A Hammerstein nonlinear dynamic system identification method and system based on industrial process

Abstract

The present invention provides a method and system for identifying a Hammerstein nonlinear dynamic system based on an industrial process, and relates to the technical field of system identification. The method comprises: collecting historical data of a Hammerstein nonlinear dynamic system as a data set; setting relevant parameters of an RBF neural network; determining the neuron center point of a hidden layer of the RBF neural network by a K-means algorithm; calculating the standard deviation of the neuron center point by a nearest neighbor Brute method; calculating the hidden layer output matrix of the hidden layer of the RBF neural network; determining the system parameters of the Hammerstein nonlinear dynamic system; calculating the output estimate of the Hammerstein nonlinear dynamic system; calculating the mean square error; judging whether the mean square error is less than a preset mean square error or whether the number of iterations reaches a preset number of iterations; if so, outputting the system parameters as the overall identification result of the Hammerstein nonlinear dynamic system, otherwise, updating the neuron center point, and restarting the calculation of the standard deviation. The present invention can improve the calculation efficiency and enhance the identification effect and stability of the algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of system identification, and in particular to an industrial process-based Hammerstein nonlinear dynamic system identification method and system. Background Art

[0002] The Hammerstein nonlinear dynamic system identification method is a technique used to identify and model nonlinear systems in industrial processes. It usually involves decomposing the system into a nonlinear static function and a linear dynamic part. The nonlinear function is usually used to describe the nonlinear relationship between the input and output of the system, while the linear dynamic part describes how input changes affect the system output over time. By combining these two parts, the Hammerstein model can effectively capture and predict the complex dynamic behavior of industrial processes, thereby helping to optimize control strategies and improve process efficiency.

[0003] As the complexity of industrial process equipment continues to increase, the application of advanced control theory is becoming more and more extensive, and the safety and economic effects it brings are very significant. However, the implementation of most advanced control technologies needs to be based on accurate mathematical models, and then analyze the system, design controllers, and improve the characteristics of the control system. Therefore, accurate system identification methods are crucial for the prediction, control, monitoring, and optimization design of industrial systems.

[0004] However, the existing identification methods use neural networks and subspace identification technology alternately, which is too complex to calculate, will cause computational loss, and the identification efficiency is too low. It is easy to fall into the local optimal solution, which limits the performance and reliability of the system. In the process of alternation, the system parameters may interfere with each other, resulting in the difficulty of converging the overall system residuals, affecting the stability of the system and the identification accuracy. Summary of the invention

[0005] In order to solve the technical problems that the existing identification methods use neural networks and subspace identification technologies alternately, the calculations are too complicated, calculation losses are generated, the identification efficiency is too low, and it is easy to fall into local optimal solutions, which limits the performance and reliability of the system. In the process of alternation, system parameters may interfere with each other, resulting in difficulty in converging the overall system residuals, affecting the stability of the system and the identification accuracy, the present invention provides a Hammerstein nonlinear dynamic system identification method and system based on industrial processes.

[0006] The technical solution provided by the embodiment of the present invention is as follows:

[0007] First aspect:

[0008] An embodiment of the present invention provides a Hammerstein nonlinear dynamic system identification method based on an industrial process, comprising:

[0009] S1: Collect historical data of Hammerstein nonlinear dynamic system as data set;

[0010] S2: Based on the data set, set the relevant parameters of the RBF neural network;

[0011] S3: Based on the data set, the RBF neural network is initialized through the K-means algorithm to determine the neuron center point of the hidden layer of the RBF neural network;

[0012] S4: Based on the neuron center point, the standard deviation of the neuron center point is calculated by the nearest neighbor Brute method;

[0013] S5: Calculate the hidden layer output matrix of the hidden layer of the RBF neural network by combining the relevant parameters and the standard deviation of the neuron center point;

[0014] S6: Determine the system parameters of the Hammerstein nonlinear dynamic system according to the hidden layer output matrix and the actual output matrix;

[0015] S7: Calculate the output estimate of the Hammerstein nonlinear dynamic system according to the system parameters;

[0016] S8: Calculate the mean square error by combining the output estimate and the actual output of the Hammerstein nonlinear dynamic system;

[0017] S9: Determine whether the mean square error is less than the preset mean square error or whether the number of iterations reaches the preset number of iterations; if so, output the system parameters as the overall identification result of the Hammerstein nonlinear dynamic system, otherwise, use the gradient descent method to update the neuron center point and return to step S4.

[0018] Second aspect:

[0019] An embodiment of the present invention provides a Hammerstein nonlinear dynamic system identification system based on an industrial process, comprising:

[0020] processor;

[0021] A memory having computer-readable instructions stored therein, wherein when the computer-readable instructions are executed by the processor, the industrial process-based Hammerstein nonlinear dynamic system identification method according to the first aspect is implemented.

[0022] The third aspect:

[0023] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the industrial process-based Hammerstein nonlinear dynamic system identification method according to the first aspect is implemented.

[0024] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0025] In the present invention, by collecting a large amount of historical data and setting the initial parameters of the RBF neural network based on these data, the complex behavior of the nonlinear dynamic system can be accurately captured and simulated. The K-means algorithm and the nearest neighbor Brute force method are combined to optimize the neuron center point and the standard deviation corresponding to the neuron center point is adaptively adjusted, so that the model can adapt to changes in system behavior, improve the adaptability and accuracy of the model, enhance the recognition effect and stability of the algorithm, and calculate the mean square error to detect and evaluate the performance of the model, reduce the prediction error, and improve the accuracy and robustness of the model. The neuron center point is updated by the gradient descent method, so that the model is continuously improved in multiple iterations, and the model parameters can be systematically adjusted and optimized to improve the accuracy of the overall recognition of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0027] Figure 1 A flowchart of a Hammerstein nonlinear dynamic system identification method based on an industrial process provided by an embodiment of the present invention;

[0028] Figure 2 A schematic structural diagram of a Hammerstein nonlinear dynamic system identification system based on an industrial process is provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0030] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0031] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0032] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are consistent.

[0033] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0034] Reference Manual Attached Figure 1 , showing a flow chart of a Hammerstein nonlinear dynamic system identification method based on industrial processes provided by an embodiment of the present invention.

[0035] The embodiment of the present invention provides a Hammerstein nonlinear dynamic system identification method based on an industrial process, the method comprising:

[0036] S1: Collect historical data of Hammerstein nonlinear dynamic system as a data set.

[0037] It should be noted that by collecting historical data, the identification method can fully learn and understand the behavior and dynamic characteristics of the system, capture the nonlinear relationships in the system, and improve the accuracy and reliability of the model in predicting future states.

[0038] In a possible implementation, the relevant parameters specifically include: the number of hidden layer neurons, input dimension, output dimension, learning rate, Hankel matrix dimension, Hankel matrix data volume, training data set size, number of training iterations and training expected indicators.

[0039] Among them, the number of hidden layer neurons determines the complexity of the hidden layer of the RBF network, affecting the expressiveness and generalization of the model. The input dimension specifies the number of features of the input data. The output dimension defines the number of nodes in the network output layer, corresponding to the dimension of the target variable. The learning rate affects the speed and stability of weight adjustment during model training. The Hankel matrix dimension determines the structure of the Hankel matrix, which is used to capture the time dynamics of data in system identification. The Hankel matrix data volume refers to the number of data points contained in the Hankel matrix, reflecting the utilization of time series data. The training data set size specifies the amount of data used to train the model. The number of training iterations sets the number of repetitions of forward and backward propagation during training. The training expectation index sets the performance standards that need to be achieved during training.

[0040] S2: Based on the data set, set the relevant parameters of the RBF neural network.

[0041] Among them, the radial basis function (RBF) neural network is a special type of artificial neural network, whose hidden layer neurons use radial basis functions as activation functions. The RBF neural network usually includes three layers: an input layer, a hidden layer, and an output layer.

[0042] It should be noted that customizing network parameters according to actual data can improve the learning efficiency of the network and the accuracy of the model, making the network better adapt to specific nonlinear dynamic system characteristics.

[0043] S3: According to the data set, the RBF neural network is initialized through the K-means algorithm to determine the neuron center point of the hidden layer of the RBF neural network.

[0044] Among them, K-means is a clustering algorithm that groups data by minimizing the distance between each point and its nearest cluster center to find the neuron center point.

[0045] It should be noted that by determining the center point of the neuron through the K-means algorithm, the initialization of the RBF network is more in line with the actual structure of the data, which can enhance the RBF network's ability to capture data features, accelerate the training process of the neural network, reduce the time required for training, and improve the convergence speed.

[0046] S4: Based on the neuron center point, the standard deviation of the neuron center point is calculated using the nearest neighbor Brute method.

[0047] Specifically, the nearest neighbor Brute method is a computational method that is usually used to determine the nearest neighbors of a data point. It finds the nearest point by calculating the distance between a point and all other points in the data set, which can ensure that the true nearest neighbor is found. Through accurate standard deviation calculation, the radial basis function of the RBF neural network can be better configured, thereby improving the learning and adaptability of the entire model to the dynamic behavior of the Hammerstein nonlinear dynamic system.

[0048] In a possible implementation, S4 specifically includes:

[0049] S401: Set the number of adjacent data points of the neuron center.

[0050] S402: Calculate the Euclidean distance between the neuron center point and each adjacent data point.

[0051] S403: Perform nearest neighbor search based on Euclidean distance.

[0052] S404: Fitting the neuron center point according to the nearest neighbor search result to obtain neuron center point fitting data.

[0053] S405: Calculate the standard deviation of the neuron center point based on the neuron center point fitting data.

[0054] S5: Calculate the hidden layer output matrix of the hidden layer of the RBF neural network by combining the relevant parameters and the standard deviation of the neuron center point.

[0055] Among them, the hidden layer output matrix refers to the matrix composed of the output values ​​of the hidden layer neurons in the RBF network. These output values ​​are the results of radial basis function processing based on the distance between the input data and the center point of each neuron.

[0056] It is necessary to explain that calculating the output matrix of the hidden layer of the RBF neural network can effectively capture and express the complex nonlinear relationship of the input data, so that the network can show a high degree of adaptability and accuracy when dealing with complex dynamic systems, thereby improving the predictive ability and stability of the entire model.

[0057] In a possible implementation, S5 is specifically:

[0058]

[0059] in, represents the output matrix of the hidden layer of the RBF neural network at time k, e represents the exponential function, c i represents the center point of the i-th neuron, r irepresents the standard deviation corresponding to the center point of the i-th neuron, i=1,2,···,h, h represents the number of neurons in the hidden layer of the RBF neural network.

[0060] In a possible implementation, it is characterized in that the Hammerstein nonlinear dynamic system is specifically:

[0061]

[0062]

[0063]

[0064]

[0065] Among them, u k ,y k and x k denote the control input, system output and system state of the Hammerstein nonlinear dynamic system at the kth moment, respectively; A, B, C and D denote the system state matrix, system input matrix, system output matrix and system direct transfer matrix of the Hammerstein nonlinear dynamic system, respectively; f denotes a static nonlinear function, represents the noise signal of the Hammerstein nonlinear dynamic system at the kth moment, K represents the Kalman gain, and w i represents the weight of the output layer, i=1,2,···,h, h represents the number of neurons in the hidden layer of the RBF neural network, represents the expanded input matrix, represents the extended transfer matrix, W represents the weight matrix, Represents the hidden layer output data of the Hammerstein nonlinear dynamic system at the kth moment.

[0066] S6: Determine the system parameters of the Hammerstein nonlinear dynamic system according to the output matrix.

[0067] It should be noted that determining the system parameters of the Hammerstein nonlinear dynamic system can accurately extract the key dynamic characteristics of the system from the complex output data, thereby making the system model more consistent with the actual working conditions.

[0068] In a possible implementation manner, the system parameters include a system state matrix, a system input matrix, an extended input influence matrix, and an extended output influence matrix.

[0069] S6 specifically includes:

[0070] S601: Perform Hankel transform on the hidden layer output matrix and the system output matrix:

[0071]

[0072] in, represents the sequence of the hidden layer output matrix at the kth moment after nonlinear transformation, represents the sequence of the system output matrix at the kth moment after nonlinear transformation, represents the past input feature matrix, Represents the future input feature matrix, Y p Represents the past output matrix, Y f represents the future output matrix, W p represents the joint matrix.

[0073] S602: Perform LQ decomposition on the result after Hankel transformation to obtain the testability matrix:

[0074]

[0075]

[0076] Among them, R ij represents the block matrix of LQ decomposition, Q1, Q2 and Q3 represent orthogonal matrices, represents the Moore-Penrose generalized inverse, and O represents the testability matrix.

[0077] S603: Perform singular value decomposition on the test matrix:

[0078]

[0079] Among them, U1 and U2 represent left singular vector matrices, S1 represents a diagonal matrix of non-zero singular values, and Represents the right singular vector matrix transpose.

[0080] S604: Calculate the extended observability matrix according to the singular value decomposition result:

[0081]

[0082] in, represents the extended observability matrix, represents the square root of the singular value matrix, represents the state estimation matrix.

[0083] S605: Calculate system parameters based on the observability matrix and the least squares method:

[0084]

[0085]

[0086]

[0087] Among them, A represents the system state matrix, B represents the system input matrix, and C represents the system output matrix. represents the expanded input matrix, represents the extended transfer matrix, Represents the matrix Remove the first m rows of Use characters It means that the matrix The last m rows of the pseudo-inverse matrix are removed, where m represents the dimension of the output data. The feature matrix representing the future state, represents the current state feature matrix, represents the system output matrix when the system state is s, K represents the gain matrix, U f represents the future input matrix, represents the state residual, represents the state residual, represents the historical input-output matrix at time s, represents the historical input and output matrix at time s-1, I s represents the identity matrix, L represents the matrix A, C, and Related parameter matrix, N represents the system combination matrix, def represents the definition, for the convenience of calculation, represents the gain submatrix composed of the first to nth rows of the K matrix, represents the gain submatrix consisting of the n+1th to n+mth rows of the K matrix, represents the submatrix consisting of the first row to the nth row of the L matrix, represents the submatrix consisting of the n+1th to n+mth rows of the L matrix, L ij represents the submatrix corresponding to the i-th block row and j-th block column of the L matrix, M j represents the submatrix corresponding to the j-th block column of the M matrix, K ij represents the submatrix corresponding to the i-th block row and the j-th block column of the K matrix, i=1 or 2, j=1,2,···,s, s represents the total number of column blocks of the submatrix divided in the matrix.

[0088] Specifically, in order to facilitate calculation, it is necessary to divide the K matrix, L matrix and M matrix. When dividing by rows, the gain submatrix composed of the first row to the nth row of the K matrix is ​​expressed as , the gain submatrix consisting of the n+1th to n+mth rows of the K matrix is ​​expressed as , the submatrix consisting of the first row to the nth row of the L matrix is ​​expressed as , the submatrix consisting of the n+1th to n+mth rows of the L matrix is ​​expressed as ; When partitioning by column, for ease of representation, the submatrix corresponding to the i-th block row and the j-th block column of the L matrix is ​​represented as L ij , the submatrix corresponding to the j-th block column of the M matrix is ​​represented as M j , the submatrix corresponding to the i-th block row and the j-th block column of the K matrix is ​​represented as K ij .

[0089] S7: Calculate the output estimate of the Hammerstein nonlinear dynamic system based on the system parameters.

[0090] It should be noted that by calculating the output estimate of the Hammerstein nonlinear dynamic system, an intuitive prediction and understanding of the system behavior can be provided, and its response can be predicted without actually intervening in the operation of the system, resulting in more efficient and accurate system adjustment and optimization.

[0091] In a possible implementation manner, S7 specifically includes:

[0092]

[0093] in, represents the output estimate at the kth moment, represents the estimated state vector at time k-1, represents the hidden layer output matrix of the Hammerstein nonlinear dynamic system at the kth moment, Represents the hidden layer output matrix of the Hammerstein nonlinear dynamic system at the k-1th moment.

[0094] S8: Calculate the mean square error by combining the output estimate and the actual output value of the Hammerstein nonlinear dynamic system.

[0095] Among them, the mean square error is a statistical indicator that measures the difference between the model prediction value and the actual observation value.

[0096] It should be noted that the mean square error can be used to intuitively understand the accuracy and reliability of the model in practical applications, identify the shortcomings of the model, and guide subsequent model adjustments and optimizations.

[0097] In a possible implementation manner, S8 specifically includes:

[0098]

[0099] Among them, MSE represents mean square error, represents the output estimate at the kth moment, represents the actual output value at the kth moment, k=1,2,···, , Represents the total output duration of the Hammerstein nonlinear dynamic system.

[0100] S9: Determine whether the mean square error is less than the preset mean square error or whether the number of iterations reaches the preset number of iterations; if so, output the system parameters as the overall identification result of the Hammerstein nonlinear dynamic system, otherwise, use the gradient descent method to update the neuron center point and return to step S4.

[0101] Among them, the gradient descent method is a commonly used optimization algorithm, which calculates the gradient of the objective function and updates the parameters along the negative direction of the gradient in the hope of reaching the minimum value of the function.

[0102] It should be noted that by iteratively updating the neuron center point, the prediction error can be gradually reduced, the accuracy and reliability of the model can be improved, the overall system identification can be more accurate, and ultimately better system performance can be achieved.

[0103] In a possible implementation, the gradient descent method is used to update the neuron center point as follows:

[0104]

[0105] Among them, min means minimization, J means optimization objective function, Indicates The standard deviation corresponding to the center point is represents the operator for partial derivative, c represents the center point, represents the derivative of the hidden layer output.

[0106] Specifically, compared with traditional identification methods, the present invention introduces a dynamic feedback adjustment mechanism based on mean square error, which makes the system identification process more accurate, can effectively adapt to and identify complex nonlinear dynamic characteristics, and significantly improves the performance of the model and the breadth of its application.

[0107] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0108] In the present invention, by collecting a large amount of historical data and setting the initial parameters of the RBF neural network based on these data, the complex behavior of the nonlinear dynamic system can be accurately captured and simulated. The K-means algorithm and the nearest neighbor Brute force method are combined to optimize the neuron center point and the standard deviation corresponding to the neuron center point is adaptively adjusted, so that the model can adapt to changes in system behavior, improve the adaptability and accuracy of the model, enhance the recognition effect and stability of the algorithm, and calculate the mean square error to detect and evaluate the performance of the model, reduce the prediction error, and improve the accuracy and robustness of the model. The neuron center point is updated by the gradient descent method, so that the model is continuously improved in multiple iterations, and the model parameters can be systematically adjusted and optimized to improve the accuracy of the overall recognition of the system.

[0109] Reference Manual Attached Figure 2 , showing a schematic structural diagram of a Hammerstein nonlinear dynamic system identification system based on industrial processes provided by the present invention.

[0110] The present invention further provides an industrial process-based Hammerstein nonlinear dynamic system identification system 20, which is applied to the above-mentioned industrial process-based Hammerstein nonlinear dynamic system identification method, comprising:

[0111] Processor 201.

[0112] The memory 202 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 201, the industrial process-based Hammerstein nonlinear dynamic system identification method of the method embodiment is implemented.

[0113] The industrial process-based Hammerstein nonlinear dynamic system identification system 20 provided by the present invention can execute the above-mentioned industrial process-based Hammerstein nonlinear dynamic system identification method and achieve the same or similar technical effects. To avoid repetition, the present invention will not go into details.

[0114] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0115] In the present invention, by collecting a large amount of historical data and setting the initial parameters of the RBF neural network based on these data, the complex behavior of the nonlinear dynamic system can be accurately captured and simulated. The K-means algorithm and the nearest neighbor Brute force method are combined to optimize the neuron center point and the standard deviation corresponding to the neuron center point is adaptively adjusted, so that the model can adapt to changes in system behavior, improve the adaptability and accuracy of the model, enhance the recognition effect and stability of the algorithm, and calculate the mean square error to detect and evaluate the performance of the model, reduce the prediction error, and improve the accuracy and robustness of the model. The neuron center point is updated by the gradient descent method, so that the model is continuously improved in multiple iterations, and the model parameters can be systematically adjusted and optimized to improve the accuracy of the overall recognition of the system.

[0116] It should be understood that the processor in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0117] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0118] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When a computer instruction or computer program is loaded or executed on a computer, a process or function according to an embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0119] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0120] In the present invention, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0121] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0122] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0123] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0124] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0125] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0126] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0127] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0128] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for identifying a Hammerstein nonlinear dynamic system based on an industrial process as described in the method embodiment is implemented.

[0129] A computer-readable storage medium provided by the present invention can implement the steps and effects of the Hammerstein nonlinear dynamic system identification method based on industrial processes in the above method embodiment. To avoid repetition, the present invention will not go into details.

[0130] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0131] In the present invention, by collecting a large amount of historical data and setting the initial parameters of the RBF neural network based on these data, the complex behavior of the nonlinear dynamic system can be accurately captured and simulated. The K-means algorithm and the nearest neighbor Brute force method are combined to optimize the neuron center point and the standard deviation corresponding to the neuron center point is adaptively adjusted, so that the model can adapt to changes in system behavior, improve the adaptability and accuracy of the model, enhance the recognition effect and stability of the algorithm, and calculate the mean square error to detect and evaluate the performance of the model, reduce the prediction error, and improve the accuracy and robustness of the model. The neuron center point is updated by the gradient descent method, so that the model is continuously improved in multiple iterations, and the model parameters can be systematically adjusted and optimized to improve the accuracy of the overall recognition of the system.

[0132] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0133] There are a few points to note:

[0134] (1) The drawings of the embodiments of the present invention only involve structures related to the embodiments of the present invention. Other structures may refer to conventional designs.

[0135] (2) For the sake of clarity, in the drawings used to describe the embodiments of the present invention, the thickness of layers or regions is exaggerated or reduced, that is, these drawings are not drawn according to the actual scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, the element may be "directly" "on" or "under" the other element or there may be intermediate elements.

[0136] (3) In the absence of conflict, the embodiments of the present invention and the features therein may be combined with each other to obtain new embodiments.

[0137] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A Hammerstein nonlinear dynamic system identification method based on industrial processes, characterized in that: include: S1: Collect historical data of Hammerstein nonlinear dynamic system as data set; S2: Based on the data set, set relevant parameters of the RBF neural network; S3: Initializing the RBF neural network according to the data set by using the K-means algorithm, and determining the neuron center point of the hidden layer of the RBF neural network; S4: Calculate the standard deviation of the neuron center point by the nearest neighbor Brute method according to the neuron center point; S5: Calculate the hidden layer output matrix of the hidden layer of the RBF neural network by combining the relevant parameters and the standard deviation of the neuron center point; S6: Determine the system parameters of the Hammerstein nonlinear dynamic system according to the output matrix; S7: Calculating an output estimate of the Hammerstein nonlinear dynamic system according to the system parameters; S8: Calculating a mean square error by combining the output estimate value and the actual output value of the Hammerstein nonlinear dynamic system; S9: Determine whether the mean square error is less than the preset mean square error or whether the number of iterations reaches the preset number of iterations; if so, output the system parameters as the overall identification result of the Hammerstein nonlinear dynamic system, otherwise, use the gradient descent method to update the neuron center point and return to step S4.

2. The industrial process-based Hammerstein nonlinear dynamic system identification method according to claim 1, characterized in that: The relevant parameters specifically include: the number of hidden layer neurons, input dimension, output dimension, learning rate, Hankel matrix dimension, Hankel matrix data volume, training data set size, number of training iterations and training expected indicators.

3. The industrial process-based Hammerstein nonlinear dynamic system identification method according to claim 1, characterized in that: The S4 specifically includes: S401: Setting the number of adjacent data points of the neuron center point; S402: Calculating the Euclidean distance between the neuron center point and each adjacent data point; S403: performing nearest neighbor search according to the Euclidean distance; S404: fitting the neuron center point according to the nearest neighbor search result to obtain neuron center point fitting data; S405: Calculate the standard deviation of the neuron center point according to the neuron center point fitting data.

4. The industrial process-based Hammerstein nonlinear dynamic system identification method according to claim 1, characterized in that: The S5 is specifically: ； in, represents the hidden layer output matrix of the Hammerstein nonlinear dynamic system at the kth moment, e represents the exponential function, c i represents the center point of the i-th neuron, r i represents the standard deviation corresponding to the center point of the i-th neuron, i=1,2,···,h, h represents the number of neurons in the hidden layer of the RBF neural network, represents the control input of the Hammerstein nonlinear dynamic system at the kth moment.

5. The industrial process-based Hammerstein nonlinear dynamic system identification method according to claim 4, characterized in that: The Hammerstein nonlinear dynamic system is specifically: ； ； ； ； Among them, u k ,y k and x k denote the control input, system output and system state of the Hammerstein nonlinear dynamic system at the kth moment, respectively; A, B, C and D denote the system state matrix, system input matrix, system output matrix and system direct transfer matrix of the Hammerstein nonlinear dynamic system, respectively; f denotes a static nonlinear function, represents the noise signal of the Hammerstein nonlinear dynamic system at the kth moment, represents the Kalman gain, w i represents the weight of the output layer, i=1,2,···,h, h represents the number of neurons in the hidden layer of the RBF neural network, represents the expanded input matrix, represents the extended transfer matrix, W represents the weight matrix, Represents the hidden layer output matrix of the Hammerstein nonlinear dynamic system at the kth moment.

6. The industrial process-based Hammerstein nonlinear dynamic system identification method according to claim 5, characterized in that: The system parameters include a system state matrix, a system input matrix, an extended input influence matrix and an extended output influence matrix; The S6 specifically includes: S601: Perform Hankel transformation on the hidden layer output matrix and the system output matrix: ； in, represents the sequence of the hidden layer output matrix at the kth moment after nonlinear transformation, represents the sequence of the system output matrix at the kth moment after nonlinear transformation, represents the past input feature matrix, Represents the future input feature matrix, Y p Represents the past output matrix, Y f represents the future output matrix, W p represents the joint matrix; S602: Perform LQ decomposition on the result after Hankel transformation to obtain the testability matrix: ； ； Among them, R ij represents the block matrix of LQ decomposition, Q1, Q2 and Q3 represent orthogonal matrices, represents the Moore-Penrose generalized inverse, O represents the testability matrix; S603: Perform singular value decomposition on the property matrix: ； Among them, U1 and U2 represent left singular vector matrices, S1 represents a diagonal matrix of non-zero singular values, and represents the right singular vector matrix transpose; S604: Calculate the extended observability matrix according to the singular value decomposition result: ； in, represents the extended observability matrix, represents the square root of the singular value matrix, represents the state estimation matrix; S605: Calculate the system parameters according to the observability matrix and the least square method: ； ； ； Among them, A represents the system state matrix, B represents the system input matrix, and C represents the system output matrix. represents the expanded input matrix, represents the extended transfer matrix, Represents the matrix Remove the first m rows of Use characters It means that the matrix The last m rows of the pseudo-inverse matrix are removed, where m represents the dimension of the output data. The feature matrix representing the future state, represents the current state feature matrix, represents the system output matrix when the system state is s, K represents the gain matrix, U f represents the future input matrix, represents the residual matrix associated with the process noise, represents the residual matrix associated with the measurement noise, represents the historical input-output matrix at time s, represents the historical input and output matrix at time s-1, I s represents the identity matrix, L represents the matrix A, C, and Related parameter matrix, N represents the system combination matrix, def represents the definition, represents the gain submatrix composed of the first to nth rows of the K matrix, represents the gain submatrix consisting of the n+1th to n+mth rows of the K matrix, represents the submatrix consisting of the first row to the nth row of the L matrix, represents the submatrix consisting of the n+1th to n+mth rows of the L matrix, L ij represents the submatrix corresponding to the i-th block row and j-th block column of the L matrix, M j represents the submatrix corresponding to the j-th block column of the M matrix, K ij represents the submatrix corresponding to the i-th block row and the j-th block column of the K matrix, i=1 or 2, j=1,2,···,s, s represents the total number of columns of the submatrices divided in the matrix.

7. The industrial process-based Hammerstein nonlinear dynamic system identification method according to claim 6, characterized in that: The S7 is specifically: ； in, represents the output estimate at the kth moment, represents the estimated state vector at time k-1, represents the hidden layer output matrix of the Hammerstein nonlinear dynamic system at the kth moment, represents the hidden layer output matrix of the Hammerstein nonlinear dynamic system at the k-1th moment.

8. The industrial process-based Hammerstein nonlinear dynamic system identification method according to claim 7, characterized in that: The S8 is specifically: ； Among them, MSE represents mean square error, represents the estimated output value at the kth moment, represents the actual output value at the kth moment, k=1,2,···, , Represents the total duration of the output of the Hammerstein nonlinear dynamic system.

9. The industrial process-based Hammerstein nonlinear dynamic system identification method according to claim 1, characterized in that: The method of updating the neuron center point by using the gradient descent method is specifically as follows: ； Among them, min means minimization, J means the optimization target, represents the standard deviation corresponding to the center point of the i-th neuron, represents the operator for partial derivative, c represents the center point of the neuron, represents the derivative of the hidden layer output matrix, e represents the exponential function, Indicates k The actual output value at the moment, Indicates k Output the estimated value at each moment. represents the system output matrix, represents the system direct transfer matrix, represents the extended transfer matrix, Denotes the Hammerstein nonlinear dynamic system in k The hidden layer output matrix at time , represents the weight of the output layer, Denotes the Hammerstein nonlinear dynamic system in k The control input at the moment, Indicates i Neuronal centers, Indicates i The standard deviation corresponding to the center point of each neuron is i =1,2,···, h , h represents the number of neurons in the hidden layer of the RBF neural network, represents the weight matrix, express k The estimated state vector at time .

10. A Hammerstein nonlinear dynamic system identification system based on industrial processes, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the industrial process-based Hammerstein nonlinear dynamic system identification method according to any one of claims 1 to 9 is implemented.

Images