A monitoring and diagnosis system for DCS control systems in thermal power plants

Abstract

The present invention relates to the technical field of fault diagnosis of thermal power plants, and specifically to a monitoring and diagnosis system for a DCS control system of a thermal power plant, comprising a signal processing unit, a modal decomposition unit, a feature clustering unit, a first diagnosis unit, a second diagnosis unit, a control diagnosis unit and a diagnosis mapping unit, and further comprising: a signal acquisition unit, the signal acquisition unit being used to perform signal acquisition on a generator set in a target thermal power plant based on a current sensor and a vibration sensor to obtain a current signal and a first vibration signal of each generator in the generator set, and to perform signal acquisition on a steam turbine set in the target thermal power plant based on a vibration sensor to obtain a second vibration signal of each steam turbine in the steam turbine set. The present invention can comprehensively obtain the operating status information of the set by performing signal acquisition on the generator set and steam turbine set in the thermal power plant based on the current sensor and the vibration sensor through the signal acquisition unit.

Description

Technical Field

[0001] The invention relates to the technical field of fault diagnosis of thermal power plants, and in particular to a monitoring and diagnosis system for a DCS control system of a thermal power plant. Background Art

[0002] The DCS control system of a thermal power plant, or distributed control system, is the core system of automated control of modern large-scale thermal power plants. It is used to distribute the various control functions of a thermal power plant to multiple controllers. Each controller is responsible for controlling a specific area or equipment, avoiding the risk of a single controller failure causing the entire system to fail, and improving the reliability and stability of the system.

[0003] Traditional systems usually rely on only a single type of sensor, such as temperature sensors or pressure sensors, which makes it difficult to fully obtain various operating status information of generator sets and steam turbine sets. For example, the collection of current signals and vibration signals may not be comprehensive and timely. Traditional systems often lack advanced data processing and feature extraction methods, and the processing of collected signals is relatively simple. In addition, they lack effective label mapping and comprehensive analysis mechanisms, making it difficult to integrate and correlate the diagnostic results of different types of signals. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the disadvantages of the above-mentioned prior art and provide a monitoring and diagnosis system for a DCS control system of a thermal power plant.

[0005] The technical solution adopted to solve the above technical problems is: a monitoring and diagnosis system for a DCS control system of a thermal power plant, comprising a signal processing unit, a modal decomposition unit, a feature clustering unit, a first diagnosis unit, a second diagnosis unit, a control diagnosis unit and a diagnosis mapping unit, and also comprising:

[0006] A signal acquisition unit, wherein the signal acquisition unit is used to perform signal acquisition on the generator set in the target thermal power plant based on the current sensor and the vibration sensor to obtain the current signal and the first vibration signal of each generator in the generator set, and perform signal acquisition on the steam turbine set in the target thermal power plant based on the vibration sensor to obtain the second vibration signal of each steam turbine in the steam turbine set, transmit the current signal of each generator in the generator set to the feature extraction unit, transmit the first vibration signal of each generator in the generator set to the signal processing unit, and transmit the second vibration signal of each steam turbine in the steam turbine set to the modal decomposition unit;

[0007] A feature extraction unit, wherein the feature extraction unit receives the current signal of each generator in the generator set transmitted by the signal acquisition unit, and performs signal processing on the current signal of each generator in the generator set based on Hilbert transform to obtain an amplitude envelope signal corresponding to the current signal, performs feature extraction on the current signal based on the amplitude envelope signal corresponding to the current signal to obtain a feature matrix corresponding to the current signal, and transmits the feature matrix corresponding to the current signal to a feature clustering unit.

[0008] Preferably, the signal processing unit receives the first vibration signal of each generator in the generator group transmitted by the signal acquisition unit, and performs spectral analysis on the first vibration signal to obtain spectral information of the first vibration signal, wherein the spectral information includes frequency components and corresponding amplitudes, the frequency components are used as nodes, and the spectral mean is used as the weight of the node, the connecting edges are determined according to the amplitude relationship between the frequency components, the graph data corresponding to the first vibration signal is constructed based on the nodes and the connecting edges, and the graph data corresponding to the first vibration signal is transmitted to the first diagnostic unit.

[0009] Preferably, the modal decomposition unit receives the second vibration signal of each turbine in the steam turbine group transmitted by the signal acquisition unit, and reconstructs the second vibration signal to obtain a first reconstructed vibration signal, obtains a feature set corresponding to the second vibration signal based on the first reconstructed vibration signal, and transmits the feature set corresponding to the second vibration signal to the second diagnostic unit.

[0010] Preferably, the feature clustering unit receives the feature matrix corresponding to the current signal transmitted by the feature extraction unit, and derives the mean in the feature matrix corresponding to the current signal to obtain a derivative sequence corresponding to the mean in the feature matrix, compares the derivative sequence corresponding to the mean in the feature matrix with a preset derivative threshold to obtain the number of points exceeding the preset derivative threshold, clusters the feature matrix corresponding to the current signal using the number of points as the K value in the K-means clustering algorithm to obtain a clustering label corresponding to the current signal, and transmits the clustering label corresponding to the current signal to the diagnostic mapping unit.

[0011] Preferably, the first diagnostic unit receives the graph data corresponding to the first vibration signal transmitted by the signal processing unit, and performs fault diagnosis on the graph data corresponding to the first vibration signal based on a trained first diagnostic model to obtain a diagnostic label corresponding to the first vibration signal, and transmits the diagnostic label corresponding to the first vibration signal to the diagnostic mapping unit.

[0012] Preferably, the second diagnostic unit receives the feature set corresponding to the second vibration signal transmitted by the modal decomposition unit, and performs fault diagnosis on the feature set corresponding to the second vibration signal based on the trained second diagnostic model to obtain the diagnostic label corresponding to the second vibration signal, that is, obtain the diagnostic label corresponding to each steam turbine in the steam turbine unit, and transmit the diagnostic label corresponding to each steam turbine in the steam turbine unit to the control diagnostic unit.

[0013] Preferably, the diagnostic mapping unit receives the clustering label corresponding to the current signal transmitted by the feature clustering unit and the diagnostic label corresponding to the first vibration signal transmitted by the first diagnostic unit, and encodes and assigns the clustering label corresponding to the current signal and the diagnostic label corresponding to the first vibration signal based on a preset label coding set to obtain a label coding sequence, matches the label coding sequence with a preset label sequence table to obtain the diagnostic label corresponding to each generator in the generator set, and transmits the diagnostic label corresponding to each generator in the generator set to the control diagnostic unit.

[0014] Preferably, the control diagnosis unit receives the diagnostic tag corresponding to each steam turbine in the steam turbine unit transmitted by the second diagnostic unit and the diagnostic tag corresponding to each generator in the generator unit transmitted by the diagnostic mapping unit, and obtains the operating status corresponding to each steam turbine in the steam turbine unit and the operating status corresponding to each generator in the generator unit in the target thermal power plant based on the operating data of the DCS control system of the target thermal power plant, and compares the operating status corresponding to each steam turbine in the steam turbine unit and the operating status corresponding to each generator in the generator unit with the diagnostic tags corresponding to each steam turbine in the steam turbine unit and the diagnostic tags corresponding to each generator in the generator unit to obtain the diagnostic tags of the DCS control system of the target thermal power plant.

[0015] Preferably, feature extraction is performed on the current signal based on an amplitude envelope signal corresponding to the current signal to obtain a feature matrix corresponding to the current signal, including:

[0016] An analytical signal of the current signal is obtained based on an amplitude envelope signal corresponding to the current signal, wherein the expression of the amplitude envelope signal is as follows:

[0017] ;

[0018] in, represents the amplitude envelope signal, represents the Hilbert transform operator, Represents the current signal, represents the integral variable;

[0019] The expression of the analytical signal is as follows:

[0020] ;

[0021] in, Represents the analytical signal, represents an imaginary unit;

[0022] The amplitude of the current signal is obtained based on the analytical signal of the current signal, and the amplitude of the current signal is processed based on a preset sliding window to obtain a set of modulus square envelope samples, wherein the expression of the amplitude is as follows:

[0023] ;

[0024] in, Indicates the amplitude of the current signal;

[0025] The peak value, mean value, slope mean value and skewness of the modulus square envelope sample set are calculated to obtain a characteristic matrix corresponding to the current signal.

[0026] Preferably, acquiring a feature set corresponding to the second vibration signal based on the first reconstructed vibration signal includes:

[0027] The first reconstructed vibration signal is decomposed based on the empirical mode decomposition technology to obtain a plurality of modal function components and a plurality of residual components corresponding to the first reconstructed vibration signal, wherein the expression of the first reconstructed vibration signal is as follows:

[0028] ;

[0029] in, represents the first reconstructed vibration signal, represents the second vibration signal, represents white noise;

[0030] Calculating average values ​​of the plurality of modal function components and the plurality of residual components;

[0031] reconstructing a second vibration signal based on an average value of the plurality of modal function components and the plurality of residual components to obtain a second reconstructed vibration signal;

[0032] The correlation between the multiple modal function components and the second reconstructed vibration signal is calculated to obtain multiple correlation coefficients, wherein the calculation formula of the correlation is as follows:

[0033] ;

[0034] in, Indicates modal function components With the second reconstructed vibration signal The relevance of represents the covariance function, represents the variance function, and , represents the total number of modal function components, Indicates Residual components;

[0035] Calculating energy characteristics of the multiple modal function components;

[0036] The multiple correlation coefficients and energy features are used as a feature set of the second vibration signal.

[0037] Preferably, the first diagnostic model adopts a graph neural network, and the first diagnostic model includes a graph attention module and a node feature classification module. The graph attention module is used to learn the attention coefficient between nodes based on the node weights in the graph data, and the attention coefficient is passed through a layer of graph attention convolution layer based on the attention coefficient to obtain the node features. The feature classification module is used to classify the nodes based on the fully connected layer and the Softmax function to obtain the diagnostic label corresponding to the first vibration signal.

[0038] Preferably, the calculation formula of the attention coefficient is as follows:

[0039] ;

[0040] in, Indicates the first Nodes and The weight of the edge connecting the nodes, represents the activation function, represents the learnable parameter vector, represents the graph filter parameter matrix, Indicates The weight corresponding to each node.

[0041] Preferably, the expression of the node feature is as follows:

[0042] ;

[0043] in, Indicates the first The node characteristics of each node.

[0044] Preferably, the second diagnostic model adopts a support vector machine.

[0045] The beneficial effects of the present invention are as follows: (1) The present invention collects signals from the generator sets and steam turbine sets in the thermal power plant through the signal acquisition unit based on the current sensor and the vibration sensor, and can comprehensively obtain the operating status information of the sets, including the current signal and vibration signal of the generator set and the vibration signal of the steam turbine set, providing a rich data basis for subsequent fault diagnosis, and the signal processing unit performs spectrum analysis on the first vibration signal of the generator set and constructs graph data, and the modal decomposition unit reconstructs and extracts features from the second vibration signal of the steam turbine set. These processing methods help to deeply explore the features and information in the vibration signal and improve the accuracy of fault diagnosis; (2) The present invention clusters the feature matrix of the current signal through the feature clustering unit, and can classify current signals with similar features into one category, providing preliminary classification information for subsequent diagnosis, and the first A diagnostic unit performs fault diagnosis on the graph data corresponding to the first vibration signal based on the trained first diagnostic model, and a second diagnostic unit performs fault diagnosis on the feature set corresponding to the second vibration signal based on the trained second diagnostic model. The two diagnostic units perform professional fault diagnosis on the vibration signals of the generator set and the steam turbine set respectively, thereby improving the pertinence and accuracy of the diagnosis. (3) The present invention encodes and assigns the clustering label of the current signal and the diagnostic label of the first vibration signal through the diagnostic mapping unit and matches them with the preset label sequence table to obtain the diagnostic label of each generator of the generator set. Finally, the control diagnostic unit compares the diagnostic label of the steam turbine set and the generator set with the operating status to obtain the diagnostic label of the DCS control system of the entire target thermal power plant, thereby realizing the complete process from signal acquisition to the final diagnostic result, and providing a reliable basis for the operation and maintenance of the thermal power plant. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A schematic diagram of the system architecture of an overall system in an embodiment of the present invention.

[0047] Figure numerals: 1. signal acquisition unit; 2. feature extraction unit; 3. signal processing unit; 4. modal decomposition unit; 5. feature clustering unit; 6. first diagnosis unit; 7. second diagnosis unit; 8. control diagnosis unit; 9. diagnosis mapping unit. DETAILED DESCRIPTION

[0048] Embodiment 1, as Figure 1 As shown, a monitoring and diagnosis system for a DCS control system of a thermal power plant proposed by the present invention includes a signal processing unit 3, a modal decomposition unit 4, a feature clustering unit 5, a first diagnosis unit 6, a second diagnosis unit 7, a control diagnosis unit 8 and a diagnosis mapping unit 9, and also includes:

[0049] The signal acquisition unit 1 is used to perform signal acquisition on the generator set in the target thermal power plant based on the current sensor and the vibration sensor to obtain the current signal and the first vibration signal of each generator in the generator set, and perform signal acquisition on the steam turbine set in the target thermal power plant based on the vibration sensor to obtain the second vibration signal of each steam turbine in the steam turbine set, transmit the current signal of each generator in the generator set to the feature extraction unit 2, transmit the first vibration signal of each generator in the generator set to the signal processing unit 3, and transmit the second vibration signal of each steam turbine in the steam turbine set to the modal decomposition unit 4;

[0050] The feature extraction unit 2 receives the current signal of each generator in the generator set transmitted by the signal acquisition unit 1, and performs signal processing on the current signal of each generator in the generator set based on the Hilbert transform to obtain an amplitude envelope signal corresponding to the current signal, performs feature extraction on the current signal based on the amplitude envelope signal corresponding to the current signal to obtain a feature matrix corresponding to the current signal, and transmits the feature matrix corresponding to the current signal to the feature clustering unit 5.

[0051] In the present invention, Hilbert transform refers to a signal processing technology used to convert a real signal into an analytical signal, so as to obtain information such as the amplitude and phase of the signal; the amplitude envelope signal represents the envelope of the signal amplitude change, reflecting the overall change trend of the signal amplitude; the feature matrix refers to the extracted features expressed in the form of a matrix, which is convenient for data processing and analysis.

[0052] In an optional embodiment, the signal processing unit 3 receives the first vibration signal of each generator in the generator group transmitted by the signal acquisition unit 1, and performs spectral analysis on the first vibration signal to obtain spectral information of the first vibration signal, wherein the spectral information includes frequency components and corresponding amplitudes, and the frequency components are used as nodes, and the spectral mean is used as the weight of the node. The connecting edges are determined according to the amplitude relationship between the frequency components, and the graph data corresponding to the first vibration signal is constructed based on the nodes and the connecting edges, and the graph data corresponding to the first vibration signal is transmitted to the first diagnostic unit 6.

[0053] It should be noted that spectrum analysis refers to the process of converting signals from the time domain to the frequency domain, and understanding the characteristics of the signal by analyzing the components and energy distribution of the signal at different frequencies; spectrum information refers to the results obtained through spectrum analysis, including the various frequency components contained in the signal and the amplitude corresponding to each frequency component.

[0054] In an optional embodiment, the modal decomposition unit 4 receives the second vibration signal of each turbine in the turbine group transmitted by the signal acquisition unit 1, and reconstructs the second vibration signal to obtain a first reconstructed vibration signal, obtains a feature set corresponding to the second vibration signal based on the first reconstructed vibration signal, and transmits the feature set corresponding to the second vibration signal to the second diagnostic unit 7.

[0055] In an optional embodiment, the feature clustering unit 5 receives the feature matrix corresponding to the current signal transmitted by the feature extraction unit 2, and derives the mean in the feature matrix corresponding to the current signal to obtain a derivative sequence corresponding to the mean in the feature matrix, compares the derivative sequence corresponding to the mean in the feature matrix with a preset derivative threshold to obtain the number of points exceeding the preset derivative threshold, clusters the feature matrix corresponding to the current signal using the number of points as the K value in the K-means clustering algorithm to obtain a clustering label corresponding to the current signal, and transmits the clustering label corresponding to the current signal to the diagnostic mapping unit 9.

[0056] It should be noted that the derivative sequence refers to a series of values ​​obtained by differentiating the mean, which reflects the change of the mean; the number of points refers to the number of values ​​in the derivative sequence that exceed the preset derivative threshold; the K-means clustering algorithm refers to an unsupervised learning algorithm that divides data into K clusters, so that the data similarity within each cluster is high, while the data similarity between different clusters is low; the K value refers to the number of clusters that need to be determined in advance in the K-means clustering algorithm; the cluster label refers to the category identifier assigned to the data by the clustering algorithm, indicating which cluster the data belongs to.

[0057] In an optional embodiment, the first diagnostic unit 6 receives the graphic data corresponding to the first vibration signal transmitted by the signal processing unit 3, and performs fault diagnosis on the graphic data corresponding to the first vibration signal based on a trained first diagnostic model to obtain a diagnostic label corresponding to the first vibration signal, and transmits the diagnostic label corresponding to the first vibration signal to the diagnostic mapping unit 9.

[0058] In an optional embodiment, the second diagnostic unit 7 receives the feature set corresponding to the second vibration signal transmitted by the modal decomposition unit 4, and performs fault diagnosis on the feature set corresponding to the second vibration signal based on the trained second diagnostic model to obtain the diagnostic label corresponding to the second vibration signal, that is, obtain the diagnostic label corresponding to each steam turbine in the steam turbine unit, and transmit the diagnostic label corresponding to each steam turbine in the steam turbine unit to the control diagnostic unit 8.

[0059] In an optional embodiment, the diagnostic mapping unit 9 receives the clustering label corresponding to the current signal transmitted by the feature clustering unit 5 and the diagnostic label corresponding to the first vibration signal transmitted by the first diagnostic unit 6, and encodes and assigns the clustering label corresponding to the current signal and the diagnostic label corresponding to the first vibration signal based on a preset label coding set to obtain a label coding sequence, matches the label coding sequence with a preset label sequence table to obtain the diagnostic label corresponding to each generator in the generator set, and transmits the diagnostic label corresponding to each generator in the generator set to the control diagnostic unit 8.

[0060] It should be noted that the preset label coding set refers to a set of coding rules set in advance for encoding and assigning values ​​to cluster labels and diagnostic labels; coding assignment refers to assigning corresponding coding values ​​to cluster labels and diagnostic labels according to the preset label coding set; label coding sequence refers to a sequence of a series of coding values ​​obtained after encoding and assigning values ​​to the cluster labels of the current signal and the diagnostic labels of the first vibration signal; the preset label sequence table refers to a table prepared in advance that contains various possible label coding sequences and their corresponding diagnostic results.

[0061] In an optional embodiment, the control diagnosis unit 8 receives the diagnostic tags corresponding to each steam turbine in the steam turbine unit transmitted by the second diagnostic unit 7 and the diagnostic tags corresponding to each generator in the generator unit transmitted by the diagnostic mapping unit 9, and obtains the operating status corresponding to each steam turbine in the steam turbine unit and the operating status corresponding to each generator in the generator unit in the target thermal power plant based on the operating data of the DCS control system of the target thermal power plant, and compares the operating status corresponding to each steam turbine in the steam turbine unit and the operating status corresponding to each generator in the generator unit with the diagnostic tags corresponding to each steam turbine in the steam turbine unit and the diagnostic tags corresponding to each generator in the generator unit to obtain the diagnostic tags of the DCS control system of the target thermal power plant.

[0062] Embodiment 2, a monitoring and diagnostic system for a DCS control system of a thermal power plant proposed by the present invention, compared with embodiment 1, this embodiment further includes: extracting features of the current signal based on the amplitude envelope signal corresponding to the current signal to obtain a feature matrix corresponding to the current signal, including:

[0063] The analytical signal of the current signal is obtained based on the amplitude envelope signal corresponding to the current signal, wherein the expression of the amplitude envelope signal is as follows:

[0064] ;

[0065] in, represents the amplitude envelope signal, represents the Hilbert transform operator, Represents the current signal, represents the integral variable;

[0066] The expression of the analytical signal is as follows:

[0067] ;

[0068] in, Represents the analytical signal, represents an imaginary unit;

[0069] The amplitude of the current signal is obtained based on the analytical signal of the current signal, and the amplitude of the current signal is processed based on a preset sliding window to obtain a set of modulus square envelope samples, wherein the expression of the amplitude is as follows:

[0070] ;

[0071] in, Indicates the amplitude of the current signal;

[0072] The peak value, mean value, slope mean value and skewness of the modulus square envelope sample set are calculated to obtain a characteristic matrix corresponding to the current signal.

[0073] In an optional embodiment, acquiring a feature set corresponding to the second vibration signal based on the first reconstructed vibration signal includes:

[0074] The first reconstructed vibration signal is decomposed based on the empirical mode decomposition technology to obtain multiple mode function components and multiple residual components corresponding to the first reconstructed vibration signal, wherein the expression of the first reconstructed vibration signal is as follows:

[0075] ;

[0076] in, represents the first reconstructed vibration signal, represents the second vibration signal, represents white noise;

[0077] Calculate the average values ​​of multiple modal function components and multiple residual components;

[0078] reconstructing a second vibration signal based on an average value of the plurality of modal function components and the plurality of residual components to obtain a second reconstructed vibration signal;

[0079] The correlation between the multiple modal function components and the second reconstructed vibration signal is calculated to obtain multiple correlation coefficients, wherein the calculation formula of the correlation is as follows:

[0080] ;

[0081] in, Indicates modal function components With the second reconstructed vibration signal The relevance of represents the covariance function, represents the variance function, and , represents the total number of modal function components, Indicates Residual components;

[0082] Calculate the energy characteristics of multiple modal function components;

[0083] The multiple correlation coefficients and energy features are used as a feature set of the second vibration signal.

[0084] In an optional embodiment, the first diagnostic model adopts a graph neural network, and the first diagnostic model includes a graph attention module and a node feature classification module. The graph attention module is used to learn the attention coefficient between nodes based on the node weights in the graph data. The attention coefficient is the weight of the connecting edge in the graph data, and a layer of graph attention convolution layer is passed based on the attention coefficient to obtain the node features. The feature classification module is used to classify the nodes based on the fully connected layer and the Softmax function to obtain the diagnostic label corresponding to the first vibration signal.

[0085] It should be noted that Graph Neural Network (GNN) is a neural network model that specifically processes graph structured data. In a graph structure, data exists in the form of nodes and edges. Nodes represent entities, and edges represent the relationships between entities. The purpose of a graph neural network is to learn the feature representations of nodes in a graph, as well as to capture the structural information of the graph and the dependencies between nodes.

[0086] In an optional embodiment, the calculation formula of the attention coefficient is as follows:

[0087] ;

[0088] in, Indicates the first Nodes and The weight of the edge connecting the nodes, represents the activation function, represents the learnable parameter vector, represents the graph filter parameter matrix, Indicates The weight corresponding to each node.

[0089] In an optional embodiment, the expression of the node feature is as follows:

[0090] ;

[0091] in, Indicates the first The node characteristics of each node.

[0092] In an optional embodiment, the second diagnostic model uses a support vector machine.

[0093] It should be noted that Support Vector Machine (SVM) is a supervised machine learning algorithm, which is often used for classification and regression problems. In classification problems, the basic idea of ​​SVM is to find a hyperplane in the feature space that can separate samples of different categories as much as possible. This hyperplane should maximize the distance from the nearest sample point (called support vector) to the hyperplane, thereby ensuring the accuracy and generalization ability of classification.

[0094] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A monitoring and diagnosis system for a DCS control system of a thermal power plant, comprising a signal processing unit (3), a modal decomposition unit (4), a feature clustering unit (5), a first diagnosis unit (6), a second diagnosis unit (7), a control diagnosis unit (8) and a diagnosis mapping unit (9), characterized in that: A signal acquisition unit (1), the signal acquisition unit (1) being used to acquire signals from a generator set in a target thermal power plant based on a current sensor and a vibration sensor to obtain a current signal and a first vibration signal of each generator in the generator set, and to acquire signals from a steam turbine set in the target thermal power plant based on a vibration sensor to obtain a second vibration signal of each steam turbine in the steam turbine set, transmit the current signal of each generator in the generator set to a feature extraction unit (2), transmit the first vibration signal of each generator in the generator set to a signal processing unit (3), and transmit the second vibration signal of each steam turbine in the steam turbine set to a modal decomposition unit (4); A feature extraction unit (2), the feature extraction unit (2) receiving the current signal of each generator in the generator set transmitted by the signal acquisition unit (1), and performing signal processing on the current signal of each generator in the generator set based on Hilbert transform to obtain an amplitude envelope signal corresponding to the current signal, performing feature extraction on the current signal based on the amplitude envelope signal corresponding to the current signal to obtain a feature matrix corresponding to the current signal, and transmitting the feature matrix corresponding to the current signal to a feature clustering unit (5); The signal processing unit (3) receives the first vibration signal of each generator in the generator set transmitted by the signal acquisition unit (1), and performs spectrum analysis on the first vibration signal to obtain spectrum information of the first vibration signal, wherein the spectrum information includes frequency components and corresponding amplitudes, the frequency components are used as nodes, and the spectrum mean is used as the weight of the node, connecting edges are determined according to the amplitude relationship between the frequency components, graph data corresponding to the first vibration signal is constructed based on the nodes and the connecting edges, and the graph data corresponding to the first vibration signal is transmitted to the first diagnosis unit (6); The modal decomposition unit (4) receives the second vibration signal of each steam turbine in the steam turbine group transmitted by the signal acquisition unit (1), reconstructs the second vibration signal to obtain a first reconstructed vibration signal, obtains a feature set corresponding to the second vibration signal based on the first reconstructed vibration signal, and transmits the feature set corresponding to the second vibration signal to a second diagnosis unit (7); The feature clustering unit (5) receives the feature matrix corresponding to the current signal transmitted by the feature extraction unit (2), and derives the mean value in the feature matrix corresponding to the current signal to obtain a derivative sequence corresponding to the mean value in the feature matrix, compares the derivative sequence corresponding to the mean value in the feature matrix with a preset derivative threshold value to obtain the number of points exceeding the preset derivative threshold value, clusters the feature matrix corresponding to the current signal using the number of points as the K value in the K-means clustering algorithm to obtain a clustering label corresponding to the current signal, and transmits the clustering label corresponding to the current signal to the diagnostic mapping unit (9); The first diagnosis unit (6) receives the graph data corresponding to the first vibration signal transmitted by the signal processing unit (3), performs fault diagnosis on the graph data corresponding to the first vibration signal based on the trained first diagnosis model to obtain a diagnosis label corresponding to the first vibration signal, and transmits the diagnosis label corresponding to the first vibration signal to the diagnosis mapping unit (9); The second diagnosis unit (7) receives the feature set corresponding to the second vibration signal transmitted by the modal decomposition unit (4), and performs fault diagnosis on the feature set corresponding to the second vibration signal based on the trained second diagnosis model to obtain a diagnosis label corresponding to the second vibration signal, that is, to obtain a diagnosis label corresponding to each steam turbine in the steam turbine unit, and transmits the diagnosis label corresponding to each steam turbine in the steam turbine unit to the control diagnosis unit (8); The diagnostic mapping unit (9) receives the clustering label corresponding to the current signal transmitted by the feature clustering unit (5) and the diagnostic label corresponding to the first vibration signal transmitted by the first diagnostic unit (6), and encodes and assigns values ​​to the clustering label corresponding to the current signal and the diagnostic label corresponding to the first vibration signal based on a preset label encoding set to obtain a label encoding sequence, matches the label encoding sequence with a preset label sequence table to obtain a diagnostic label corresponding to each generator in the generator set, and transmits the diagnostic label corresponding to each generator in the generator set to the control diagnostic unit (8); The control diagnosis unit (8) receives the diagnostic tags corresponding to the individual steam turbines in the steam turbine unit transmitted by the second diagnosis unit (7) and the diagnostic tags corresponding to the individual generators in the generator unit transmitted by the diagnosis mapping unit (9), and obtains the operating status corresponding to the individual steam turbines in the steam turbine unit and the operating status corresponding to the individual generators in the generator unit in the target thermal power plant based on the operating data of the DCS control system of the target thermal power plant, and compares the operating status corresponding to the individual steam turbines in the steam turbine unit and the operating status corresponding to the individual generators in the generator unit with the diagnostic tags corresponding to the individual steam turbines in the steam turbine unit and the diagnostic tags corresponding to the individual generators in the generator unit to obtain the diagnostic tags of the DCS control system of the target thermal power plant.

2. A monitoring and diagnostic system for a DCS control system of a thermal power plant according to claim 1, characterized in that: Extracting features of the current signal based on an amplitude envelope signal corresponding to the current signal to obtain a feature matrix corresponding to the current signal includes: An analytical signal of the current signal is obtained based on an amplitude envelope signal corresponding to the current signal, wherein the expression of the amplitude envelope signal is as follows: ； in, represents the amplitude envelope signal, represents the Hilbert transform operator, Represents the current signal, represents the integral variable; The expression of the analytical signal is as follows: ； in, Represents the analytical signal, represents an imaginary unit; The amplitude of the current signal is obtained based on the analytical signal of the current signal, and the amplitude of the current signal is processed based on a preset sliding window to obtain a set of modulus square envelope samples, wherein the expression of the amplitude is as follows: ； in, Indicates the amplitude of the current signal; The peak value, mean value, slope mean value and skewness of the modulus square envelope sample set are calculated to obtain a characteristic matrix corresponding to the current signal.

3. A monitoring and diagnostic system for a DCS control system of a thermal power plant according to claim 2, characterized in that: Acquiring a feature set corresponding to the second vibration signal based on the first reconstructed vibration signal includes: The first reconstructed vibration signal is decomposed based on the empirical mode decomposition technology to obtain a plurality of modal function components and a plurality of residual components corresponding to the first reconstructed vibration signal, wherein the expression of the first reconstructed vibration signal is as follows: ； in, represents the first reconstructed vibration signal, represents the second vibration signal, represents white noise; Calculating average values ​​of the plurality of modal function components and the plurality of residual components; reconstructing a second vibration signal based on an average value of the plurality of modal function components and the plurality of residual components to obtain a second reconstructed vibration signal; The correlation between the multiple modal function components and the second reconstructed vibration signal is calculated to obtain multiple correlation coefficients, wherein the calculation formula of the correlation is as follows: ； in, Indicates modal function components With the second reconstructed vibration signal The relevance of represents the covariance function, represents the variance function, and , represents the total number of modal function components, Indicates Residual components; Calculating energy characteristics of the multiple modal function components; The multiple correlation coefficients and energy features are used as a feature set of the second vibration signal.

4. A monitoring and diagnostic system for a DCS control system of a thermal power plant according to claim 3, characterized in that: The first diagnostic model adopts a graph neural network, and the first diagnostic model includes a graph attention module and a node feature classification module. The graph attention module is used to learn the attention coefficient between nodes based on the node weights in the graph data, and the attention coefficient is passed through a layer of graph attention convolution layer based on the attention coefficient to obtain the node features. The feature classification module is used to classify the nodes based on the fully connected layer and the Softmax function to obtain the diagnostic label corresponding to the first vibration signal.

5. A monitoring and diagnostic system for a DCS control system of a thermal power plant according to claim 4, characterized in that: The calculation formula of the attention coefficient is as follows: ； in, Indicates the first Nodes and The weight of the edge connecting the nodes, represents the activation function, represents the learnable parameter vector, represents the graph filter parameter matrix, Indicates The weight corresponding to each node.

6. A monitoring and diagnostic system for a DCS control system of a thermal power plant according to claim 5, characterized in that: The expression of the node feature is as follows: ； in, Indicates the first The node characteristics of each node.

7. A monitoring and diagnostic system for a DCS control system of a thermal power plant according to claim 6, characterized in that: The second diagnostic model uses a support vector machine.

Images