scope of the claims of the present application shall belong to the scope of the claims of the present application.

Claims

1. A method for controlling a reactor startup separator, characterized in that: The following steps are involved: Get input and output data; The input and output data include: historical data and online data; the historical data and the online data both include: pressure data, liquid level data and operating data of the actuator; Performing spatial identification based on the input and output data to obtain a state space model; Solving the state space model to obtain a predictive control instruction; Performing feedforward compensation on the predicted control instruction based on the disturbance estimation value to obtain a target control instruction; The actuator is driven and controlled according to the target control instruction.

2. The control method of a reactor startup separator according to claim 1, characterized in that: The performing spatial identification based on the input and output data to obtain a state space model includes: Based on the input and output data, the N4-SID subspace identification method is used to perform space identification to obtain the state space model.

3. The control method of the reactor startup separator according to claim 2, characterized in that: The step of performing space identification based on the input and output data using the N4-SID subspace identification method to obtain the state space model includes: Building a model based on the historical data and the online data to obtain an initial model; A preset method is used in combination with a forgetting factor to balance the historical data and the online data in the initial model; The model after the balancing process is reduced in order to obtain the state space model.

4. The control method for a reactor startup separator according to claim 3, characterized in that: The constructing of the model based on the historical data and the online data to obtain the initial model includes: Based on the historical data and the online data, projection calculation is performed through the Hankel matrix to obtain the initial model.

5. The control method for a reactor startup separator according to claim 3, characterized in that: The balancing process of the historical data and the online data in the initial model by using a preset method and combining a forgetting factor includes: A sliding window algorithm is used in combination with the forgetting factor to balance the historical data and online data in the initial model.

6. The control method for a reactor startup separator according to claim 1, characterized in that: Solving the state space model to obtain a predictive control instruction includes: The state space model is solved using an MPC model predictive control algorithm to obtain predictive control instructions.

7. The control method for a reactor startup separator according to claim 6, characterized in that: The adopting of the MPC model predictive control algorithm to solve the state space model to obtain the predictive control instruction comprises: Calculating the state space model to obtain a prediction model; determining a model predictive control function; The prediction model is dynamically solved based on the model predictive control function to obtain the predictive control instruction.

8. The control method for a reactor startup separator according to claim 7, characterized in that: The method further comprises: In the process of dynamically solving the prediction model, the output prediction value of the intermediate process is corrected.

9. The control method for a reactor startup separator according to any one of claims 1 to 8, characterized in that: The disturbance estimate is obtained by the following steps: Obtain observation data; The disturbance estimation value is obtained by performing calculation based on the observation data and in combination with a disturbance estimation expression.

10. A control system for a reactor startup separator, characterized in that: include: N4-SID subspace identification system and MPC model predictive control system; The N4-SID subspace identification system includes: A data acquisition module, the data acquisition module is used to obtain input and output data; the input and output data include: historical data and online data; the historical data and the online data both include: pressure data, liquid level data and operating data of the actuator; A space identification module, configured to perform space identification based on the input and output data to obtain a state space model; The MPC model predictive control system includes: A model predictive control module is configured to solve the state space model and obtain predictive control instructions; a robust compensation module, configured to perform feedforward compensation on the predictive control instruction based on a disturbance estimation value to obtain a target control instruction; A drive control module is used to drive and control the actuator according to the target control instruction.

11. The control system for a reactor startup separator according to claim 10, characterized in that: The space recognition module includes: A model building submodule, wherein the model building submodule is used to build a model based on the historical data and the online data to obtain an initial model; A data balancing processing submodule is used to balance the historical data and online data in the initial model using a preset method in combination with a forgetting factor; The order reduction processing submodule is used to perform order reduction processing on the model after the balancing processing is completed to obtain the state space model.

12. The control system for a reactor startup separator according to claim 11, characterized in that: The model building submodule performs projection calculation through the Hankel matrix to obtain the initial model.

13. The control system of the reactor startup separator according to claim 11, characterized in that: The data balancing processing submodule uses a sliding window algorithm in combination with the forgetting factor to perform balancing processing on the historical data and online data in the initial model.

14. The control system for a reactor startup separator according to claim 10, characterized in that: The model predictive control module uses the MPC model predictive control algorithm to solve the state space model and obtain predictive control instructions.

15. The control system for a reactor startup separator according to claim 14, characterized in that: The model predictive control module includes: A prediction model calculation submodule, the prediction model calculation submodule is used to calculate the state space model to obtain a prediction model; A control function determination submodule, wherein the control function determination submodule is used to determine a model prediction control function; A dynamic solution submodule is used to dynamically solve the prediction model based on the model predictive control function to obtain the predictive control instruction.

Images