MPC control method of industrial aromatic hydrocarbon combination unit

Abstract

The invention relates to an MPC control method of an industrial aromatic hydrocarbon combination unit. The MPC control method comprises the following steps: establishing a model of the industrial aromatic hydrocarbon combination unit; a state space model is adopted to describe: # imgabs0 # design parameter identification algorithm, and parameters A, B and C of the model are accurately estimated; designing an MPC controller based on the established state space model; a coordination control mechanism of the MPC is designed, the industrial aromatic hydrocarbon combination device is divided into a plurality of subsystems, each subsystem is controlled by a local MPC controller, and the subsystems are coordinated through information interaction; by improving the accuracy and adaptability of the model, optimizing a parameter identification algorithm and designing an effective coordination control mechanism, accurate control of the industrial aromatic hydrocarbon combination unit is realized, the production efficiency, the product quality and the stability are improved, and the energy consumption is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial automation control, and particularly to an MPC control method for an industrial aromatic hydrocarbon combined unit. Background Art

[0002] An industrial aromatic hydrocarbon combined unit is a complex chemical production system, including multiple unit operations such as reforming, extraction, distillation, isomerization, and adsorption separation. Each unit is interrelated and interacts with each other. Traditional control methods are difficult to adapt to the complexity and dynamic characteristics of the aromatic hydrocarbon combined unit, and cannot achieve precise control of the entire production process, resulting in problems such as low production efficiency, unstable product quality, and high energy consumption.

[0003] Model Predictive Control (MPC), as an advanced control strategy, has the advantages of being able to handle multivariable constraints and predicting the future behavior of the system, and has been widely applied in the chemical production process. However, when the existing MPC methods are applied to an industrial aromatic hydrocarbon combined unit, there are still some problems. For example, the accuracy and adaptability of the model are insufficient, and it is difficult to accurately describe the complex nonlinear and time-varying characteristics of the aromatic hydrocarbon combined unit; the parameter identification method is not efficient and accurate enough, resulting in the model parameters not being able to reflect the changes in the production process in a timely manner; there is a lack of an effective coordinated control mechanism, and the synergistic effect between each unit cannot be fully exerted. Therefore, it is necessary to develop an advanced MPC control method suitable for an industrial aromatic hydrocarbon combined unit. Summary of the Invention

[0004] The purpose of the present invention is to provide an MPC control method for an industrial aromatic hydrocarbon combined unit, which realizes precise control of the industrial aromatic hydrocarbon combined unit, improves production efficiency, product quality and stability, and reduces energy consumption by improving the accuracy and adaptability of the model, optimizing the parameter identification algorithm, and designing an effective coordinated control mechanism.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is: an MPC control method for an industrial aromatic hydrocarbon combined unit, including the following steps:

[0006] Step 1: Establish a model of the industrial aromatic hydrocarbon combined unit; described by using a state-space model:

[0007]

[0008] Wherein, is the state vector of the system, is the control input vector, is the output vector of the system and are the coefficient matrices of the system, is the process noise vector, is the measurement noise vector;

[0009] Step 2: Design a parameter identification algorithm to accurately estimate the parameters A, B, and C of the model;

[0010] Step 3: Based on the established state-space model, design an MPC controller;

[0011] Step 4: Design a coordinated control mechanism based on distributed MPC. Divide the industrial aromatics combined unit into multiple subsystems. Each subsystem is controlled by a local MPC controller, and the subsystems are coordinated through information interaction. Define a coordination variable z to describe the correlation between subsystems.

[0012] Preferably, the state vector x(k) contains 5 elements, namely the reforming reactor temperature x1(k), the distillation column pressure x2(k), the aromatic hydrocarbon concentration x3(k) in the reforming reactor, the distillation column feed flow rate x4(k), and the distillation column liquid level x5(k).

[0013] Preferably, the control input vector u(k) contains 3 elements, namely the feed flow rate u1(k) into the reforming reactor, the reflux ratio u2(k) of the distillation column, and the cooling medium flow rate u3(k).

[0014] Preferably, the output vector y(k) contains 3 elements, namely the purity y1(k) of the finally produced benzene, the overall energy consumption y2(k) of the unit, and the toluene content y3(k) in the top product of the distillation column.

[0015] Preferably, the parameter identification algorithm is the recursive least squares algorithm (RLS) based on an adaptive genetic factor; define the error vector where is the predicted output of the model;

[0016] The iterative formula of the recursive least squares method is as follows:

[0017]

[0018] where, is the parameter vector to be identified (including the elements of A, B, C), K(k) is the gain matrix, P(k) is the covariance matrix, λ ∈ (0, 1) is the forgetting factor, is the regression vector.

[0019] Preferably, introduce an adaptive forgetting factor adjustment strategy; according to the dynamic characteristics of the system and the error change situation, adjust the value of the forgetting factor λ in real time. When the system is in a stage of large dynamic changes, reduce the value of the forgetting factor to make the algorithm pay more attention to recent data; when the system tends to be stable, increase the value of the forgetting factor to improve the stability of the algorithm. The implementation steps of the adaptive forgetting factor adjustment strategy are as follows:

[0020] ① Calculate the variance of the error vector e(k) Calculate the variance of recent errors through the sliding window method, that is where M is the window size;

[0021] ② Define a dynamic index D(k), for example where is the moving average of the error variance; D(k) reflects the relative magnitude of the current error variance and the average error variance, and is used to measure the degree of dynamic change of the system;

[0022] ③ Adjust the forgetting factor λ(k) according to the dynamic index D(k), and set the thresholds D th1 and D th2 (D th1 < D th2 ), when D(k) > D th2 , it indicates that the system has a large dynamic change, and reduce the forgetting factor λ(k) = λ(k - 1) - Δλ, where Δλ is the forgetting factor adjustment step size; when D(k) < D th1 , the system tends to be stable, and increase the forgetting factor λ(k) = λ(k - 1) + Δλ; when D th1 ≤ D(k) ≤ D th2 , keep the forgetting factor unchanged λ(k) = λ(k - 1).

[0023] Preferably, the goal of the MPC controller is to minimize a performance index function under the condition of satisfying the system constraints:

[0024]

[0025] where, N p is the prediction horizon, N c is the control horizon, y(k + i|k) is the output prediction of the system at time k for the future time k + i, r(k + i) is the reference trajectory, and are the weight matrices, Δu(k + i|k) = u(k + i|k) - u(k + i - 1|k) is the increment of the control input. At the same time, consider the system constraint conditions, including the input constraint u min ≤ u(k + i|k) ≤ u max and the output constraint

[0026] y min ≤ y(k + i|k) ≤ y max

[0027] By solving the above optimization problem, the optimal control input sequence u * (k), u* (k + 1), …, u * (k + N c - 1), and apply the first control input u * (k) to the system.

[0028] Preferably, the performance index function of each subsystem is modified to:

[0029]

[0030] where j represents the number of the subsystem, y j , u j and r j are respectively the output, control input and reference trajectory of subsystem j, Q j , R j and S j are weight matrices, is the desired value of the coordination variable of subsystem j; through the iterative optimization method, continuously adjust the value of the coordination variable z to make the optimization results of each subsystem reach the global optimum.

[0031] The technical effects of the present invention are as follows: The present invention focuses on the MPC control of industrial aromatic hydrocarbon combined units. Through a series of innovative designs and technical improvements, it has significant advantages in improving the production efficiency, product quality and stability of the units. In the model establishment link, the state - space model is used to describe the complex dynamic characteristics of industrial aromatic hydrocarbon combined units, laying a solid foundation for the implementation of subsequent control strategies. The parameter identification algorithm combines the recursive least - squares method with an adaptive forgetting factor, which can track the system changes in real - time, accurately estimate the model parameters, and ensure a high degree of fit between the model and the actual operation of the unit.

[0032] The MPC controller design effectively realizes the optimal control of the unit by minimizing the carefully constructed performance index function under the premise of meeting the system constraints. The controller fully considers the output tracking error and the cost of control input changes, and can generate the optimal control input sequence according to the current state and prediction information of the system, improving the control accuracy and response speed of the unit. The coordinated control mechanism based on distributed MPC is a major highlight of the present invention. The unit is divided into multiple subsystems. Through information interaction and the setting of coordination variables, while optimizing their own performance, each subsystem fully considers global coordination, enabling the operation of the entire industrial aromatic hydrocarbon combined unit to reach the global optimum state and enhancing the unit's ability to cope with complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic flow chart of an MPC control method for an industrial aromatic hydrocarbon combined unit. DETAILED DESCRIPTION OF THE INVENTION

[0034] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0035] The specific technical solution of the present invention is as follows:

[0036] 1. Model establishment of industrial aromatics combined unit

[0037] Considering the multivariable, nonlinear and time-varying characteristics of the industrial aromatics combined unit, a state-space model is used to describe the system

[0038]

[0039] Among them, is the state vector of the system, is the control input vector, is the output vector of the system and are the coefficient matrices of the system, is the process noise vector, is the measurement noise vector.

[0040] Combined with the specific process operations of industrial aromatics, the state vector covers the following key elements:

[0041] ① Temperature-related: The temperatures at different positions in each key reactor (such as reforming reactor, isomerization reactor, etc.), because temperature has a decisive impact on the chemical reaction rate, product selectivity, etc. in aromatics production, and is an important parameter reflecting the internal reaction state of the unit; the internal temperature state of heat source equipment such as heating furnaces, which directly affects the heat provided to the reaction or separation process.

[0042] ② Pressure-related: The pressure conditions in the reactor, as pressure affects the reaction equilibrium and direction; the pressures at different trays or positions in separation equipment such as distillation columns and extraction columns, and the pressure conditions affect the gas-liquid equilibrium of each component in the mixture, thereby affecting the separation effect.

[0043] ③ Substance composition-related: The concentrations of various reactants and products in the reactor, for example, the concentrations of different hydrocarbons such as alkanes, cycloalkanes, and aromatics in the reforming reactor, which determine the reaction process and product yield; the contents of each aromatic component (benzene, toluene, xylene, etc.) and non-aromatic components in the feed, discharge and different positions in the distillation column, reflecting the state of the separation process.

[0044] ④Flow related: The flow status of each material in the device, such as the flow rate of raw materials entering the reactor, the feed and discharge flow rates of each column, etc. Changes in flow rate will affect the material balance of reaction and separation.

[0045] ⑤Liquid level related: The liquid level height in equipment such as storage tanks and columns. The liquid level status affects the storage and flow balance of materials and is one of the important indicators for the stable operation of the device.

[0046] The control input vector mainly includes:

[0047] ①Heating / cooling medium flow control: Control the fuel flow rate entering the heating furnace to adjust the heat provided by the heating furnace, thereby controlling the temperature required for the reaction or separation process; adjust the flow rate of the cooling medium (such as water, air, etc.) to control the temperature of equipment such as reactors and coolers and keep it within the appropriate operating range.

[0048] ②Feed flow control: Adjust the flow rate of raw materials entering the reactor to control the scale and rate of the reaction; control the feed flow rate entering separation equipment such as distillation columns and extraction columns, which affects the separation effect and production capacity.

[0049] ③Pressure regulation control: Control the pressure in equipment such as reactors and columns by adjusting the operating status of equipment such as compressors and vacuum pumps to meet the process requirements.

[0050] ④Reflux ratio control: In the distillation process, adjust the ratio of the liquid amount returned to the top of the column to the distillate amount (reflux ratio) to optimize the separation efficiency and product purity of the distillation column.

[0051] ⑤Solvent flow control: In the aromatics extraction process, control the flow rate of the extraction solvent (such as sulfolane, etc.) to ensure the effective extraction and separation of aromatics.

[0052] The output vector mainly includes:

[0053] ①Product quality index related: The purity and yield of aromatic products such as benzene, toluene, and xylene produced finally, which are the key indicators to measure the production effect of the device; the content of impurities in the product, such as the content of sulfur, nitrogen, etc. in the product, reflects the quality level of the product.

[0054] ②Process operation parameter related: The temperature, pressure, and composition of the material at the reactor outlet, which reflect the results and status of the reaction; the temperature, pressure, and composition of the products at the top and bottom of the distillation column, which reflect the separation effect of the distillation process; the overall energy consumption index of the device, such as the energy consumption per unit output, which reflects the operation efficiency and economy of the device.

[0055] To build a reasonable system model and facilitate the design of the data set collection and construction plan, let's assume:

[0056] ① State vector x(k): It contains 5 elements, namely the reforming reactor temperature x1(k), the distillation column pressure x2(k), the aromatic hydrocarbon concentration x3(k) in the reforming reactor, the distillation column feed flow rate x4(k), and the distillation column liquid level x5(k). These elements can comprehensively reflect the key operating states of the unit and provide important basis for subsequent control decisions.

[0057] ② Control input vector u(k): It contains 3 elements, namely the feed flow rate u1(k) into the reforming reactor, the reflux ratio u2(k) of the distillation column, and the cooling medium flow rate u3(k). By adjusting these control inputs, the operating performance of the unit can be directly affected, and effective control of the production process can be achieved.

[0058] ③ Output vector y(k): It contains 3 elements, namely the purity y1(k) of the finally produced benzene, the overall energy consumption y2(k) of the unit, and the toluene content y3(k) in the top product of the distillation column. These output vectors are directly related to the product quality and the operating efficiency of the unit, and are important manifestations of the control objectives.

[0059] 2. Parameter Identification Algorithm Design

[0060] To accurately estimate the parameters A, B, and C of the model, a recursive least squares (RLS) algorithm based on an adaptive genetic factor is designed. Define the error vector where is the predicted output of the model.

[0061] The iterative formula of the recursive least squares method is as follows:

[0062]

[0063] where, is the parameter vector to be identified (including the elements of A, B, C), K(k) is the gain matrix, P(k) is the covariance matrix, λ ∈ (0, 1) is the forgetting factor, is the regression vector.

[0064] To improve the accuracy and adaptability of parameter identification, an adaptive forgetting factor adjustment strategy is introduced. According to the dynamic characteristics of the system and the change of errors, the value of the forgetting factor λ is adjusted in real time. When the system is in a stage of large dynamic changes, the value of the forgetting factor is reduced to make the algorithm pay more attention to recent data; when the system tends to be stable, the value of the forgetting factor is increased to improve the stability of the algorithm. In practical applications, the implementation steps of the adaptive forgetting factor adjustment strategy are as follows:

[0065] ① Calculate the variance of the error vector e(k) The variance of recent errors can be calculated by the sliding window method, that is where M is the window size.

[0066] ② Define a dynamic index D(k), for example where is the moving average of the error variance. D(k) reflects the relative magnitude of the current error variance and the average error variance, and is used to measure the degree of dynamic change of the system.

[0067] ③ Adjust the forgetting factor λ(k) according to the dynamic index D(k). Set thresholds D th1 and D th2 (D th1 < D th2 ), when D(k) > D th2 , it indicates that the system has a large dynamic change, and the forgetting factor λ(k) is decreased as λ(k) = λ(k - 1) - Δλ, where Δλ is the forgetting factor adjustment step size; when D(k) < D th1 , the system tends to be stable, and the forgetting factor λ(k) is increased as λ(k) = λ(k - 1) + Δλ; when D th1 ≤ D(k) ≤ D th2 , the forgetting factor remains unchanged as λ(k) = λ(k - 1).

[0068] 3. MPC Controller Design

[0069] Based on the established state - space model, design an MPC controller. The goal of the MPC controller is to minimize a performance index function under the condition of satisfying system constraints:

[0070]

[0071] where, N p is the prediction horizon, N c is the control horizon, y(k + i|k) is the output prediction of the system at time k for the future time k + i, r(k + i) is the reference trajectory, and are the weight matrices, Δu(k + i|k) = u(k + i|k) - u(k + i - 1|k) is the increment of the control input. At the same time, consider the system constraints, including the input constraint u min ≤ u(k + i|k) ≤ u max and the output constraint

[0072] y min ≤ y(k + i|k) ≤ y max

[0073] By solving the above optimization problem, obtain the optimal control input sequence u * (k), u * (k + 1), …, u * (k + N c-1), and apply the first control input u * (k) to the system.

[0074] 4. Coordinated Control Mechanism Design

[0075] To give full play to the synergy between units of the industrial aromatics combined plant, a coordinated control mechanism based on distributed MPC is designed. The industrial aromatics combined plant is divided into multiple subsystems, and each subsystem is controlled by a local MPC controller. The subsystems are coordinated through information interaction. Define a coordination variable z to describe the correlation between subsystems. In the optimization process, the MPC controller of each subsystem not only considers its own performance index and constraint conditions but also the influence of the coordination variable. Specifically, the performance index function of each subsystem is modified to:

[0076]

[0077] where j represents the number of the subsystem, y j , u j and r j are the output, control input, and reference trajectory of subsystem j respectively, Q j , R j and S j are weight matrices, is the desired value of the coordination variable for subsystem j.

[0078] Through iterative optimization, continuously adjust the value of the coordination variable z to make the optimization results of each subsystem reach the global optimum.

[0079] Specific implementation method of applying the above scheme

[0080] ① System initialization

[0081] a. Collect data of the industrial aromatics combined plant to obtain the input and output data of the system for model establishment and parameter identification. b. Initialize the parameters A, B, and C of the state-space model, as well as the covariance matrix P(0), forgetting factor λ(0), weight matrices Q, R, and S, etc.

[0082] c. Determine the prediction horizon N p , control horizon N c and input-output constraint conditions u min , u max , y min and y max .

[0083] ② Parameter identification

[0084] a. At each sampling time k, collect the input u(k) and output y(k) data of the system.

[0085] b. Calculate the regression vector according to the iterative formula of the recursive least squares method Gain matrix K(k) and covariance matrix P(k).

[0086] c. Calculate the error vector e(k) and update the parameter vector to be identified

[0087] d. Adjust the value of the forgetting factor λ(k) according to the dynamic characteristics of the system and the change of the error

[0088] ③ MPC control

[0089] a. According to the system state x(k) at the current moment and the identified model parameters, predict the future output of the system to obtain y(k+i|k) (i = 0, 1, …, N p -1).

[0090] b. Construct the performance index function J of the MPC controller and consider the input-output constraint conditions of the system

[0091] c. Use the quadratic programming interior point algorithm to solve the optimization problem to obtain the optimal control input sequence

[0092] u * (k), u * (k + 1), …, u * (k + N c -1).

[0093] The specific steps are as follows

[0094] · Transform the optimization problem of MPC into the standard quadratic programming form The constraint condition is A cq u = b cq A ineq u ≤ b ineq . Among them, H, g, A eq , b eq , A ineq , b ineq Are derived according to the performance index function and constraint conditions of MPC

[0095] · Introduce slack variables s and barrier functions Transform the inequality constraint into an equality constraint and construct an augmented

[0096] Objective function Where μ is the Lagrange multiplier

[0097] · Use the Newton method to solve the optimal solution of the augmented objective function. Calculate the gradient of the augmented objective function and the Hessian matrix

[0098] By iteratively updating u, s, and μ: [u t+1 , s t+1 , μ t+1 = [u t , s t , μ t + α[Δu, Δs, Δμ], where α is the step size,

[0099] [Δu, Δs, Δμ] is obtained by solving the linear equations Get.

[0100] · When the convergence condition is met (such as ∈ is the preset convergence accuracy), stop the iteration and obtain the optimal control input sequence.

[0102] d. Apply the first control input u * (k) to the industrial aromatics combined unit.

[0103] ④ Coordinated control

[0104] a. Divide the industrial aromatics combined unit into multiple subsystems, and set local MPC controllers for each subsystem. b. Information interaction is carried out between subsystems to obtain the value of the coordination variable z.

[0105] c. Each subsystem performs local optimization and solution according to its own performance index function J j and the influence of the coordination variable. d. By means of iterative optimization, continuously adjust the value of the coordination variable z to make the optimization results of each subsystem reach the global optimum.

[0106] The present invention focuses on the MPC control of the industrial aromatics combined unit. Through a series of innovative designs and technical improvements, it has significant advantages in improving the production efficiency, product quality, and stability of the unit. In the model establishment section, a state-space model is used to describe the complex dynamic characteristics of the industrial aromatics combined unit, laying a solid foundation for the implementation of subsequent control strategies. The parameter identification algorithm combines the recursive least squares method with an adaptive forgetting factor, which can track system changes in real time, accurately estimate model parameters, and ensure a high degree of fit between the model and the actual operation of the unit.

[0107] The design of the MPC controller can effectively achieve the optimal control of the device by minimizing a carefully constructed performance index function while satisfying the system constraints. This controller fully considers the output tracking error and the cost of control input changes, and can generate an optimal control input sequence based on the current state and prediction information of the system, improving the control accuracy and response speed of the device. The coordinated control mechanism based on distributed MPC is a major highlight of the present invention. The device is divided into multiple subsystems. Through information interaction and the setting of coordination variables, each subsystem not only optimizes its own performance but also fully considers global cooperation, enabling the operation of the entire industrial aromatics combined device to reach the global optimal state and enhancing the device's ability to handle complex working conditions.

[0108] From the perspective of practical applications, the technical solution of the present invention can effectively solve many problems existing in traditional control methods in industrial aromatics combined devices, and has good adaptability and scalability. It not only helps to improve product quality and production efficiency, reduce energy consumption, but also enhances the stability and reliability of the device operation, bringing significant economic and social benefits to enterprises and having broad application prospects in the field of industrial aromatics production.

[0109] It should be noted that in this article, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0110] In this article, specific examples are used to elaborate on the principle and implementation manner of the present invention. The description of the above examples is only used to help understand the method of the present invention and its core idea. The above is only the preferred implementation manner of the present invention. It should be pointed out that due to the limited nature of written expression and the objectively infinite specific structures, for those of ordinary skill in the art in this technical field, without departing from the principle of the present invention, several improvements, refinements or changes can be made, or the above technical features can be combined in an appropriate manner; these improvements, refinements, changes or combinations, or directly applying the concept and technical solution of the invention to other occasions without improvement, should all be regarded as the protection scope of the present invention.

Claims

1. An MPC control method for an industrial aromatics complex, characterized in that: The steps include: Step 1: Establish a model of the industrial aromatics complex; use the state space model to describe: in, is the state vector of the system, is the control input vector, is the output vector of the system and is the coefficient matrix of the system, is the process noise vector, is the measurement noise vector; Step 2: Design a parameter identification algorithm to accurately estimate the model parameters A, B and C; Step 3: Design the MPC controller based on the established state space model; Step 4: Design a coordinated control mechanism based on distributed MPC, divide the industrial aromatics complex into multiple subsystems, each subsystem is controlled by a local MPC controller, and the subsystems are coordinated through information exchange. A coordination variable z is defined to describe the relationship between the subsystems.

2. The MPC control method for an industrial aromatics complex according to claim 1, characterized in that: State vector x(k): contains 5 elements, namely reforming reactor temperature x1(k), distillation tower pressure x2(k), aromatics concentration in reforming reactor x3(k), distillation tower feed flow x4(k), and distillation tower liquid level x5(k).

3. The MPC control method for an industrial aromatics complex according to claim 1, characterized in that: Control input vector u(k): contains 3 elements, namely the raw material flow rate u1(k) entering the reforming reactor, the reflux ratio u2(k) of the distillation tower, and the cooling medium flow rate u3(k).

4. The MPC control method for an industrial aromatics complex according to claim 1, characterized in that: Output vector y(k): contains 3 elements, namely the purity of the final benzene output y1(k), the overall energy consumption of the device y2(k), and the toluene content in the top product of the distillation tower y3(k).

5. The MPC control method for an industrial aromatics complex according to claim 1, characterized in that: The parameter identification algorithm is a recursive least squares algorithm (RLS) based on adaptive genetic factors; the error vector is defined as in is the predicted output of the model; The iterative formula of the recursive least squares method is as follows: in, is the parameter vector to be identified (including elements of A, B, and C), K(k) is the gain matrix, P(k) is the covariance matrix, λ∈(0,1) is the forgetting factor, is the regression vector.

6. The MPC control method for an industrial aromatics complex according to claim 5, characterized in that: Introduce an adaptive forgetting factor adjustment strategy; adjust the value of the forgetting factor λ in real time according to the dynamic characteristics of the system and the change of errors. When the system is in a stage of large dynamic changes, reduce the value of the forgetting factor so that the algorithm pays more attention to recent data; when the system tends to be stable, increase the value of the forgetting factor to improve the stability of the algorithm. The implementation steps of the adaptive forgetting factor adjustment strategy are as follows: ① Calculate the variance of the error vector e(k) The variance of recent errors is calculated by the sliding window method, that is, Where M is the window size; ② Define a dynamic index D(k), for example in is the moving average of the error variance; D(k) reflects the relative size of the current error variance and the average error variance, and is used to measure the degree of dynamic change of the system; ③ According to the dynamic index D(k), adjust the forgetting factor λ(k) and set the threshold D th1 and D th2 (D th1 <D th2 ), when D(k)>D th2 When D(k) is smaller than λ(k-1)-Δλ, it indicates that the system changes dynamically. Reduce the forgetting factor λ(k)=λ(k-1)-Δλ, where Δλ is the forgetting factor adjustment step size. <D th1 When D th1 ≤D(k)≤D th2 When , the forgetting factor remains unchanged λ(k)=λ(k-1).

7. The MPC control method for an industrial aromatics complex according to claim 1, characterized in that: The goal of the MPC controller is to minimize a performance indicator function while satisfying system constraints: Among them, N p is the prediction time domain, N c is the control time domain, y(k+i|k) is the output prediction of the system at time k for the future time k+i, r(k+i) is the reference trajectory, and is the weight matrix, Δu(k+i|k)=u(k+i|k)-u(k+i-1|k) is the increment of the control input. At the same time, consider the constraints of the system, including the input constraint u min ≤u(k+i|k)≤u max and output constraints y min ≤y(k+i|k)≤y max By solving the above optimization problem, we can obtain the optimal control input sequence u * (k),u * (k+1),…,u * (k+N c -1), and input the first control * (k) Application to the system.

8. The MPC control method for an industrial aromatics complex according to claim 1, characterized in that: The performance indicator function of each subsystem is modified as follows: Where j represents the number of the subsystem, y j 、u j and r j are the output, control input and reference trajectory of subsystem j, Q j , R j and S j is the weight matrix, is the desired coordination variable value of subsystem j; through iterative optimization, the value of the coordination variable z is continuously adjusted to make the optimization results of each subsystem reach the global optimum.

Images