CN108562835A - A kind of fault arc detection method based on BP neural network - Google Patents

A kind of fault arc detection method based on BP neural network Download PDF

Info

Publication number
CN108562835A
CN108562835A CN201810227332.1A CN201810227332A CN108562835A CN 108562835 A CN108562835 A CN 108562835A CN 201810227332 A CN201810227332 A CN 201810227332A CN 108562835 A CN108562835 A CN 108562835A
Authority
CN
China
Prior art keywords
current
neural network
fault
arc detection
denoted
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
Application number
CN201810227332.1A
Other languages
Chinese (zh)
Inventor
梁昆
张轩铭
王利强
陈龙
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.)
HANGZHOU TOP TECHNOLOGY Co Ltd
Original Assignee
HANGZHOU TOP TECHNOLOGY 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 HANGZHOU TOP TECHNOLOGY Co Ltd filed Critical HANGZHOU TOP TECHNOLOGY Co Ltd
Priority to CN201810227332.1A priority Critical patent/CN108562835A/en
Publication of CN108562835A publication Critical patent/CN108562835A/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01RMEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/12Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computer systems based on biological models
    • G06N3/02Computer systems based on biological models using neural network models
    • G06N3/04Architectures, e.g. interconnection topology
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computer systems based on biological models
    • G06N3/02Computer systems based on biological models using neural network models
    • G06N3/08Learning methods

Abstract

The invention discloses a kind of fault arc detection methods based on BP neural network, include the following steps:One, sample characteristics extract:Extract electric current flat shoulder time, current-jump value ratio, electric current substantially this four main features of the standard deviation of current average in the number, 1s of saltus step in 1s;Two, neural metwork training;Three, online judgment step:Judge whether to belong to fault electric arc.The fault arc detection method based on BP neural network of the present invention, takes full advantage of large sample, avoids the mode of artificial selected threshold.

Description

A kind of fault arc detection method based on BP neural network
Technical field
The invention belongs to faults to judge field, specifically a kind of fault electric arc detection side based on BP neural network Method.
Background technology
The fault judgment method of mainstream is typically the artificial feature for choosing current waveform at present, then by manually choosing fixation Threshold decision whether belong to fault electric arc.Fault electric arc can carry out the simulation under various scenes, various loadtypes, compare It is easy to generate a larger sample database.But by the method for manual analysis selected threshold, it is difficult to make full use of full-page proof This.
Invention content
In order to solve the above technical problems existing in the prior art, the present invention provides a kind of based on BP neural network Fault arc detection method includes the following steps:
One, sample characteristics extract:Extract electric current flat shoulder time, current-jump value ratio, in 1s electric current substantially saltus step time This four main features of the standard deviation of current average in number, 1s;
Two, neural metwork training;
Three, online judgment step:Judge whether to belong to fault electric arc.
Further, the time of the current-jump in a power frequency period is denoted as tpulse, calculate from tpulseExtreme value afterwards Point, extreme point time are denoted as textreme, then the electric current flat shoulder time be:
tflat=textreme-tpulse
Further, Δ ipulseIt is denoted as current-jump value, the sampling number in a power frequency period is P, in power frequency period Average current is poor:Then current-jump ratio calculates as follows:
Further, the number of electric current substantially saltus step is denoted as N in 1spulse_num
Further, if the current average in each period isAverage value (the N being denoted as in N number of period =50), then the standard deviation standard deviation of current average is in 1s:
Further, step 2 is specially:
(1) network configuration is defined
A. input layer:4 nodes;
B.2 a hidden layer:Respectively 5 nodes, 3 nodes;
C.Softmax layers:SoftMax functions
D. output layer:2 nodes;
E. transmission function:sigmoid
F. learning rate 0.8;
(2) training numerical example acquisition
T={ (x1,y1),(x2,y2),...,(xm,ym),
WhereinN=4
The input of BP neural network is the four-dimensional eigen vector x of above-mentioned 4 main features, and output is y;
(3) training and model tuning obtain the output model of BP neural network.
Further, step 3 is specially:
(1) fault electric arc judgement is carried out to one section of current waveform, its characteristic information vector is denoted as x;
(2) characteristic information vector x is input to the probability [p for being predicted to obtain softmax layers of output in model1 p2]T
(3) if p1>p2, then judging result is fault electric arc;If p1<=p2Then judging result is non-faulting electric arc.
The fault arc detection method based on BP neural network of the present invention, takes full advantage of large sample, avoids artificial The mode of selected threshold.
Description of the drawings
Fig. 1 is the implementing procedure figure of the present invention.
Specific implementation mode
The invention will be further described below.
As shown in Figure 1, the fault arc detection method based on BP neural network of the present invention, includes the following steps:
1. sample characteristics extract
Extract 4 main features such as following A, B, C, D:
The flat shoulder time (feature A) of electric current:Since most of scene has other loads in practice, electric current can not be directly measured The flat shoulder time.The time of current-jump in one power frequency period (20ms) is denoted as t by the present inventionpulse, then calculate from tpulseAfterwards Extreme point, the extreme point time is denoted as textreme.Then the electric current flat shoulder time is:
tflat=textreme-tpulse
Current-jump value ratio (feature B):ΔipulseIt is denoted as current-jump value, the sampling number in a power frequency period is P, the average current in power frequency period it is poor:Then saltus step ratio calculates as follows,
The number (feature C) of electric current substantially saltus step in 1s:It is each during due to the electric appliances normal use such as Switching Power Supply Period all can there is a situation where current-jumps.And a feature of usually fault electric arc is that the duration is short, intermittently, therefore Current-jump number is as a feature:Npulse_num
The standard deviation (feature D) of current average in 1s:If the current average in each period is Note For the average value (N=50) in N number of period, then standard deviation is:
2. neural metwork training
(4) network configuration is defined
G. input layer:4 nodes;
H.2 a hidden layer:Respectively 5 nodes, 3 nodes;
I.Softmax layers:SoftMax functions
J. output layer:2 nodes;
K. transmission function:sigmoid
L. learning rate 0.8.
(5) training numerical example acquisition
T={ (x1,y1),(x2,y2),...,(xm,ym),
WhereinN=4
The input of BP neural network is the four-dimensional eigen vector x of above-mentioned 4 main features, and output is y;
(6) training and model tuning obtain the output model of BP neural network.
3. online judgment step
(4) assume to need to carry out fault electric arc judgement to one section of current waveform, its characteristic information vector is denoted as x;
(5) characteristic information vector x is input to the probability [p for being predicted to obtain softmax layers of output in model1 p2]T
(6) if p1>p2, then judging result is fault electric arc;If p1<=p2Then judging result is non-faulting electric arc.

Claims (7)

1. a kind of fault arc detection method based on BP neural network, includes the following steps:
One, sample characteristics extract:Extract electric current flat shoulder time, current-jump value ratio, in 1s electric current substantially saltus step number, 1s This four main features of the standard deviation of interior current average;
Two, neural metwork training;
Three, online judgment step:Judge whether to belong to fault electric arc.
2. the fault arc detection method based on BP neural network as described in claim 1, it is characterised in that:
The time of current-jump in one power frequency period is denoted as tpulse, calculate from tpulseExtreme point afterwards, extreme point time It is denoted as textreme, then the electric current flat shoulder time be:
tflat=textreme-tpulse
3. the fault arc detection method based on BP neural network as described in claim 1, it is characterised in that:
ΔipulseIt is denoted as current-jump value, the sampling number in a power frequency period is P, and the average current in power frequency period is poor:Then current-jump ratio calculates as follows:
4. the fault arc detection method based on BP neural network as described in claim 1, it is characterised in that:Electric current is big in 1s The number of width saltus step is denoted as Npulse_num
5. the fault arc detection method based on BP neural network as described in claim 1, it is characterised in that:If each period Interior current average is The average value (N=50) being denoted as in N number of period, then in 1s current average mark Quasi- difference standard deviation is:
6. the fault arc detection method based on BP neural network as described in claim 1, it is characterised in that:
Step 2 is specially:
(1) network configuration is defined
A. input layer:4 nodes;
B.2 a hidden layer:Respectively 5 nodes, 3 nodes;
C.Softmax layers:SoftMax functions
D. output layer:2 nodes;
E. transmission function:sigmoid
F. learning rate 0.8;
(2) training numerical example acquisition
T={ (x1,y1),(x2,y2),...,(xm,ym),
Wherein xi∈!n, n=4
The input of BP neural network is the four-dimensional eigen vector x of above-mentioned 4 main features, and output is y;
(3) training and model tuning obtain the output model of BP neural network.
7. the fault arc detection method based on BP neural network as claimed in claim 6, it is characterised in that:
Step 3 is specially:
(1) fault electric arc judgement is carried out to one section of current waveform, its characteristic information vector is denoted as x;
(2) characteristic information vector x is input to the probability [p for being predicted to obtain softmax layers of output in model1 p2]T
(3) if p1>p2, then judging result is fault electric arc;If p1<=p2Then judging result is non-faulting electric arc.
CN201810227332.1A 2018-03-19 2018-03-19 A kind of fault arc detection method based on BP neural network Pending CN108562835A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810227332.1A CN108562835A (en) 2018-03-19 2018-03-19 A kind of fault arc detection method based on BP neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810227332.1A CN108562835A (en) 2018-03-19 2018-03-19 A kind of fault arc detection method based on BP neural network

Publications (1)

Publication Number Publication Date
CN108562835A true CN108562835A (en) 2018-09-21

Family

ID=63531771

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810227332.1A Pending CN108562835A (en) 2018-03-19 2018-03-19 A kind of fault arc detection method based on BP neural network

Country Status (1)

Country Link
CN (1) CN108562835A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109615070A (en) * 2018-12-06 2019-04-12 浙江巨磁智能技术有限公司 Electric power artificial intelligence chip and power failure recognition methods
CN110058133A (en) * 2019-04-15 2019-07-26 杭州拓深科技有限公司 A kind of electrical circuit fault electric arc wrong report optimization method based on feedback mechanism
CN110097259A (en) * 2019-04-15 2019-08-06 杭州拓深科技有限公司 A kind of concentrating type electrical safety hidden danger pre-judging method based on artificial neural network
CN110244198A (en) * 2019-05-09 2019-09-17 山东优柏电子科技有限公司 Resistive load serial arc detection method and application based on compound criterion
CN110398669A (en) * 2019-06-11 2019-11-01 深圳供电局有限公司 Method for detecting arc
WO2021057107A1 (en) * 2019-09-23 2021-04-01 华为技术有限公司 Direct-current arc detection method, apparatus, device and system and storage medium

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH07181219A (en) * 1993-12-24 1995-07-21 Tokyo Electric Power Co Inc:The Method and apparatus for detecting partial discharge
CN104569684A (en) * 2015-01-14 2015-04-29 上海和伍新材料科技有限公司 Fault electric arc detection method based on electric arc spectrum signals
CN104635133A (en) * 2015-03-14 2015-05-20 北京芯同汇科技有限公司 Novel fault arc detection method and device
CN104678265A (en) * 2015-01-30 2015-06-03 广东雅达电子股份有限公司 Detection device and detection method for series arc faults
CN105425118A (en) * 2015-10-29 2016-03-23 山东建筑大学 Multi-information fusion fault arc detection method and device
CN107340459A (en) * 2016-11-24 2017-11-10 安徽江淮汽车集团股份有限公司 A kind of DC Line Fault arc method for measuring and system

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH07181219A (en) * 1993-12-24 1995-07-21 Tokyo Electric Power Co Inc:The Method and apparatus for detecting partial discharge
CN104569684A (en) * 2015-01-14 2015-04-29 上海和伍新材料科技有限公司 Fault electric arc detection method based on electric arc spectrum signals
CN104678265A (en) * 2015-01-30 2015-06-03 广东雅达电子股份有限公司 Detection device and detection method for series arc faults
CN104635133A (en) * 2015-03-14 2015-05-20 北京芯同汇科技有限公司 Novel fault arc detection method and device
CN105425118A (en) * 2015-10-29 2016-03-23 山东建筑大学 Multi-information fusion fault arc detection method and device
CN107340459A (en) * 2016-11-24 2017-11-10 安徽江淮汽车集团股份有限公司 A kind of DC Line Fault arc method for measuring and system

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
刘鹏等: "基于神经网络的故障电弧检测装置的研究", 《低压电器》 *

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109615070A (en) * 2018-12-06 2019-04-12 浙江巨磁智能技术有限公司 Electric power artificial intelligence chip and power failure recognition methods
CN110058133A (en) * 2019-04-15 2019-07-26 杭州拓深科技有限公司 A kind of electrical circuit fault electric arc wrong report optimization method based on feedback mechanism
CN110097259A (en) * 2019-04-15 2019-08-06 杭州拓深科技有限公司 A kind of concentrating type electrical safety hidden danger pre-judging method based on artificial neural network
CN110058133B (en) * 2019-04-15 2021-03-02 杭州拓深科技有限公司 Feedback mechanism-based electric circuit fault arc false alarm optimization method
CN110244198A (en) * 2019-05-09 2019-09-17 山东优柏电子科技有限公司 Resistive load serial arc detection method and application based on compound criterion
CN110398669A (en) * 2019-06-11 2019-11-01 深圳供电局有限公司 Method for detecting arc
WO2021057107A1 (en) * 2019-09-23 2021-04-01 华为技术有限公司 Direct-current arc detection method, apparatus, device and system and storage medium

Similar Documents

Publication Publication Date Title
CN105846780B (en) A kind of photovoltaic module method for diagnosing faults based on decision-tree model
WO2015158198A1 (en) Fault recognition method and system based on neural network self-learning
Illias et al. Hybrid modified evolutionary particle swarm optimisation-time varying acceleration coefficient-artificial neural network for power transformer fault diagnosis
Zeng et al. The design of top-hat morphological filter and application to infrared target detection
CN105093122B (en) Emergency light battery SOC method of estimation based on the adaptive SQKF of strong tracking
CN105975504A (en) Recurrent neural network-based social network message burst detection method and system
CN101975910B (en) Intelligent fault classification and location method for ultra-high voltage direct current transmission line
CN104809722A (en) Electrical device fault diagnosis method based on infrared thermography
CN107391852B (en) Transient stability real-time evaluation method and device based on deep belief network
CN104155574A (en) Power distribution network fault classification method based on adaptive neuro-fuzzy inference system
CN106093713A (en) A kind of power distribution network section Fault Locating Method containing distributed power source and system
CN104523268B (en) Electroencephalogram signal recognition fuzzy system and method with transfer learning ability
CN104242267B (en) A kind of wind-power electricity generation sends out transmission line distance protecting method
Amraee Loss-of-field detection in synchronous generators using decision tree technique
CN102809718A (en) Ultra-high-frequency partial discharge signal identification method for gas insulated switchgear (GIS)
CN108089099A (en) The diagnostic method of distribution network failure based on depth confidence network
Barbeiro et al. State estimation in distribution smart grids using autoencoders
CN105426908B (en) A kind of substation&#39;s attributive classification method based on convolutional neural networks
CN104218672B (en) A kind of power distribution network operation risk early warning system evaluated based on real-time running state
CN107422272B (en) Intelligent detection device for SOC of power battery of electric automobile
CN101477172A (en) Analogue circuit fault diagnosis method based on neural network
Wei et al. Waveform matching approach for fault diagnosis of a high-voltage transmission line employing harmony search algorithm
CN104934968A (en) Multi-agent based distribution network disaster responding recovery coordinate control method and multi-agent based distribution network disaster responding recovery coordinate control device
Yunpeng et al. Multi-step ahead time series forecasting for different data patterns based on LSTM recurrent neural network
CN103886374A (en) Cable joint wire temperature prediction method based on RBF neural network

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