Model adaptive predictive control method and system based on disturbance state compensation

Abstract

The invention relates to the technical field of predictive control, and discloses a model adaptive predictive control method and system based on disturbance state compensation, and the method can recognize and compensate the system disturbance in real time through introducing an adaptive disturbance identifier based on deep learning, and improves the predictive control accuracy. The control precision and robustness of the system in a changeable disturbance environment are obviously improved; according to the method, the disturbance identifier is constructed by using a deep learning technology, so that the disturbance identification process under data driving is more intelligent, fine changes of disturbance can be captured more accurately, and the accuracy of disturbance identification is improved; a set of performance evaluation mechanism is established, an optimal controller is allowed to be automatically selected from a controller set, and control optimization of the system under disturbance is ensured; and through accurate error evaluation, disturbance compensation and prediction model correction, the accuracy and effectiveness of the control signal are further ensured. According to the technology and the method, the control effect is optimized, the service life of equipment is prolonged, and the maintenance cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of predictive control, and in particular to a model adaptive predictive control method and system based on disturbance state compensation. Technical Background

[0002] With the rapid development of industrial automation and intelligent manufacturing, predictive control technology, as an important modern control strategy, has been widely used in many fields. However, actual industrial systems often face disturbances of various types and sizes, such as changes in the external environment, load fluctuations, etc. These disturbance factors directly affect the stability and control accuracy of the system. Therefore, in order to improve the robustness of the system in complex environments, adaptive predictive control methods that can effectively identify and compensate for disturbances have gradually become a research hotspot.

[0003] Traditional predictive control methods usually rely on an accurate description of the system model. However, due to the unpredictability and complexity of disturbance factors, it becomes difficult to build an accurate system model. Especially in the case of complex and changeable disturbances, the control effect of the system is often difficult to achieve the expected goal. Summary of the invention

[0004] The present invention aims to disclose a model adaptive predictive control method and system based on disturbance state compensation to solve the problems existing in the prior art.

[0005] To achieve the above-mentioned object, in a first aspect, the present invention discloses a model adaptive predictive control method based on disturbance state compensation, comprising:

[0006] S1: Obtain historical data of system operation without disturbance, and obtain historical data of system operation under different disturbance types and sizes, build a data set, and divide the data set into a training subset and a test subset;

[0007] S2: Construct the basic model of adaptive disturbance identifier;

[0008] S3: constructing a parameter-optimal adaptive disturbance identifier model based on the training subset, the validation subset and the adaptive disturbance identifier basic model;

[0009] S4: constructing a system controller set; constructing a prediction model of the system under the nth controller, constructing a system prediction output matrix at time t+1 based on the prediction model, identifying a disturbance compensation vector of the system at time t+1 based on the prediction output matrix and the parameter optimal adaptive disturbance identifier model, and constructing a disturbance compensator based on the disturbance compensation vector;

[0010] S5: Calculate the error between the system expected output vector and the system predicted output vector at time t+1 based on the disturbance compensator, construct a controller evaluation function based on the error, and perform performance evaluation on all controllers; obtain the optimal controller of the system at time t+1 by minimizing the performance index; wherein t represents the time;

[0011] S6: Inputting the trigger signal corresponding to the optimal controller into the controller switcher, switching the controller, and performing prediction and control switching at the next moment.

[0012] To achieve the above-mentioned purpose, in a second aspect, the present invention also discloses a model adaptive predictive control system based on disturbance state compensation, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the first aspect when executing the computer program.

[0013] The present invention has the following beneficial effects

[0014] The model adaptive predictive control method based on disturbance state compensation provided by the present application can identify and compensate system disturbances in real time by introducing an adaptive disturbance identifier based on deep learning, significantly improving the control accuracy and robustness of the system in the face of a variable disturbance environment; the invention uses deep learning technology to construct a disturbance identifier, making the disturbance identification process under data drive more intelligent, and can more accurately capture the subtle changes of disturbances and improve the accuracy of disturbance identification; a set of performance evaluation mechanisms is established to allow the automatic selection of the optimal controller in the controller set to ensure the optimal control of the system under disturbance; through accurate error evaluation, disturbance compensation and correction of the prediction model, the accuracy and effectiveness of the control signal are further ensured. These technologies and methods not only optimize the control effect, but also extend the service life of the equipment and reduce maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of a model adaptive predictive control method based on disturbance state compensation according to a preferred embodiment of the present invention;

[0016] Figure 2 is a flowchart of the implementation steps of a preferred embodiment of the present invention;

[0017] Figure 3 is a control block diagram of a model adaptive predictive control system based on disturbance state compensation according to a preferred embodiment of the present invention;

[0018] Figure 4 It is an output effect diagram of a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0019] The technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the usual meanings understood by persons with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one" or "one" do not indicate quantity restrictions, but indicate the existence of at least one. Words such as "connect" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship also changes accordingly.

[0021] like Figure 1 As shown, this embodiment provides a model adaptive predictive control method based on disturbance state compensation, including:

[0022] S1: Obtain historical data of system operation without disturbance, and obtain historical data of system operation under different disturbance types and sizes, build a data set, and divide the data set into a training subset and a test subset;

[0023] S2: Construct the basic model of adaptive disturbance identifier;

[0024] S3: constructing a parameter-optimal adaptive disturbance identifier model based on the training subset, the validation subset and the adaptive disturbance identifier basic model;

[0025] S4: constructing a system controller set; constructing a prediction model of the system under the nth controller, constructing a system prediction output matrix at time t+1 based on the prediction model, identifying a disturbance compensation vector of the system at time t+1 based on the prediction output matrix and the parameter optimal adaptive disturbance identifier model, and constructing a disturbance compensator based on the disturbance compensation vector;

[0026] S5: Calculate the error between the system expected output vector and the system predicted output vector at time t+1 based on the disturbance compensator, construct a controller evaluation function based on the error, and perform performance evaluation on all controllers; obtain the optimal controller of the system at time t+1 by minimizing the performance index; wherein t represents the time;

[0027] S6: Inputting the trigger signal corresponding to the optimal controller into the controller switcher, switching the controller, and performing prediction and control switching at the next moment.

[0028] The above-mentioned model adaptive predictive control method based on disturbance state compensation can identify and compensate system disturbances in real time by introducing an adaptive disturbance identifier based on deep learning, significantly improving the control accuracy and robustness of the system in the face of variable disturbance environments; using deep learning technology to build a disturbance identifier makes the data-driven disturbance identification process more intelligent, and can more accurately capture subtle changes in disturbances and improve the accuracy of disturbance identification; a set of performance evaluation mechanisms is established to allow the automatic selection of the optimal controller from the controller set to ensure the optimal control of the system under disturbance; through precise error evaluation, disturbance compensation and correction of the prediction model, the accuracy and effectiveness of the control signal are further ensured. These technologies and methods not only optimize the control effect, but also extend the service life of the equipment and reduce maintenance costs.

[0029] In a complete example, first, obtain the system operation history data under the condition of no disturbance; obtain the system operation history data under different disturbance types and sizes, build a data set, and divide it into training subset and test subset; build an adaptive disturbance identifier basic model; use the training subset to train the disturbance identifier basic model; use the test subset to evaluate the disturbance identifier basic model; build the optimal parameter adaptive disturbance identifier model, specifically:

[0030] Obtain the historical data of system operation without disturbance, and record the system output sequence as m is the sequence dimension (number of sensors), L is the sequence length (number of sampling points);

[0031] Obtain the historical data of system operation under different disturbance types and sizes, and construct a data set, which is recorded as:

[0032]

[0033] Among them, S is a data set constructed from historical system operation data under different disturbance types and sizes, S j =(y j ,d j ) represents the j-th output sequence-disturbance vector pair (i.e., the j-th data subset) composed of the j-th system output sequence under the j-th disturbance (including disturbance type and disturbance size) and the j-th disturbance vector, j = 1, 2, …, J, J is the total number of disturbances (of different types and sizes), that is, the total number of output sequence-disturbance vector pairs (i.e., data subsets) in the data set S, is the jth system output sequence, represents the disturbance vector under the jth disturbance (including disturbance type and disturbance size), I is the total number of disturbance types, and the jth disturbance vector d j Satisfies the following expression:

[0034]

[0035] Among them, d ji ∈{d1,d2,…,d i ,…,d I} is the jth perturbation vector d j The i-th component (i.e., the i-th disturbance type) of ji =0 represents the jth disturbance vector d j There is no type i disturbance in (the disturbance size of the type i disturbance is 0), It means that there is an i-th type of disturbance in the j-th disturbance vector, and the disturbance size is q=1,2,…,Q, Q is the total number of disturbance sizes; d ji There are Q+1 possible values, perturbations (of different types and sizes) or perturbation vectors d j The total number of is J = I·(Q+1); in particular, when all d ji = 0, there is d j =[0,0,…,0,…,0], which means that all disturbances do not exist, that is, there is no disturbance.

[0036] The dataset S is randomly divided into training subsets according to the ratio of α:1. and validation subset They are used for model training and performance evaluation respectively. t , J v The training subsets S are train , validation subset S val The number of data subsets in J t +J v = J. α is the variable for dividing the data set.

[0037] Construct the basic model of adaptive disturbance identifier;

[0038] The input of the basic model of the adaptive disturbance identifier is the sequence y0 and the data set S; the basic model includes an input layer, a hidden layer, a fully connected layer and an output layer; the output layer generates a disturbance compensation vector.

[0039] The basic model of the adaptive disturbance identifier is G(y0,S;φ), where represents the parameter set of the adaptive disturbance identifier model, and A is the total number of model parameters.

[0040] Using the training subset S train, perform P rounds of training on the basic model of the disturbance identifier; use the validation subset S val Evaluate the basic model of disturbance identifier; construct the optimal parameter adaptive disturbance identifier model;

[0041] Set the number of training rounds p, where p = 1, 2, ..., P, and P is the total number of training rounds; set the initial regularization strength hyperparameter for the pth round of training Conduct the pth round of model training;

[0042] Set the number of iterations j in the p-th round of model training, j = 1, 2, ..., J t ;

[0043] Construct the loss function of the jth iteration of the pth round of training to satisfy the following expression:

[0044]

[0045] in, Indicates the system output sequence without disturbance in the pth round of training and the jth input data subset S j =(y j ,d j )∈S train The set of model parameters for the jth iteration The loss function is is the model parameter set of the jth iteration of the pth round of training, |·| represents the calculation symbol of the set size of the set within the symbol, Indicates the system output sequence without disturbance in the pth round of training and the jth input data subset S j ∈S train The jth system output sequence y j The set of model parameters for the jth iteration The disturbance compensation vector outputted below is The model parameter set for the jth iteration of the pth round of training The regularization strength hyperparameter under , Represents the model parameter set of the jth iteration of the pth round of training The regularization term under Ω(·) represents the regularization operation on the parameter set within the symbol;

[0046] Calculate the gradient of the loss function of the jth iteration of the pth round of training with respect to the model parameters. The expression is as follows (4):

[0047]

[0048] in, Represents the loss function For model parameters The gradient of is the disturbance compensation vector For model parameters The partial derivative of .

[0049] The gradient descent method is used to update the model parameters, and the update rule satisfies the following expression:

[0050]

[0051] in is the model parameter set for the j+1th iteration of the pth round of training, The learning rate for the kth iteration of the pth round of training;

[0052] Let j = j + 1, when j < J t season Go back to step "Construct the loss function of the jth iteration of the pth round of training" to execute;

[0053] When j ≥ J t When , the optimal model parameters of the p-th round of training are calculated by minimizing the loss function, satisfying the following expression:

[0054]

[0055] in, Indicates that the total number of J in the p-th round of training t The optimal model parameters obtained by iteration are Indicates that the total number of J in the p-th round of training t The set of model parameters for the iteration, Indicates that the independent variable φ p The value of the independent variable in the value set that makes the function in the brackets reach the minimum value;

[0056] Use the optimal model parameters from the pth round of training Constructing the p-th parameter optimal adaptive disturbance identifier model Using the validation subset S val Model Conduct assessments;

[0057] Using the pth parameter optimal adaptive disturbance identifier model The system output sequence y0 and the jth input data subset S in the absence of disturbance j =(y j ,d j )∈S val The jth system output sequence y j , calculate the optimal model parameters The disturbance compensation vector of the next j-th output Recorded as j=1,2,…,J v , a total of J v A disturbance compensation vector:

[0058] Construct model evaluation indicators to satisfy the following expressions:

[0059]

[0060] Among them, MSE p represents the p-th parameter optimal adaptive disturbance identifier model Evaluation indicator, Δy j =y j -y0 represents the current j-th input data subset S j =(y j ,d j ) in which the system outputs the sequence y j The difference between the system output sequence y0 in the unperturbed case; ||·|| 2 It means to perform square sum operation on the inner values ​​of symbols;

[0061] Set the threshold MSE of the model evaluation indicator th , if MSE p <MSE th , then the optimal adaptive disturbance identifier model for the current pth parameter is That is, the optimal parameter adaptive disturbance identifier model constructed after P = p rounds of training

[0062] If MSE p ≥MSE th , then let p=p+1 and go back to step "Set the number of iterations for the p-th round of model training" to execute.

[0063] Then, a system controller set is constructed; a prediction model of the system under the nth controller is constructed to obtain the system prediction output vector at time t+1; a system prediction output matrix at time t+1 is constructed; the disturbance compensation vector of the system at time t+1 is identified using the parameter optimal adaptive disturbance identifier model; and a disturbance compensator is constructed:

[0064] Design N system controllers respectively and construct a system controller set to satisfy the following expression:

[0065] Γ={Γ1,Γ2,…,Γ n ,…,Γ N} (8)

[0066] Among them, Γ is the system controller set, Γ nis the nth controller, n=1,2,…,N, N is the total number of system controllers;

[0067] Construct the system in the nth controller Γ n The prediction model below is used to predict the system output at time t+1. The prediction model satisfies the following expression:

[0068]

[0069] in, are the system state vectors at time t and time t+1 respectively, q is the system state vector dimension, are the system control input vectors at time t and time t+1 respectively, For x (t) , μ (t) , μ (t+1) The predicted system output vector at time t+1, m is the dimension of the system control input vector and output vector; is the nth controller Γ n The state transition matrix, is the nth controller Γ n The control input matrix, is the nth controller Γ n The output matrix of is the nth controller Γ n The direct transfer matrix of .

[0070] The L output vectors are collected in real time to form the system output matrix at the current time t, which satisfies the following expression:

[0071]

[0072] In the formula, is the nth controller Γ n The system output matrix at time t, L is the number of columns in the output matrix, that is, the length of the output sequence (number of sampling points); are the nth controller Γ n The system output vector (measured value) at t=1,2,3,…,t-1,t;

[0073] Construct the system prediction output matrix at time t+1, which satisfies the following expression:

[0074]

[0075] in, is the system prediction output matrix at time t+1; represents the nth controller Γ n The system output vector at time t is: Indicated by x(t) , μ (t) , μ (t+1) The predicted system output vector at time t+1.

[0076] Specifically, the system prediction output matrix at time t+2 is constructed as:

[0077]

[0078] in, is the system output vector (sampling value) at time t+1, According to the prediction model (9), x (t +1) , μ (t+1) , μ (t+2) The predicted system output vector at time t+2;

[0079] Use the optimal parameter adaptive disturbance identifier model trained in the first step System prediction output matrix at time t+1 The disturbance compensation vector at time t+1 is predicted, and the prediction result satisfies the following expression:

[0080]

[0081] in, is the adaptive disturbance identifier model The predicted disturbance compensation vector at time t+1; Indicates that in the nth controller Γ n The system output sequence is And the system prediction output matrix at time t+1 As input, after is the predicted disturbance compensation vector output by the adaptive disturbance identifier model of the model parameters;

[0082] Construct a disturbance compensator to satisfy the following expression:

[0083]

[0084] in, is the nth controller Γ n The predicted output vector of the system at time t+1 after disturbance compensation.

[0085] Furthermore, the error between the system expected output vector and the system predicted output vector at time t+1 is calculated, and the controller evaluation function is constructed to evaluate the performance of all controllers. By minimizing the performance index, the optimal controller of the system at time t+1 is obtained:

[0086] Calculate the error between the system expected output vector and the system predicted output vector at time t+1, which satisfies the following expression:

[0087]

[0088] In the formula represents the expected output y of the system at time t+1 *(t+1) With the nth controller Γ n The system predicts the output vector The error vector between *(t+1) is the expected output vector of the system at time t+1, y *(t+1) =r (t+1) , r (t+1) is the value of the system's given task curve r(t) at time t+1;

[0089] For the nth controller Γ n Perform performance evaluation and construct an evaluation function that satisfies the following expression:

[0090]

[0091] in, is the nth controller Γ n The performance index of is, τ is a counting variable;

[0092] Take n = 1, 2, ..., N, and calculate the performance index set of N controllers:

[0093]

[0094] By minimizing the performance index, the optimal controller at time t+1 is calculated to satisfy the following expression:

[0095]

[0096] In the formula represents the optimal controller selected at time t+1, It represents the value n* of the independent variable when the function in the brackets reaches the minimum value in the set of values ​​of the independent variable n. n* is the subscript of the optimal controller, n*∈{1,2,…,N}.

[0097] Furthermore: the control system switches to the optimal controller at time t+1:

[0098] The optimal controller is selected Corresponding trigger signal Input to the controller switcher, the system switches to the nth controller at time t+1 Run below;

[0099] Let n*=n, t=t+1, and go to execute the build system in the nth controller Γ n The prediction model under is used to predict the steps of the system output at time t+1, and perform prediction and control switching at the next moment.

[0100] In other words, equations (1) to (16) constitute a model adaptive predictive control method based on disturbance state compensation.

[0101] Specifically, taking the aircraft control system as an example, the parameters of this embodiment are specifically shown in Table 1.

[0102] Table 1 Parameters of the embodiment

[0103]

[0104] The control block diagram of the aircraft control system is as follows Figure 3 As shown in the figure, when the system detects a disturbance, the optimal controller is selected for switching by predicting the output of the model at the next moment and compensating the disturbance state to ensure the robustness and rapid response of the system under different disturbance environments.

[0105] The response curve of the aircraft control system is as follows: Figure 4 As shown in the figure, from the response curve, when there is no disturbance (no disturbance curve), the system responds smoothly. When there is a disturbance but the controller is not switched (disturbance-no switching controller curve), the response is slow, there are large fluctuations, and the convergence performance is poor. By using the model adaptive predictive control method based on disturbance state compensation, when the controller is switched (disturbance-switching controller curve), the system can quickly adjust when a disturbance occurs, showing good dynamic response and faster convergence speed, and finally stabilizes near the target value.

[0106] The embodiment of the present application also provides a model adaptive predictive control system based on disturbance state compensation, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the method. The model adaptive predictive control system based on disturbance state compensation can implement various embodiments of the above-mentioned model adaptive predictive control method and achieve the same beneficial effects, which will not be described in detail here.

[0107] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A model adaptive predictive control method based on disturbance state compensation, characterized in that: include: S1: Obtain historical data of system operation without disturbance, and obtain historical data of system operation under different disturbance types and sizes, build a data set, and divide the data set into a training subset and a test subset; S2: Construct the basic model of adaptive disturbance identifier; S3: constructing a parameter-optimal adaptive disturbance identifier model based on the training subset, the validation subset and the adaptive disturbance identifier basic model; S4: build a system controller set; Constructing a prediction model of the system under the nth controller, constructing a prediction output matrix of the system at time t+1 based on the prediction model, identifying a disturbance compensation vector of the system at time t+1 based on the prediction output matrix and the parameter optimal adaptive disturbance identifier model, and constructing a disturbance compensator based on the disturbance compensation vector; S5: Calculate the error between the system expected output vector and the system predicted output vector at time t+1 based on the disturbance compensator, construct a controller evaluation function based on the error, and perform performance evaluation on all controllers; obtain the optimal controller of the system at time t+1 by minimizing the performance index; wherein t represents the time; S6: Inputting the trigger signal corresponding to the optimal controller into the controller switcher, switching the controller, and performing prediction and control switching at the next moment.

2. The model adaptive predictive control method based on disturbance state compensation according to claim 1, characterized in that: The S1 includes: S11: Obtain the historical data of the system operation without disturbance, and record the system output sequence as represents the real number domain, m is the sequence dimension, and L is the sequence length; S12: Obtain historical data of system operation under different disturbance types and sizes, and construct a data set, recorded as: Among them, S is a data set constructed from historical system operation data under different disturbance types and sizes, S j =(y j ,d j ) represents the j-th output sequence-disturbance vector pair composed of the j-th system output sequence under the j-th disturbance and the j-th disturbance vector, where the disturbance includes the disturbance type and the disturbance size, j = 1, 2, ..., J, J is the total number of disturbances, is the jth system output sequence, L represents the sequence length, represents the disturbance vector under the jth disturbance, I is the total number of disturbance types, and the jth disturbance vector d j Satisfies the following expression: Among them, d ji ∈{d1,d2,…,d i ,…,d I } is the jth perturbation vector d j The i-th disturbance type, i=1,2,…,I,d ji =0 indicates the jth perturbation vector d j There is no type i disturbance in It means that there is an i-th type of disturbance in the j-th disturbance vector, and the disturbance size is q=1,2,…,Q, Q is the total number of disturbance sizes; d ji There are Q+1 possible values, perturbation or perturbation vector d j The total number is J = I·(Q+1); The dataset S is randomly divided into training subsets according to the ratio of α:

1. and validation subset α is the data set partition variable, which is used for model training and performance evaluation respectively, J t , J v The training subsets S are train , validation subset S val The number of data subsets in J t +J v =J.

3. The model adaptive predictive control method based on disturbance state compensation according to claim 1, characterized in that: The S2 includes: The input of the basic model of the adaptive disturbance identifier is set as the sequence y0 and the data set S; the basic model of the adaptive disturbance identifier includes an input layer, a hidden layer, a fully connected layer and an output layer; the output layer generates a disturbance compensation vector; The basic model of the adaptive disturbance identifier satisfies the following relationship: G(y0,S;φ); in, represents the parameter set of the adaptive disturbance identifier model, and A is the total number of model parameters.

4. The model adaptive predictive control method based on disturbance state compensation according to claim 1, characterized in that: The S3 includes: S31: Using training subset S train The basic model of the adaptive disturbance identifier is trained for P rounds; the validation subset S is used val The basic model of the adaptive disturbance identifier is evaluated and the optimal parameter adaptive disturbance identifier model is constructed; The evaluation index for evaluating the basic model of the adaptive disturbance identifier satisfies the following expression: Among them, MSE p represents the p-th parameter optimal adaptive disturbance identifier model Evaluation indicator, Δy j Represents the current j-th input data subset S j =(y j ,d j ) in which the system outputs the sequence y j The difference between the system output sequence y0 in the unperturbed case, Δy j =y j -y0, represents the disturbance compensation vector, ||·|| 2 It represents the operation of summing the squares of the values ​​within the symbol.

5. The model adaptive predictive control method based on disturbance state compensation according to claim 1, characterized in that: The S4 includes: S41: Design N system controllers respectively and construct a system controller set to satisfy the following expression: C={C1,C2,…,C n ,…,G N } (4) Among them, Γ is the system controller set, Γ n is the nth controller, n=1,2,…,N, N is the total number of system controllers; S42: Constructing the system in the nth controller Γ n The prediction model below is used to predict the system output at time t+1. The prediction model satisfies the following expression: in, are the system state vectors at time t and time t+1 respectively, q is the system state vector dimension, are the system control input vectors at time t and time t+1 respectively, For x (t) , μ (t) , μ (t+1) The predicted system output vector at time t+1, m is the dimension of the system control input vector and output vector; is the nth controller Γ n The state transition matrix, is the nth controller Γ n The control input matrix, is the nth controller Γ n The output matrix of is the nth controller Γ n The direct transfer matrix of S43: Real-time collection of L output vectors to form the system output matrix at the current time t, satisfying the following expression: In the formula, is the nth controller Γ n The system output matrix at time t, L is the number of columns in the output matrix; are the nth controller Γ n The system output vector at time t=1,2,3,…,t-1,t; S44: Construct the system prediction output matrix at time t+1, satisfying the following expression: in, is the system prediction output matrix at time t+1, represents the nth controller Γ n The system output vector at time t is: Indicated by x (t) , μ (t) , μ (t+1) The predicted system output vector at time t+1; S45: Utilizing the optimal parameter adaptive disturbance identifier model System prediction output matrix at time t+1 The disturbance compensation vector at time t+1 is predicted, and the prediction result satisfies the following expression: in, is the parameter-optimal adaptive disturbance identifier model The predicted disturbance compensation vector at time t+1; Indicates that in the nth controller Γ n The system output sequence is And the system prediction output matrix at time t+1 As input, after is the predicted disturbance compensation vector output by the adaptive disturbance identifier model of the model parameters; S46: Construct a disturbance compensator to satisfy the following expression: in, is the nth controller Γ n The predicted output vector of the system at time t+1 after disturbance compensation.

6. The model adaptive predictive control method based on disturbance state compensation according to claim 1, characterized in that: The S5 includes: S51: Calculate the error between the system expected output vector and the system predicted output vector at time t+1, satisfying the following expression: In the formula, represents the expected output y of the system at time t+1 *(t+1) With the nth controller Γ n The predicted output vector of the system at time t+1 after disturbance compensation is The error vector between *(t+1) is the expected output vector of the system at time t+1, y *(t+1) =r (t+1) , r (t+1) is the value of the system's given task curve r(t) at time t+1, is the nth controller Γ n The predicted output vector of the system at time t+1 after disturbance compensation; S52: For the nth controller Γ n Perform performance evaluation and construct an evaluation function that satisfies the following expression: in, is the nth controller Γ n The performance index of is, τ is a counting variable; Take n = 1, 2, ..., N, and calculate the performance index set of N controllers, which satisfies the following relationship: S53: By minimizing the performance index, the optimal controller at time t+1 is calculated to satisfy the following expression: In the formula, represents the optimal controller selected at time t+1, It represents the value n* of the independent variable when the function in the brackets reaches the minimum value in the set of values ​​of the independent variable n. n* is the subscript of the optimal controller, n*∈{1,2,…,N}.

7. The model adaptive predictive control method based on disturbance state compensation according to claim 5 is characterized in that: The S6 includes: The optimal controller is selected Corresponding trigger signal Input to the controller switcher, the system switches to the nth controller at time t+1 Run below; Let n*=n, t=t+1, and go to S42 to execute prediction and control switching at the next moment.

8. A model adaptive predictive control system based on disturbance state compensation, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method described in any one of claims 1 to 7 are implemented.

Images