CN111458599A - Series arc fault detection method based on one-dimensional convolutional neural network - Google Patents
Series arc fault detection method based on one-dimensional convolutional neural network Download PDFInfo
- Publication number
- CN111458599A CN111458599A CN202010302286.4A CN202010302286A CN111458599A CN 111458599 A CN111458599 A CN 111458599A CN 202010302286 A CN202010302286 A CN 202010302286A CN 111458599 A CN111458599 A CN 111458599A
- Authority
- CN
- China
- Prior art keywords
- data
- fault
- load
- neural network
- current
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
- G01R31/085—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution lines, e.g. overhead
-
- 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
- G01R31/12—Testing 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
- G01R31/1227—Testing 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 of components, parts or materials
- G01R31/1263—Testing 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 of components, parts or materials of solid or fluid materials, e.g. insulation films, bulk material; of semiconductors or LV electronic components or parts; of cable, line or wire insulation
- G01R31/1272—Testing 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 of components, parts or materials of solid or fluid materials, e.g. insulation films, bulk material; of semiconductors or LV electronic components or parts; of cable, line or wire insulation of cable, line or wire insulation, e.g. using partial discharge measurements
-
- 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
Abstract
The invention relates to a series arc fault detection method based on a one-dimensional convolutional neural network, which comprises the following steps of: 1) acquiring current data of three types of loads, namely resistive load, inductive load and nonlinear load, in normal and fault states, and dividing the current data into five types of data, namely normal resistive or inductive load, fault of resistive load, fault of inductive load, normal nonlinear load and fault of nonlinear load, so as to form a data set; 2) constructing a convolutional neural network by taking current data as input and five types of labels as output; inputting data in the data set into a convolutional neural network for training to obtain an arc fault detection model; 3) the method comprises the steps of collecting a current waveform of a line to be detected in a specified length at a set time interval, inputting current data into an arc fault detection model, calculating a type label corresponding to the current data so as to detect the line in real time, and disconnecting the line through a tripping device when continuous faults are detected. The method is beneficial to improving the accuracy of arc fault detection.
Description
Technical Field
The invention belongs to the technical field of arc fault detection, and particularly relates to a series arc fault detection method based on a one-dimensional convolutional neural network.
Background
With the increasing of electric equipment, the problem of electric safety is not ignored. According to 2009-2018 national fire data counted by the fire department of the Ministry of public Security, 24.98 thousands of fires occur each year on average, and the direct property loss is 32.55 million yuan. From the viewpoint of the cause of fire, the number of electrical fires accounts for about 30% of the total number of fires per year, and electrical fires are the leading cause of fire. The electrical fault sources that trigger electrical fires are short circuits, overloads, leakage, and arc faults, among others. Currently, circuit breakers, fuses and earth leakage protectors can effectively protect against short circuits, overloads and earth leakage faults, respectively. For parallel arc faults and earth arc faults, the rapid increase in current induced by them tends to cause the circuit breaker and fuse to operate. For series arc faults, the leakage protector cannot detect such faults because no leakage branch exists; the presence of the impedance of the series arc itself causes the load current to be less than the normal operating current, and the circuit breaker and fuse are likewise unable to detect such faults. Series arc fault detection is a weak link of current power supply and distribution system fault monitoring.
Arcing is a gas ionization discharge phenomenon. When an arc occurs, physical phenomena such as arc sound, arc light, heat generation, noise, and electromagnetic field change are involved, and these physical characteristics can be used as a basis for the occurrence of an arc fault. In addition, the load current and the load terminal voltage can also show special characteristics, such as high-frequency noise and zero-break phenomenon of the current and distortion of the voltage waveform, and the electric characteristics can also be used as the basis for the occurrence of the arc fault.
The current arc fault detection technology mainly has two main research directions: the arc fault detection technology based on the arc physical characteristics is used for carrying out arc fault detection by capturing arc physical characteristic signals through a sensing device placed at a specific position. The method is easily limited by the position of the fault arc, and is only suitable for arc fault fixed-point detection of a small space and a short line. The other is an arc fault detection technology based on arc electrical characteristics, which analyzes the time-frequency domain characteristics of current signals or voltage signals (mainly current signals) in a circuit and selects proper characteristic quantity as the criterion of arc faults. The method is not limited by the spatial position of the fault arc, and is the main research direction of the current arc fault detection technology. Arc current based arc fault detection technology has become the main research direction of current arc fault detection technology in view of feasibility and simplicity of arc fault detection.
The traditional arc fault detection technology based on current usually analyzes current signals by comparing the difference of current waveforms in normal and arc fault states by means of a time domain analysis method, a frequency domain analysis method and a time-frequency domain analysis method, selects proper characteristic quantities through or without signal transformation (Fourier transformation, wavelet transformation and the like), and artificially selects proper characteristic threshold values as fault criteria through experimental comparison. The method has the problems that the characteristic quantity is difficult to select, and the characteristic threshold value is easily influenced by load property, fault reason, combustion working condition and the like.
In order to overcome the defect that the traditional characteristic threshold value method is easily influenced by loads, working conditions and the like, part of scholars consider that machine learning related intelligent algorithms are used for fault arc detection. Conventional machine learning algorithms include decision trees, neural networks, support vector machines, bayesian classifiers, random forests, clustering algorithms, and the like. The emergence of machine learning intelligent algorithm provides a new research direction for arc detection, which can enable us to get rid of the selection of characteristic threshold, but still is difficult to avoid the core problem of fault arc detection, namely the selection of characteristic quantity.
The characteristic selection in the arc fault detection technology is a process of distinguishing whether a current waveform belongs to a normal state or an arc fault state according to specific characteristics provided by self or other people by related professionals according to experience. The characteristic quantity of the current time domain signal can be a characteristic quantity of the current time domain signal, such as a valid value, a slope, a kurtosis, a half-wave asymmetry, a slope mutation rate, a zero rest duration and the like; or characteristic quantities of the current frequency domain signal, such as the content and phase of each subharmonic, the total harmonic distortion, the inter-harmonic content, etc.; or characteristic quantities of the current time-frequency domain signal, such as wavelet energy entropy, wavelet coefficients and the like.
Although the detection effect of the arc fault detection technology is continuously improved, the technology is still difficult to be applied to engineering practice. This is mainly because current arc fault detection methods do not depart from a manual, subjective feature selection process. However, the characteristic amount applicable to one load may fail in another load. For a complex power supply and distribution system, characteristic quantities with high accuracy and good universality need to be continuously searched for fault detection, and the process is time-consuming and labor-consuming and is difficult to break through.
The deep learning is a feature extraction tool based on data driving, and has strong data feature mining capability. The method takes the original form of data as input, and realizes automatic extraction of original data features through layer-by-layer abstraction of an algorithm, thereby getting rid of dependence on artificial feature extraction and expert knowledge. Arc fault detection techniques rely primarily on fault feature extraction of current sequences, and manual feature extraction can limit the feature expression capability of current signals to some extent. Deep learning is a branched area of machine learning. Convolutional neural networks are one of its representative algorithms. The convolution operation is carried out on the input original information through a multilayer network, the low-dimensional information is nonlinearly mapped into the high-dimensional information, and finally the complex mapping from the input to the output target is established, so that the method is an ideal feature extraction tool.
Disclosure of Invention
The invention aims to provide a series arc fault detection method based on a one-dimensional convolutional neural network, which is favorable for improving the accuracy of arc fault detection.
In order to achieve the purpose, the invention adopts the following technical scheme: a series arc fault detection method based on a one-dimensional convolutional neural network comprises the following steps:
1) acquiring current data of normal and fault states of resistive, inductive and nonlinear loads, and dividing the current data into five types of data: resistance or inductive load normal data, resistance load fault data, inductive load fault data, nonlinear load normal data and nonlinear load fault data form a data set for training the convolutional neural network;
2) constructing a convolutional neural network by taking current data as input and five types of labels of normal resistance or inductive load, resistive load fault, inductive load fault, normal nonlinear load and nonlinear load fault as output; inputting the data in the data set into the constructed convolutional neural network for network training to obtain an arc fault detection model;
3) the method comprises the steps of collecting a current waveform of a line to be detected with a specified length at a set time interval, inputting collected current data into a trained arc fault detection model, calculating a type label corresponding to the current data, detecting the line in real time, and disconnecting the line through a tripping device when faults occur in N time slices.
Further, in step 1, the method for acquiring current data of different types of loads in normal and fault states includes: each type of load is respectively connected with an arc generator, an alternating current power supply and a switch to form an arc fault generating circuit, current data in the circuit are collected in real time through a current sensor installed in the circuit, a data acquisition card and a computer, the number of fault data is increased through artificial simulated arc discharge, and current waveforms of the type of load in normal and arc fault states are obtained.
Further, in step 2, the convolutional neural network mainly includes an input layer, alternating convolutional layers and pooling layers, a full-link layer, and an output layer, the input layer preprocesses input data, the convolutional layers and the pooling layers respectively correspond to convolutional operation and pooling operation, the full-link layer formalizes a target task as a target function, maps learned features to a sample type space, and the output layer outputs a predicted value of the convolutional neural network for the type of the input sample.
Further, the training method of the convolutional neural network comprises the following steps: dividing a data set into a training data set and a testing data set, inputting the training data set and the testing data set in batches, repeatedly updating the weights and the offsets of the convolution layer and the full connection layer of the network, and finally obtaining a group of parameter values which enable the loss value of the loss function to be minimum.
Further, in step 3, after the type tag corresponding to the current data is calculated, it is determined whether an arc fault occurs, if no fault occurs, no action is performed, if a fault occurs, the number of faults is accumulated, and when it is detected that faults occur in all of the N consecutive time slices, the line is disconnected through the trip device.
Compared with the prior art, the invention has the following beneficial effects:
1) the method solves the problem that the characteristic quantity and the characteristic threshold value of the traditional characteristic quantity method are difficult to select, does not need to artificially extract the characteristic quantity, does not need to set the characteristic threshold value, and is not influenced by the magnitude of the load current.
2) The method has the advantages that by means of the characteristic self-extraction capability of the convolutional neural network, the method is more applicable to various loads, the accuracy rate of arc fault detection is obviously improved, and the range of arc working conditions which can be detected by the arc detection device is expanded.
3) The method can detect whether the arc fault occurs or not, can detect the type of the fault load, is beneficial to related personnel to quickly lock the fault object, timely remove the fault load, reduce the probability of fire and protect the safety of life and property.
4) The method only takes the current signal as input, and is simpler and more convenient and more feasible compared with a multisource fusion arc fault detection technology adopting physical quantities such as voltage, current, even temperature, electromagnetic field and the like.
Drawings
FIG. 1 is a flow chart of a method implementation of an embodiment of the present invention.
FIG. 2 is a schematic diagram of an arc fault generation circuit in an embodiment of the present invention.
FIG. 3 is a schematic diagram of a convolutional neural network structure in an embodiment of the present invention.
FIG. 4 is a schematic diagram of a convolutional neural network training process in an embodiment of the present invention.
FIG. 5 is a table of over parameters for an arc fault detection model in an embodiment of the invention.
FIG. 6 is a table of the structure of an arc fault detection model in an embodiment of the invention.
Fig. 7 is a schematic structural diagram of the differential high-frequency current sensor according to the embodiment of the present invention (with the front cover removed).
Detailed Description
The invention is described in further detail below with reference to the figures and the embodiments.
The invention provides a series arc fault detection method based on a one-dimensional convolutional neural network, which comprises the following steps of:
1) acquiring current data of normal and fault states of resistive, inductive and nonlinear loads, and dividing the current data into five types of data: resistive or inductive load normal data, resistive load fault data, inductive load fault data, nonlinear load normal data, and nonlinear load fault data, constitute a data set used to train the convolutional neural network.
Electrical apparatus is diversified in the low voltage distribution system, and the current waveform that different types of load produced may differ greatly, for the convenience of detection, classifies the trouble: resistive/inductive load normal (tag 0), resistive load fault (tag 1), inductive load fault (tag 2), nonlinear load normal (tag 3) and nonlinear load fault (tag 4), for five types of tags. The normal states of the resistive load and the inductive load are classified into the same type because the current waveforms of the resistive load and the inductive load are sine waves in the normal operation state.
Since the probability of the arc fault occurring in the conventional line is very small, the collected current data contains a large amount of normal data and a small amount of fault data, which may cause the fault detection model to be prone to detect a normal state and to be difficult to detect a fault state. Therefore, the method for acquiring the current data of the loads in normal and fault states of different types comprises the following steps: each type of load is connected to an arc generator, an ac power source, and a switch, respectively, to form an arc fault generation circuit as shown in fig. 2. The current sensor installed in the circuit, the data acquisition card and the computer are used for acquiring current data in the circuit in real time, and the number of fault data is increased by artificially simulating arc discharge, so that current waveforms under the normal load and arc fault states are obtained.
In this embodiment, a differential high-frequency current sensor is adopted to collect a series arc fault signal, as shown in fig. 7, the differential high-frequency current sensor includes a casing 6, a magnetic ring 1 and a secondary winding 2 wound on the magnetic ring 1 are arranged in the casing 6, a primary conductor through hole 4 penetrating through the middle of the magnetic ring 1 is formed in the casing 6 to allow a differential primary current-carrying conductor to pass through, and an output terminal outlet groove 3 is formed in the casing 6 to allow an output terminal 5 of the secondary winding 2 to pass through. In this embodiment, the housing 6 is composed of a main housing and a front cover, and fig. 7 is a schematic structural view of the sensor with the front cover removed.
The differential primary current-carrying conductor is provided with two current-carrying conductors, the currents on the two current-carrying conductors are in equal and opposite directions, the two current-carrying conductors vertically penetrate through the sensor, and the eccentricity of the two current-carrying conductors influences the mutual inductance of the primary conductor and the secondary winding of the sensor. The primary conductor perforation 4 comprises two through holes arranged in parallel so as to simultaneously pass through a zero line and a live line when in use, namely two current-carrying conductors of the differential primary current-carrying conductor.
In this embodiment, the magnetic ring 1 is a high-frequency magnetic ring made of a high-frequency soft magnetic material meeting the bandwidth requirement, and the magnetic ring is magnetized by the measured current on the primary current-carrying conductor and alternates with the alternation of the measured current. The high-frequency soft magnetic material may be ferrite, a magnetic powder core, or the like. The secondary winding 2 is made of copper core enameled wire, the winding angle of the copper core enameled wire is between 0 degree and 360 degrees, namely the copper core enameled wire is not fully wound around the magnetic ring 1, the preferred value is 180 degrees, but the copper core enameled wire is not limited to the angle, and the copper core enameled wire is used for generating induced voltage along with alternating magnetic flux on the magnetic ring. And an easily-cured insulating material such as epoxy resin or thermoplastic material is filled between the magnetic ring 1 and the shell 6 so as to cast and mold the sensor. The output terminal wire outlet groove 3 is arranged below the shell 6, and the output terminal 5 penetrates out to be connected with the sampling resistor to output sampling signals. And a non-inductive or low-inductive sampling resistor is connected in parallel to the double output terminals 5 led out from the output terminal wire outlet groove 3 so as to reduce the influence of the self inductance of the sampling resistor on the output response of the sensor. Annular metal sheets can be arranged on two sides of the vertical section of the sensor to serve as shielding covers so as to avoid possible interference of external magnetic fields.
The differential high-frequency current sensor overcomes the problems of small response of the hollow Rogowski coil and saturation of the magnetic core Rogowski coil, so that the output response of the secondary winding is larger, and the magnetic core cannot be saturated. The sensor can also overcome the problem of fault signal shielding caused by a high-power load to a low-power fault branch. The sensor can be effectively used for collecting series fault arc signals of a low-voltage distribution network, a subsequent signal conditioning circuit (an integrator and a filter circuit) is not needed, the requirement on hardware is low, and the cost is effectively reduced.
In the embodiment, 500 groups of normal and fault waveforms are collected for each load, namely 1000 groups of waveforms are collected for each load, the sampling time T of each group of waveforms is 100ms, and the length L of the waveforms is 5970 points, namely the sampling frequency f is 59.7 kHz.
2) Constructing a convolutional neural network by taking current data as input and five types of labels of normal resistance or inductive load, resistive load fault, inductive load fault, normal nonlinear load and nonlinear load fault as output; and inputting the data in the data set into the constructed convolutional neural network for network training to obtain an arc fault detection model.
As shown in fig. 3, the convolutional neural network is a layered model with multiple layers stacked, and mainly includes an input layer, alternating convolutional layers and pooling layers, a full-link layer, and an output layer. The input layer preprocesses input data. The convolution layer and the pooling layer correspond to convolution operation and pooling operation respectively, nonlinear activation operation is usually arranged behind the convolution layer and the pooling layer, and through a series of operations, original data are mapped to a hidden layer feature space, and features of original signals are extracted layer by layer to play a role in feature extraction. The full-connection layer formalizes the target task into a target function, maps the learned characteristics to a sample type space, and the output layer outputs the predicted value of the convolutional neural network to the type of the input sample.
As shown in fig. 4, the training method of the convolutional neural network includes: dividing a data set into a training data set and a testing data set, inputting the training data set and the testing data set in batches, repeatedly updating the weights and the offsets of the convolution layer and the full connection layer of the network, and finally obtaining a group of parameter values which enable the loss value of the loss function to be minimum. In this embodiment, the hyper-parameters corresponding to the arc fault detection model are shown in fig. 5, and the established model structure is shown in fig. 6. The setting of the hyper-parameters is gradually established in the network training process, but is not the only solution, and other similar hyper-parameter settings are regarded as the extension of the invention. The selection of the network layer and the placement sequence of the network layer are not unique, and other similar network stacks are regarded as the extension of the invention.
3) Collecting a current waveform of a line to be detected with a specified length at a set time interval, inputting collected current data into a trained arc fault detection model, and calculating a class label corresponding to the current data so as to detect the line in real time; and after calculating the type label corresponding to the current data, judging whether an arc fault occurs, if the arc fault does not occur, not acting, if the arc fault occurs, accumulating the number of the faults, and when detecting that faults occur in all the N continuous time slices, disconnecting the line through a tripping device. Wherein, N can be adjusted according to actual needs.
In order to improve the arc fault detection effect, the invention provides a series arc fault detection method based on a one-dimensional convolutional neural network. The method does not need to manually select characteristic quantity and set a characteristic threshold value, reduces the manual intervention process, automatically extracts various arc fault characteristics hidden behind arc current data through a network, has stronger applicability and realizes intelligent identification of series arc faults.
The above are preferred embodiments of the present invention, and all changes made according to the technical scheme of the present invention that produce functional effects do not exceed the scope of the technical scheme of the present invention belong to the protection scope of the present invention.
Claims (5)
1. A series arc fault detection method based on a one-dimensional convolutional neural network is characterized by comprising the following steps:
1) acquiring current data of normal and fault states of resistive, inductive and nonlinear loads, and dividing the current data into five types of data: resistance or inductive load normal data, resistance load fault data, inductive load fault data, nonlinear load normal data and nonlinear load fault data form a data set for training the convolutional neural network;
2) constructing a convolutional neural network by taking current data as input and five types of labels of normal resistance or inductive load, resistive load fault, inductive load fault, normal nonlinear load and nonlinear load fault as output; inputting the data in the data set into the constructed convolutional neural network for network training to obtain an arc fault detection model;
3) the method comprises the steps of collecting a current waveform of a line to be detected with a specified length at a set time interval, inputting collected current data into a trained arc fault detection model, calculating a type label corresponding to the current data, detecting the line in real time, and disconnecting the line through a tripping device when faults occur in N time slices.
2. The method for detecting the series arc fault based on the one-dimensional convolutional neural network as claimed in claim 1, wherein in the step 1, the method for acquiring the current data under the normal and fault states of different types of loads comprises the following steps: each type of load is respectively connected with an arc generator, an alternating current power supply and a switch to form an arc fault generating circuit, current data in the circuit are collected in real time through a current sensor installed in the circuit, a data acquisition card and a computer, the number of fault data is increased through artificial simulated arc discharge, and current waveforms of the type of load in normal and arc fault states are obtained.
3. The method according to claim 1, wherein in step 2, the convolutional neural network is mainly composed of an input layer, alternating convolutional layers and pooling layers, a fully-connected layer and an output layer, the input layer preprocesses input data, the convolutional layers and pooling layers correspond to convolutional operation and pooling operation, respectively, the fully-connected layer formalizes a target task as an objective function, maps learned features to a sample type space, and the output layer outputs a predicted value of the convolutional neural network for the type of an input sample.
4. The method for detecting the series arc fault based on the one-dimensional convolutional neural network as claimed in claim 3, wherein the training method of the convolutional neural network is as follows: dividing a data set into a training data set and a testing data set, inputting the training data set and the testing data set in batches, repeatedly updating the weights and the offsets of the convolution layer and the full connection layer of the network, and finally obtaining a group of parameter values which enable the loss value of the loss function to be minimum.
5. The method according to claim 1, wherein in step 3, after the class label corresponding to the current data is calculated, it is determined whether an arc fault has occurred, if no arc fault has occurred, no action is performed, if a fault has occurred, the number of faults is accumulated, and when a fault has occurred in all of N consecutive time slices, the line is disconnected by a tripping device.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010302286.4A CN111458599A (en) | 2020-04-16 | 2020-04-16 | Series arc fault detection method based on one-dimensional convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010302286.4A CN111458599A (en) | 2020-04-16 | 2020-04-16 | Series arc fault detection method based on one-dimensional convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111458599A true CN111458599A (en) | 2020-07-28 |
Family
ID=71684599
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010302286.4A Pending CN111458599A (en) | 2020-04-16 | 2020-04-16 | Series arc fault detection method based on one-dimensional convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111458599A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112345895A (en) * | 2020-10-28 | 2021-02-09 | 宁波立新科技股份有限公司 | Series arc fault detection method based on discriminant analysis strategy |
| CN112748318A (en) * | 2021-01-22 | 2021-05-04 | 中国矿业大学 | Series fault arc detection method based on improved PSO-BP neural network |
| CN112904226A (en) * | 2021-01-19 | 2021-06-04 | 国网河北省电力有限公司 | Method for rapidly judging short-circuit fault of high-voltage bus based on induced electricity |
| CN113030789A (en) * | 2021-04-12 | 2021-06-25 | 辽宁工程技术大学 | Series arc fault diagnosis and line selection method based on convolutional neural network |
| CN113933636A (en) * | 2021-10-29 | 2022-01-14 | 国网湖北省电力有限公司电力科学研究院 | Power distribution network fault test system based on arc generator |
| CN114113926A (en) * | 2021-11-01 | 2022-03-01 | 广东电网有限责任公司广州供电局 | Series arc fault diagnosis method and device, computer equipment and storage medium |
| CN114999095A (en) * | 2022-05-23 | 2022-09-02 | 山东建筑大学 | Building electrical fire monitoring method and system based on time and space fusion |
| WO2024018119A1 (en) * | 2022-07-19 | 2024-01-25 | Hager-Electro Sas | Method, device and equipment for detecting an electric arc fault |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020118022A1 (en) * | 2001-02-27 | 2002-08-29 | Dring Jonathan Samuel | Arc detection |
| CN101154800A (en) * | 2007-09-29 | 2008-04-02 | 吴为麟 | Method for detecting trouble electric arc and its protecting equipment |
| CN108061832A (en) * | 2017-12-04 | 2018-05-22 | 辽宁工程技术大学 | Tandem type fault electric arc emulation mode based on neutral net black-box model |
| CN108090658A (en) * | 2017-12-06 | 2018-05-29 | 河北工业大学 | Arc fault diagnostic method based on time domain charactreristic parameter fusion |
| CN108627745A (en) * | 2018-07-02 | 2018-10-09 | 华立科技股份有限公司 | Realize the intelligent electric energy meter and its implementation of fault electric arc detecting function |
| 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 |
| CN110320452A (en) * | 2019-06-21 | 2019-10-11 | 河南理工大学 | A kind of series fault arc detection method |
| CN110765858A (en) * | 2019-09-14 | 2020-02-07 | 杭州拓深科技有限公司 | Non-invasive fault arc monitoring method based on convolutional neural network |
-
2020
- 2020-04-16 CN CN202010302286.4A patent/CN111458599A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020118022A1 (en) * | 2001-02-27 | 2002-08-29 | Dring Jonathan Samuel | Arc detection |
| CN101154800A (en) * | 2007-09-29 | 2008-04-02 | 吴为麟 | Method for detecting trouble electric arc and its protecting equipment |
| CN108061832A (en) * | 2017-12-04 | 2018-05-22 | 辽宁工程技术大学 | Tandem type fault electric arc emulation mode based on neutral net black-box model |
| CN108090658A (en) * | 2017-12-06 | 2018-05-29 | 河北工业大学 | Arc fault diagnostic method based on time domain charactreristic parameter fusion |
| 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 |
| CN108627745A (en) * | 2018-07-02 | 2018-10-09 | 华立科技股份有限公司 | Realize the intelligent electric energy meter and its implementation of fault electric arc detecting function |
| CN110320452A (en) * | 2019-06-21 | 2019-10-11 | 河南理工大学 | A kind of series fault arc detection method |
| CN110765858A (en) * | 2019-09-14 | 2020-02-07 | 杭州拓深科技有限公司 | Non-invasive fault arc monitoring method based on convolutional neural network |
Non-Patent Citations (2)
| Title |
|---|
| 刘德军 等: "基于峰值指标和脉冲指标的串联电弧故障检测研究", 《电器与能效管理技术》 * |
| 周新城 等: "基于1D-CNN的串联故障电弧检测", 《消防科学与技术》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112345895A (en) * | 2020-10-28 | 2021-02-09 | 宁波立新科技股份有限公司 | Series arc fault detection method based on discriminant analysis strategy |
| CN112345895B (en) * | 2020-10-28 | 2023-02-10 | 宁波立新科技股份有限公司 | Series arc fault detection method based on discriminant analysis strategy |
| CN112904226A (en) * | 2021-01-19 | 2021-06-04 | 国网河北省电力有限公司 | Method for rapidly judging short-circuit fault of high-voltage bus based on induced electricity |
| CN112748318A (en) * | 2021-01-22 | 2021-05-04 | 中国矿业大学 | Series fault arc detection method based on improved PSO-BP neural network |
| CN113030789A (en) * | 2021-04-12 | 2021-06-25 | 辽宁工程技术大学 | Series arc fault diagnosis and line selection method based on convolutional neural network |
| CN113933636A (en) * | 2021-10-29 | 2022-01-14 | 国网湖北省电力有限公司电力科学研究院 | Power distribution network fault test system based on arc generator |
| CN114113926A (en) * | 2021-11-01 | 2022-03-01 | 广东电网有限责任公司广州供电局 | Series arc fault diagnosis method and device, computer equipment and storage medium |
| CN114999095A (en) * | 2022-05-23 | 2022-09-02 | 山东建筑大学 | Building electrical fire monitoring method and system based on time and space fusion |
| CN114999095B (en) * | 2022-05-23 | 2023-11-14 | 山东建筑大学 | Building electrical fire monitoring method and system based on time and space fusion |
| WO2024018119A1 (en) * | 2022-07-19 | 2024-01-25 | Hager-Electro Sas | Method, device and equipment for detecting an electric arc fault |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111458599A (en) | Series arc fault detection method based on one-dimensional convolutional neural network | |
| Cui et al. | Hilbert-transform-based transient/intermittent earth fault detection in noneffectively grounded distribution systems | |
| CA2554962C (en) | High impedance fault detection | |
| Xiong et al. | Incipient fault identification in power distribution systems via human-level concept learning | |
| CN105425118B (en) | A kind of Multi-information acquisition fault arc detection method and device | |
| Wei et al. | High impedance arc fault detection based on the harmonic randomness and waveform distortion in the distribution system | |
| CN106707094B (en) | The classifying identification method of low pressure power and distribution supply cable arc fault | |
| Wei et al. | Distortion-based detection of high impedance fault in distribution systems | |
| Gu et al. | High impedance fault detection in overhead distribution feeders using a DSP-based feeder terminal unit | |
| Qin et al. | Research on distribution network fault recognition method based on time-frequency characteristics of fault waveforms | |
| CN205941736U (en) | Trouble arc detection device | |
| Jiang et al. | A coupling method for identifying arc faults based on short-observation-window SVDR | |
| Jiang et al. | Machine learning approach to detect arc faults based on regular coupling features | |
| CN107870281B (en) | Method and device for detecting electric leakage and fault arc | |
| Grassetti et al. | Low cost arc fault detection in aerospace applications | |
| Yang et al. | Multi-Frequency bands based Pole-to-Ground fault detection method for MMC-Based radial DC distribution systems | |
| Jiang et al. | AC series arc fault detection based on RLC arc model and convolutional neural network | |
| Paydarnia et al. | Improved structure of PNN using PCA in transformer fault diagnostic | |
| Cui et al. | Modeling study for high impedance fault detection in MV distribution system | |
| Vahidi et al. | A novel approach to adaptive single phase autoreclosure scheme for EHV power transmission lines based on learning error function of ADALINE | |
| CN113917294B (en) | Intelligent self-adaptive arc detection method based on wavelet decomposition and application device thereof | |
| Sirojan et al. | High Impedance Fault Detection by Convolutional Deep Neural Network | |
| de Souza et al. | High impedance fault detection in distribution systems: An approach based on fourier transform and artificial neural networks | |
| Hamatwi et al. | Comparative analysis of high impedance fault detection techniques on distribution networks | |
| Zhou et al. | Method of cable incipient faults detection and identification based on wavelet transform and gray correlation analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200728 |
|
| RJ01 | Rejection of invention patent application after publication |