CN114355110A - Fault current mode identification method based on convolutional neural network - Google Patents
Fault current mode identification method based on convolutional neural network Download PDFInfo
- Publication number
- CN114355110A CN114355110A CN202210268900.9A CN202210268900A CN114355110A CN 114355110 A CN114355110 A CN 114355110A CN 202210268900 A CN202210268900 A CN 202210268900A CN 114355110 A CN114355110 A CN 114355110A
- Authority
- CN
- China
- Prior art keywords
- neural network
- convolutional neural
- power frequency
- method based
- current
- 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
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
- G01R31/086—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution networks, i.e. with interconnected conductors
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R19/00—Arrangements for measuring currents or voltages or for indicating presence or sign thereof
- G01R19/0092—Arrangements for measuring currents or voltages or for indicating presence or sign thereof measuring current only
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention relates to a fault current mode identification method based on a convolutional neural network, S1, obtaining a data sample; s2, preprocessing power frequency current data in the data sample, and generating a two-dimensional oscillogram by taking power frequency current acquisition time and current amplitude as horizontal and vertical coordinates; s4, taking the two-dimensional oscillogram picture of the known label as input, taking the corresponding known label as output, and importing the input into a convolutional neural network for training; s5, preprocessing power frequency current data acquired by an actual circuit, and generating a two-dimensional oscillogram by taking power frequency current acquisition time and current amplitude as horizontal and vertical coordinates; s6, importing the two-dimensional oscillogram picture of the unknown label in the S5 as input into the convolutional neural network obtained in the S4; and S7, outputting a pattern recognition result. The power frequency current mode recognition rate is improved, the training and recognition speed is improved, and the calculated amount is reduced.
Description
Technical Field
The invention relates to the field of power distribution networks, in particular to a fault current mode identification method based on a convolutional neural network.
Background
In order to ensure that the fault position of the power transmission line is found quickly and the stable operation of the line is maintained, a fault precise positioning system based on fault traveling waves is applied to the power grid in a large scale at present, the fault position can be precisely positioned through the time difference of the traveling waves reaching two monitoring points and the distance between the two devices, and help is provided for operation and maintenance personnel to find and process the fault in time. In an actual power transmission line, load shedding exists, other lines are in series interference, and the line power frequency current changes due to the situations of switching-off, switching-on and the like, so that generally, the fault power frequency current needs to be manually judged, but the fault can occur at any time in one day, and a large amount of manpower is consumed for manual judgment.
At present, a Support Vector Machine (SVM) mode and a BP neural network mode are commonly used for pattern recognition of the collected data set. The SVM is a two-classification model, the basic model of the SVM is a linear classifier with the maximum interval defined on a feature space, namely a separation hyperplane which can correctly divide a training data set and has the maximum geometric interval is solved, for part of nonlinear classification problems, the SVM can be converted into a linear classification problem in a certain dimensional feature space through nonlinear transformation, a linear support vector machine is learned in a high-dimensional feature space, but the nonlinear transformation is required to be carried out by utilizing a kernel function in the nonlinear SVM. In actual life, most of classification problems of linear SVM can not be solved, a kernel function needs to be selected by using a nonlinear SVM, and the problem that the linearity is inseparable in an original space is solved by mapping data to a high-dimensional space. In the process, the proper kernel function can effectively realize the classification problem which cannot be solved by a low-dimensional space. However, from the perspective of practical application effect, the fault power frequency current mode identification effect of the SVM is general, and a large number of cases of wrong classification may occur, because the classification still cannot be realized when a part of arrays are mapped into a high-dimensional space, and it is difficult to select a proper kernel function as the mapping, so the practical application effect is not good.
The principle of the BP neural network as a mode of machine learning is a multi-layer feedforward neural network trained according to an error back propagation algorithm, and the BP neural network is one of the most widely applied neural network models. However, the method still has more problems in the implementation process, firstly, the characteristic quantity is difficult to extract, if the characteristic quantity is not extracted, the collected array is directly calculated, 10000 points are collected when the 10KHz sampling frequency is collected for 1s of fault power frequency, so that the weight is too much, the calculated quantity is huge, the training and classification effects are influenced, the quantity of neurons can be effectively reduced by extracting the characteristic quantity, the training speed is accelerated, but the characteristic quantity is very difficult to extract, the huge array is often difficult to extract into simple characteristic quantity, and the characteristics of the array are inevitably lost in the extraction process; secondly, the training requirement is large in sample amount, a large amount of training samples are needed to train the neural network due to excessive weights, but a large amount of fault power frequency current data is usually difficult to obtain, so that the neural network training is difficult, and meanwhile, the BP neural network algorithm has a plurality of defects: the convergence speed of the learning training process of the BP algorithm is low, the local convergence condition is easy to cause, the learning rate of the BP network is unstable, and the BP neural network is difficult to adapt to the pattern classification of large-scale complex arrays.
An SVM classifier and a BP neural network can not effectively and quickly realize mode classification for dealing with the fault current waveform which is a waveform containing a large amount of data in a single array, and various problems can be caused when the classifier is established and classified or used, so that the classification effect is low, and a method for quickly and effectively realizing classification for dealing with the fault current waveform which is a large amount of data is urgently needed at present.
Disclosure of Invention
The technical problem to be solved by the present invention is to provide a fault current pattern recognition method based on a convolutional neural network, so as to overcome the above-mentioned deficiencies in the prior art.
The technical scheme for solving the technical problems is as follows: a fault current mode identification method based on a convolutional neural network comprises the following steps:
s1, acquiring a data sample;
s2, preprocessing power frequency current data in the data sample, and generating a two-dimensional oscillogram by taking power frequency current acquisition time and current amplitude as horizontal and vertical coordinates;
s3, marking all the two-dimensional oscillogram pictures, and determining the label of each picture;
s4, taking the two-dimensional oscillogram picture of the known label as input, taking the corresponding known label as output, and importing the input into a convolutional neural network for training;
s5, preprocessing power frequency current data acquired by an actual circuit, and generating a two-dimensional oscillogram by taking power frequency current acquisition time and current amplitude as horizontal and vertical coordinates;
s6, importing the two-dimensional oscillogram picture of the unknown label in the S5 as input into the convolutional neural network obtained in the S4;
and S7, outputting a pattern recognition result.
On the basis of the technical scheme, the invention can be further improved as follows.
Further, the data samples are: real data collected in the actual line and data of power frequency current obtained by building the actual line by simulation software under various transient conditions.
Further, the simulation software is ATP-EMPT transient simulation software.
Further, the pretreatment in S2 specifically includes:
s21, taking the power frequency current acquisition time and the current amplitude as horizontal and vertical coordinates, and connecting all current amplitude points by a smooth curve to form a two-dimensional oscillogram;
and S22, normalizing all the two-dimensional waveform diagrams to unify the current amplitude and sampling frequency and unify the resolution and size of the drawn two-dimensional waveform diagram pictures to generate two-dimensional waveform diagram pictures with unified specifications.
Further, the method for normalizing the current amplitude comprises the following steps:
assume that the current amplitude at the k point isi kThen, the normalized amplitude of the k point is:
wherein the content of the first and second substances,i maxandi minthe maximum value and the minimum value of the current amplitude are respectively, and k is a natural number.
Further, the method for normalizing the sampling frequency comprises the following steps:
the method comprises the steps that two groups of fault power frequency current sensors with sampling frequencies of MkHz and NkHz respectively are assumed to acquire data of 1 s;
the normalized sampling frequency is nkHz, M is more than 1, N is more than 1, and M is not equal to N;
the MkHz and NkHz down-sampling frequency is processed, namely, every M/N points or N/N points are averaged to be processed as amplitude points.
Further, the training convolutional neural network specifically comprises:
determining a convolutional neural network structure and parameters;
determining the number of labels needing to be divided and giving numbers;
and taking the two-dimensional oscillogram picture of the known label as an input layer to enter a convolutional neural network, and then classifying through an internal convolutional layer, a pooling layer, a full-link layer and softmax to finally output a classification result.
And further, testing whether the resolution error rate of the convolutional neural network meets an error threshold value or not by using the test set, if not, adjusting the structure and parameters of the convolutional neural network until the resolution error rate of the test set is lower than the threshold value, and if so, determining that the network passes the test.
Still further, the error threshold is 5%.
Further, the mode of preprocessing the power frequency current data collected by the actual line in S5 is the same as the mode of preprocessing the power frequency current data in the data sample in S2.
The invention has the beneficial effects that:
the method realizes the identification of the power frequency fault current by a convolutional neural network mode, and compared with the traditional mode identification mode, the method presents the power frequency current acquisition time and the current amplitude in a two-dimensional oscillogram (plane graph) mode, but does not traditionally calculate by taking the whole long array as input, connects all amplitude points by smooth curves, takes the obtained two-dimensional oscillogram image as the input of the neural network, and the convolutional neural network can learn the image characteristics of the two-dimensional oscillogram and capture various characteristics in data from the shape angle;
in the initial training process, in order to ensure enough training samples, simulation data and real data collected in an actual line are mixed to be used as training samples, and fault or brake-off current waveforms under various ideal conditions can be obtained through simulation;
in addition, the convolutional neural network reduces the number of parameters to be trained by the neural network through weight sharing (the weight sharing is the characteristics of the convolutional neural network), greatly reduces the calculated amount and accelerates the operation speed.
Drawings
FIG. 1 is a flow chart of a fault current pattern recognition method based on a convolutional neural network according to the present invention;
fig. 2 is a diagram illustrating the correspondence of the labeling result.
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.
Example 1
As shown in fig. 1 and fig. 2, a fault current pattern recognition method based on a convolutional neural network includes:
s1, acquiring a data sample;
s2, preprocessing power frequency current data in the data sample, and generating a two-dimensional oscillogram by taking power frequency current acquisition time and current amplitude as horizontal and vertical coordinates;
s3, marking all the two-dimensional oscillogram pictures, and determining the label of each picture;
the label contains mode information including fault current, line load change, line brake separating, line normal operation and the like;
s4, taking the two-dimensional oscillogram picture of the known label as input, taking the corresponding known label as output, and importing the input into a convolutional neural network for training;
s5, preprocessing power frequency current data acquired by an actual circuit, and generating a two-dimensional oscillogram by taking power frequency current acquisition time and current amplitude as horizontal and vertical coordinates;
s6, importing the two-dimensional oscillogram picture of the unknown label in the S5 as input into the convolutional neural network obtained in the S4;
and S7, outputting a pattern recognition result.
Example 2
This embodiment is a further improvement on embodiment 1, and specifically includes the following steps:
acquisition of data samples in S1:
because the conditions of actual line fault power frequency current, load rise, line switching-off and the like cause that the current data which can be obtained are very limited and are difficult to meet the training samples under all conditions, firstly, the actual line can be set up by using simulation software and the data of the power frequency current under various transient conditions can be obtained, the simulation software can quickly obtain a large amount of data under ideal conditions, and the data are used as the training samples together with the real data collected in part of the actual line;
of course, if the real data is enough, the data obtained by the simulation software may not be used, and the specific situation may be determined according to the actual situation.
Example 3
This embodiment is a further improvement on the basis of embodiment 2, and specifically includes the following steps:
the simulation software may be ATP-EMPT transient simulation software, although other software is not excluded.
Example 4
This embodiment is a further improvement on the embodiment 1, 2 or 3, and specifically includes the following steps:
because the input of the convolutional neural network is a picture, after a data sample is obtained, power frequency current data in the data sample is preprocessed, power frequency current acquisition time and current amplitude are used as horizontal and vertical coordinates, and then a smooth curve is used for connecting all amplitude points, so that a two-dimensional waveform diagram is generated, the colors of all the amplitude points and connecting lines in the diagram are different from those of other areas, for example, all the amplitude points and connecting lines in the diagram are black, and other areas are white, so that the difference between the amplitude points and the connecting lines is represented by 0 and 1 in subsequent (referring to convolutional neural network training and identification) processing, and the calculated amount is reduced;
meanwhile, considering that the sampling frequencies of different data are different, all two-dimensional oscillograms need to be normalized to unify the current amplitude and the sampling frequency, unify the resolution and the size of the drawn two-dimensional oscillogram pictures, and generate the two-dimensional oscillogram pictures with unified specifications.
Example 5
This embodiment is a further improvement on the basis of embodiment 4, and specifically includes the following steps:
the method for normalizing the current amplitude comprises the following steps:
assume that the current amplitude at the k point isi kThen, the normalized amplitude of the k point is:
wherein the content of the first and second substances,i maxandi minthe maximum value and the minimum value of the current amplitude are respectively, and k is a natural number, such as 0-n.
In addition, the method for normalizing the sampling frequency comprises the following steps:
the method comprises the steps that two groups of fault power frequency current sensors with sampling frequencies of MkHz and NkHz respectively are assumed to acquire data of 1 s;
the normalized sampling frequency is nkHz, M is more than 1, N is more than 1, and M is not equal to N;
the MkHz and NkHz down-sampling frequency is processed, that is, the average value is taken for every M/N points or N/N points to be the processed amplitude value point, in this embodiment, N can be 1, M can be 10, and N can be 5, which is only an exemplary description, but other values can be taken, and it is possible to ensure that M/N and N/N are both integers as much as possible.
Example 6
This embodiment is a further improvement on embodiment 1, and specifically includes the following steps:
the training convolutional neural network specifically comprises the following steps:
determining a convolutional neural network structure and parameters;
determining the number of labels needing to be divided and giving numbers;
and taking the two-dimensional oscillogram picture of the known label as an input layer to enter a convolutional neural network, and then classifying through an internal convolutional layer, a pooling layer, a full-link layer and softmax to finally output a classification result.
After the convolutional neural network is trained, a test set can be used for testing whether the resolution error rate of the convolutional neural network meets an error threshold value, if not, the structure and parameters of the convolutional neural network are adjusted until the resolution error rate of the test set is lower than the threshold value, and if so, the network is considered to pass the test and has the fault power frequency current mode identification capability.
The specific value of the error threshold may be determined according to actual requirements, for example, 5%.
The structure and parameters of the convolutional neural network can be determined by: the size of the two-dimensional oscillogram picture determines the structure of the convolutional neural network and the size of the convolutional kernel.
Example 7
The embodiment is a further improvement on any embodiment of embodiments 1 to 6, and specifically includes the following steps:
the way of preprocessing the power frequency current data collected by the actual line in S5 is the same as the way of preprocessing the power frequency current data in the data sample in S2, so the preprocessing way in S5 is not described in detail herein.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.
Claims (10)
1. A fault current mode identification method based on a convolutional neural network is characterized by comprising the following steps:
s1, acquiring a data sample;
s2, preprocessing power frequency current data in the data sample, and generating a two-dimensional oscillogram by taking power frequency current acquisition time and current amplitude as horizontal and vertical coordinates;
s3, marking all the two-dimensional oscillogram pictures, and determining the label of each picture;
s4, taking the two-dimensional oscillogram picture of the known label as input, taking the corresponding known label as output, and importing the input into a convolutional neural network for training;
s5, preprocessing power frequency current data acquired by an actual circuit, and generating a two-dimensional oscillogram by taking power frequency current acquisition time and current amplitude as horizontal and vertical coordinates;
s6, importing the two-dimensional oscillogram picture of the unknown label in the S5 as input into the convolutional neural network obtained in the S4;
and S7, outputting a pattern recognition result.
2. The fault current pattern recognition method based on the convolutional neural network as claimed in claim 1, wherein:
the data samples were: real data collected in the actual line and data of power frequency current obtained by building the actual line by simulation software under various transient conditions.
3. The fault current pattern recognition method based on the convolutional neural network as claimed in claim 2, wherein: the simulation software is ATP-EMPT transient simulation software.
4. The fault current pattern recognition method based on the convolutional neural network as claimed in claim 1, wherein:
the pretreatment in S2 specifically includes:
s21, taking the power frequency current acquisition time and the current amplitude as horizontal and vertical coordinates, and connecting all current amplitude points by a smooth curve to form a two-dimensional oscillogram;
and S22, normalizing all the two-dimensional waveform diagrams to unify the current amplitude and sampling frequency and unify the resolution and size of the drawn two-dimensional waveform diagram pictures to generate two-dimensional waveform diagram pictures with unified specifications.
5. The fault current pattern recognition method based on the convolutional neural network as claimed in claim 4, wherein:
the method for normalizing the current amplitude comprises the following steps:
assume that the current amplitude at the k point isi kThen, the normalized amplitude of the k point is:
wherein the content of the first and second substances,i maxandi minthe maximum value and the minimum value of the current amplitude are respectively, and k is a natural number.
6. The fault current pattern recognition method based on the convolutional neural network as claimed in claim 4, wherein:
the method for normalizing the sampling frequency comprises the following steps:
the method comprises the steps that two groups of fault power frequency current sensors with sampling frequencies of MkHz and NkHz respectively are assumed to acquire data of 1 s;
the normalized sampling frequency is nkHz, M is more than 1, N is more than 1, and M is not equal to N;
the MkHz and NkHz down-sampling frequency is processed, namely, every M/N points or N/N points are averaged to be processed as amplitude points.
7. The fault current pattern recognition method based on the convolutional neural network as claimed in claim 1, wherein:
the training convolutional neural network specifically comprises the following steps:
determining a convolutional neural network structure and parameters;
determining the number of labels needing to be divided and giving numbers;
and taking the two-dimensional oscillogram picture of the known label as an input layer to enter a convolutional neural network, and then classifying through an internal convolutional layer, a pooling layer, a full-link layer and softmax to finally output a classification result.
8. The fault current pattern recognition method based on the convolutional neural network as claimed in claim 7, wherein:
and testing whether the resolution error rate of the convolutional neural network meets an error threshold value or not by using the test set, if not, adjusting the structure and parameters of the convolutional neural network until the resolution error rate of the test set is lower than the threshold value, and if so, determining that the network passes the test.
9. The fault current pattern recognition method based on the convolutional neural network as claimed in claim 8, wherein: the error threshold is 5%.
10. The fault current pattern recognition method based on the convolutional neural network as claimed in claim 1, wherein:
the mode of preprocessing the power frequency current data collected by the actual line in the S5 is the same as the mode of preprocessing the power frequency current data in the data sample in the S2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210268900.9A CN114355110A (en) | 2022-03-18 | 2022-03-18 | Fault current mode identification method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210268900.9A CN114355110A (en) | 2022-03-18 | 2022-03-18 | Fault current mode identification method based on convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114355110A true CN114355110A (en) | 2022-04-15 |
Family
ID=81095205
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210268900.9A Pending CN114355110A (en) | 2022-03-18 | 2022-03-18 | Fault current mode identification method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114355110A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2024052183A1 (en) * | 2022-09-06 | 2024-03-14 | Mercedes-Benz Group AG | Circuit arrangement and method for detecting a health status of electrical loads |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102445603A (en) * | 2010-10-04 | 2012-05-09 | 台湾积体电路制造股份有限公司 | Method for measuring capacitances of capacitors |
| CN103499769A (en) * | 2013-09-23 | 2014-01-08 | 武汉大学 | Self-adaptive line selection method for single-phase earth fault of resonant earthed system |
| CN103837795A (en) * | 2014-02-18 | 2014-06-04 | 国网山东省电力公司 | Dispatching end grid fault diagnosis method based on wide-area fault recording information |
| CN107656179A (en) * | 2017-09-21 | 2018-02-02 | 北京国华世纪电子科技有限公司 | A kind of series fault arc detecting system and method |
| CN108732465A (en) * | 2018-05-30 | 2018-11-02 | 广东电网有限责任公司 | A kind of electrical power distribution network fault location method based on wavelet transformation and CNN |
-
2022
- 2022-03-18 CN CN202210268900.9A patent/CN114355110A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102445603A (en) * | 2010-10-04 | 2012-05-09 | 台湾积体电路制造股份有限公司 | Method for measuring capacitances of capacitors |
| CN103499769A (en) * | 2013-09-23 | 2014-01-08 | 武汉大学 | Self-adaptive line selection method for single-phase earth fault of resonant earthed system |
| CN103837795A (en) * | 2014-02-18 | 2014-06-04 | 国网山东省电力公司 | Dispatching end grid fault diagnosis method based on wide-area fault recording information |
| CN107656179A (en) * | 2017-09-21 | 2018-02-02 | 北京国华世纪电子科技有限公司 | A kind of series fault arc detecting system and method |
| CN108732465A (en) * | 2018-05-30 | 2018-11-02 | 广东电网有限责任公司 | A kind of electrical power distribution network fault location method based on wavelet transformation and CNN |
Non-Patent Citations (1)
| Title |
|---|
| 林翔宇 等: "基于多源信号输入卷积神经网络的高压开关故障诊断方法", 《电气时代》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2024052183A1 (en) * | 2022-09-06 | 2024-03-14 | Mercedes-Benz Group AG | Circuit arrangement and method for detecting a health status of electrical loads |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107784661B (en) | Transformer substation equipment infrared image classification and identification method based on region growing method | |
| CN107507194B (en) | Insulator string fault detection method based on infrared image temperature distribution rule and BP neural network | |
| CN111864896B (en) | Power load monitoring method and system | |
| CN107038416B (en) | Pedestrian detection method based on binary image improved HOG characteristics | |
| CN112233073A (en) | Real-time detection method for infrared thermal imaging abnormity of power transformation equipment | |
| CN113449594B (en) | Multilayer network combined remote sensing image ground semantic segmentation and area calculation method | |
| CN113205063A (en) | Visual identification and positioning method for defects of power transmission conductor | |
| CN112287784B (en) | Radar signal classification method based on deep convolutional neural network and feature fusion | |
| CN110598693A (en) | Ship plate identification method based on fast-RCNN | |
| CN111008641B (en) | Power transmission line tower external force damage detection method based on convolutional neural network | |
| CN105894041A (en) | Method of extracting substation information in power distribution system based on hyperspectral remote sensing images | |
| CN107679453A (en) | Weather radar electromagnetic interference echo recognition methods based on SVMs | |
| CN110175571A (en) | The intellectual monitoring of substation equipment state and recognition methods | |
| CN109886937A (en) | Defects of insulator detection method based on super-pixel segmentation image recognition | |
| CN115908142B (en) | Visual identification-based damage inspection method for tiny contact net parts | |
| CN114355110A (en) | Fault current mode identification method based on convolutional neural network | |
| CN114241425B (en) | Training method and device of garbage detection model, storage medium and equipment | |
| CN112131956A (en) | Voltage sag source classification method based on difference hash algorithm | |
| CN111507249A (en) | Transformer substation nest identification method based on target detection | |
| CN110703006A (en) | Three-phase power quality disturbance detection method based on convolutional neural network | |
| CN112801332B (en) | Short-term wind speed prediction method based on gray level co-occurrence matrix | |
| CN113221614A (en) | Power grid insulator damage image identification method based on hybrid neural network | |
| CN115601747A (en) | Method and system for calculating confluency of adherent cells | |
| Dong et al. | An Automatic Detection Method of Bird's Nest on Electric Tower Based on Attention Full Convolutional Neural Networks | |
| CN114359948A (en) | Power grid wiring diagram primitive identification method based on overlapping sliding window mechanism and YOLOV4 |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220415 |
|
| RJ01 | Rejection of invention patent application after publication |