CN110133455A - The electrical failure sparking discrimination method of tandem type low-voltage alternating-current - Google Patents

The electrical failure sparking discrimination method of tandem type low-voltage alternating-current Download PDF

Info

Publication number
CN110133455A
CN110133455A CN201910317906.9A CN201910317906A CN110133455A CN 110133455 A CN110133455 A CN 110133455A CN 201910317906 A CN201910317906 A CN 201910317906A CN 110133455 A CN110133455 A CN 110133455A
Authority
CN
China
Prior art keywords
deep learning
sparking
current
tandem type
discrimination
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
CN201910317906.9A
Other languages
Chinese (zh)
Inventor
刘寒
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.)
Kean Electric Polytron Technologies Inc
Original Assignee
Kean Electric Polytron Technologies Inc
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 Kean Electric Polytron Technologies Inc filed Critical Kean Electric Polytron Technologies Inc
Priority to CN201910317906.9A priority Critical patent/CN110133455A/en
Publication of CN110133455A publication Critical patent/CN110133455A/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

Abstract

The present invention provides a kind of electrical failure sparking discrimination methods of tandem type low-voltage alternating-current, include: step 1: loop current signals when based on normal operation circuit in the case of Several Typical Load and generating the sparking of tandem type electric fault construct the deep learning model decomposed based on wavelet energy;Step 2: obtaining the loop current signals in power utilization environment in real time, the deep learning Model Distinguish electric fault sparking based on building.The present invention is based on the loop current signals of Several Typical Load in power utilization environment to construct the deep learning model decomposed based on wavelet energy, and then electric disaster hidden-trouble investigation and early warning are carried out to every route according to the deep learning model, electric fault sparking is recognized, the accurate monitoring of electrical fire is realized.

Description

The electrical failure sparking discrimination method of tandem type low-voltage alternating-current
Technical field
The invention belongs to electrical safety protection technology fields, and in particular to a kind of electrical failure sparking of tandem type low-voltage alternating-current Discrimination method.
Background technique
Electrical fire generally refers to release due to electric wiring, electrical equipment, utensil and power supplying and distributing equipment failure The thermal energy put: the energy discharged such as high temperature, electric arc, electric spark and non-faulting;Such as the hot surface of electric heating appliance, having Ignite ontology or other combustibles under burning condition and caused by fire, also include the fire as caused by thunder and lightning and electrostatic.Each In class electrical fire, leakage fire, short circuit and fire hazard are relatively conventional.
Leakage fire
Some place of route because of certain reason (natural cause or artificial origin, such as expose to wind and rain, humidity, high temperature, Bruising is scratched, is rubbed, corroding) make the insulation of electric wire or the insulating capacity decline of timbering material, cause between electric wire and electric wire (reinforcing bar, black sheet iron etc. that electric wire passes through cement wall) has one between (passing through the insulation of damage, bracket etc.), conducting wire and the earth Portion of electrical current passes through, and this phenomenon is exactly to leak electricity.When electric leakage occurs, the electric current sewed is flowing into the earth on the way, such as resistance When biggish position, localized hyperthermia can be generated or generate electric leakage spark, cause neighbouring combustible to catch fire, so as to cause fire.
Short circuit and fire hazard
Bare conductor in electric wiring or after the insulation breakdown of insulated conductor, firewire and zero curve or firewire and ground wire (packet Include ground connection and be subordinated to the earth) short circuit is just cried the phenomenon that certain point is touched together, electric current is caused to increase significantly suddenly, it is commonly called as touching Line, swinging cross or even electric.Resistance is reduced suddenly when due to short circuit, and electric current increases suddenly, and the calorific value of moment is also very big, significantly super Calorific value when route works normally has been crossed, and has been also easy to produce strong spark and electric arc in short dot, insulating layer can not only be made fast Quick burning is burnt, and can make metal molten, is caused neighbouring inflammable combustible combustion, is caused fire.
To monitor electrical fire, the electrical fire monitoring equipment used in the prior art includes temperature sensor, remains Aftercurrent mutual inductor, electric fault sparking detector, fire-fighting Internet of Things data gateway, disappears at Combined electric appliance fire disaster monitoring probe The components such as anti-the Internet transmission module.
Temperature sensor monitors the data such as cable temperature, built-in monitoring distribution box temperature by the mode of tying up, and is transmitted to mixed Box-like detector for electric fire protection;Residual current transformer is monitored in route by the way that route is passed through coil by electric current in the same direction Residual current data;Combined electric appliance fire disaster monitoring probe provides 8 addresses, for docking temperature sensor, residual current Mutual inductor realizes the display and local alarm of line temperature, residual current data.
Existing electrical fire monitoring equipment is mainly used for data monitoring, has deficiency to security protection and monitoring:
1, installation equipment is more, execution conditions are limited, is typically only used for second level distribution box, cannot achieve end Electrical Safety Management;
2, monitoring index lacks for route sparking monitoring, can not effectively accomplish early warning in advance.
3, high temperature, residual current, electric arc alarm data can be found by monitoring, but to be only used for fire pre- for monitoring data It is alert, it cannot achieve hidden danger positioning and exclude.
Summary of the invention
The object of the present invention is to provide a kind of electrical failure sparking discrimination methods of tandem type low-voltage alternating-current, are based on power utilization environment The loop current signals of middle Several Typical Load construct the deep learning model decomposed based on wavelet energy, and then according to the deep learning Model carries out electric disaster hidden-trouble investigation and early warning to every route, and identification electric fault sparking realizes that electrical fire is precisely supervised Control.
The present invention provides a kind of electrical failure sparking discrimination methods of tandem type low-voltage alternating-current, comprising:
Step 1: circuit when based on normal operation circuit in the case of Several Typical Load and generating the sparking of tandem type electric fault Current signal constructs the deep learning model decomposed based on wavelet energy;
Step 2: obtaining the loop current signals in power utilization environment in real time, the deep learning Model Distinguish electricity based on building The sparking of gas failure.
Further, step 1 includes:
Loop current signals based on Several Typical Load, with the average and standard deviation of each layer detail signal energy of wavelet decomposition As deep learning input feature vector amount, the deep learning model is constructed.
Further, step 1 further include:
Before being trained to the deep learning model, based on particle swarm optimization algorithm to deep learning weight and threshold value Initial value carry out optimizing, to accelerate deep learning convergence speed.
Further, step 1 further include:
It is updated during deep learning based on adaptive regularized learning algorithm rate and adjusts the weight and threshold value.
Further, step 2 includes:
The loop current signals of Several Typical Load are obtained based on current transformer and signal conditioning circuit.
Compared with prior art the beneficial effects of the present invention are:
Loop current signals based on Several Typical Load in power utilization environment construct the deep learning mould decomposed based on wavelet energy Type, and then electric disaster hidden-trouble investigation and early warning are carried out to every route according to the deep learning model, identification electric fault is beaten Fire realizes the accurate monitoring of electrical fire.
Detailed description of the invention
Fig. 1 is the flow chart of the electrical failure sparking discrimination method of tandem type low-voltage alternating-current of the present invention;
Fig. 2 is the current signal wavelet transformation energy profile of two kinds of Several Typical Loads;
Fig. 3 is the overall structure figure of the deep learning model decomposed the present invention is based on wavelet energy.
Specific embodiment
The present invention is described in detail for each embodiment shown in reference to the accompanying drawing, but it should be stated that, these Embodiment is not limitation of the present invention, those of ordinary skill in the art according to these embodiments made by function, method, Or equivalent transformation or substitution in structure, all belong to the scope of protection of the present invention within.
Join shown in Fig. 1, present embodiments provide a kind of electrical failure sparking discrimination method of tandem type low-voltage alternating-current, comprising:
Step S1: circuit when based on normal operation circuit in the case of Several Typical Load and generating the sparking of tandem type electric fault Current signal constructs the deep learning model decomposed based on wavelet energy;
Step S2: obtaining the loop current signals in power utilization environment in real time, the deep learning Model Distinguish electricity based on building The sparking of gas failure.
The electrical failure sparking discrimination method of the tandem type low-voltage alternating-current, the loop current based on Several Typical Load in power utilization environment Signal constructs the deep learning model decomposed based on wavelet energy, and then carries out electricity to every route according to the deep learning model Gas fire hazard investigation and early warning, identification electric fault sparking, realize the accurate monitoring of electrical fire.
In the present embodiment, step S1 includes:
Loop current signals based on Several Typical Load, with the average and standard deviation of each layer detail signal energy of wavelet decomposition As deep learning input feature vector amount, the deep learning model is constructed.
In the present embodiment, step S1 further include:
Before being trained to the deep learning model, based on particle swarm optimization algorithm to deep learning weight and threshold value Initial value carry out optimizing, to accelerate deep learning convergence speed.
In the present embodiment, step S1 further include:
It is updated during deep learning based on adaptive regularized learning algorithm rate and adjusts the weight and threshold value.
In the present embodiment, step S2 includes:
The loop current signals of Several Typical Load are obtained based on current transformer and signal conditioning circuit.
Loop current amplitude is smaller when the present invention occurs for the sparking of AC series type electric fault, traditional circuit protection dress The problem of cannot effectively detecting is set, proposes one kind based on wavelet transformation energy in conjunction with deep learning and suitable for a variety of typical cases The electrical failure sparking discrimination method of the tandem type low-voltage alternating-current of load.It is simulated using homemade arc generating device and generates low pressure friendship Electric fault sparking is flowed, circuit when obtaining normal operation circuit in the case of Several Typical Load and generating the sparking of tandem type electric fault Current signal.Wavelet decomposition is carried out to the signal of acquisition, the average and standard deviation of each layer detail signal energy is inputted into BP Small echo deep learning is constituted after deep learning, realizes the identification to different loads test sample.Using particle swarm optimization algorithm meter Deep learning training initial value is calculated, improves training speed using autoadapted learning rate method.Algorithm exports result clear, Input layer characteristic quantity is chosen reasonable.The experimental results showed that being reached using the accuracy rate that this method carries out electric fault sparking identification 95% or more.
Invention is further described in detail below.
1, tandem type electric fault sparking mechanism of production and examination criteria
Arc phenomenon is related with the factors such as the type of gas and pressure, temperature, interelectrode distance and voltage.Tandem type Electric arc is typically due to switch and disconnects or poor wire connection, generate when loosening, and with load is series relationship.Tandem type is electrical Failure sparking is equivalent to when occurring increases the element with certain voltage in the loop, at this time loop current Amplitude Ration circuit It is small when normal operation.Traditional protective device (stoppage circuit breaker etc.) can not generally detect the sparking of tandem type electric fault.
2, electric fault sparking Wave data acquisition
The Several Typical Load in power utilization environment can be obtained using the current transformer and signal conditioning circuit of data collection system Loop current signals, include the case where operate normally with circuit in there are electric fault sparkings.The case where electric fault is struck sparks It can be generated by arc generator.
3, typical current signal waveform analysis
By many experiments, obtains when circuit in the case of a large amount of Several Typical Loads works normally and occur electric fault sparking and return Current signal in road.By experimental result: as nearly resistive load, loop current signals when electric fault sparking occurs With typical " flat shoulder " feature, but for the light modulation lamp load of normal operation (switch property), current waveform also has class Like feature.Currently, since switching mode load use is increasingly frequent, using the shape of current waveform or simple Fourier analysis etc. Method can not generally adapt to the testing requirements of electric fault sparking and easily lead to erroneous judgement.
4, the signal energy distribution decomposed based on discrete dyadic wavelet
Since loop current waveform has apparent mutation and a glitch noise when electric fault, which is struck sparks, to be occurred, and wavelet analysis Method has the function of " school microscop ", is suitable for analyzing jump signal and non-stationary signal, by adjusting time-frequency window Size and location can meet the requirement of analysis signal local feature.Therefore, the present invention is electric as extracting using wavelet analysis method The preprocessing means of gas failure sparking feature.If two into discrete wavelet function are as follows:
ψJ, k(t)=2-j/2ψ(2-jT-k) j, k ∈ Z (1)
The discrete wavelet transformer of respective function f (t) is changed to
If discrete wavelet family of functions ψJ, k(t) L is constituted2(R) orthonormal base, then any f (t) ∈ L2(R)
The deployable linear combination for orthonormal base
Under the conditions of orthogonal basis, wavelet transform is to maintain the constant mapping of norm, and information redundancy is not present in transformation.
After wavelet transform, original signal is broken down into different frequency ranges.If signal sampling frequencies are Fs, then in the decomposition of kth layer, the frequency range of approximation signal is [0, fs/2k+1], the frequency range of detail signal is [fs/2k+1, fs/2k].The quadratic power of coefficient and identical as the gross energy of time domain waveform after wavelet transformation.It therefore, can be from the angle analysis of energy The feature of current signal when electric fault sparking occurs.
5 layers of small wavelength-division are done using various current sampling signals of Symmlets2 (abbreviation Sym2) wavelet basis function to acquisition Solution.Wherein, each layer detail signal energy of the current sampling data of one group of typical electrical equipment load accounts for waveform gross energy Ratio is as shown in Figure 2, wherein (a) is that hair dryer top grade operates normally and when fault electric arc, (b) operate normally for desk lamp with dimmer switch and When fault electric arc.In figure, ☆ indicates that first layer details energy, Δ indicate that second layer details energy, ◇ indicate third layer details energy Amount, indicate the 4th layer of details energy, and zero indicates layer 5 details energy.
The normal operation waveform of electrical equipment is approximately sine wave, detail signal energy very little.Electric fault occurs to beat When fiery, there are catastrophe point and glitch noise signal in current waveform, there are maximum points in the detail signal after wavelet decomposition, make Energy distribution changes.And dimming lamp load is switching mode load, in the detail signal of wavelet decomposition just in normal use There are Energy distribution variations.
Since sparking arc burning is influenced by external environment, electrode material etc., the Energy distribution of adjacent inaction interval becomes Change it is larger, different loads operate normally and electric fault sparking occur when range of energy distribution it is overlapped.Pass through acquisition Mass data and after in the above way making energy profile, even electric fault occurs for the same load of discovery, different moments There is also differents for sample of signal when sparking.Therefore, the analysis method based on wavelet energy be only capable of extract electric fault beat Energy-distributing feature when fire occurs, and be difficult to merely through the electric fault under these a variety of loading conditions of feature accurate recognition Sparking.
4, the electric fault based on small echo deep learning, which is struck sparks, recognizes
The method that " deep learning " refers to the artificial neural network of multilayer and train it.One layer of neural network can be a large amount of squares Battle array number goes weight by nonlinear activation method as input, then generates another data acquisition system as output.This is just as life Juju is the same through the working mechanism of brain, and by the matrix quantity of verification, multilayer tissue is linked together, and it is " big to form neural network Brain " carries out precisely complicated processing, just as people identify object mark picture.Deep learning have self study, it is adaptive, Nonlinear Mapping and good generalization ability, if can in conjunction with wavelet transformation obtain energy-distributing feature, can effectively solve compared with Complicated problem.
Deep learning training process is specific as follows:
(1) use is from lower rising unsupervised learning
Using no each layer parameter of nominal data order training method, unsupervised training is carried out.First with no nominal data training first The parameter of layer first learns the parameter of first layer, limitation and sparsity constraints due to model capacity when training, so that obtain Model can learn the structure to data itself, to obtain the feature for having more expression ability than inputting;Is obtained in study After n-1 layers, by n-1 layers of output as the input of n-th layer, thus training n-th layer respectively obtains each layer parameter.
(2) top-down supervised learning
Each layer parameter obtained based on the first step further finely tunes the parameter of entire multilayered model, carries out Training mistake Journey;The first step is led to similar to the random initializtion disposal process of neural network due to the first step not instead of random initializtion of DL What the structure of overfitting input data obtained, thus entire disposition more tightly global optimum, so as to obtain better effect;
The average value of 5 Periodic decomposition energy is taken after wavelet transformation by layerAnd standard deviationAs BP depth The input feature vector amount of study, the higher-dimension that original signal is constituted measure space reflection to low-dimensional feature space.
The input feature vector amount of BP deep learning, it is empty to low-dimensional feature that the higher-dimension that original signal is constituted measures space reflection Between.
In formula, j is decomposition level;K is current period;N is total periodicity.
To make small echo deep learning model be easy to be transplanted in hardware device, differentiate result using " electric fault sparking " and The output of " normal operation " two-value, therefore, output layer only needs a node, is in the value of network output quantity in [0,1] range.
There is certain optimizing space when to make network training, the desired output of sample is converted in the following manner.
When predicting test set sample, output is boundary with 0.5, and [0.5,1] is determined as that electric fault is struck sparks;[0,0.5] is sentenced It is set to normal operation.
Hidden layer transmission function f () selects non-linear Sigmoid function, makes to have between input layer and output layer non-thread Property map feature, target error function E (t) are defined as sample desired output Y 'kIt is exported with each neuron of network output layer The difference e of valuek=Y 'k-YkQuadratic power and.
In the present embodiment, before being trained to BP deep learning, particle group optimizing (ParticleSwarm is used Optimization, PSO) initial value progress optimizing of the algorithm to deep learning weight and threshold value, to accelerate deep learning training Convergence rate.
A potential optimal solution of each particle representing optimized problem in algorithm, with position, speed and fitness value 3 The feature of the index expression particle.It include the whole weights and threshold value of deep learning in the population of design.In each iteration, Particle updates the position of oneself by tracking individual extreme value and global extremum, to find optimal solution.
The particle populations scale of algorithm construction is 30, and inertia weight is ω=0.7, and acceleration factor c1=c2=2 is maximum The number of iterations is 200, fitness function be all sample output errors of training set quadratic power and.Utilize Particle Swarm Optimization The weight of one group of deep learning and the optimal solution of threshold value can be obtained in method.
In the present embodiment, to improve learning efficiency, adaptive regularized learning algorithm rate is introduced in BP deep learning and updates tune Whole weight and threshold value.
Adaptively regularized learning algorithm rate formula is
In formula, μ1、μ2For learning rate acceleration, general μ1Value is μ between 1.01~1.32Value be 0.7~0.9 it Between;E (t) is the error of the t times training output, and it is excessively frequent to avoid learning rate variation to introduce k1, k2.
Join shown in Fig. 3, Fig. 3 is the deep learning model overall structure figure decomposed based on wavelet energy.Deep learning has 10 A input node does normalized using minimax method and is supported on difference to eliminate same type before data input network Influence under power.
Wavelet decomposition has multi-resolution characteristics, and deep learning has Nonlinear Mapping, adaptive learning and fault-tolerant ability. The present embodiment is based on loop current sampled signal, using the average and standard deviation of each layer detail signal energy of wavelet decomposition as deeply Degree study input feature vector amount constructs the small echo deep learning mould of the electric fault sparking identification suitable for a variety of Several Typical Loads Type.Initial value optimizing using particle swarm optimization algorithm to deep learning weight and threshold value, avoids random starting values bring Uncertainty effectively improves the pace of learning of network using the method that autoadapted learning rate adjusts.Using Mean Impact Value side Method demonstrates the validity of deep learning input layer characteristic quantity, can be avoided the redundancy of deep learning input information.
It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie In the case where without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims Variation is included within the present invention.

Claims (5)

  1. The discrimination method 1. a kind of electrical failure of tandem type low-voltage alternating-current is struck sparks characterized by comprising
    Step 1: loop current when based on normal operation circuit in the case of Several Typical Load and generating the sparking of tandem type electric fault Signal constructs the deep learning model decomposed based on wavelet energy;
    Step 2: the loop current signals in power utilization environment, the electrically event of the deep learning Model Distinguish based on building are obtained in real time Barrier sparking.
  2. The discrimination method 2. the electrical failure of tandem type low-voltage alternating-current according to claim 1 is struck sparks, which is characterized in that the step Rapid one includes:
    Loop current signals based on Several Typical Load, using the average and standard deviation of each layer detail signal energy of wavelet decomposition as Deep learning input feature vector amount constructs the deep learning model.
  3. The discrimination method 3. the electrical failure of tandem type low-voltage alternating-current according to claim 2 is struck sparks, which is characterized in that the step Rapid one further include:
    Before being trained to the deep learning model, based on particle swarm optimization algorithm to deep learning weight and threshold value just Initial value carries out optimizing, to accelerate deep learning convergence speed.
  4. The discrimination method 4. the electrical failure of tandem type low-voltage alternating-current according to claim 3 is struck sparks, which is characterized in that the step Rapid one further include:
    It is updated during deep learning based on adaptive regularized learning algorithm rate and adjusts the weight and threshold value.
  5. The discrimination method 5. the electrical failure of tandem type low-voltage alternating-current according to claim 4 is struck sparks, which is characterized in that the step Rapid two include:
    The loop current signals of Several Typical Load are obtained based on current transformer and signal conditioning circuit.
CN201910317906.9A 2019-04-19 2019-04-19 The electrical failure sparking discrimination method of tandem type low-voltage alternating-current Pending CN110133455A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910317906.9A CN110133455A (en) 2019-04-19 2019-04-19 The electrical failure sparking discrimination method of tandem type low-voltage alternating-current

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910317906.9A CN110133455A (en) 2019-04-19 2019-04-19 The electrical failure sparking discrimination method of tandem type low-voltage alternating-current

Publications (1)

Publication Number Publication Date
CN110133455A true CN110133455A (en) 2019-08-16

Family

ID=67570519

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910317906.9A Pending CN110133455A (en) 2019-04-19 2019-04-19 The electrical failure sparking discrimination method of tandem type low-voltage alternating-current

Country Status (1)

Country Link
CN (1) CN110133455A (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106326918A (en) * 2016-08-12 2017-01-11 国网山东省电力公司滨州供电公司 Multiscale energy feature linear recognition method for partial discharge ultrahigh frequency signal of transformer
CN107064752A (en) * 2017-03-22 2017-08-18 北京航空航天大学 A kind of distinguished number of aviation fault electric arc detection
CN108646149A (en) * 2018-04-28 2018-10-12 国网江苏省电力有限公司苏州供电分公司 Fault electric arc recognition methods based on current characteristic extraction

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106326918A (en) * 2016-08-12 2017-01-11 国网山东省电力公司滨州供电公司 Multiscale energy feature linear recognition method for partial discharge ultrahigh frequency signal of transformer
CN107064752A (en) * 2017-03-22 2017-08-18 北京航空航天大学 A kind of distinguished number of aviation fault electric arc detection
CN108646149A (en) * 2018-04-28 2018-10-12 国网江苏省电力有限公司苏州供电分公司 Fault electric arc recognition methods based on current characteristic extraction

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
张士文 等: "一种基于小波变换能量与神经网络结合的串联型故障电弧辨识方法", 《电工技术学报》 *

Similar Documents

Publication Publication Date Title
Gururajapathy et al. Fault location and detection techniques in power distribution systems with distributed generation: A review
Pillai et al. A comprehensive review on protection challenges and fault diagnosis in PV systems
Wang et al. A novel deep learning method for the classification of power quality disturbances using deep convolutional neural network
US20190067941A1 (en) Methods and apparatus for determining conditions of power lines
Ghaderi et al. High-impedance fault detection in the distribution network using the time-frequency-based algorithm
Alipio et al. Modeling the frequency dependence of electrical parameters of soil
CN105891629B (en) A kind of discrimination method of transformer equipment failure
Khokhar et al. A new optimal feature selection algorithm for classification of power quality disturbances using discrete wavelet transform and probabilistic neural network
Alipio et al. Frequency dependence of soil parameters: Effect on the lightning response of grounding electrodes
Hines et al. The topological and electrical structure of power grids
Ghanizadeh et al. ANN and cross-correlation based features for discrimination between electrical and mechanical defects and their localization in transformer winding
CN103091609B (en) A kind of GIS local discharge on-line monitoring device performance detecting system and method thereof
Pillai et al. A comparative evaluation of advanced fault detection approaches for PV systems
JP5444343B2 (en) Utility instrument
Baqui et al. High impedance fault detection methodology using wavelet transform and artificial neural networks
Chakravorti et al. Recent trends in the condition monitoring of transformers
EP1714366B1 (en) High impedance fault detection
Lu et al. A comprehensive review on DC arc faults and their diagnosis methods in photovoltaic systems
Griffiths et al. The invisible fray: A critical analysis of the use of reflectometry for fray location
CN106199305B (en) Underground coal mine power supply system dry-type transformer insulation health state evaluation method
Alam et al. A comprehensive review of catastrophic faults in PV arrays: types, detection, and mitigation techniques
Berges et al. Training load monitoring algorithms on highly sub-metered home electricity consumption data
Kong et al. A Novel Traveling-Wave-Based Main Protection Scheme for $\pm $800 kV UHVDC Bipolar Transmission Lines
Li et al. Contamination level prediction of insulators based on the characteristics of leakage current
Tripathy Power transformer differential protection using neural network principal component analysis and radial basis function 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