[go: up one dir, main page]

CN117517897A - Arc fault detection method based on edge calculation - Google Patents

Arc fault detection method based on edge calculation Download PDF

Info

Publication number
CN117517897A
CN117517897A CN202311524316.6A CN202311524316A CN117517897A CN 117517897 A CN117517897 A CN 117517897A CN 202311524316 A CN202311524316 A CN 202311524316A CN 117517897 A CN117517897 A CN 117517897A
Authority
CN
China
Prior art keywords
current
arc
data
detection
fault
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
CN202311524316.6A
Other languages
Chinese (zh)
Inventor
韩庆轩
杨昌
崔建平
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.)
Tianjin Aviation Mechanical and Electrical Co Ltd
Original Assignee
Tianjin Aviation Mechanical and Electrical 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 Tianjin Aviation Mechanical and Electrical Co Ltd filed Critical Tianjin Aviation Mechanical and Electrical Co Ltd
Priority to CN202311524316.6A priority Critical patent/CN117517897A/en
Publication of CN117517897A publication Critical patent/CN117517897A/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
    • BPERFORMING OPERATIONS; TRANSPORTING
    • B64AIRCRAFT; AVIATION; COSMONAUTICS
    • B64FGROUND OR AIRCRAFT-CARRIER-DECK INSTALLATIONS SPECIALLY ADAPTED FOR USE IN CONNECTION WITH AIRCRAFT; DESIGNING, MANUFACTURING, ASSEMBLING, CLEANING, MAINTAINING OR REPAIRING AIRCRAFT, NOT OTHERWISE PROVIDED FOR; HANDLING, TRANSPORTING, TESTING OR INSPECTING AIRCRAFT COMPONENTS, NOT OTHERWISE PROVIDED FOR
    • B64F5/00Designing, manufacturing, assembling, cleaning, maintaining or repairing aircraft, not otherwise provided for; Handling, transporting, testing or inspecting aircraft components, not otherwise provided for
    • B64F5/60Testing or inspecting aircraft components or systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Manufacturing & Machinery (AREA)
  • Transportation (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Testing Of Short-Circuits, Discontinuities, Leakage, Or Incorrect Line Connections (AREA)

Abstract

The invention belongs to the field of fault arc detection of aviation power distribution systems, and relates to an arc fault detection method based on edge calculation, which comprises the following steps: the electric arc detection terminal collects current information in the circuit in real time, judges whether various current characteristics exceed respective current characteristic thresholds for each current period, and if so, the current period is an electric arc period. And judging whether the accumulated arc period in one moving time window exceeds a set threshold value, if so, generating an arc fault, and reporting an arc fault alarm and current data in one moving time window to an arc detection far end. And the arc detection far end performs arc identification, if the arc fault is identified, the arc fault alarm information is reported to the upper computer, and an arc protection command is issued to the arc detection terminal. And the arc detection terminal performs arc fault protection. The technical scheme of the invention reduces the interference of nonlinear loads and the like and reduces the false alarm rate while having higher arc detection rate.

Description

Arc fault detection method based on edge calculation
Technical Field
The invention belongs to the technical field of aviation power supply and distribution systems and electric safety detection, and particularly relates to an arc fault detection method based on edge calculation.
Background
The arc is a gas free discharge phenomenon in which gas has good insulation in a normal state, but when a sufficiently large electric field is applied to both ends of a gas gap, the gas free discharge occurs, and an arc is generated. The arc is characterized by high energy and high temperature, which can reach thousands of degrees celsius in a few microseconds. The electric arc generates a large amount of heat, so that the aging and damage of the insulation layers of surrounding cables are accelerated, the contact and the insulation materials are burnt, and the safety of circuits and equipment is endangered; in severe cases, cables and equipment of the electrical system may ignite, causing fires and explosions.
The electric arc brings great threat to the safe operation of the aircraft, and students at home and abroad conduct a great deal of research on an electric arc fault detection method, for example, in the patent 'an aircraft fault electric arc detection method', db3 wavelet which adopts an energy ratio method to select the characteristics of the aircraft fault electric arc is used as a basic wavelet of a decomposition signal, the acquired current signal is subjected to wavelet decomposition, d2 signal is extracted, and the standard deviation of the cycle of the wavelet high-frequency component is used as the identification characteristics of the aviation fault electric arc, so that electric arc identification is conducted.
In the existing arc fault identification, the actual influence of complex load types and different current sizes on the arc faults is often ignored, so that the arc fault identification method is only suitable for arc types of individual types, the interference of complex loads cannot be eliminated, the existing aircraft arc fault detection terminal has limited calculation, the existing arc fault detection method is low in detection accuracy, and the detection false alarm rate is high.
Disclosure of Invention
The invention solves the technical problems that in aviation arc fault detection, the types of nonlinear loads are more, the types of arcs are complex, fault identification is seriously interfered, the adopted arc detection algorithm has limited detection capability, and the false alarm rate is high, and provides an arc fault detection method based on edge calculation.
The aim of the invention can be achieved by the following technical scheme, which comprises the following steps:
the invention provides an arc fault arc detection method based on edge calculation, which comprises the following steps of: an arc detection distal end and a plurality of arc detection terminals, said method comprising the steps of:
s1, any arc detection terminal collects current in a main circuit of a power supply and distribution network in real time at a specific sampling frequency, and performs validity check, data preprocessing and current frequency detection on collected current data;
s2, the arc detection terminal determines an ADC sampling rate according to a current frequency detection result, performs ADC sampling according to the ADC sampling rate, and performs various current characteristic extraction on each current period;
s3, the arc detection terminal judges whether the current characteristic of each current period exceeds the respective characteristic threshold value, if so, the current period is considered as an arc period, and the accumulated quantity of the arc period is added by 1;
s4, in a preset moving time window, the arc detection terminal judges whether the arc period number exceeds a set first threshold value of the arc period accumulation amount, if yes, the arc fault is considered to be detected, the arc period accumulation amount is cleared, and otherwise, the detection of the next current period in the moving time window is continued according to S3;
s5, the arc detection terminal detects that an arc fault exists in the circuit, and reports an arc fault alarm and all current data in the moving time window to an arc detection far end;
s6, the arc detection far end receives arc fault alarm information reported by the arc detection terminal and all current data in the movement detection time window;
s7, the electric arc detection far end performs feature extraction on all current data in the moving detection time window according to the same current period as the electric arc detection terminal, and extracts various current features;
s8, the arc detection far end takes the extracted various current characteristics as input, and inputs the extracted various current characteristics into an arc fault neural network identifier to identify an arc fault;
s9, if the arc fault neural network identifier judges that the section of current data has an arc fault, the arc detection remote end reports arc fault alarm information to the upper computer, stores the section of current data and sends an arc protection command to the arc detection terminal; otherwise, if the arc fault neural network identifier identifies that the segment of data does not have arc faults, the arc detection remote end reports arc fault false alarm information to the upper computer and stores the segment of current data;
s10, the arc detection terminal receives an arc protection command issued by the arc detection remote end, and performs arc protection.
Further, the arc detection terminal data is preprocessed to perform data validity verification, specifically: comparing the acquired current signal with a preset filtering threshold value, if the acquired current signal is larger than the filtering threshold value, considering the acquired current data as effective data, and transferring to the next step; and if the current data is lower than the filtering threshold value, the acquired current data is considered to be invalid, and current sampling is repeated.
Further, the current frequency detection of the arc detection terminal determines whether the frequency of the current is direct current or alternating current according to the collected current data, selects the ADC sampling rate based on the current frequency, and selects the ADC sampling rate based on the frequency of 400HZ if the current is direct current.
Further, the current characteristic extraction performed by the arc detection terminal per current period specifically includes: carrying out multiple current characteristic extraction on the cache current data sampled by each ADC, calculating a current characteristic value of each current period, wherein the calculated characteristic value is numbered as current characteristic 1 and current characteristic 2, current characteristics n and n are positive integers larger than 2, and the current characteristics are as follows: the current average value, the current peak value, the current standard deviation, the current two-period average amplitude difference, the current uniformity, the current change rate average value, the current bias state, the current cosine similarity, the current front n times harmonic wave and/or the pearson correlation coefficient of each frequency component of the current front window and the current rear window; the arc detection terminal judges whether each current period is an arc period or not, specifically: and the arc detection terminal judges whether each current characteristic i exceeds the current characteristic i identification threshold of the terminal, if all the current characteristics i exceed the current characteristic i identification threshold of the terminal, wherein i=1, 2,..n, the current period is judged to be the arc period, and the arc period accumulation amount is added by 1.
Further, the arc detection remote end calculates current characteristics 1,2, and m of each current period according to the current period, wherein m is a positive integer greater than or equal to 4, and the calculated current characteristics 1,2, and m of each current period are led into a trained neural network arc fault identifier to perform arc fault identification.
Further, the current characteristics are selected by respectively extracting current characteristics from arc current data, normal current data, interference current data and load current data to obtain arc current characteristics, normal current characteristics, interference current characteristics and load current characteristics, analyzing the obtained characteristics in a characteristic value range, determining the maximum value and the minimum value of the characteristics to obtain characteristic value ranges [ characteristic minimum value, characteristic maximum value ], comparing the arc current characteristic value range with the normal current characteristic value range, the interference current characteristic value range and the load current characteristic value range, and if the coincidence degree of the arc current characteristic value range and other characteristic value ranges is smaller than a preset characteristic value range coincidence threshold, identifying an arc fault by the current characteristics; if the arc current characteristic value range and other value ranges do not overlap, the characteristic value range overlap ratio is zero, and the current characteristic is preferably used for arc identification.
Further, the window length of the arc detection moving window is determined according to an arc detection standard and a current system;
the arc detection moving window length and the arc current period have the following relationship:
arc detection moving window length = current period number that arc detection needs to identify;
the current period is determined by the voltage frequency f, the current period = 1/voltage frequency f; in the case of a DC power supply, the current period can be selected as required, and can be selected to be 2.5ms.
Further, the neural network arc fault identifier is a pre-trained BP neural network:
(1) Setting nodes of an input layer, a hidden layer and an output layer of the neural network, and setting learning rate and an activation function;
(2) And performing feature extraction through pre-acquired arc data, normal data, load data and interference data, respectively extracting corresponding features of four types of data of the feature i to obtain corresponding arc current feature i, normal current feature i, load current feature i and interference current feature i, and performing neural network training by taking an array formed by the four types of features as a training set, wherein the output result of the neural network is arc fault and non-arc fault.
By adopting the technical scheme, the invention has the following advantages:
according to the scheme, the arc detection terminal and the arc detection far end are combined, and the high-sensitivity rapid detection of the terminal and the high-accuracy arc detection secondary rechecking of the far end are combined, so that the accuracy of the arc detection is improved, and the false alarm rate is reduced. The arc detection far end adopts an intelligent algorithm, and arc characteristic threshold value setting is not needed, so that the restriction of a single arc characteristic threshold value on the arc detection rate is avoided, the false alarm rate is reduced, and the detection rate is improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only some embodiments of the invention, and that it is within the scope of the invention to obtain other drawings from these drawings without inventive faculty for a person skilled in the art.
FIG. 1 is a flow chart of an arc detection method identification based on edge-based computation;
FIG. 2 is a flowchart of an identification of an arc detection terminal based on edge-based calculations;
FIG. 3 is a flowchart of an arc detection method identification based on edge calculation according to an embodiment of the present invention;
fig. 4 is a flowchart of the identification of an arc detection terminal based on edge calculation according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to specific embodiments of the present invention and corresponding drawings. It will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The basic principle of the technical scheme of the invention is shown in fig. 1 and 2.
Firstly, arc current data is acquired through an arc detection terminal conditioning circuit, and the acquired arc current data is subjected to data preprocessing, so that weak signals are eliminated, and the noise disturbance influence when no current exists in a line is eliminated. Frequency detection is carried out, whether alternating current or direct current is carried out when current in the main circuit is determined, and if the alternating current or the direct current is the direct current, data sampling is carried out according to a preset sampling rate; if the alternating current is the alternating current, corresponding periodic sampling is carried out according to the frequency of the alternating current. If the arc characteristics are selected to have the frequency domain characteristics, the corresponding sampling frequency needs to contain the characteristic frequency band of the arc frequency domain.
And after data sampling, extracting current characteristics, and calculating each current characteristic of each sampling period. Judging whether all the current characteristics exceed respective current characteristic thresholds, if all the current characteristics exceed respective current characteristic thresholds, judging the current period as an arc period, and carrying out arc period accumulation quantity +1.
If the accumulated arc period amount in a certain moving time window is larger than a preset threshold value, judging that an arc occurs in the line, and alarming, wherein the arc detection terminal uploads arc fault alarming information, stored current data and current characteristics in one moving time window to an arc detection far end; otherwise, judging that no arc fault occurs, and continuing to perform arc detection.
After receiving alarm information and current data transmitted by an arc detection terminal, an arc detection far end processes the current data, extracts current characteristics according to a current period, inputs the extracted current characteristics as input to a trained arc fault neural network identifier for arc identification, reports the arc fault alarm information to an upper computer by the arc detection far end if the current data is identified as arc data, stores arc fault current data, and generates an arc protection command to the arc protection terminal, and the arc detection terminal receives the arc detection far end command for arc fault protection. Otherwise, the false detection information and the current data are recorded, and the false detection early warning information is reported to the upper computer.
Step one: the arc detection terminal collects and processes circuit current data in real time and detects whether an arc fault exists in a circuit or not, as shown in fig. 4.
(1) The arc detection terminal collects current signals in a main circuit of the power supply and distribution network through a sampling circuit (such as a low-pass filtering sampling circuit) in a specific sampling frequency and uninterrupted sampling mode.
(2) And preprocessing the data of the arc detection terminal, and checking the validity of the data. Comparing the collected current signal with a preset filtering threshold value, and if the current signal is larger than the filtering threshold value, considering the collected current data as effective data, and transferring to the next step. And if the current data is lower than the filtering threshold value, the acquired current data is considered to be invalid, and current sampling is repeated. This process is mainly for filtering the effects of unwanted small signal noise.
(3) Arc detection terminal current frequency detection, the frequency of the current is determined by the collected current data, whether the current is direct current or alternating current (such as 400Hz, 360Hz, 800Hz or variable frequency). The ADC sampling rate is selected based on the current frequency, and if dc, the ADC sampling rate is selected based on a frequency of 400HZ.
(4) And the arc detection terminal performs periodic ADC sampling according to the calculated sampling rate and caches the current data.
(5) The arc detection terminal performs current feature extraction per current period. And (3) carrying out various current characteristic extraction on the buffer current data sampled by each ADC, and calculating a current characteristic value of each current period, wherein the calculated characteristic value is numbered as current characteristic 1 and current characteristic 2.
(6) The arc detection terminal judges whether each current period is an arc period. The arc detection terminal judges whether each current characteristic i exceeds the current characteristic i identification threshold of the arc detection terminal itself. If all current characteristics i exceed respective current characteristic i identification thresholds (where i=1, 2,..n), then the current period is determined to be an arc period, and an arc period accumulation amount +1 is performed.
(7) In an arc detection moving window, the arc detection terminal judges whether the arc period accumulation amount exceeds a preset first threshold value of the arc period accumulation amount, if so, the arc detection terminal considers that an arc fault is generated, carries out fault alarm, and clears the arc period accumulation amount. If not, repeating the steps (4) - (6).
Step two: the arc detection terminal detects the occurrence of arc faults and transmits stored arc current data, current characteristics and arc warning information in an arc detection moving window to the arc fault detection remote equipment through a data bus.
Step three: and the arc detection far end receives and stores arc alarm information reported by the arc detection terminal, arc current data and arc current characteristics in a window, and performs secondary detection on the arc current data.
(1) The arc detection remote end divides arc current data into a plurality of current periods according to the current periods, and calculates current characteristics 1,2 and m (m is a positive integer greater than or equal to 4) of each current period according to the current periods.
(2) And (3) leading the calculated current cycle current characteristics 1, the current characteristics 2, the current characteristics m into a trained neural network arc fault identifier to identify arc faults.
Step four: if the arc detection remote end judges that the reported current data detects an arc fault, the arc detection remote end considers that the arc fault exists in the main circuit, and issues an arc fault protection command to an arc fault detection terminal, and the arc fault detection terminal protects the main circuit.
Step five: if the arc detection remote end judges that the reported current data does not detect the arc fault, the detection result is recorded, and the suspected false alarm information of the arc detection remote end is reported to the upper computer system.
In step one, the arc current characteristics herein refer to current characteristics that have a distinct distinction between arc faults and non-arc faults. The current characteristics are selected by respectively extracting current characteristics from arc current data, normal current data, interference current data and load current data to obtain arc current characteristics, normal current characteristics, interference current characteristics and load current characteristics, analyzing the obtained characteristics in characteristic value fields, determining the maximum value and the minimum value of the respective characteristics to obtain characteristic value fields [ characteristic minimum value, characteristic maximum value ], comparing the arc current characteristic value fields with normal current characteristic value fields, interference current characteristic value fields and load current characteristic value fields, and if the coincidence ratio of the arc current characteristic value fields and other characteristic value fields is smaller than a preset characteristic value field coincidence threshold, the current characteristics can be used for arc fault identification. If the characteristic value range of the arc current is not overlapped with other value ranges, the overlapping degree of the characteristic value range is zero, and the current characteristic is used for the arc identification effect to be optimal.
In the first step, the current characteristic may be any characteristic, such as a mean value, a peak-to-peak value, a standard deviation, a two-period average amplitude difference, a current uniformity, a current change rate mean value, a bias state, a cosine similarity, a first n-order harmonic sum, and pearson correlation coefficients of frequency components of front and rear windows.
In the first step, the arc detection terminal may determine whether the current characteristic i exceeds the current characteristic i threshold value when arc fault recognition is performed, and if so, consider the current period as an arc period.
In the first step, the current characteristic threshold of the arc detection terminal is set to detect as many arc faults as possible, and the reduction of the false detection rate can be further verified through the arc detection far end.
In step one, the arc detection moving window may be set by: the window length of the arc detection moving window is determined according to the arc detection standard and the current system. The arc detection moving window length and the arc current period have the following relationship:
arc detection moving window length = current period number of cycles arc detection needs to identify.
The current period is determined by the voltage frequency f, current period=1/voltage frequency f. As for 230VAC/800Hz, the current period=1.25 ms. In the case of a DC power supply, the current period can be selected as required, and is usually 2.5ms.
If the single current period is 2.5ms and the arc detection time is required to be 100ms for the 115VAC/400HZ voltage system, the arc detection moving window length is selected to be 100ms, and 40 current periods are included.
In step one, it may be determined whether the current period is an arc period by determining whether the arc characteristic or characteristics exceeds an arc characteristic threshold. Whether the current period is an arc period may also be determined synthetically by determining whether the arc characteristic or characteristics exceed an arc characteristic threshold, and determining whether the arc characteristic or characteristics exceed a combination of certain thresholds.
In the first step, the preset first threshold value of the arc period accumulation amount may be determined according to an industry arc determination standard, or may be determined according to a specific protection requirement, for example, the arc detection moving window may be set to 100ms, or may be other reasonable duration.
In step three, the neural network arc fault identifier is a pre-trained BP neural network:
(1) Setting the input layer, hidden layer and output layer nodes of the neural network, and setting learning rate, activation function and other parameters.
(2) And performing feature extraction through pre-acquired arc data, normal data, load data and interference data, respectively extracting corresponding features of four types of data of a feature i (i is a positive integer greater than or equal to 4), obtaining corresponding arc current features i, normal current features i, load current features i and interference current features i, performing neural network training by taking an array formed by the four types of features as a training set, and outputting the results of the neural network as arc faults and non-arc faults.
To verify the feasibility of the present invention, the specific example provided by the present invention is to use arc data collected under the power supply of 115v 400hz ac power as an input signal, and process the arc data according to the identification flow of fig. 3.
The first step, arc current data are collected through low-pass filtering, high-frequency noise signals are filtered through low-pass filtering, and the influence of high-frequency interference signals on arc characteristic signals is discharged. Since the arc fault characteristic frequency band is concentrated in the range of tens of KHz to 50KHz, the cut-off frequency of 50KHz low-pass filtering can be selected, and other cut-off frequencies can be selected.
And secondly, data preprocessing is carried out, the collected current signals are compared with a filtering threshold value 1, and when the current data is larger than the filtering threshold value 1, the sampling filtering super-threshold value count accumulation amount Ncount is accumulated. And in one sampling period, if the filtering super-threshold counting accumulation amount Ncount exceeds the filtering threshold 2, the acquired current data is considered to be effective data and data caching is performed, otherwise, the periodic current data is considered to be invalid data, the periodic data is removed, and the next periodic current data is acquired again.
And thirdly, detecting the frequency of the current signal, if the current signal can be subjected to zero crossing detection, judging whether the current is alternating current or direct current, if the current is alternating current, further judging the frequency of the alternating current, such as aviation 115V 400Hz.
Fourth, setting ADC sampling rate based on current frequency, wherein the principle is that the sampling rate setting can periodically and completely collect the data of current period. If the frequency is 400Hz and the sampling rate is 80K, one current period is 2.5ms, namely 200 times of sampling is carried out every 2.5ms, and data of one complete current period can be acquired.
And fifthly, sampling the periodic ADC according to the calculated period and storing data. The method can be used for sampling and calculating simultaneously, and the calculation result is stored; or can be stored and calculated.
A sixth step of starting a moving window, which may be determined according to arc criteria or according to other criteria, where the moving window is given 100ms, each current period being 2.5ms, i.e. one arc detection moving window being 40 current periods.
And seventhly, calculating the characteristics of each current period, and calculating the current uniformity and the frequency band amplitude mean value of each current period.
Eighth, arc determination is performed for each current cycle. Firstly judging whether the current uniformity exceeds a current uniformity threshold, and if so, considering that the first condition is satisfied.
And ninth, if the first condition is met, judging whether the current frequency band amplitude average value exceeds a current frequency band amplitude average value threshold value, and if the current frequency band amplitude average value exceeds the current frequency band amplitude average value threshold value, judging that the second condition is met.
And tenth, if the first condition and the second condition are satisfied at the same time, the current period is considered as an arc period, and the arc period accumulation amount +1 is performed.
And eleventh step, judging whether the arc period accumulated quantity exceeds a preset arc judgment threshold value, if so, judging that an arc occurs, giving out fault alarm, resetting the arc period accumulated quantity, and reporting arc fault alarm information, current data in a moving window and current characteristics to an arc detection far end.
If not, further judging whether the current mobile recognition window is finished, and if so, starting the next mobile recognition window; if the current movement identification window is not finished, the characteristic calculation and the judgment of the next current period are continued.
Twelfth, the arc detection far end receives arc fault information and current data reported by an arc detection terminal, the arc current data is extracted according to a period of 2.5ms, and current peak-to-peak values, current standard deviation, current change rate mean value, alternating current component effective value, first 20 harmonic sums, high-low frequency difference mean value and frequency spectrum standard deviation of all the data are extracted.
And thirteenth, the electric arc detection far end takes the extracted current peak-to-peak value, current standard deviation, current change rate mean value, alternating current component effective value, the first 20 times of harmonic sum, high-low frequency difference mean value and frequency spectrum standard deviation as input, and inputs the input into the trained electric arc fault neural network identifier for electric arc fault identification.
And fourteenth step, if the arc detection far end identifies an arc fault, the arc detection far end reports arc fault information to the upper computer, stores the arc fault information and current data, and generates an arc protection command to the arc detection terminal. Otherwise, the arc detection remote end reports the arc false detection information to the upper computer and stores the false detection information and current data.
Fifteenth, the arc detection terminal receives an arc fault protection command issued by the arc detection remote end to perform arc protection.
The technical scheme of the invention provides an arc fault detection method based on edge calculation, which comprises the following steps: the arc detection terminal collects current information in a circuit in real time at a specific frequency, performs current data preprocessing and current frequency detection, determines a current period according to the detected current frequency, performs ADC period sampling, extracts various current characteristics for each current period, judges whether the various current characteristics exceed respective current characteristic thresholds, considers the current period as an arc period if the current characteristics exceed respective current characteristic thresholds, and performs arc period accumulation +1. After the arc detection terminal judges the arc period of each current period, judging whether the accumulated amount of the arc period in one moving time window exceeds a set threshold value, if so, considering that an arc fault occurs, and reporting an arc fault alarm and current data in one moving time window to an arc detection far end. The arc detection remote end receives arc fault alarm information and current data of the arc detection terminal, extracts various current characteristics according to a current period, inputs the extracted current characteristics into the arc fault neural network identifier for arc identification, reports the arc fault alarm information to the upper computer if the arc fault exists in the identified current data, stores the arc fault alarm information and the current data to the local, and sends an arc protection command to the arc detection terminal. And the arc detection terminal receives an arc protection command issued by the arc detection remote end and performs arc fault protection. The fault arc identification method can be applied to fault arc online detection of various types of arcs and loads, and has high arc detection rate, meanwhile, the interference of nonlinear loads and the like is reduced, and the false alarm rate is reduced due to the adoption of multi-feature fusion detection and intelligent algorithm.

Claims (8)

1.一种基于边缘计算的电弧故障电弧检测方法,其特征在于,供配电网络包含:一个电弧检测远端和多个电弧检测终端,所述的方法包括如下步骤:1. An arc fault arc detection method based on edge computing, characterized in that the power supply and distribution network includes: an arc detection remote end and multiple arc detection terminals. The method includes the following steps: S1,任意一个电弧检测终端以特定采样频率实时采集供配电网络主电路中的电流,并对采集的电流数据进行有效性检验、数据预处理和电流频率检测;S1, any arc detection terminal collects the current in the main circuit of the power supply and distribution network in real time at a specific sampling frequency, and performs validity verification, data preprocessing and current frequency detection on the collected current data; S2,所述电弧检测终端按照电流频率检测的结果确定ADC采样率,按照ADC采样率进行ADC采样,并对每个电流周期进行多种电流特征提取;S2, the arc detection terminal determines the ADC sampling rate according to the current frequency detection result, performs ADC sampling according to the ADC sampling rate, and extracts multiple current features for each current cycle; S3,电弧检测终端判断每个电流周期的电流特征是否超过各自的特征阈值,若超过,则认定此电流周期为电弧周期,并进行电弧周期累积量加1;S3, the arc detection terminal determines whether the current characteristics of each current cycle exceed their respective characteristic thresholds. If it exceeds, the current cycle is deemed to be an arc cycle, and the arc cycle accumulation amount is increased by 1; S4,在预设的移动时间窗口内,电弧检测终端判断电弧周期数是否超过设定的电弧周期累积量第一阈值,若超过则认为检测到电弧故障,并清除电弧周期累积量,否则,按照S3继续进行移动时间窗口内下一个电流周期的检测;S4, within the preset moving time window, the arc detection terminal determines whether the number of arc cycles exceeds the set first threshold of arc cycle accumulation. If it exceeds, it is considered that an arc fault is detected and the arc cycle accumulation is cleared. Otherwise, according to S3 continues to detect the next current cycle within the moving time window; S5,电弧检测终端检测到电路中存在电弧故障,将电弧故障报警和所述移动时间窗口内的所有电流数据上报给电弧检测远端;S5, the arc detection terminal detects an arc fault in the circuit, and reports the arc fault alarm and all current data within the moving time window to the arc detection remote end; S6,电弧检测远端接收到电弧检测终端上报的电弧故障报警信息和所述移动检测时间窗口内的所有电流数据;S6. The arc detection remote end receives the arc fault alarm information reported by the arc detection terminal and all current data within the movement detection time window; S7,电弧检测远端按照和电弧检测终端相同的电流周期,对所述移动检测时间窗口内的所有电流数据进行特征提取,提取多种电流特征;S7, the arc detection remote end performs feature extraction on all current data within the movement detection time window according to the same current cycle as the arc detection terminal, and extracts multiple current features; S8,电弧检测远端将提取的多种电流特征作为输入,输入到电弧故障神经网络识别器中,进行电弧故障识别;S8, the arc detection remote end uses the extracted various current characteristics as input to the arc fault neural network identifier for arc fault identification; S9,若电弧故障神经网络识别器中判断此段电流数据存在电弧故障,则电弧检测远端将电弧故障报警信息上报给上位机、存储该段电流数据,并将电弧保护命令下发给电弧检测终端;否则,若电弧故障神经网络识别器识别此段数据不存在电弧故障,则电弧检测远端将电弧故障误报信息上报给上位机、存储该段电流数据;S9, if the arc fault neural network identifier determines that there is an arc fault in this section of current data, the arc detection remote end will report the arc fault alarm information to the host computer, store the section of current data, and issue the arc protection command to the arc detection terminal; otherwise, if the arc fault neural network identifier identifies that there is no arc fault in this section of data, the arc detection remote end will report the arc fault false alarm information to the host computer and store the current data of this section; S10,电弧检测终端接收到电弧检测远端下发的电弧保护命令,进行电弧保护。S10: The arc detection terminal receives the arc protection command issued by the arc detection remote end and performs arc protection. 2.根据权利要求1所述的一种基于边缘计算的电弧故障检测方法,其特征在于:电弧检测终端数据预处理,进行数据有效性校验,具体为:将采集到的电流信号与预先设定的滤波阈值进行比较,若大于滤波阈值则认为采集到的电流数据为有效数据,并转入下一个步骤;若低于滤波阈值,则认为采集到的电流数据无效,重复进行电流采样。2. An arc fault detection method based on edge computing according to claim 1, characterized in that: arc detection terminal data is preprocessed to perform data validity verification, specifically: combining the collected current signal with a preset Compare with a certain filter threshold. If it is greater than the filter threshold, the collected current data is considered to be valid data, and the next step is entered; if it is lower than the filter threshold, the collected current data is considered invalid, and current sampling is repeated. 3.根据权利要求1所述的一种基于边缘计算的电弧故障检测方法,其特征在于:电弧检测终端电流频率检测,通过采集的电流数据确定电流的频率,是直流电还是交流电,基于电流频率选择ADC采样率,若为直流电,则以400HZ的频率为基准选择ADC采样率。3. An arc fault detection method based on edge computing according to claim 1, characterized in that: the arc detection terminal current frequency is detected, and the frequency of the current is determined through the collected current data, whether it is direct current or alternating current, and is selected based on the current frequency. ADC sampling rate, if it is DC, select the ADC sampling rate based on the frequency of 400HZ. 4.根据权利要求1所述的一种基于边缘计算的电弧故障检测方法,其特征在于:电弧检测终端进行每电流周期的电流特征提取具体为:对每个ADC采样的缓存电流数据,进行多种电流特征提取,计算每个电流周期的电流特征值,计算的特征值编号为电流特征1,电流特征2,...,电流特征n,n为大于2的正整数,电流特征为:电流均值、电流峰峰值、电流标准差、电流两周期平均幅值差、电流均匀度、电流变化率均值、电流偏态、电流余弦相似度、电流前n次谐波和或电流前后窗口各频率分量皮尔逊相关系数;电弧检测终端判断每个电流周期是否为电弧周期,具体为:电弧检测终端判断每个电流特征i是否超过本身的的电流特征i识别阈值,若所有电流特征i均超过各自的电流特征i识别阈值,其中i=1,2,..n,则判定当前电流周期为电弧周期,并进行电弧周期累积量加1。4. An arc fault detection method based on edge computing according to claim 1, characterized in that: the arc detection terminal performs current feature extraction for each current cycle, specifically: performing multiple operations on the cached current data sampled by each ADC. A kind of current feature extraction, calculate the current feature value of each current cycle, the calculated feature value number is current feature 1, current feature 2,..., current feature n, n is a positive integer greater than 2, the current feature is: current Average value, current peak-to-peak value, current standard deviation, current average amplitude difference between two cycles, current uniformity, current change rate average, current skewness, current cosine similarity, the first n harmonics of the current and each frequency component of the current window before and after Pearson correlation coefficient; the arc detection terminal determines whether each current cycle is an arc cycle, specifically: the arc detection terminal determines whether each current feature i exceeds its own current feature i identification threshold, if all current features i exceed their respective If the current characteristic i identification threshold is used, where i=1, 2, ..n, the current current cycle is determined to be an arc cycle, and the arc cycle accumulation amount is increased by 1. 5.根据权利要求1所述的一种基于边缘计算的电弧故障检测方法,其特征在于:电弧检测远端将电弧电流数据按照电流周期,计算每个电流周期电流特征1、电流特征2、...、电流特征m,m为大于等于4的正整数,并将计算得到的电流周期电流特征1、电流特征2、...、电流特征m导入训练好的神经网络电弧故障识别器,进行电弧故障识别。5. An arc fault detection method based on edge computing according to claim 1, characterized in that: the arc detection remote end uses the arc current data according to the current cycle to calculate the current characteristics 1, current characteristics 2, of each current cycle. .., current characteristic m, m is a positive integer greater than or equal to 4, and the calculated current period current characteristic 1, current characteristic 2, ..., current characteristic m is imported into the trained neural network arc fault identifier for Arc fault identification. 6.根据权利要求1所述的一种基于边缘计算的电弧故障检测方法,其特征在于:电流特征的选取是通过对电弧电流数据、正常电流数据、干扰电流数据和负载电流数据分别进行电流特征提取,得到电弧电流特征、正常电流特征、干扰电流特征和负载电流特征,并对得到的特征进行特征值域分析,确定各自特征的最大值和最小值,得到特征值域[特征最小值,特征最大值],并对电弧电流特征值域和正常电流特征值域、干扰电流特征值域、负载电流特征值域进行比较,若电弧电流特征值域和其他特征值域的重合度小于预设的特征值域重合阈值,则该电流特征可以用于进行电弧故障识别;若电弧电流特征值域和其他值域没有重合,则特征值域重合度为零,则优选该电流特征用于电弧识别。6. An arc fault detection method based on edge computing according to claim 1, characterized in that: the current characteristics are selected by performing current characteristics on arc current data, normal current data, interference current data and load current data respectively. Extract, obtain arc current characteristics, normal current characteristics, interference current characteristics and load current characteristics, conduct characteristic value domain analysis on the obtained characteristics, determine the maximum and minimum values of respective characteristics, and obtain the characteristic value range [characteristic minimum value, characteristic value maximum value], and compare the arc current characteristic value domain with the normal current characteristic value domain, interference current characteristic value domain, and load current characteristic value domain. If the overlap between the arc current characteristic value domain and other characteristic value domains is less than the preset If the characteristic value domain coincides with the threshold, then the current feature can be used for arc fault identification; if the arc current characteristic value domain does not overlap with other value domains, then the coincidence degree of the characteristic value domain is zero, then the current feature is preferably used for arc identification. 7.根据权利要求1所述的一种基于边缘计算的电弧故障检测方法,其特征在于:7. An arc fault detection method based on edge computing according to claim 1, characterized in that: 电弧检测移动窗口的窗口长度根据电弧检测标准和电流体制确定;The window length of the arc detection moving window is determined according to the arc detection standard and current system; 电弧检测移动窗口长度和电弧电流周期具有如下关系:The arc detection moving window length and arc current period have the following relationship: 电弧检测移动窗口长度=电流周期*电弧检测需要识别的周期数;Arc detection moving window length = current cycle * number of cycles that need to be identified for arc detection; 电流周期由电压频率f确定,电流周期=1/电压频率f;若为直流供电体制,则电流周期可根据需要选择,可选2.5ms。The current period is determined by the voltage frequency f, current period = 1/voltage frequency f; if it is a DC power supply system, the current period can be selected as needed, and 2.5ms is optional. 8.根据权利要求1所述的一种基于边缘计算的电弧故障检测方法,其特征在于:神经网络电弧故障识别器是预先训练好的BP神经网络:8. An arc fault detection method based on edge computing according to claim 1, characterized in that: the neural network arc fault identifier is a pre-trained BP neural network: (1)设置神经网络的输入层、隐藏层、输出层节点,设置学习率及激活函数;(1) Set the input layer, hidden layer, and output layer nodes of the neural network, and set the learning rate and activation function; (2)通过预先采集的电弧数据、正常数据、负载数据和干扰数据,进行特征提取,分别提取特征i的四类数据的相应特征,得到对应的电弧电流特征i、正常电流特征i、负载电流特征i和干扰电流特征i,将四类特征组成的数组作为训练集进行神经网络训练,神经网络输出结果为电弧故障和非电弧故障。(2) Feature extraction is performed through the pre-collected arc data, normal data, load data and interference data, and the corresponding features of the four types of data of feature i are extracted respectively to obtain the corresponding arc current feature i, normal current feature i, and load current Feature i and interference current feature i, an array composed of four types of features is used as a training set for neural network training. The output results of the neural network are arc faults and non-arc faults.
CN202311524316.6A 2023-11-15 2023-11-15 Arc fault detection method based on edge calculation Pending CN117517897A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311524316.6A CN117517897A (en) 2023-11-15 2023-11-15 Arc fault detection method based on edge calculation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311524316.6A CN117517897A (en) 2023-11-15 2023-11-15 Arc fault detection method based on edge calculation

Publications (1)

Publication Number Publication Date
CN117517897A true CN117517897A (en) 2024-02-06

Family

ID=89741458

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311524316.6A Pending CN117517897A (en) 2023-11-15 2023-11-15 Arc fault detection method based on edge calculation

Country Status (1)

Country Link
CN (1) CN117517897A (en)

Similar Documents

Publication Publication Date Title
CN113589105B (en) Method and system for identification of serial fault arcs in building electrical fires based on the Internet of Things
CN101187687B (en) High resistance ground fault detection method based on transient traveling wave
RU2484487C2 (en) Method and device for determination of electric arc occurrence on at least one electric cable
EP1939995A2 (en) Methods and systems for detecting series arcs in electrical systems
CN105738764B (en) Fault Section Location of Distribution Network based on transient information Whole frequency band
CN110007197B (en) Series arc fault detection system and detection method
CN104569683B (en) A kind of detection method of fault electric arc
US9869709B2 (en) Electric arc detection method and device
CN104678265A (en) Detection device and detection method for series arc faults
CN110320434A (en) High resistive fault discrimination method and system based on zero-sequence current waveform Interval Slope curve
CN107870281B (en) A leakage and fault arc detection method and device thereof
CN112183628A (en) Alternating current arc fault detection method and system based on multiple linear time-frequency transformations
WO2019127440A1 (en) Method and device for handling direct current arc
CN210111945U (en) DC side arc fault detection system of photovoltaic system
CN205941736U (en) Trouble arc detection device
Gupta et al. Series arc fault detection in low voltage distribution system with signal processing and machine learning approach
CN117517897A (en) Arc fault detection method based on edge calculation
CN106950482B (en) A kind of corona interference elimination method based on alternate signal map similarity relation
KR102449161B1 (en) Arc detection method and device using frequency analysis
KR102243313B1 (en) PARTIAL DISCHARGE JUDGING METHOD and DIAGNOSTIC SYSTEM
CN114062880A (en) Direct-current fault arc detection method based on adaptive signal processing and CART tree ensemble learning
US12248034B2 (en) Series arc detection firmware
CN117590181A (en) Low-voltage alternating current series arc fault identification method utilizing arc voltage randomness
Chothani et al. Modern Method to Discriminate High Resistance Internal Winding Faults and Inrush Condition/External Faults in Power Transformer
Ali et al. Studies on the application of wavelet families for a high impedance fault location algorithm in a distribution 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