Model identification method, device, equipment and storage medium for industrial control system

Abstract

The present application provides a model identification method, apparatus, device, and storage medium for an industrial control system, belonging to the field of model calculation technology. The method includes: obtaining input data and output data during the operation of the industrial control system; determining an estimated model of the industrial control system based on the input data and output data, detecting external disturbances in the estimated model of the industrial control system, and removing external disturbances from the input data and output data; regularizing the input data and output data of the industrial control system after removing external disturbances to obtain a regularized high-order model of the industrial control system; and performing model reduction and model correction on the regularized high-order model of the industrial control system to obtain a model identification result of the industrial control system. The present application can improve the accuracy of the identification results.

Description

Technical Field

[0001] The present application relates to the field of model calculation technology, and in particular to a model identification method, apparatus, device and storage medium for an industrial control system. Background Art

[0002] In the actual use of various systems, it is usually necessary to identify the mathematical model of the system through system identification, such as the specific model formula, etc. System identification can also be called model identification or model recognition.

[0003] When the model is identified in the existing technology, the influence of external factors on the model is not taken into account, resulting in poor accuracy and robustness of the results obtained by the recognition method adopted by the existing technology. Therefore, when the model is identified for a model with large noise and interference, it is easily affected by noise and interference, resulting in low accuracy of the recognition results obtained after model recognition based on the existing technology. Summary of the Invention

[0004] The purpose of this application is to provide a model recognition method, device, equipment and storage medium for an industrial control system, which can improve the accuracy of the recognition results.

[0005] The embodiment of the present application is implemented as follows:

[0006] In one aspect of an embodiment of the present application, a model identification method for an industrial control system is provided, comprising:

[0007] Obtain input and output data during the operation of industrial control systems;

[0008] Determining an estimation model of the industrial control system based on the input data and the output data, detecting external disturbances in the estimation model of the industrial control system, and removing external disturbances in the input data and the output data;

[0009] Regularization is performed on the input data and output data after removing external disturbances to obtain a regularized high-order model of the industrial control system;

[0010] The regularized high-order model of the industrial control system is subjected to model reduction and model correction to obtain the model identification result of the industrial control system.

[0011] Optionally, performing external disturbance detection on an estimation model of the industrial control system and removing external disturbances in input data and output data includes:

[0012] Decouple the estimation model of the industrial control system and calculate the model parameters to obtain the initial high-order model;

[0013] Determine whether the estimated model has external perturbations based on the range of zeros and poles of the initial high-order model;

[0014] If so, differential processing is performed on the input data and the output data to remove external disturbances in the input data and the output data.

[0015] Optionally, decoupling the estimation model of the industrial control system and calculating model parameters to obtain an initial high-order model include:

[0016] Decoupling estimation models for industrial control systems;

[0017] Convert the decoupled estimation model into a target matrix;

[0018] Model parameters in a target matrix are determined, and an initial high-order model is determined based on the model parameters.

[0019] Optionally, performing model order reduction and model correction processing on the regularized high-order model of the industrial control system to obtain a model identification result of the industrial control system includes:

[0020] Performing parameter estimation on a regularized high-order model of the industrial control system, and determining at least one low-order model corresponding to the regularized high-order model based on the parameter estimation result;

[0021] Perform model evaluation processing on each low-order model to obtain the frequency domain error of each low-order model;

[0022] Perform parameter correction on low-order models whose frequency domain errors meet the model evaluation requirements;

[0023] The low-order model after parameter correction is subjected to hysteresis iteration processing to obtain the model identification result of the industrial control system.

[0024] Optionally, performing model evaluation processing on each low-order model to determine the frequency domain error of each low-order model includes:

[0025] Determining an evaluation frequency interval for frequency domain error, the evaluation frequency interval including at least two intervals;

[0026] Model evaluation processing is performed on each low-order model in each evaluation frequency interval to obtain the frequency domain error of each low-order model.

[0027] Optionally, parameter correction processing is performed on the low-order model whose frequency domain error meets the model evaluation requirements, including:

[0028] Determine the poles and zeros of the low-order model whose errors in each frequency domain meet the model evaluation requirements;

[0029] Discard the poles and zeros that do not meet the preset requirements to obtain a low-order model after initial correction;

[0030] The low-order model after initial correction is corrected based on preset gain constraints and process speed constraints.

[0031] Optionally, a hysteresis iterative process is performed on the parameter-corrected low-order model to obtain a model identification result of the industrial control system, including:

[0032] Performing order reduction processing on the low-order model after parameter correction to obtain a new low-order model, which is a second-order model or a first-order model;

[0033] The new low-order model is subjected to delay shift correction and iterative calculation to obtain the model identification result of the industrial control system.

[0034] Another aspect of an embodiment of the present application provides a model identification device for an industrial control system, comprising: an acquisition module, a disturbance processing module, a regularization module, and an identification module;

[0035] An acquisition module is used to obtain input data and output data during the operation of the industrial control system;

[0036] A disturbance processing module is used to determine an estimation model of the industrial control system based on the input data and the output data, detect external disturbances on the estimation model of the industrial control system, and remove external disturbances in the input data and the output data;

[0037] Regularization module, used to regularize the input and output data after removing external disturbances to obtain a regularized high-order model of the industrial control system;

[0038] The identification module is used to perform model reduction and model correction processing on the regularized high-order model of the industrial control system to obtain the model identification result of the industrial control system.

[0039] Optionally, a disturbance processing module is specifically used to decouple the estimation model of the industrial control system and calculate the model parameters to obtain an initial high-order model; determine whether the estimation model has external disturbances based on the range of zeros and poles of the initial high-order model; if so, perform differential processing on the input data and output data to remove the external disturbances in the input data and output data.

[0040] Optionally, the disturbance processing module is specifically used to decouple the estimation model of the industrial control system; convert the decoupled estimation model into a target matrix; determine the model parameters in the target matrix, and determine the initial high-order model based on the model parameters.

[0041] Optionally, the identification module is specifically used to perform parameter estimation on the regularized high-order model of the industrial control system, determine at least one low-order model corresponding to the regularized high-order model based on the result of the parameter estimation; perform model evaluation processing on each low-order model to obtain the frequency domain error of each low-order model; perform parameter correction processing on the low-order model whose frequency domain error meets the model evaluation requirements; perform lag iterative processing on the low-order model after parameter correction to obtain the model identification result of the industrial control system.

[0042] Optionally, the identification module is specifically used to determine the evaluation frequency interval of the frequency domain error, and the evaluation frequency interval includes at least two intervals; model evaluation processing is performed on each low-order model within each evaluation frequency interval to obtain the frequency domain error of each low-order model.

[0043] Optionally, the identification module is specifically used to determine the poles and zeros of the low-order model whose frequency domain errors meet the model evaluation requirements; discard the poles and zeros that do not meet the preset requirements to obtain the low-order model after initial correction; and correct the low-order model after initial correction based on preset gain constraints and process speed constraints.

[0044] Optionally, the identification module is specifically used to reduce the order of the low-order model after parameter correction to obtain a new low-order model, which is a second-order model or a first-order model; perform delay shift correction and iterative calculation on the new low-order model to obtain the model identification result of the industrial control system.

[0045] Another aspect of an embodiment of the present application provides a computer device including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps of the model recognition method of the industrial control system are implemented.

[0046] In another aspect of an embodiment of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the model recognition method for an industrial control system are implemented.

[0047] The beneficial effects of the embodiments of the present application include:

[0048] In the model identification method, device, equipment and storage medium of an industrial control system provided by the embodiments of the present application, external disturbance detection can be performed during the process of identifying the model of the industrial control system, and external disturbances in the input data and output data can be removed, thereby regularizing the input data and output data of the industrial control system with the external disturbance removed to obtain a regularized high-order model of the industrial control system, and after the regularization processing, the regularized high-order model of the industrial control system is subjected to model order reduction and model correction processing to obtain a model identification result of the industrial control system. A model identification result with higher accuracy can be obtained, and since the error caused by the disturbance is removed, the precision and accuracy of the model identification can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0050] Figure 1 A schematic diagram of the model structure of the system model in the model identification method of the industrial control system provided in an embodiment of the present application;

[0051] Figure 2 A flow chart of a model identification method for an industrial control system provided in an embodiment of the present application;

[0052] Figure 3 Another flowchart of the model identification method for an industrial control system provided in an embodiment of the present application;

[0053] Figure 4 Another flowchart of the model identification method for an industrial control system provided in an embodiment of the present application;

[0054] Figure 5 Another flowchart of the model identification method for an industrial control system provided in an embodiment of the present application;

[0055] Figure 6 Another flowchart of the model identification method for an industrial control system provided in an embodiment of the present application;

[0056] Figure 7 Another flowchart of the model identification method for an industrial control system provided in an embodiment of the present application;

[0057] Figure 8 Another flowchart of the model identification method for an industrial control system provided in an embodiment of the present application;

[0058] Figure 9 A schematic diagram of the structure of a model recognition device for an industrial control system provided in an embodiment of the present application;

[0059] Figure 10 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] To make the objectives, technical solutions, and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Generally, the components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0061] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present application. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments in the present application without creative work are within the scope of protection of the present application.

[0062] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not require further definition or explanation in subsequent drawings.

[0063] In the description of this application, it should be noted that the terms "first", "second", "third", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0064] It should be noted that system identification is a method for estimating the mathematical model of an object based on its input and output data under certain criteria functions. Specifically, the object may be the industrial control system involved in the embodiments of this application. The industrial control system may be an automated production system used in industries such as oil refining and petrochemicals. In actual work, it is often necessary to determine the mathematical model of the system through system identification. For example, it can meet the following requirements: PID (proportional-integral-derivative) tuning, trend prediction, loop performance evaluation, and fault diagnosis.

[0065] However, in the prior art, when identifying a model, the influence of external factors on the model is not taken into consideration, resulting in poor accuracy and robustness of the results obtained by the identification method adopted by the prior art. Therefore, when performing model identification on a model with large noise and interference, it is easily affected by noise and interference, resulting in low accuracy of the identification results obtained after model identification based on the prior art.

[0066] In order to solve the above problems existing in the prior art, an embodiment of the present application provides a model identification method for an industrial control system. The specific structure of the system model used in the process of implementing this method is explained in detail below.

[0067] Figure 1 For a schematic diagram of the model structure of the system model in the model identification method of the industrial control system provided in the embodiment of the present application, please refer to Figure 1 The system model involved in the embodiment of the present application may include multiple input data and multiple output data, wherein y(t) is multiple output data, u(t) is multiple input data, G0(q) is the process transfer function matrix, and v(t) is the disturbance term, which can be an external disturbance. Figure 1 The corresponding system model can be converted into the following formula:

[0068] y(t)=G0(q)u(t)+v(t);

[0069] Among them, G0(q), u(t) and v(t) can be expressed as:

[0070] v(t)=H0(q)e(t);

[0071]

[0072]

[0073] Where y(t) is a p×1 dimensional vector, u(t) is an m×1 dimensional vector; G0(q) is a p×m polynomial matrix; g0(k) is the impulse response coefficient matrix at time k, with dimensions p×m; v(t) is a p×1 dimensional stationary random process with mean 0; e(t) is a white noise vector; H0(q) is a p×p dimensional disturbance transfer function matrix, which is assumed to be stable and has minimum phase (i.e., the inverse matrix is ​​also stable); I p is the unit matrix of the corresponding dimension; h0(k) is the impulse response coefficient matrix of the perturbation transfer function at time k; q is the shift operator, satisfying q -1 u(t)=u(t-1).

[0074] Assume that the data is a uniformly sampled time series with a length of N, a sampling period of T, and data ZN Expressed as:

[0075] Z N ={y(t),u(t)|t=1,…,N};

[0076] In the above formula y(t)=G0(q)u(t)+v(t), G0(q) and H0(q) represent the real system model. However, due to the influence of noise, disturbance, model mismatch, etc., the data Z N The estimated model may have errors, which can be expressed as follows:

[0077] y(t)=G(q,θ)u(t)+H(q,θ)e(t);

[0078] Where θ is the model parameter, which is composed of the related parameters of G(q,θ) and H(q,θ). The prediction error of the model is ε(t,θ), with a dimension of p×1, and its expression is:

[0079]

[0080] in, is the one-step model prediction output at time t. The model parameters are:

[0081]

[0082] Among them, V N (θ) is the prediction error loss function, and the specific formula is as follows:

[0083]

[0084] By transforming the formula y(t) = G(q,θ)u(t) + H(q,θ)e(t), we can obtain another form of the formula, as follows:

[0085] y(t)=A -1 (q)B(q)u(t)+D -1 (q)C(q)e(t);

[0086] Where A(q), B(q), C(q), and D(q) are polynomial matrices, and it is assumed that the order of all non-zero polynomials in the matrix is ​​n. Where G(q,θ)=A -1 (q)B(q),H(q,θ)=D -1 (q)C(q).

[0087] According to Ljung's asymptotic theory, when the model order n satisfies the following formula:

[0088]

[0089] And the input is a continuous excitation signal of finite order, and it is assumed that for all n and N, the loss function V N The global minimum of (θ) can be obtained, then the following formula is obtained:

[0090]

[0091]

[0092] in,

[0093]

[0094] When the loop is open, the cross spectrum Φ eu (ω)=Φ ue (ω)=0. Where R is the covariance matrix of the white noise vector, denotes the Kronecker product, and col(.) denotes the matrix vector operator. Assuming that the matrix M is an m×n matrix, then:

[0095] col(M)=[M1,M2,…,M n ] T ;

[0096] The dimension of col(M) is mn×1, where M i is the i-th column of the matrix M. Since A(q), C(q), and D(q) in the above formula are all diagonal polynomial matrices, and the non-zero polynomials of the matrices are all first-unit polynomials, the above multi-input multi-output model (including multiple input data and multiple output data) can be decomposed into p multi-input single-output models (including multiple input data and one output data).

[0097] The specific implementation process of the model identification method of the industrial control system involved in the embodiment is explained in detail below.

[0098] Figure 2 For a flow chart of the model identification method for an industrial control system provided in the embodiment of the present application, please refer to Figure 2 ,Model identification methods for industrial control systems, including:

[0099] S210: Acquire input data and output data during operation of the industrial control system.

[0100] Optionally, the execution subject of the method may be a computer device, for example, a software program in the computer device, etc., which is not specifically limited here.

[0101] Among them, input data and output data can be data generated during the operation of the industrial control system, which can be obtained by detection. The input data is also Figure 1 u(t) in the output data is Figure 1 y(t) in .

[0102] For different types of industrial control systems, different methods can be used to obtain input data and output data. For example, detection equipment is respectively set at the input and output ports of the industrial control system to obtain the above input data and output data.

[0103] S220: Determine an estimation model of the industrial control system based on the input data and the output data, perform external disturbance detection on the estimation model of the industrial control system, and remove the external disturbance in the input data and the output data.

[0104] Optionally, after obtaining the input data and the output data, an estimation model of the industrial control system can be obtained. For an industrial control system including multiple output data, when the i-th output data y i When (t) is an integral system, it can be considered that there is only one integral link, and y i (t) is processed by difference, and the following formula is obtained:

[0105]

[0106] Where T is the sampling period. For the integral system, the input u1(t)~u m (t) and Δy i (t) model, replace Δy in the above formula i The value of (t) is given by the above formula y(t) = A -1 (q)B(q)u(t)+D -1 The y(t) in (q)C(q)e(t) can be used to obtain the estimation model of the industrial control system. After integration and approximate estimation, the estimation model can be obtained.

[0107] Specifically, after integration and approximate estimation, the estimation model of the industrial control system is as follows:

[0108] A(q)y(t)=B(q)u(t)+e(t);

[0109] Specifically, A(q) may be a diagonal polynomial matrix.

[0110] Optionally, external disturbance detection is the process of determining the disturbance term in the model, that is, the process of obtaining the aforementioned v(t). External disturbance detection can be performed based on the estimation model to obtain the specific value of the disturbance term, thereby removing the external disturbance in the estimation model. The external disturbance can specifically be a low-frequency disturbance.

[0111] After determining the existence of low-frequency disturbances, differential processing can be performed to eliminate the influence of low-frequency disturbances, that is, to remove external disturbances in the estimation model.

[0112] It should be noted that the external disturbance may be a low-frequency disturbance that changes relatively slowly.

[0113] S230: Regularizing the input data and output data of the industrial control system after removing external disturbances to obtain a regularized high-order model of the industrial control system.

[0114] Alternatively, in industrial control systems, disturbance refers to errors caused by the model itself or external factors, in addition to noise. Common disturbances in control systems include slowly changing factors such as drift and trend changes. The above process only removes external disturbances. Besides external disturbances, the influence of the model itself, also known as internal disturbances, is also included, also known as model mismatch. Internal disturbances can be removed by introducing regularization.

[0115] Optionally, regularization processing can specifically be a process of introducing additional information into the model when the training data is insufficient, so as to prevent overfitting and improve the generalization performance of the model. After regularization processing, internal disturbances can be removed.

[0116] S240: Performing model order reduction and model correction processing on the regularized high-order model of the industrial control system to obtain a model identification result of the industrial control system.

[0117] Optionally, after obtaining the regularized high-order model of the above-mentioned industrial control system, parameter estimation, order reduction and model correction processing can be performed sequentially or separately based on the model to obtain a model identification result of the industrial control system.

[0118] Among them, model reduction can specifically be the reduction of the order of the regularized high-order model, for example: reducing the fifth-order model to the second-order, etc.; model correction processing can be further correction of the parameters after the model is determined to ensure the accuracy of the parameters, and then the model identification results of the industrial control system can be obtained.

[0119] In a model identification method for an industrial control system provided in an embodiment of the present application, external disturbance detection can be performed during the process of identifying the model of the industrial control system, and external disturbances of input data and output data can be removed, thereby performing regularization processing on the input data and output data without external disturbances to obtain a regularized high-order model of the industrial control system, and after the regularization processing, the regularized high-order model of the industrial control system is subjected to model order reduction and model correction processing to obtain a model identification result of the industrial control system. A model identification result with higher accuracy can be obtained, and since the error caused by the disturbance is removed, the precision and accuracy of the model identification can be improved.

[0120] Another specific implementation process of the model identification method for the industrial control system involved in this embodiment is explained in detail below.

[0121] Figure 3 For another flow chart of the model identification method of the industrial control system provided in the embodiment of the present application, please refer to Figure 3 , detect external disturbances in the estimation model of industrial control systems and remove external disturbances in input data and output data, including:

[0122] S310: Decoupling the estimation model of the industrial control system and calculating model parameters to obtain an initial high-order model.

[0123] Optionally, after obtaining the aforementioned estimation model A(q)y(t)=B(q)u(t)+e(t), the model can be decoupled and the model parameters can be calculated to obtain a high-order model, which can be an initial high-order model.

[0124] In order to further illustrate the above process, the specific implementation process of decoupling the estimation model of the industrial control system and calculating the model parameters is explained in detail below.

[0125] Figure 4 For another flow chart of the model identification method of the industrial control system provided in the embodiment of the present application, please refer to Figure 4 , decouple the estimation model of the industrial control system and calculate the model parameters to obtain the initial high-order model, including:

[0126] S410: Decoupling an estimation model of the industrial control system.

[0127] Optionally, the aforementioned estimation model can be decoupled, that is, the formula A(q)y(t) = B(q)u(t) + e(t) can be decoupled, and the formula can be converted into p multi-input single-output loops, as follows:

[0128]

[0129] Take the i-th output as an example:

[0130] A ii (q)y i (t)=[B i1 (q) … B im (q)]u(t)+e i (t).

[0131] S420: Convert the decoupled estimation model into a target matrix.

[0132] Assuming that the hysteresis of the loop is nk = 1, then:

[0133] A ii (q)=1+a ii,1 q -1 +…+a ii,n q -n ;

[0134] B ij (q) = b ij,1 q -1 +…+b ij,n q -n ;

[0135] The following definitions can be made:

[0136]

[0137] θ i =[a ii,1 …a ii,n b i1,1 …b i1,n …b im,1 …b im,n ] T ;

[0138] Then the model parameter θ is θ=[θ i …θ p ] T ;

[0139] Formula A can be ii (q)y i (t)=[B i1 (q) … B im (q)]u(t)+e i (t) is converted into the target matrix, which is expressed as follows:

[0140] Y i =Φ i θ i +Λ i ;

[0141] Among them, Y i =[y i (n+1) … y i (N)] T ; Λ i =[e i (n+1) … e i (N)] T .

[0142] S430: Determine model parameters in the target matrix, and determine an initial high-order model based on the model parameters.

[0143] Optionally, the specific process of determining the model parameters is as follows:

[0144] First determine the loss function of the multi-input single-output loop corresponding to the i-th output:

[0145] V i (θ i ,P i )=(Y i -Φ i θ i ) T (Y i -Φ i θ i );

[0146] Based on the least squares method, we can get:

[0147]

[0148] Optionally, determine the above model parameters After that, the initial high-order model can be determined, that is, the formula Y i =Φ i θ i +Λ i The parameter θ i The result after confirmation.

[0149] Optionally, after S310: decoupling the estimation model of the industrial control system and calculating model parameters to obtain an initial high-order model, the method further includes:

[0150] S320: Determine whether the estimated model has external disturbances according to the range of the zero points and the poles of the initial high-order model.

[0151] Alternatively, for the initial high-order model, the multinomial matrix will usually produce zeros or poles close to the unit circle z = 1. For example:

[0152] A ii The extreme point of (q) is p i,k , B ijThe zero point of (q) is z ij,k , where k=1,…,n,j=1,…,m,when c1 <p i,k ≤1, and c1 <z ij,k ≤1,p i,k ,z ij,k When ∈R holds true, it can be determined that there is external disturbance in the estimation model.

[0153] S330: If yes, perform differential processing on the input data and the output data to remove external disturbances in the input data and the output data.

[0154] Optionally, after determining that the estimated model has external disturbances, y i (t) and input u1(t)~u m (t) Perform differential processing at the same time to eliminate the influence of low-frequency disturbances.

[0155] In the model identification method of the industrial control system provided in the embodiment of the present application, external disturbances existing in the model can be identified and processed, thereby reducing the errors and impacts caused by external disturbances in the model and improving the accuracy of model identification.

[0156] Optionally, after removing the external perturbations, regularization can be performed. The specific process is as follows:

[0157] Based on the estimation model of the industrial control system without external disturbances, the regularization term is introduced to obtain the loss function of the multi-input single-output loop corresponding to the i-th output:

[0158] V i (θ i ,P i )=(Y i -Φ i θ i ) T (Y i -Φ i θ i )+θ i T P i θ i ;

[0159] Based on the least squares method, we can get:

[0160]

[0161] Among them, the regularization matrix P i The dimension is (m+1)n×(m+1)n, and the parameter θ i is the maximum a posteriori estimate, the regularization matrix P i Contains the parameters θ iPrior distribution knowledge. When the regularization matrix P is selected i P i =α i I, where α i is a hyperparameter, then the formula For ridge regression, if is ill-conditioned, and the parameter θ obtained by regularization i More realistic.

[0162] For the formula θ i =[a ii,1 …a ii,n b i1,1 …b i1,n …b im,1 …b im,n ] T For example, θ i Parameter a in ii,k and b ij,k are gradually attenuated, so θ i Contains (m+1) segments of parameters (corresponding to y i The regression variables and u1~u m The length of each parameter is n (corresponding to the model order).

[0163] Assume that represents θ i The fth parameter in the model can be assumed to be Satisfies the following Gaussian distribution:

[0164]

[0165] Correspondingly, the disturbance term v i (t) satisfies the following Gaussian distribution:

[0166]

[0167] The parameter θ can be obtained i The conditional probability distribution of is:

[0168]

[0169] Through derivation, it can be obtained that when When , the posterior estimate Equal to the above formula middle The fth segment parameter parameter The covariance matrix of the probability distribution Can be set to:

[0170]

[0171] Where c and μ are hyperparameters, and c ≥ 0, μ ≥ 0. By choosing appropriate hyperparameters, the variance of model parameters caused by noise can be reduced even if the model order n is selected to be relatively large.

[0172] Then we can get a regularized high-order model of the industrial control system, which can be used as an approximate estimate of the true model:

[0173]

[0174] in, During the estimation process, for example, the power spectrum of the disturbance term can be estimated as:

[0175]

[0176] in, is the covariance matrix obtained from the prediction error of the regularized high-order model of the industrial control system; the specific relationship is as follows:

[0177]

[0178] Since the input power spectrum and mutual spectrum It can also be calculated, then according to the above formula The power spectrum can be calculated

[0179] Another specific implementation process of the model identification method for the industrial control system involved in the embodiment will be explained in detail below.

[0180] Figure 5 For another flow chart of the model identification method of the industrial control system provided in the embodiment of the present application, please refer to Figure 5 , the regularized high-order model of the industrial control system is reduced in order and corrected to obtain the model identification results of the industrial control system, including:

[0181] S510: Perform parameter estimation on a regularized high-order model of the industrial control system, and determine at least one low-order model corresponding to the regularized high-order model according to the parameter estimation result.

[0182] Optionally, during the parameter estimation process, the i-th output y can be determined first. i (t) The corresponding frequency domain loss function is as follows:

[0183]

[0184] Based on the solution method of the minimization problem, low-order models of order 1 to n can be calculated in sequence, and the order that minimizes the following formula is selected as the identified low-order model:

[0185]

[0186] The low-order model can be expressed as:

[0187]

[0188] Based on the above formula, the low-order model of each order can be determined in turn.

[0189] S520: Performing model evaluation processing on each low-order model to obtain a frequency domain error of each low-order model.

[0190] Optionally, the evaluation process for the model may specifically be to determine the simulation output of the model, calculate the simulation error based on the simulation output, and then obtain the frequency domain error of each low-order model.

[0191] S530: Perform parameter correction processing on the low-order model whose frequency domain error meets the model evaluation requirements.

[0192] Optionally, after determining the frequency domain error of each low-order model, it can be determined whether the frequency domain error meets the model evaluation requirements. If so, parameter correction processing can be performed on the low-order model; accordingly, if not, parameter correction processing is not performed.

[0193] S540: Performing hysteresis iterative processing on the low-order model after parameter correction to obtain a model identification result of the industrial control system.

[0194] Optionally, after obtaining the low-order model after parameter correction, a delayed iterative process can be performed to optimize the low-order model. The specific optimization process may include quadratic order reduction, delay shift correction, iterative calculation and other processes, and finally the model identification result of the industrial control system is obtained.

[0195] The specific implementation process of determining the frequency domain error in the model identification method of the industrial control system involved in the embodiment is explained in detail below.

[0196] Figure 6 For another flow chart of the model identification method of the industrial control system provided in the embodiment of the present application, please refer to Figure 6 , perform model evaluation processing on each low-order model to determine the frequency domain error of each low-order model, including:

[0197] S610: Determine an evaluation frequency interval for frequency domain error.

[0198] The evaluation frequency interval includes at least two intervals.

[0199] Optionally, the evaluation frequency interval of the frequency domain error can be divided into at least two intervals, for example: [0,ω c ] and (ω c ,ω T ] Among them, ω c represents the cutoff frequency; ω T represents the Nyquist frequency.

[0200] That is, the divided intervals may be from the 0 frequency to the cutoff frequency and from the cutoff frequency to the Nyquist frequency. It should be noted that these two intervals may be pre-configured. During implementation, more intervals may be set according to actual needs or the intervals may be divided in other ways, without specific limitation herein.

[0201] S620: Performing model evaluation processing on each low-order model within each evaluation frequency interval to obtain a frequency domain error of each low-order model.

[0202] The model evaluation process is as follows.

[0203] First determine the simulation output of each low-order model:

[0204]

[0205] The simulation error of the simulation output is obtained:

[0206]

[0207] Among them, d represents the d-norm, which can be 2, representing the two-norm.

[0208] According to the gradual theory, the definition The 3σ bound of is:

[0209]

[0210] Defining frequency domain error uncertainty for:

[0211]

[0212] Define u j (t)~y i The frequency domain error of the model (t) is:

[0213]

[0214] The frequency domain error is determined based on the aforementioned corresponding frequency range to obtain the frequency domain error of each low-order model.

[0215] In the model identification method of the industrial control system provided in the embodiment of the present application, the model can be evaluated, and the different frequency domain errors in different frequency ranges can be fully considered to obtain a model that better meets the needs and improve the accuracy of obtaining frequency domain errors.

[0216] The specific implementation process of the parameter correction processing in the model identification method of the industrial control system involved in the embodiment is explained in detail below.

[0217] Figure 7 For another flow chart of the model identification method of the industrial control system provided in the embodiment of the present application, please refer to Figure 7 , perform parameter correction processing on low-order models whose frequency domain errors meet the model evaluation requirements, including:

[0218] S710: Determine the poles and zeros of the low-order model whose frequency domain errors meet the model evaluation requirements.

[0219] Optionally, the poles and zeros of the low-order model whose frequency domain errors meet the model evaluation requirements can be determined separately, which can be specifically expressed as:

[0220] The extreme point is The zero point is

[0221] S720: discarding poles and zeros that do not meet preset requirements to obtain a low-order model after initial correction.

[0222] Optionally, the corresponding poles and zeros can be discarded if they meet the following conditions:

[0223] and

[0224] Where c1 is a threshold constant close to 1.

[0225] S730: Correcting the initially corrected low-order model based on a preset gain constraint and process speed constraint.

[0226] Optionally, after obtaining the initially calibrated low-order model, the model may be further calibrated based on preset gain constraints and process speed constraints.

[0227] The specific implementation process of the hysteresis iteration process in the model identification method of the industrial control system involved in the embodiment is explained in detail below.

[0228] Figure 8 For another flow chart of the model identification method of the industrial control system provided in the embodiment of the present application, please refer to Figure 8 , the low-order model after parameter correction is subjected to hysteresis iteration processing to obtain the model identification results of the industrial control system, including:

[0229] S810: performing order reduction processing on the low-order model after parameter correction to obtain a new low-order model.

[0230] Among them, the new low-order model is a second-order model or a first-order model.

[0231] Optionally, when the data excitation level is insufficient or there is disturbance, the calculation process assumes that the inputs u1(t) to u m (t) relative to y i The lag of (t) is 1. Since the regularized high-order model has a relatively strong linear generalization ability, part of the lag will also be fitted into the model parameter θ. This may lead to lag errors in the model. In order to remove this error, the low-order model after parameter correction can be reduced in order.

[0232] Specifically, the low-order model after parameter correction can be reduced to a first-order model or a second-order model.

[0233] S820: Perform delay shift correction and iterative calculation on the new low-order model to obtain a model identification result of the industrial control system.

[0234] Optionally, after obtaining the new low-order model, the new low-order model can be subjected to delay shift correction, and then iterative calculation is performed until the lag difference before and after the iteration is less than a preset threshold or the upper limit of the number of iterations set by the user is reached.

[0235] The result obtained after the above processing is the model identification result of the industrial control system.

[0236] In the model identification method for the industrial control system provided in the embodiment of the present application, the model lag problem is taken into consideration, and an iterative optimization method is adopted to continuously approach the actual model lag, so that the model identification result of the industrial control system can be obtained more accurately.

[0237] The following describes the devices, equipment, storage media, etc. corresponding to the model identification method for the industrial control system provided by this application. The specific implementation process and technical effects are described above and will not be repeated below.

[0238] Figure 9 For a structural diagram of a model recognition device for an industrial control system provided in an embodiment of the present application, please refer to Figure 9 , a model identification device for an industrial control system, comprising: an acquisition module 910, a disturbance processing module 920, a regularization module 930 and an identification module 940;

[0239] An acquisition module 910 is used to acquire input data and output data during operation of the industrial control system;

[0240] A disturbance processing module 920 is configured to determine an estimation model of the industrial control system based on the input data and the output data, detect external disturbances in the estimation model of the industrial control system, and remove external disturbances from the input data and the output data;

[0241] Regularization module 930, used to perform regularization processing on the input data and output data after removing external disturbances to obtain a regularized high-order model of the industrial control system;

[0242] The identification module 940 is used to perform model order reduction and model correction processing on the regularized high-order model of the industrial control system to obtain a model identification result of the industrial control system.

[0243] Optionally, the disturbance processing module 920 is specifically used to decouple the estimation model of the industrial control system and calculate the model parameters to obtain an initial high-order model; determine whether there is external disturbance in the estimation model based on the range of zero points and poles of the initial high-order model; if so, perform differential processing on the input data and output data to remove the external disturbance in the estimation model.

[0244] Optionally, the disturbance processing module 920 is specifically used to decouple the estimation model of the industrial control system; convert the decoupled estimation model into a target matrix; determine model parameters in the target matrix, and determine an initial high-order model based on the model parameters.

[0245] Optionally, the identification module 940 is specifically used to perform parameter estimation on the regularized high-order model of the industrial control system, determine at least one low-order model corresponding to the regularized high-order model based on the result of the parameter estimation; perform model evaluation processing on each low-order model to obtain the frequency domain error of each low-order model; perform parameter correction processing on the low-order model whose frequency domain error meets the model evaluation requirements; perform lag iterative processing on the low-order model after parameter correction to obtain the model identification result of the industrial control system.

[0246] Optionally, the identification module 940 is specifically used to determine the evaluation frequency interval of the frequency domain error, and the evaluation frequency interval includes at least two intervals; each low-order model is evaluated within each evaluation frequency interval to obtain the frequency domain error of each low-order model.

[0247] Optionally, the identification module 940 is specifically used to determine the poles and zeros of the low-order model whose frequency domain errors meet the model evaluation requirements; discard the poles and zeros that do not meet the preset requirements to obtain the low-order model after initial correction; and correct the low-order model after initial correction based on the preset gain constraints and process speed constraints.

[0248] Optionally, the identification module 940 is specifically used to reduce the order of the low-order model after parameter correction to obtain a new low-order model, which is a second-order model or a first-order model; perform delay shift correction and iterative calculation on the new low-order model to obtain a model identification result of the industrial control system.

[0249] In a model identification device for an industrial control system provided in an embodiment of the present application, external disturbance detection can be performed during the process of identifying the model of the industrial control system, and external disturbances of input data and output data can be removed, thereby performing regularization processing on the input data and output data with the external disturbance removed to obtain a regularized high-order model of the industrial control system, and after the regularization processing, the regularized high-order model of the industrial control system is subjected to model order reduction and model correction processing to obtain a model identification result of the industrial control system. A model identification result with higher accuracy can be obtained, and since the error caused by the disturbance is removed, the precision and accuracy of the model identification can be improved.

[0250] The above-mentioned device is used to execute the method provided in the above-mentioned embodiment. Its implementation principle and technical effect are similar and will not be repeated here.

[0251] The above modules can be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), one or more microprocessors, or one or more field programmable gate arrays (FPGAs). For another example, when a module is implemented by scheduling program code through a processing element, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules can be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0252] Figure 10 For a schematic diagram of the computer device provided in this application embodiment, please refer to Figure 10 The computer device includes: a memory 960 and a processor 970. The memory 960 stores a computer program that can be run on the processor 970. When the processor 970 executes the computer program, the steps of the model recognition method of the industrial control system are implemented.

[0253] In another aspect of an embodiment of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the model recognition method for an industrial control system are implemented.

[0254] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interface, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0255] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0256] In addition, the functional units in various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or hardware plus software functional units.

[0257] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor (English: processor) to perform some steps of the method of each embodiment of the present invention. The aforementioned storage medium includes: a USB flash drive, a mobile hard disk, a read-only memory (English: Read-Only Memory, abbreviated: ROM), a random access memory (English: Random Access Memory, abbreviated: RAM), a magnetic disk or an optical disk, and other media that can store program code.

[0258] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited to them. Any changes or substitutions that can be easily conceived by any person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0259] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A model identification method for an industrial control system, characterized in that: include: Obtain input and output data during the operation of industrial control systems; Determining an estimation model of the industrial control system based on the input data and the output data, performing external disturbance detection on the estimation model of the industrial control system and removing the external disturbance in the input data and the output data; performing regularization processing on the input data and the output data of the industrial control system from which external disturbances have been removed, to obtain a regularized high-order model of the industrial control system; Performing model order reduction and model correction processing on the regularized high-order model of the industrial control system to obtain a model identification result of the industrial control system; The performing model order reduction and model correction processing on the regularized high-order model of the industrial control system to obtain a model identification result of the industrial control system includes: Performing parameter estimation on the regularized high-order model of the industrial control system, and determining at least one low-order model corresponding to the regularized high-order model according to the parameter estimation result; Performing model evaluation processing on each low-order model to obtain a frequency domain error of each low-order model; Perform parameter correction on low-order models whose frequency domain errors meet the model evaluation requirements; The low-order model after parameter correction is subjected to hysteresis iteration processing to obtain the model identification result of the industrial control system.

2. The model identification method for an industrial control system according to claim 1, wherein: The performing external disturbance detection on the estimation model of the industrial control system and removing the external disturbance in the input data and the output data includes: Decoupling the estimation model of the industrial control system and calculating model parameters to obtain an initial high-order model; determining whether the estimated model has external disturbances based on the range of zeros and poles of the initial high-order model; If so, differential processing is performed on the input data and the output data to remove external disturbances in the input data and the output data.

3. The model identification method for an industrial control system according to claim 2, wherein: Decoupling the estimation model of the industrial control system and calculating model parameters to obtain an initial high-order model includes: Decoupling an estimation model of the industrial control system; Converting the decoupled estimation model into a target matrix; Model parameters in the target matrix are determined, and the initial high-order model is determined based on the model parameters.

4. The model identification method for an industrial control system according to claim 1, wherein: The performing model evaluation processing on each low-order model to determine the frequency domain error of each low-order model includes: Determining an evaluation frequency interval for the frequency domain error, wherein the evaluation frequency interval includes at least two intervals; Model evaluation processing is performed on each of the low-order models within each evaluation frequency interval to obtain a frequency domain error of each of the low-order models.

5. The model identification method for an industrial control system according to claim 1, wherein: The parameter correction processing of the low-order model whose frequency domain error meets the model evaluation requirements includes: Determine the poles and zeros of the low-order model whose errors in each frequency domain meet the model evaluation requirements; Discard the poles and zeros that do not meet the preset requirements to obtain a low-order model after initial correction; The initially corrected low-order model is corrected based on a preset gain constraint and a process speed constraint.

6. The model identification method for an industrial control system according to claim 1, wherein: The method of performing hysteresis iterative processing on the parameter-corrected low-order model to obtain a model identification result of the industrial control system includes: Performing order reduction processing on the low-order model after parameter correction to obtain a new low-order model, wherein the new low-order model is a second-order model or a first-order model; Delay shift correction and iterative calculation are performed on the new low-order model to obtain a model identification result of the industrial control system.

7. A model recognition device for an industrial control system, characterized in that: include: Acquisition module, disturbance processing module, regularization module and recognition module; The acquisition module is used to acquire input data and output data when the industrial control system is running; The disturbance processing module is configured to determine an estimation model of the industrial control system based on the input data and the output data, perform external disturbance detection on the estimation model of the industrial control system, and remove external disturbances from the input data and the output data; The regularization module is used to perform regularization processing on the input data and the output data of the industrial control system after removing external disturbances, so as to obtain a regularized high-order model of the industrial control system; The identification module is used to perform model order reduction and model correction processing on the regularized high-order model of the industrial control system to obtain a model identification result of the industrial control system; The identification module is specifically configured to perform parameter estimation on the regularized high-order model of the industrial control system, determine at least one low-order model corresponding to the regularized high-order model based on the parameter estimation result, perform model evaluation processing on each low-order model to obtain a frequency domain error of each low-order model, and perform parameter correction processing on the low-order model whose frequency domain error meets the model evaluation requirement; The low-order model after parameter correction is subjected to hysteresis iteration processing to obtain the model identification result of the industrial control system.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium, characterized in that The storage medium stores a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 6.

Images