CN111898446A - Single-phase earth fault studying and judging method based on multi-algorithm normalization analysis - Google Patents

Single-phase earth fault studying and judging method based on multi-algorithm normalization analysis Download PDF

Info

Publication number
CN111898446A
CN111898446A CN202010612126.XA CN202010612126A CN111898446A CN 111898446 A CN111898446 A CN 111898446A CN 202010612126 A CN202010612126 A CN 202010612126A CN 111898446 A CN111898446 A CN 111898446A
Authority
CN
China
Prior art keywords
model
fault
layer
dbn
data
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.)
Withdrawn
Application number
CN202010612126.XA
Other languages
Chinese (zh)
Inventor
时金媛
张蓓蓓
苏标龙
苏光
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nari Technology Co Ltd
NARI Nanjing Control System Co Ltd
Original Assignee
Nari Technology Co Ltd
NARI Nanjing Control System Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nari Technology Co Ltd, NARI Nanjing Control System Co Ltd filed Critical Nari Technology Co Ltd
Priority to CN202010612126.XA priority Critical patent/CN111898446A/en
Publication of CN111898446A publication Critical patent/CN111898446A/en
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • 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
    • 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
    • G01R31/088—Aspects of digital computing
    • 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
    • G01R31/50—Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
    • G01R31/52—Testing for short-circuits, leakage current or ground faults
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Signal Processing (AREA)
  • Supply And Distribution Of Alternating Current (AREA)

Abstract

The invention discloses a single-phase earth fault studying and judging method based on multi-algorithm normalization analysis, which utilizes a pre-trained deep learning model to carry out fault identification on working condition data of a selected time period under a selected bus so as to determine whether a fault occurs; and performing multi-working-condition type identification on the working condition data which is determined to have faults by utilizing a pre-trained decision tree. The method helps the dispatching personnel to quickly position the single-phase grounding of the power distribution network, forms an operable result, and provides a basis for improving the fault processing efficiency of the dispatching personnel and guaranteeing the operation stability of the power distribution network.

Description

Single-phase earth fault studying and judging method based on multi-algorithm normalization analysis
Technical Field
The invention relates to the technical field of power distribution networks, in particular to a single-phase earth fault studying and judging method based on multi-algorithm normalization analysis.
Background
The medium voltage distribution network in China widely adopts a neutral point non-effective grounding operation mode, and mainly comprises a neutral point non-grounding mode and an arc suppression coil grounding mode. Compared with a high-voltage transmission network, the probability of faults of a medium-voltage distribution network is much higher, particularly single-phase earth faults occur frequently, and statistics show that the single-phase earth faults account for about 80% of the total faults of the distribution network. When a single-phase earth fault occurs in a power distribution network with a neutral point non-effective earth operation mode, a short circuit loop cannot be formed, and only small earth fault current is caused by distributed capacitance of a system, so that the power distribution network is also called as a small-current earth fault. Because the line voltage between the three phases of the system is basically kept unchanged, the load power supply is not influenced, and theoretically, the system can operate for a period of time with a fault. However, due to the non-fault phase-to-ground voltage rise during grounding, especially intermittent arc grounding faults, arc overvoltage can be generated, and system insulation and equipment safety are seriously damaged. Meanwhile, an overvoltage may cause a ground fault to be converted into an interphase short-circuit fault, so that a line trips, and a user is powered off. Therefore, an effective single-phase earth fault determination method becomes an important means for solving the existing problems.
Disclosure of Invention
The invention aims to research and determine the problem of fault types when a power distribution network is subjected to single-phase grounding, and provides a single-phase grounding fault studying and judging method based on multi-algorithm normalization analysis.
In order to achieve the above purpose, the invention adopts the following technical scheme: a single-phase earth fault studying and judging method based on multi-algorithm normalization analysis comprises the following steps:
carrying out fault identification on the working condition data of a selected time period under a selected bus by using a pre-trained deep learning model to determine whether a fault occurs; and performing multi-working-condition type identification on the working condition data which is determined to have faults by utilizing a pre-trained decision tree.
Further, the deep learning model comprises a CNN model and a DBN model, and the training method of the deep learning model is as follows:
pre-training the CNN model by using the training set data to obtain parameters of a convolutional layer and a full link layer in the CNN model;
fixing parameters of a CNN model convolutional layer, and performing unsupervised pre-training on the DBN model by using output data of a training set after passing through the CNN model convolutional layer to obtain parameters of each RBM layer in the DBN model;
and taking the full connection layer of the CNN model as an RBM layer of the DBN model, migrating the parameter of the convolutional layer of the CNN model and the parameter of the RBM layer in the DBN model, which are acquired after pre-training, and performing supervised fine tuning on the whole network by using training set data.
Still further, the method further comprises a step of verifying the result of the deep learning model, and specifically comprises the following steps:
introducing a fault identification rate as an identification evaluation index, wherein the calculation method comprises the following steps:
Figure RE-GDA0002667055540000021
in the formula: t is an input known fault test sample set; c is the number of correctly identified samples in the set;
an error rate or leakage rate evaluation index is introduced, and the calculation mode is as follows:
Figure RE-GDA0002667055540000022
in the formula: n is an input multiple fault test sample set; m is the number of samples in the set which are wrongly or overlooked.
Further, the method for acquiring the training set data is as follows:
the selected input signal is the working condition data of the selected time period under the selected bus, the input data is picturized, and the waveform sequence is matrixed.
Further, the deep learning model comprises a CNN model and a DBN model, the DBN model comprises 3 RBMs and a Softmax layer which are stacked in series, a full connection layer of the CNN model is used as the RBM layer of the DBN model, a Dropout mechanism is added in each RBM layer, and the Softmax layer maps hidden layer output of the last RBM into a (0,1) interval to obtain probability values of all categories, so that multi-label classification is performed.
Further, the overall cost c of the DBN model adopts a cross entropy function, and the calculation process of c is
Figure RE-GDA0002667055540000031
Wherein y isiRepresenting the true classification result, aiThe resulting probability value calculated for the previous step.
Further, each node in the decision tree is a classifier.
Further, single-phase earth fault location is carried out by adopting a dyeing principle after multi-working-condition type identification is completed.
The present invention also provides a computer readable storage medium, which stores a computer program, and the computer program when executed by a processor implements the steps of the single-phase ground fault studying and judging method based on multi-algorithm normalization analysis provided by the above technical solution.
The beneficial technical effects are as follows:
the invention discloses a single-phase earth fault studying and judging method based on multi-algorithm normalization analysis. The method helps the dispatching personnel to quickly position the single-phase grounding of the power distribution network so as to form an operable result, improve the fault processing efficiency of the dispatching personnel and guarantee the operation stability of the power distribution network.
Drawings
FIG. 1 is a waveform sequence imaging process according to an embodiment of the present invention;
FIG. 2 illustrates waveforms corresponding to types of faults in accordance with an embodiment of the present invention;
FIG. 3 is a schematic diagram of fault location in an embodiment of the present invention;
FIG. 4 is a schematic diagram of positioning logic in an embodiment of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
The embodiment I provides a single-phase earth fault studying and judging method based on multi-algorithm normalization analysis, which comprises the following steps: step one, carrying out fault identification on working condition data (namely the working condition data obtained by data imaging feature extraction) of a selected time period under a selected bus by using a pre-trained deep learning model, and determining whether a fault occurs; and step two, performing multi-working-condition type identification on the working condition data which is determined to have faults by using a decision tree which is trained in advance.
1) Input data pictured feature extraction
The selected input signal is A, B, C three-phase current waveform on the line, and the waveform length is data from 4 power frequency periods before the fault occurrence time to 8 power frequency periods after the fault occurrence time. To make the input data into a picture, the waveform sequence is matrixed. The input waveform sequence is [1 × 3072] dimensional data calculated by 1K sampling rate, and each data is the amplitude of the point position signal. It can be determined that the number of columns of the matrix is 3072. Taking Max (Max) Max (a), Min (Min) (a), the number of rows of the matrix is (Max-Min)/0.1, the resolution of the taken data is 0.1A, and the dimension of the transformed data matrix is [ (Max-Min)/0.1, 3072 ]. The formed new matrix is a binary matrix, each column of the original waveform sequence data has only one data, the row is obtained by the data value/0.1, and after the row and the column are determined, the value of the position in the matrix is marked as 1.
The specific waveform sequence imaging process is shown in fig. 1.
2) The deep learning model comprises a CNN model and a DBN model, and the training and process analysis of the deep learning model in this embodiment includes:
firstly, pre-training a CNN model by using training set data to obtain parameters of a convolutional layer and a full link layer in the CNN model. With the increase of the iteration times, the loss of the network gradually decreases and tends to be stable, the accuracy of the generated model gradually increases and tends to 93%, the good data processing capability of the CNN model is embodied, and the identification accuracy of the model is to be improved.
And then, fixing the parameters of the convolutional layer of the CNN model, and performing unsupervised pre-training on the DBN model by using the output data of the convolutional layer of the training set to obtain the parameters of each RBM layer in the DBN model. With the increase of the iteration times, the loss of the network gradually decreases and tends to be stable, the accuracy of the generated model gradually increases and tends to 97%, the identification accuracy of the network is greatly improved, and the good processing capability of the DBN model on the one-dimensional data is reflected.
And finally, the full connection layer of the CNN model is an RBM layer of the DBN model, the parameter of the convolutional layer of the CNN model and the parameter of the RBM layer in the DBN model after pre-training are migrated, and the whole network is subjected to supervised fine tuning by using training set data. The loss of the network continues to decrease and tends to be stable, and the accuracy of the generated model gradually increases and tends to be 99%. Compared with the recognition accuracy of the CNN model pre-training, the recognition accuracy after model fusion is greatly improved.
The deep learning in the invention refers to: and a DBN probability generation model is adopted, the whole network generates training data according to the maximum probability through the weight among RBM layer neurons in the training network, and high-level abstract features are formed, so that the classification performance of the model is improved. The learning of the DBN model can be divided into two processes, namely unsupervised layer-by-layer pre-training RBM and supervised Back Propagation (BP) algorithm fine adjustment, the two processes are combined to ensure that parameters are not easy to fall into local optimum, and the disadvantage of long training time is compensated to a certain extent.
Considering the design of a full connection layer in a project CNN model, a DBN model structure built by the project comprises the CNN model and the DBN model, the DBN model comprises 3 RBMs stacked in series and a Softmax layer, the full connection layer of the CNN model is used as the RBM layer of the DBN model, a Dropout mechanism is added in each RBM layer, the Softmax layer maps the hidden layer output of the last RBM into a (0,1) interval, and the probability of each category is obtained, so that multi-label classification is carried out.
Suppose there are K categories, SiRepresenting the probability value of the output of the ith unit, namely the category, and the calculation process of the Softmax layer is
Figure RE-GDA0002667055540000061
ZiZkRepresenting i, k data in the sample, respectively.
The overall cost c of the DBN model adopts a cross entropy function, and y is assumediRepresenting the true classification result, c is calculated as
Figure RE-GDA0002667055540000062
In the training process of the DBN, the DBN firstly trains the 1 st RBM to obtain a proper hidden layer, the activation probability of a hidden layer unit is used as the apparent layer input of the 2 nd RBM, and the training processes of the 2 nd RBM and the 3 rd RBM are analogized in the same way. And performing unsupervised training of RBMs on the DBN layer by adopting unmarked single and composite voltage sag signal data, taking output obtained after a plurality of RBMs are stacked as characteristic parameters, and transmitting the characteristic parameters to the Softmax layer for multi-label classification, thereby forming a complete DBN. And finally, carrying out supervised fine adjustment on the whole network by using the marked single and composite voltage sag signal data and a BP algorithm, and obtaining a pre-trained DBN model when the cost of the network structure is minimum.
The training method of the decision tree in this embodiment includes: according to the analysis of the actual fault waveform, the corresponding fault waveforms show different modes according to different impedance sizes, impedance types, transition modes, repetition modes and the like corresponding to different fault reasons, and the modes can be distinguished and classified by adopting a classification algorithm learned by a machine in the prior art to determine the discrimination of working conditions such as grounding, short circuit, excitation surge current, lightning stroke, power restoration, power failure, other working conditions and the like.
In order to distinguish the above multiple abnormal conditions, a decision tree concept and an iterative method (OVR) are fused in this embodiment, where each small node of the decision tree is a two-classifier of a three-layer BP neural network model, and an ANN classifier is optionally adopted in a specific embodiment.
In a second embodiment, on the basis of the first embodiment, after the obtained multi-condition type identification result, the method further includes a step of verifying the result, including:
the evaluation index of the fault recognition rate in the conventional pattern recognition is introduced, and the calculation method comprises the following steps:
Figure RE-GDA0002667055540000071
in the formula: t is an input known fault test sample set; c is the number of correctly identified samples in the set.
In order to reflect the identification stability of the model to various fault signals, an evaluation index of error (leakage) rate is introduced, and the calculation method is as follows:
Figure RE-GDA0002667055540000072
in the formula: n is an input multiple fault test sample set; m is the number of samples in the set that are misclassified (missed).
The test adopts a plurality of different types of faults (including grounding, short circuit, excitation inrush current, accumulation, power restoration and power failure)) to verify the model generated by training, and introduces noise to carry out iterative training on the network.
In this embodiment, the multi-condition type identification of the condition data determined to have a fault by using the pre-trained decision tree includes:
selecting a total 2924 group of working condition data, wherein a grounding 522 group, a short circuit 236 group, an excitation inrush current 560 group, a lightning stroke 601 group, a power restoration 524 group, a power failure 293 group and other working condition 188 groups, distributing a training set and a test set according to a ratio of 7:3, wherein the training set is 2046 groups, the test set is 878 groups, regular terms are added to prevent a performance function in an overfitting model training process, and the coefficient of the regular terms is set to be 0.00001. Since the condition data is randomly mixed redistribution, the following results show the average results of 10 experiments.
The error of the classifier used finally in the training and testing process, and the total error of the whole discrimination process are shown in the following table,
TABLE 1 working Condition discrimination error results
Figure RE-GDA0002667055540000081
Wherein, when the last multi-working condition discrimination is carried out, the comprehensive training is not carried out, and the error is calculated by the whole data set. From the results, it can be seen that the decision tree model-based multi-condition classification process combines the ANN classifiers of the various conditions, that is, each node in the decision tree is the ANN classifier of the three-layer BP neural network model.
The error of multi-working condition discrimination can be controlled below 6 percent, and the requirement of not misjudging the fault working condition as much as possible is met.
Based on the method, more than twenty common fault types can be distinguished, one or a plurality of most matched standard fault waveforms can be quickly found for a certain fault waveform, and a historical reference case is given. Fig. 2 shows 6 common fault waveforms.
Based on the above embodiments, the third embodiment provides a single-phase earth fault studying and judging method based on multi-algorithm normalization analysis, and further includes the following steps after the multi-operating-condition type discrimination is completed:
and (3) according to the models and classifiers formed in the first step and the second step, carrying out one-by-one matching analysis on the incoming waveforms, simultaneously superposing the steady-state data of the real-time system to carry out multi-algorithm normalization analysis, carrying out comprehensive analysis by using the fault identification results obtained by classifying the steps 1 and 2 and utilizing steady-state information such as voltage, current, network topological relation and the like to form a fault interval and form an operable fault processing strategy, thereby completing all fault studying and judging logics.
The specific analysis method is as follows:
a) and acquiring a fault indicator waveform file collected by the power distribution automation system.
b) And identifying faults by using a fault model for the waveform file, classifying fault conditions by adopting a rapid classifier, and acquiring standard fault waveforms of corresponding classes.
c) And performing topology analysis according to the fault indicators with the fault waveforms, and analyzing a topology model of the communication area where the equipment is located (starting from the main network bus, forming a topological upstream-downstream connection relation of all the fault indicators where the communication area is located).
d) And acquiring waveform files of all fault indicators under the same bus (if the files exist, the equipment is eliminated if the files do not exist), and identifying and classifying the faults of the waveform files of all fault indicators under the bus.
e) And classifying all the equipment with fault waveforms according to fault types, acquiring the classification with the largest action number as a total classification for judging faults, and recording other classification results as suspected classification results.
f) The equipment with fault waveform is used as a fault signal, positioning analysis is carried out on the current earth fault, the analysis principle adopts the traditional positioning logic (dyeing principle) to carry out positioning (as shown in figure 4), and a final fault positioning section is formed.
In this embodiment, a multi-algorithm normalization analysis is performed according to the algorithm calculation result in combination with the steady-state data of the real-time system, and the fault identification result obtained in the first step and the second step is comprehensively analyzed by using the steady-state information such as voltage, current, network topology relation, and the like, so as to form a fault section and an operable fault processing strategy, thereby completing all fault studying and judging logics. The fault identification model formed in the step one is combined with the step two to finally form a fast classifier which is combined with the input waveform to carry out fast fault identification (whether the waveform data has faults or not); fault conditions (grounding, short circuit, etc.) are quickly defined.
In a specific embodiment, an operable fault handling strategy can be generated by adopting the prior art according to a fault positioning result; and forming corresponding information according to the fault positioning result and the fault processing strategy to complete fault processing.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While the present invention has been described with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, which are illustrative and not restrictive, and it will be apparent to those skilled in the art that various changes and modifications can be made therein without departing from the spirit and scope of the invention as defined in the appended claims.

Claims (9)

1. A single-phase earth fault studying and judging method based on multi-algorithm normalization analysis is characterized by comprising the following steps:
carrying out fault identification on the working condition data of a selected time period under a selected bus by using a pre-trained deep learning model to determine whether a fault occurs; and performing multi-working-condition type identification on the working condition data which is determined to have faults by utilizing a pre-trained decision tree.
2. The method as claimed in claim 1, wherein the deep learning model includes a CNN model and a DBN model, and the deep learning model is trained by the following method:
pre-training the CNN model by using the training set data to obtain parameters of a convolutional layer and a full link layer in the CNN model;
fixing parameters of a CNN model convolutional layer, and performing unsupervised pre-training on the DBN model by using output data of a training set after passing through the CNN model convolutional layer to obtain parameters of each RBM layer in the DBN model;
and taking the full connection layer of the CNN model as an RBM layer of the DBN model, migrating the parameter of the convolutional layer of the CNN model and the parameter of the RBM layer in the DBN model, which are acquired after pre-training, and performing supervised fine tuning on the whole network by using training set data.
3. The method for studying and judging the single-phase earth fault based on the multi-algorithm normalization analysis as claimed in claim 2, further comprising the step of verifying the result of the deep learning model, and specifically comprising:
introducing a fault identification rate as an identification evaluation index, wherein the calculation method comprises the following steps:
Figure FDA0002562407180000011
in the formula: t is an input known fault test sample set; c is the number of correctly identified samples in the set;
an error rate or leakage rate evaluation index is introduced, and the calculation mode is as follows:
Figure FDA0002562407180000021
in the formula: n is an input multiple fault test sample set; m is the number of samples in the set which are wrongly or overlooked.
4. The method for studying and judging the single-phase earth fault based on the multi-algorithm normalization analysis as claimed in claim 2, wherein the training set data is obtained by the following steps:
the selected input signal is the working condition data of the selected time period under the selected bus, the input data is picturized, and the waveform sequence is matrixed.
5. The method as claimed in claim 2, wherein the deep learning model includes a CNN model and a DBN model, the DBN model includes 3 RBMs stacked in series and a Softmax layer, a fully-connected layer of the CNN model is used as an RBM layer of the DBN model, a Dropout mechanism is added in each RBM layer, and the Softmax layer maps hidden layer output of the last RBM into a (0,1) interval to obtain probability values of each category, so as to perform multi-label classification.
6. The method as claimed in claim 5, wherein the overall cost c of the DBN model is a cross entropy function, and c is calculated by
Figure FDA0002562407180000022
Wherein y isiRepresenting the true classification result, aiThe probability value obtained for the previous step is calculated.
7. The method as claimed in claim 1, wherein each node in the decision tree is a classifier.
8. The method as claimed in claim 1, further comprising locating the single-phase earth fault by using a dyeing principle after the multi-condition type discrimination is completed.
9. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 8.
CN202010612126.XA 2020-09-04 2020-09-04 Single-phase earth fault studying and judging method based on multi-algorithm normalization analysis Withdrawn CN111898446A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010612126.XA CN111898446A (en) 2020-09-04 2020-09-04 Single-phase earth fault studying and judging method based on multi-algorithm normalization analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010612126.XA CN111898446A (en) 2020-09-04 2020-09-04 Single-phase earth fault studying and judging method based on multi-algorithm normalization analysis

Publications (1)

Publication Number Publication Date
CN111898446A true CN111898446A (en) 2020-11-06

Family

ID=73207926

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010612126.XA Withdrawn CN111898446A (en) 2020-09-04 2020-09-04 Single-phase earth fault studying and judging method based on multi-algorithm normalization analysis

Country Status (1)

Country Link
CN (1) CN111898446A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115545085A (en) * 2022-11-04 2022-12-30 南方电网数字电网研究院有限公司 Weak fault current fault type identification method, device, equipment and medium
CN116008733A (en) * 2023-03-21 2023-04-25 成都信息工程大学 Single-phase grounding fault diagnosis method based on integrated deep neural network

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115545085A (en) * 2022-11-04 2022-12-30 南方电网数字电网研究院有限公司 Weak fault current fault type identification method, device, equipment and medium
CN116008733A (en) * 2023-03-21 2023-04-25 成都信息工程大学 Single-phase grounding fault diagnosis method based on integrated deep neural network

Similar Documents

Publication Publication Date Title
CN104052612B (en) A kind of Fault Identification of telecommunication service and the method and system of positioning
CN107219457A (en) Frame-type circuit breaker fault diagnosis and degree assessment method based on operation annex electric current
CN110726898B (en) Power distribution network fault type identification method
CN109444667A (en) Power distribution network initial failure classification method and device based on convolutional neural networks
CN109409444B (en) Multivariate power grid fault type discrimination method based on prior probability
CN110702966B (en) Fault arc detection method, device and system based on probabilistic neural network
CN113376516A (en) Medium-voltage vacuum circuit breaker operation fault self-diagnosis and early-warning method based on deep learning
CN111898446A (en) Single-phase earth fault studying and judging method based on multi-algorithm normalization analysis
CN113496440B (en) User abnormal electricity consumption detection method and system
CN110794254B (en) Power distribution network fault prediction method and system based on reinforcement learning
CN114325256A (en) Power equipment partial discharge identification method, system, equipment and storage medium
CN114091549A (en) Equipment fault diagnosis method based on deep residual error network
CN117272143A (en) Power transmission line fault identification method and device based on gram angle field and residual error network
CN112633315A (en) Electric power system disturbance classification method
Moravej et al. Power transformer protection scheme based on time‐frequency analysis
CN114626433A (en) Fault prediction and classification method, device and system for intelligent electric energy meter
CN111062569A (en) Low-current fault discrimination method based on BP neural network
CN114415056A (en) Distributed power supply island fault detection method
CN113610119A (en) Method for identifying power transmission line developmental fault based on convolutional neural network
CN113945862A (en) Method, device and equipment for identifying high-resistance grounding fault of power distribution network
CN114252725A (en) HHT and ResNet 18-based single-phase earth fault type comprehensive identification method
Pan et al. Study on intelligent anti–electricity stealing early-warning technology based on convolutional neural networks
CN115392710A (en) Wind turbine generator operation decision method and system based on data filtering
CN114755529A (en) Multi-feature fusion single-phase earth fault type identification method based on deep learning
CN115758235A (en) Open-circuit fault diagnosis method and device based on direct-current charging pile charging module

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
WW01 Invention patent application withdrawn after publication
WW01 Invention patent application withdrawn after publication

Application publication date: 20201106