CN115456013A - Wind turbine generator converter power module fault diagnosis method based on operation data - Google Patents
Wind turbine generator converter power module fault diagnosis method based on operation data Download PDFInfo
- Publication number
- CN115456013A CN115456013A CN202211029876.XA CN202211029876A CN115456013A CN 115456013 A CN115456013 A CN 115456013A CN 202211029876 A CN202211029876 A CN 202211029876A CN 115456013 A CN115456013 A CN 115456013A
- Authority
- CN
- China
- Prior art keywords
- node
- graph
- wind turbine
- fault diagnosis
- matrix
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/327—Testing of circuit interrupters, switches or circuit-breakers
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Wind Motors (AREA)
Abstract
The invention discloses a wind turbine generator converter power module fault diagnosis method based on operation data, which comprises the following steps of: identifying a fault mode according to historical operating data of the wind turbine generator; carrying out normalization processing on the data; 3) Decomposing the fault signal, extracting features, and representing by using a feature vector; 4) Calculating the similarity between different feature vectors by taking each feature vector as a node, selecting the nodes with the similarity larger than k to establish an adjacency relation, and constructing an adjacency matrix; 5) Inputting the nodes, the adjacency matrix and the degree matrix of the graph into a graph convolution neural network model to realize fault diagnosis; 6) And evaluating the graph convolution neural network model. In the invention, the MRSVD noise reduction method with a multi-partition structure is utilized to effectively eliminate the noise in the sample data, solve the problem of mode aliasing to a certain extent, establish a graph structure, convert the time sequence problem into the problem of graph data on time and space, utilize a graph convolution neural network to realize fault diagnosis and improve the accuracy.
Description
Technical Field
The invention belongs to the technical field of wind power generation, and particularly relates to a wind turbine generator converter power module fault diagnosis method based on operation data.
Background
The converter is used as one of core components of the wind turbine generator, and the fault rate is high. The main equipment failures of the converter power module are short circuit and open circuit failures of the IGBT module. And the short-circuit fault of the IGBT module can be judged by detecting the conduction voltage drop of the D pole and the S pole through the auxiliary circuit. When the IGBT module of the converter is in an open circuit, the extraction of the fault characteristic quantity is difficult, and the accurate positioning of the IGBT module of the converter with the open circuit fault has certain difficulty.
In the existing wind turbine generator system fault diagnosis research, two categories, namely qualitative analysis method and quantitative analysis method, can be generally distinguished. The qualitative analysis method mainly comprises an expert system method, a symbol directed graph method and a fault tree analysis method. The quantitative analysis method comprises an analytic model method and a data-driven method, wherein the analytic model method comprises a state estimation method and a parameter estimation method, and the data-driven method mainly comprises an intelligent fault diagnosis method, a mathematical diagnosis method and a signal processing method. With the rise of artificial intelligence, fault diagnosis based on an intelligent algorithm is gradually favored by learners. And C, S, tsai and the like utilize wavelets to analyze the fault characteristics of the damaged blades of the wind turbine generator, so that the fault state of the wind turbine generator is diagnosed. Zhang and the like are combined with a back propagation neural network and a genetic algorithm to construct a wind turbine generator anomaly identification model. Zhujie et al propose a mechanical failure diagnosis method based on a translation invariant convolutional neural network. The ginger, the shin-shine and the like perform wavelet threshold denoising on the acquired signals, extract fault characteristics by a MEEMD decomposition method, and finally perform mode identification on the main shaft bearing fault by using a method of combining a CPSO-BBO algorithm and a support vector machine. Although the methods such as the wavelet analysis can realize the processing of noise signals, the methods all have the inevitable defect of modal aliasing, and the methods such as the BBO algorithm can improve the diagnosis speed to a certain extent, but have characteristics which are not learned, and have the problems of insufficient network performance and complex network structure.
Disclosure of Invention
In order to solve the technical problems in the prior art, the invention aims to provide a wind turbine generator converter power module fault diagnosis method based on operation data.
In order to achieve the purpose and achieve the technical effect, the invention adopts the technical scheme that:
the wind turbine generator converter power module fault diagnosis method based on the operation data comprises the following steps:
1) Identifying N fault modes according to historical operating data of the wind turbine generator;
2) Normalizing the data obtained in the step 1), and eliminating adverse effects caused by singular sample data;
3) Decomposing the fault signal, extracting features, and representing different features by using feature vectors;
4) Taking each eigenvector as a node, calculating the similarity between different eigenvectors by using the Langmuir distance, selecting the node with the similarity larger than k to establish an adjacency relation, and constructing an adjacency matrix A;
5) Inputting the nodes, the adjacent matrix A and the degree matrix of the graph into a graph convolution neural network model to realize fault diagnosis of a wind turbine generator converter power module;
6) And evaluating the graph convolution neural network model by using the accuracy, precision and recall rate as evaluation indexes.
Further, in the step 2), the data obtained in the step 1) is normalized according to the following formula:
where x and z are data before and after processing, respectively, μ is the mean of the data set, and σ is the standard deviation of the data set.
Further, in step 3), decomposing the fault signal to extract features, wherein the specific steps of expressing different features by using feature vectors include:
31 By a multi-division-structured MRSVD noise reduction method, a signal is decomposed into M layers, and two approximate signals A of each layer in a decomposition model used for noise reduction are analyzed j,2 ,A j,1 Reconstructing to obtain a reconstructed signal A j Thus, M components can be obtained by decomposing the M layers, and M reconstructed signals can be obtained;
32 Calculate an energy value for each component
33 Construct a feature vector X' = [ E ] 1 ,E 2 ,…,E M ]；
34 Normalizing the feature vectors, so that the operation is more convenient and the error can be reduced; order to
The normalized feature vector is X = [ E = 1 /E,E 2 /E,…,E M /E]。
Further, in step 4), each feature vector is used as a node, the similarity between different feature vectors is calculated by using the distance between the nodes in the lanches, the nodes with the similarity larger than k are selected to establish an adjacency relation, and the specific steps of constructing the adjacency matrix a include:
41 Computing the Langmuir distance between nodes, the Langmuir distance D of two discrete random variables x and y L (x, y) is represented as:
42 Setting different threshold values k by adopting an experimental error method, and training to obtain an optimal k value; selecting nodes with similarity larger than k as adjacent nodes, constructing edges between the nodes, and finally constructing an adjacent matrix A by using the points and the edges, wherein the adjacent matrix A is expressed as:
wherein if the node J is adjacent to the node M, J is 1,2,3.. . . . . M, then I xJxM Is the similarity value between node J and node M; if node J is not adjacent to node M, then I xJxM =0,m represents the number of eigenvectors.
Further, in step 5), the graph convolution neural network model extracts implicit graph information by using structural information of connection between edges and vertexes of the graph and attribute information attached to the graph structure;
the graph integration layer propagation formula in the graph integration neural network model is expressed as:
in the formula, sigma is an activation function;
is a matrix of degrees of the graph,
a is the adjacency matrix of the figure,
I N is an n-order identity matrix; x (l) Is a characteristic of the l-th layer; w is the weight matrix to be trained.
Further, in step 5), before fault diagnosis, the number of GCN layers of the graph convolution neural network model is determined, an experimental error method is adopted in the training process, an output matrix of the previous layer of GCN becomes a new node feature matrix of the next layer of GCN, and feature information fusion and dimension transformation are performed through a plurality of GCN layers, so that each node feature is fused with the adjacent node feature.
Furthermore, the number of the GCN layers is three.
Further, in step 6), the calculation formulas of the accuracy, the precision and the recall rate are respectively:
accuracy＝(TP+TN)/(P+N)
precision＝TP/(TP+FP)
recall＝TP/(TP+FN)＝TP/P
in the formula, accuracy is the correct rate; precision is precision; recall is the recall rate; TP is the number correctly divided into positive examples, i.e. the number of examples (number of samples) that are actually positive examples and are divided into positive examples by the classifier; FP is the number wrongly divided into positive cases, i.e. the number of cases that are actually negative but divided into positive cases by the classifier; FN is the number of instances that are wrongly divided into negative cases, i.e. the number of instances that are actually positive cases but are divided into negative cases by the classifier; TN is the number of instances correctly divided into negative cases, i.e., actually negative cases and divided into negative cases by the classifier.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides a wind turbine generator converter power module fault diagnosis method based on operating data, which is characterized in that a fault signal is decomposed layer by layer through a multi-resolution singular value decomposition (MRSVD) noise reduction method of a multi-partition structure to realize noise reduction processing, noise in sample data is eliminated, fault characteristics are highlighted, a characteristic vector is constructed, and the occurrence of modal aliasing is reduced; and then, calculating the similarity among all the characteristic vectors by using the Langmuir distance, constructing a graph structure by using the data characteristics as vertexes and the similarity as edges, converting a time sequence problem into a graph data problem on time and space, extracting complete space-time characteristics of a graph, and finally, realizing fault category classification by using a graph convolution neural network model to accurately identify the fault category. According to the MRSVD noise reduction method, the mutation information can be clearly and effectively detected under the background of strong noise by virtue of good numerical robustness of the MRSVD noise reduction method, the MRSVD noise reduction method has the advantages of zero phase shift, small waveform distortion and high signal-to-noise ratio, compared with other existing methods, the noise in sample data can be effectively eliminated, the problem of mode aliasing is solved to a certain extent, meanwhile, in the fault classification process, similarity among characteristic components is utilized, a Langerhans distance improved adjacency matrix is adopted, a time sequence problem is converted into a time and space graph data problem, fault classification is realized by virtue of the strong advantages of a graph structure, the fault identification accuracy is improved, and a direction is provided for fault diagnosis of a power module of a converter of a wind turbine generator.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a schematic of the configuration of the energy feature of the present invention;
FIG. 3 is a schematic diagram of the present invention of a graph convolution neural network model to extract spatial features.
Detailed Description
The present invention is described in detail below so that the advantages and features of the present invention can be more easily understood by those skilled in the art, and thus the scope of the present invention can be clearly and clearly defined.
The following presents a simplified summary of one or more aspects in order to provide a basic understanding of such aspects. This summary is not an extensive overview of all contemplated aspects, and is intended to neither identify key or critical elements of all aspects nor delineate the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed description that is presented later.
As shown in fig. 1 to 3, the wind turbine converter power module fault diagnosis method based on the operation data includes the following steps:
1) Identifying N fault modes according to historical operation data of the wind turbine generator;
2) Normalizing the data obtained in the step 1). In order to eliminate adverse effects caused by singular sample data, normalization processing is carried out on the collected operation data;
3) Decomposing the fault signal through the multi-component MRSVD to obtain a plurality of approximate signals and reconstructed signals, then extracting energy characteristics of the component signals and carrying out normalization processing to obtain relative values of the energy, so as to reflect the characteristic distribution of the current signals of the power converter in a frequency domain, and constructing a characteristic vector and carrying out normalization processing to obtain a normalized characteristic vector;
4) Taking each eigenvector as a node, calculating the similarity between different eigenvectors by using the Langmuir distance, selecting the node with the similarity larger than k to establish an adjacency relation, and constructing an adjacency matrix A;
5) Inputting the nodes, the adjacent matrix A and the degree matrix of the graph into a graph convolution neural network model, realizing the classification of the fault categories of the power module of the wind turbine converter, accurately identifying the fault categories and completing the fault diagnosis;
6) And evaluating the model by using accuracy (accuracycacy), precision (precision) and recall (call) as evaluation indexes.
And 4), abstracting each feature vector obtained by decomposing the MRSVD into nodes of the graph.
In the step 4), the concrete step of obtaining the adjacency relation between the construction nodes according to the Langmuir distance comprises the following steps:
setting different threshold values k by adopting an experimental error method, training to obtain an optimal k value, calculating the similarity between different characteristic vectors by utilizing the Langmuir distance, taking the nodes with the similarity larger than k as adjacent nodes, and determining the adjacent relation between the nodes;
the adjacency matrix a is represented as:
wherein if the node J is adjacent to the node M, J is 1,2,3.. . . . . M, then I xJxM Is the similarity value between node J and node M; if node J is not adjacent to node M, then I xJxM =0,m represents the number of eigenvectors.
Example 1
As shown in fig. 1 to 3, the wind turbine converter power module fault diagnosis method based on the operation data includes the following steps:
1) And identifying N fault modes by taking a certain type of wind turbine generator as an analysis object according to historical operation data of the wind turbine generator.
2) And (6) normalizing the data. In order to eliminate adverse effects caused by singular sample data, the collected operation data is normalized according to the following formula:
where x and z are data before and after processing, respectively, μ is the mean of the data set, and σ is the standard deviation of the data set.
3) Decomposing each level of signals through the MRSVD with a multi-division structure, extracting features, representing different features by using feature vectors, and extracting fault features by the following specific steps:
31 M-layer decomposition of the signal by means of a multi-component MRSVD, two approximation signals A for each layer in a decomposition model for noise reduction j,2 ,A j,1 Reconstructing to obtain a reconstructed signal A j Thus, M layers are decomposed to obtain M components (feature components), i.e., M reconstructed signals, as shown in fig. 2;
determination of the number of decomposition layers M:
different numbers of energy characteristics can be obtained by different decomposition layer numbers M, and the diagnostic result is influenced, so that the invention adopts an experimental error method to simulate under different decomposition layer numbers to obtain the optimal energy characteristic number and achieve the best diagnostic result;
32 Calculate an energy value for each component
33 Construct a feature vector X' = [ E ] 1 ,E 2 ,…,E M ]；
34 The feature vectors are normalized, so that the operation is more convenient, and the error can be reduced. Order to
The normalized eigenvector is X = [ E ] 1 /E,E 2 /E,…,E M /E]。
4) Taking each feature vector as a node, calculating the similarity between different feature vectors by using the Langmuir distance, selecting the nodes with the similarity larger than k to construct edges between the nodes, and finally constructing an adjacency matrix A by using the points and the edges, wherein the method specifically comprises the following steps:
41 Computing the Langmuir distance between nodes, the Langmuir distance D of two discrete random variables x and y L (x, y) is represented as:
42 Selecting nodes with similarity larger than k as adjacent nodes, and establishing an adjacent relation to obtain an adjacent matrix A; the k value is different in size, and the connection tightness degree of the whole graph structure is also different. When the value of k is large, the adjacency relation of nodes on the graph structure is sparse, and the characteristic information among the related nodes cannot be deeply fused; when the value of k is small, the adjacency relation is too dense, and the feature information between the variables with weak correlation is excessively fused. Therefore, different threshold values k are set by adopting an experimental error method, training is carried out, a reasonable k value is selected by calculating an evaluation index, and an adjacency matrix A is constructed as shown in the formula:
the adjacency matrix a is represented as:
wherein if the node J is adjacent to the node M, J is 1,2,3.. . . . . M, then I xJxM Is the similarity value between node J and node M; if node J is not adjacent to node M, then I xJxM =0,m represents the number of eigenvectors.
5) And inputting the nodes, the adjacency matrix A and the degree matrix of the graph into a Graph Convolution Neural (GCN) network model, realizing the classification of the fault categories of the power module of the wind turbine converter, accurately identifying the fault categories and completing the fault diagnosis.
The graph convolution neural network model extracts the implicit graph information by utilizing the structure information of the connection of the graph edges and the vertexes and the attribute information attached to the graph structure. The volume integral layer propagation formula in the volume neural network model can be expressed as:
wherein σ is an activation function;
is a matrix of degrees of the graph,
a is the adjacency matrix of the figure,
I N is an n-order identity matrix; x (l) Is a feature of the l-th layer; w is the weight matrix to be trained.
Before fault diagnosis is carried out, the number of GCN layers is required to be determined, preferably, 3 GCN layers are arranged, an output matrix of a first layer of GCN becomes a new node feature matrix of a second layer of GCN, an experimental error method is adopted in the training process, feature information fusion and dimension transformation are carried out through a plurality of layers of GCN networks, after each node feature is fused with adjacent node features, the result is used as the input of a full-connection matrix, a classification result is finally obtained, indexes such as classification accuracy and loss rate are calculated, and a reasonable number of layers are selected.
The graph convolution neural network focuses on information in a K-order neighbor centered on a node in the graph, and filter parameters on the information are shared for each node. A single-layer GCN can only extract information of first-order neighbors. To extract information of a wider range of nodes in the graph, the method can be realized by stacking a plurality of layers of graph convolution neural networks. As shown in the schematic diagram of GCN extraction space characteristics of fig. 3, a single-layer GCN can be used to extract information on the vertices (2), (3), (4), and (5) near the center point (1), and the convolutional layer sensing area becomes larger with the increase of convolutional layers through multi-layer stacking and obtains a more abstract information representation. Through two layers of GCN, center point (1) obtains information on adjacent vertexes (2), (3), (4), (5), (6), (7), (8), (9), and through three layers of GCN, center point (1) obtains information on adjacent vertexes (2), (3), (4), (5), (6), (7), (8), (9) and (c).
6) The model is evaluated by taking accuracy (accuracycacy), precision (precision) and recall (call) as evaluation indexes, and the calculation formula is as follows:
accuracy＝(TP+TN)/(P+N)
precision＝TP/(TP+FP)
recall＝TP/(TP+FN)＝TP/P
in the formula, TP is the number of instances (sample number) which are correctly divided into positive examples, i.e., are actually positive examples and are divided into positive examples by the classifier; FP is the number wrongly divided into positive cases, i.e. the number of cases that are actually negative but divided into positive cases by the classifier; FN is the number of instances that are wrongly divided into negative cases, i.e. the number of instances that are actually positive cases but are divided into negative cases by the classifier; TN is the number of instances that are correctly divided into negative cases, i.e., the number of instances that are actually negative and are divided into negative cases by the classifier.
The parts or structures of the invention which are not described in detail can be the same as those in the prior art or the existing products, and are not described in detail herein.
The above description is only an embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by the present specification, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (8)
1. The wind turbine generator converter power module fault diagnosis method based on the operation data is characterized by comprising the following steps of:
1) Identifying N fault modes according to historical operation data of the wind turbine generator;
2) Normalizing the data obtained in the step 1) to eliminate adverse effects caused by singular sample data;
3) Decomposing the fault signal, extracting features, and representing different features by using feature vectors;
4) Taking each eigenvector as a node, calculating the similarity between different eigenvectors by using the Langmuir distance, selecting the node with the similarity larger than k to establish an adjacency relation, and constructing an adjacency matrix A;
5) Inputting the nodes, the adjacent matrix A and the degree matrix of the graph into a graph convolution neural network model to realize fault diagnosis of a power module of the wind turbine converter;
6) And evaluating the graph convolution neural network model by using the accuracy, precision and recall rate as evaluation indexes.
2. The wind turbine converter power module fault diagnosis method based on the operation data according to claim 1, characterized in that in step 2), the data obtained in step 1) is normalized according to the following formula:
where x and z are data before and after processing, respectively, μ is the mean of the data set, and σ is the standard deviation of the data set.
3. The wind turbine converter power module fault diagnosis method based on the operation data as claimed in claim 1, wherein in step 3), the fault signal is decomposed to extract features, and the specific steps of different features expressed by feature vectors include:
31 By a multi-division-structured MRSVD noise reduction method, a signal is decomposed into M layers, and two approximate signals A of each layer in a decomposition model used for noise reduction are analyzed j,2 ,A j,1 Reconstructing to obtain a reconstructed signal A j Thus, M components can be obtained by decomposing the M layers, and M reconstructed signals can be obtained;
32 Calculate an energy value for each component
33 Construct a feature vector X' = [ E ] 1 ,E 2 ,…,E M ]；
34 Normalizing the feature vectors, so that the operation is more convenient and the error can be reduced; order to
The normalized feature vector is X = [ E = 1 /E,E 2 /E,…,E M /E]。
4. The wind turbine generator converter power module fault diagnosis method based on the operation data as claimed in claim 1, wherein in the step 4), each eigenvector is used as a node, the similarity between different eigenvectors is calculated by using the Langmuir distance, the node with the similarity larger than k is selected to establish the adjacency relation, and the specific step of constructing the adjacency matrix A comprises:
41 Calculates the Langmuir distance between nodes, the Langmuir distance D of two discrete random variables x and y L (x, y) is represented as:
42 Setting different threshold values k by adopting an experimental error method, and training to obtain an optimal k value; selecting nodes with similarity larger than k as adjacent nodes, constructing edges between the nodes, and finally constructing an adjacent matrix A by using the points and the edges, wherein the adjacent matrix A is expressed as:
wherein if the node J is adjacent to the node M, J is 1,2,3.. . . . . M, then I xJxM Is the similarity value between node J and node M; if node J is not adjacent to node M, then I xJxM ＝0。
5. The wind turbine generator converter power module fault diagnosis method based on the operation data according to claim 1, characterized in that in step 5), the graph convolution neural network model extracts implicit graph information by using structural information of connection between graph edges and vertices and attribute information attached to a graph structure;
the graph integration layer propagation formula in the graph integration neural network model is expressed as:
in the formula, sigma is an activation function;
is a matrix of degrees of the graph,
a is the adjacency matrix of the figure,
I N is an n-order identity matrix; x (l) Is a characteristic of the l-th layer; w is the weight matrix to be trained.
6. The wind turbine generator converter power module fault diagnosis method based on the operation data according to claim 1, wherein in the step 5), before fault diagnosis is performed, the number of GCN layers of the graph convolution neural network model is determined, an experimental error method is adopted in a training process, an output matrix of a previous layer of GCN becomes a new node feature matrix of a next layer of GCN, and feature information fusion and dimension transformation are performed through the multiple GCN layers, so that each node feature is fused with adjacent node features.
7. The wind turbine converter power module fault diagnosis method based on operating data of claim 6, wherein the number of GCN layers is three.
8. The wind turbine converter power module fault diagnosis method based on the operation data as claimed in claim 6, wherein in step 6), the calculation formulas of the accuracy, precision and recall rate are respectively:
accuracy＝(TP+TN)/(P+N)
precision＝TP/(TP+FP)
recall＝TP/(TP+FN)＝TP/P
in the formula, accuracy is the correct rate; precision is precision; recall is the recall rate; TP is the number correctly divided into positive examples, i.e. the number of examples (number of samples) that are actually positive examples and are divided into positive examples by the classifier; FP is the number wrongly divided into positive cases, i.e. the number of cases that are actually negative but divided into positive cases by the classifier; FN is the number of instances that are wrongly divided into negative cases, i.e. the number of instances that are actually positive cases but are divided into negative cases by the classifier; TN is the number of instances that are correctly divided into negative cases, i.e., the number of instances that are actually negative and are divided into negative cases by the classifier.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211029876.XA CN115456013A (en) | 2022-08-25 | 2022-08-25 | Wind turbine generator converter power module fault diagnosis method based on operation data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211029876.XA CN115456013A (en) | 2022-08-25 | 2022-08-25 | Wind turbine generator converter power module fault diagnosis method based on operation data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115456013A true CN115456013A (en) | 2022-12-09 |
Family
ID=84299908
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211029876.XA Pending CN115456013A (en) | 2022-08-25 | 2022-08-25 | Wind turbine generator converter power module fault diagnosis method based on operation data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115456013A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117235465A (en) * | 2023-11-15 | 2023-12-15 | 国网江西省电力有限公司电力科学研究院 | Transformer fault type diagnosis method based on graph neural network wave recording analysis |
-
2022
- 2022-08-25 CN CN202211029876.XA patent/CN115456013A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117235465A (en) * | 2023-11-15 | 2023-12-15 | 国网江西省电力有限公司电力科学研究院 | Transformer fault type diagnosis method based on graph neural network wave recording analysis |
| CN117235465B (en) * | 2023-11-15 | 2024-03-12 | 国网江西省电力有限公司电力科学研究院 | Transformer fault type diagnosis method based on graph neural network wave recording analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109635928B (en) | Voltage sag reason identification method based on deep learning model fusion | |
| CN113962259B (en) | Multi-mode double-layer fault diagnosis method for fuel cell system | |
| CN110579709B (en) | Fault diagnosis method for proton exchange membrane fuel cell for tramcar | |
| CN105631596A (en) | Equipment fault diagnosis method based on multidimensional segmentation fitting | |
| CN114676742A (en) | Power grid abnormal electricity utilization detection method based on attention mechanism and residual error network | |
| CN112596016A (en) | Transformer fault diagnosis method based on integration of multiple one-dimensional convolutional neural networks | |
| CN110866366A (en) | XGboost algorithm-based island detection method for photovoltaic microgrid containing PHEV | |
| CN111398811A (en) | PMSM demagnetization fault diagnosis method based on terminal current cost sensitive learning | |
| CN114548199A (en) | Multi-sensor data fusion method based on deep migration network | |
| CN116089857A (en) | Transformer fault identification method based on CEEMDAN-DBN | |
| CN115456013A (en) | Wind turbine generator converter power module fault diagnosis method based on operation data | |
| Afrasiabi et al. | Two-stage deep learning-based wind turbine condition monitoring using SCADA data | |
| CN115757103A (en) | Neural network test case generation method based on tree structure | |
| CN116842459B (en) | Electric energy metering fault diagnosis method and diagnosis terminal based on small sample learning | |
| CN113642255A (en) | Photovoltaic power generation power prediction method based on multi-scale convolution cyclic neural network | |
| CN115564075B (en) | Main and auxiliary integrated fault collaborative diagnosis method and system for urban power grid | |
| CN116400244B (en) | Abnormality detection method and device for energy storage battery | |
| CN112595918A (en) | Low-voltage meter reading fault detection method and device | |
| CN116167277A (en) | Elman network-based power equipment fault diagnosis method, system and medium | |
| CN114415056A (en) | Distributed power supply island fault detection method | |
| AGUIAR et al. | Features extraction and selection with the scattering transform for electrical load classification | |
| CN117390508B (en) | Hydroelectric generating set signal state identification method based on time-shifting multi-scale cosine similarity entropy | |
| CN114136565B (en) | Fault diagnosis system and method for multi-vibration-source system | |
| CN112666483B (en) | Lithium battery residual life prediction method for improving ARMA (autoregressive moving average) | |
| CN117371623B (en) | Electric energy meter running state early warning method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |