CN109270384A - A kind of method and system of the electric arc of electrical equipment for identification - Google Patents
A kind of method and system of the electric arc of electrical equipment for identification Download PDFInfo
- Publication number
- CN109270384A CN109270384A CN201811342742.7A CN201811342742A CN109270384A CN 109270384 A CN109270384 A CN 109270384A CN 201811342742 A CN201811342742 A CN 201811342742A CN 109270384 A CN109270384 A CN 109270384A
- Authority
- CN
- China
- Prior art keywords
- electrical equipment
- data
- voltage
- layer
- arc
- 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.)
- Granted
Links
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of methods that electrical equipment for identification uses route electric arc, it include: the voltage data and current data of real-time acquisition electrical equipment under steady state conditions, and the voltage and current data are filtered and sampling processing, with the voltage data and current data for the electrical equipment that respectively obtains that treated, the voltage-to-current geometric locus of electrical equipment is obtained according to the voltage data and current data of treated electrical equipment, the voltage-to-current geometric locus is normalized and image binarization is handled, to obtain the bianry image of electrical equipment and be backed up, the bianry image similarity for determining whether the electrical equipment for finding and obtaining in arc data library reaches 80% or more bianry image, if it is show that the route of electrical equipment in use produces electric arc.The present invention is able to solve the technical problem that electric arc recognition accuracy present in existing arc method for measuring is not high, calculating process is many and diverse, judges time length.
Description
Technical field
The invention belongs to fire safety and mode identification technologies, it particularly relates to which a kind of use for identification
The method and system of electric equipment electric arc.
Background technique
Under the development and the increasing environment of living standards of the people of science and technology, electric energy is wide as the important energy
General application.While electric energy is widely applied to people's lives every aspect, very big security risk, many use are also brought
The insecurity factor of electricity results in the continuous generation of electrical fire.The statistics of the Fire Department of Ministry of Public Security is shown, from 2011 to 2016
Year, China occurs electrical fire 52.4 ten thousand altogether, causes 3261 people death, and two over thousands of people are injured, and direct economic loss is up to 92
Hundred million yuan.
According to statistics, there are about 40% electrical fire be as electrical equipment failure generate electric arc caused by, therefore, arc-detection
Technology has been widely adopted at present.Common arc method for measuring domestic at present has: (1) analyzing electricity when fault electric arc generates
The features such as arc current waveforms amplitude, zero crossing or waveform, by two-by-two or multiple combined time domain approach are realized to electric arc
Identification；(2) collected data are subjected to multilayer decomposition using wavelet analysis, by the characteristic quantity after decomposition and under normal circumstances
Characteristic quantity after decomposition compares, to judge whether there is electric arc generation.
However, above-mentioned arc method for measuring there is technical issues that: the first, carrying out electricity in the time domain
In the deterministic process of arc current, due to the generation of serial arc, the electric current of circuit entirety can decline, but different electrical equipments are producing
After raw electric arc, the amplitude of circuital current decline is not consistent, therefore needs to be correspondingly arranged different threshold values, this causes this method pervasive
Property it is not strong, in addition, some electrical equipments when in use the phase generate arc current waveform it is relatively similar, therefore time domain is caused to examine
The method discrimination of survey is not high；The second, for current data when carrying out wavelet analysis, calculation amount is excessive, judges that the time is relatively long,
Therefore, the real-time of arc-detection can be impacted.
Summary of the invention
Aiming at the above defects or improvement requirements of the prior art, the present invention provides a kind of electrical equipment electric arcs for identification
Method and system, it is intended that solving present in existing arc method for measuring that electric arc recognition accuracy is not high, calculated
Journey is many and diverse, judges the technical problem of time length.
To achieve the above object, according to one aspect of the present invention, providing one kind, electrical equipment uses line for identification
The method of road electric arc, comprising the following steps:
(1) voltage data and current data of electrical equipment under steady state conditions are acquired in real time, and to the voltage and current
Data are filtered and sampling processing, with the voltage data and current data for the electrical equipment that respectively obtains that treated；
(2) the voltage-to-current rail of electrical equipment is obtained according to the voltage data and current data of treated electrical equipment
Trace curve, is normalized the voltage-to-current geometric locus and image binarization is handled, to obtain the two-value of electrical equipment
Image is simultaneously backed up；
(3) it determines whether to find in arc data library and the bianry image of step (2) obtained electrical equipment
Similarity reaches 80% or more bianry image, if it is shows that the route of electrical equipment in use produces electricity
Arc sends warning message to electrical equipment user at this time, and cuts off the circuit of electrical equipment, and then process terminates, and otherwise enters
Step (4)；
(4) bianry image of electrical equipment is input in trained electric arc discrimination model, to obtain electric arc identification knot
Fruit.
Preferably, filtering uses the way of median average filter, the i.e. N number of electric current of continuous sampling and N in step (1)
The data of a voltage remove a maximum value and minimum value respectively, then calculate the arithmetic mean of instantaneous value of N-2 data, wherein N value
Between preferably 4 to 15.
Preferably, arc data library is established in initial phase, and the every number of n tables of data is shared in the arc data library
According to table reflect electrical equipment obtained using arc generator measurement, different under different working condition electric current, voltage and
Mapping relations between corresponding bianry image three, wherein n indicates that the classification for the electrical equipment that the method can be applicable in is total
Number.
Preferably, the electric arc discrimination model is generated by following procedure:
(a) multiple bianry images are obtained from arc data library every time, using convolutional neural networks to this several binary map
As being trained, duplicate process of laying equal stress on, until bianry images all in arc data library all are trained to finish, to obtain
Trained convolutional neural networks, and counter i=1 is set；
(b) the trained convolutional neural networks obtained using step (a) are to institute in i-th of tables of data in arc data library
There is each of k bianry image bianry image aijIt is handled, to obtain the corresponding eigenvectors matrix of the bianry image,
It is provided with j ∈ (1, k)；
(c) each two-value in i-th of tables of data is calculated in arc data library according to the corresponding eigenvectors matrix of image
Image aijThe average value d of Euclidean distance between other all bianry imagesij, it is taken out the maximum value d of average valueimax,
And it is carried out using eigenvectors matrix of the PCA method to the corresponding k 1*n size of k bianry images all in i-th of tables of data
Dimension-reduction treatment, to obtain the eigenvalue matrix of 1*s, wherein s is the integer between 3 to 5；
(d) i=i+1 is set, and judges whether i is equal to n, if yes then enter step (e), otherwise return step (b)；
(e) to k obtained maximum value dimaxIt is arranged in maximum value array according to sequence from small to large, it will be in the array
Adjacent element is subtracted each other two-by-two, obtains new array [Diff1, Diff2..., Diffk-1], it is selected from new array absolutely
To the maximum element of value, and obtain smaller value dist of the element in maximum value array in corresponding two elements；
(f) eigenvalue matrix that step (c) obtains is clustered using density-based algorithms, to obtain n kind
The corresponding cluster result of electrical equipment.
Preferably, convolutional neural networks are 8 layers of structure, and wherein first layer is input layer, and the image of input is the figure of p*p*1
Picture, the second layer are convolutional layers, receive the image of the p*p*1 from input layer, and wherein convolution kernel is 3*3*6, which uses full 0
Filling, step-length 1, this layer of output matrix size are p*p*6, and third layer is pond layer, and convolution kernel is long and wide having a size of 2*2
Step-length is 2, this layer of output matrix is (p/2) * (p/2) * 6, and the 4th layer is convolutional layer, and convolution kernel is having a size of 5*5*16, step-length
It is 1, this layer is filled without using full 0, and the matrix of output is (p/2-4) * (p/2-4) * 16, and layer 5 is pond layer, convolution kernel ruler
Very little is 2*2, and step-length 2, the matrix size of output is (p/4-2) * (p/4-2) * 16；Layer 6 is full articulamentum, output section
Point number is 120；Layer 7 is full articulamentum, and output node number is 64, and the 8th layer is output layer, and output node number is
10.
Preferably, loss function used in convolutional neural networks is cross entropy loss function.
Preferably, step (f) is then arranged close specifically, obtained eigenvalue matrix is mapped in n-dimensional space first
Angle value minpts is the integer between 3 to 5, and cluster core point maximum radius is equal to the smaller value dist that above-mentioned steps (e) obtain.
Preferably, step (4) is specifically, input electric arc discrimination model for the bianry image of electrical equipment first, to obtain
Then corresponding eigenvalue matrix calculates corresponding points of the eigenvalue matrix in the n-dimensional space of electric arc discrimination model to n kind electricity consumption
Equipment clusters the distance l of core point, finally compares distance l and dist, if l is less than dist, then it represents that the electrical equipment generates
Otherwise electric arc indicates that electrical equipment does not generate electric arc.
It is another aspect of this invention to provide that providing a kind of system that electrical equipment for identification uses route electric arc, packet
It includes:
First module, for acquiring the voltage data and current data of electrical equipment under steady state conditions in real time, and to this
Voltage and current data are filtered and sampling processing, with the voltage data and electric current number of the electrical equipment that respectively obtains that treated
According to；
Second module obtains the electricity of electrical equipment for the voltage data and current data according to treated electrical equipment
Piezo-electric trajectory trace curve, is normalized the voltage-to-current geometric locus and image binarization is handled, to obtain electricity consumption
The bianry image of equipment is simultaneously backed up；
Third module is used to determine whether to find in arc data library and the obtained electrical equipment of the second module
Bianry image similarity reach 80% or more bianry image, if it is show the route of electrical equipment in use
Electric arc is produced, sends warning message to electrical equipment user at this time, and cut off the circuit of electrical equipment, then process terminates,
Otherwise enter the 4th module；
4th module, for the bianry image of electrical equipment to be input in trained electric arc discrimination model, to obtain
Electric arc recognition result.
In general, through the invention it is contemplated above technical scheme is compared with the prior art, can obtain down and show
Beneficial effect:
1, the present invention is able to solve the low technical problem of recognition accuracy present in existing arc method for measuring: due to
It using step (2) and step (3), is collected into a large amount of arc characteristic data and establishes arc data library, and in subsequent detection mistake
Voltage data and current data of the electrical equipment under steady state conditions in a period of time are acquired in journey, it therefore, being capable of area very well
The indicatrix similar to arc current for dividing incandescent lamp, fluorescent lamp etc. to generate in switch temporal program, and pass through
Learn a large amount of training data and effectively improves recognition accuracy；
2, the present invention is able to solve the technology that computation complexity present in existing arc method for measuring is high, the calculating time is long
Problem: step (5) are arrived due to using step (4), first training convolutional neural networks model is carried out followed by feature vector
The feature space for clustering and establishing carries out electric arc identification, and therefore, it is only necessary to collected data are carried out intermediate value when detecting
Filtering and image binarization processing can carry out accurate electric arc identification using established electric arc discrimination model, greatly drop
Low computation complexity, and reduce the time of calculating；
3, compared to traditional detection, the bianry image of the voltage and current track by drawing electric arc constructs the present invention
About the two-dimensional surface of arc characteristic, therefore, than the information and feature that one-dimensional signal remains more electric arcs；
4, present invention utilizes the clustering algorithms for being based on depth learning technology (i.e. convolutional neural networks) and machine learning
(OPTICS) electric arc discrimination model generated analyzes the bianry image of the voltage and current track of electric arc, meets artificial
The development trend of intelligence.
Detailed description of the invention
Fig. 1 is flow chart of the electrical equipment for identification of the invention using the method for route electric arc.
Fig. 2 is to obtain electrical equipment according to the voltage data and current data of treated electrical equipment in the method for the present invention
Voltage-to-current geometric locus schematic diagram.
The schematic diagram of the bianry image of electrical equipment obtained in the step of Fig. 3 is the method for the present invention (2).
Fig. 4 is the schematic diagram of convolutional neural networks used in the method for the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.As long as in addition, technical characteristic involved in the various embodiments of the present invention described below
Not constituting a conflict with each other can be combined with each other.
As shown in Figure 1, the present invention for identification electrical equipment using route electric arc method the following steps are included:
(1) voltage data and current data of electrical equipment under steady state conditions are acquired in real time, and to the voltage and current
Data are filtered and sampling processing, with the voltage data and current data for the electrical equipment that respectively obtains that treated；
Specifically, the time span acquired every time in this step is 2 seconds, sample frequency is 25 to 35KHz, preferably
30KHz。
Filtering is used using the way of median average filter (also known as anti-impulse disturbances average filter in this step
Method), the data for continuously sampling N number of electric current and N number of voltage are equivalent to, removes a maximum value and minimum value respectively, then counts
Calculate the arithmetic mean of instantaneous value of N-2 data.N value is typically chosen 4~15.The filtering method can effectively overcome alternating current contingency
Impulse disturbances.And elimination may be due to the sampled value deviation caused by impulse disturbances.
Time interval in this step in sampling process is 1 second, it is therefore an objective to slow down the calculating pressure of server, and prevent
There may be the situations that noise jamming influences data exception.
(2) the voltage-to-current rail of electrical equipment is obtained according to the voltage data and current data of treated electrical equipment
Trace curve (as shown in Figure 2), is normalized the voltage-to-current geometric locus and image binarization is handled, to be used
The bianry image (as shown in Figure 3) of electric equipment is simultaneously backed up；
Specifically, image binarization treated bianry image is having a size of p*p in this step, depth 1, wherein p
For the integer between 10 to 50；
(3) it determines whether to find in arc data library and the bianry image of step (2) obtained electrical equipment
Similarity reaches 80% or more bianry image, if it is shows that the route of electrical equipment in use produces electricity
Arc sends warning message to electrical equipment user at this time, and cuts off the circuit of electrical equipment, and then process terminates, and otherwise enters
Step (4)；
In this step, arc data library is established in initial phase, and n tables of data is shared in the arc data library
(the classification sum for the electrical equipment that wherein n can be applicable in for the present invention, in the present embodiment, n=12), every tables of data is anti-
Electrical equipment obtain using arc generator measurement, different electric current, voltage and corresponding under different working condition are reflected
Mapping relations between bianry image three.
12 tables of data used in the present invention include: air-conditioning tables of data, fluorescent lamp tables of data, fan data table, electric ice
Case tables of data, hair dryer tables of data, immersion heater tables of data, desk lamp tables of data, laptop tables of data, data from microwave oven table,
Dust catcher tables of data, electromagnetic oven tables of data and washing machine tables of data, it should be understood that the present invention is not limited to above-mentioned use
Electric equipment, any electrical equipment are all included into protection scope of the present invention.
For example, the working condition stored in air-conditioning tables of data has purification air, refrigeration, heating, dehumidifier, sleep, blowing shape
State etc..
(4) bianry image of electrical equipment is input in trained electric arc discrimination model, to obtain electric arc identification knot
Fruit, and recognition result is sent to user.
Specifically, final electric arc recognition result includes that the electrical equipment produces electric arc, electric arc is not still generated.
Electric arc discrimination model in the present invention is generated by following procedure:
(a) multiple bianry images are obtained from arc data library every time (the bianry image quantity wherein read is 2m, m 1
Integer between to 7), this several bianry image is trained using convolutional neural networks, duplicate process of laying equal stress on, until electric arc
Until all bianry images are all trained to finish in database, to obtain trained convolutional neural networks, and counting is set
Device i=1；
Specifically, output is the one-dimension array of 1*n；
As shown in figure 4, convolutional neural networks of the invention include input layer, output layer, two convolutional layers, two ponds
Layer and two full articulamentums.
First layer is input layer, and the image of input is the image of p*p*1.
The second layer is convolutional layer, receives the image of the p*p*1 from input layer, and wherein convolution kernel is 3*3*6, which makes
It is filled with full 0, step-length 1, this layer of output matrix size is p*p*6；
Third layer is pond layer, and convolution kernel is having a size of 2*2, and long and wide step-length is 2, this layer of output matrix is (p/2) *
(p/2)*6；
4th layer is convolutional layer, and convolution kernel is filled having a size of 5*5*16, step-length 1, this layer without using full 0, the square of output
Battle array is (p/2-4) * (p/2-4) * 16；
Layer 5 is pond layer, and convolution kernel is (p/4-2) * (p/4- having a size of 2*2, step-length 2, the matrix size of output
2)*16；
Layer 6 is full articulamentum, and output node number is 120；
Layer 7 is full articulamentum, and output node number is 64；
8th layer is output layer, and output node number is 10.
Wherein, the loss function that convolutional neural networks utilize is cross entropy loss function.
(b) the trained convolutional neural networks obtained using step (a) are to institute in i-th of tables of data in arc data library
There is each of k bianry image bianry image aijIt is handled, to obtain the corresponding eigenvectors matrix of the bianry image,
Wherein (1, k) j ∈；
Specifically, the matrix that the corresponding feature vector of each bianry image is 1*n；
(c) each two-value in i-th of tables of data is calculated in arc data library according to the corresponding eigenvectors matrix of image
Image aijThe average value d of Euclidean distance between other all bianry imagesij, and it is taken out the maximum value of average value
dimax, using Principal Component Analysis (Primary Component Analysis, abbreviation PCA) to all k in i-th of tables of data
The eigenvectors matrix of the corresponding k 1*n size of a bianry image carries out dimension-reduction treatment, to obtain eigenvalue matrix (its of 1*s
Middle s is the integer between 3 to 5)；
The specific formula that this step uses are as follows:
(d) i=i+1 is set, and judges whether i is equal to n, if yes then enter step (e), otherwise return step (b)；
(e) to k obtained maximum value dimaxIt is arranged in maximum value array according to sequence from small to large, it will be in the array
Adjacent element is subtracted each other two-by-two, obtains new array [Diff1, Diff2..., Diffk-1], it is selected from new array absolutely
To the maximum element of value, and obtain smaller value dist of the element in maximum value array in corresponding two elements；
(f) (Ordering Points To is used in this step using density-based algorithms
Identify the Clustering Structure, abbreviation OPTICS algorithm) eigenvalue matrix that step (c) is obtained into
Row cluster, to obtain the corresponding cluster result of n kind electrical equipment；
Specifically, obtained eigenvalue matrix is mapped in n-dimensional space first, then density of setting value minpts is
Integer between 3 to 5, cluster core point maximum radius are equal to the smaller value dist that above-mentioned steps (e) obtain.
The working principle of the invention is: the bianry image of electrical equipment being inputted electric arc discrimination model first, to obtain pair
Then the eigenvalue matrix answered calculates corresponding points of the eigenvalue matrix in the n-dimensional space of electric arc discrimination model and sets to n kind electricity consumption
The distance l of standby cluster core point, finally compares distance l and dist, if l is less than dist, then it represents that the electrical equipment produces
Electric arc.Otherwise expression does not generate electric arc with point device.
As it will be easily appreciated by one skilled in the art that the foregoing is merely illustrative of the preferred embodiments of the present invention, not to
The limitation present invention, any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should all include
Within protection scope of the present invention.
Claims (9)
1. a kind of method that electrical equipment for identification uses route electric arc, which comprises the following steps:
(1) voltage data and current data of electrical equipment under steady state conditions are acquired in real time, and to the voltage and current data
It is filtered and sampling processing, with the voltage data and current data for the electrical equipment that respectively obtains that treated；
(2) the voltage-to-current track for obtaining electrical equipment according to the voltage data and current data of treated electrical equipment is bent
Line, is normalized the voltage-to-current geometric locus and image binarization is handled, to obtain the bianry image of electrical equipment
And it is backed up；
(3) it determines and detects whether to find in arc data library and the bianry image of step (2) obtained electrical equipment
Similarity reaches 80% or more bianry image, if it is shows that the route of electrical equipment in use produces electricity
Arc sends warning message to electrical equipment user at this time, and cuts off the circuit of electrical equipment, and then process terminates, and otherwise enters
Step (4)；
(4) bianry image of electrical equipment is input in trained electric arc discrimination model, to obtain electric arc recognition result.
2. the method according to claim 1, wherein filtering uses median average in step (1)
The data of filter method, i.e. continuous sampling N number of electric current and N number of voltage remove a maximum value and minimum value respectively, then calculate N-
The arithmetic mean of instantaneous value of 2 data, wherein N value is preferably between 4 to 15.
3. the method according to claim 1, wherein arc data library is in initial phase foundation, the electric arc
Shared every tables of data of n tables of data is reflected electrical equipment obtained using arc generator measurement, different and existed in database
Mapping relations under different working condition between electric current, voltage and corresponding bianry image three, wherein n indicates the side
The classification sum for the electrical equipment that method can be applicable in.
4. the method according to claim 1, wherein the electric arc discrimination model is generated by following procedure
:
(a) multiple bianry images are obtained from arc data library every time, using convolutional neural networks to this several bianry image into
Row training, duplicate process of laying equal stress on, until bianry images all in arc data library all are trained to finish, to be trained
Good convolutional neural networks, and counter i=1 is set；
(b) the trained convolutional neural networks obtained using step (a) are to all k in i-th of tables of data in arc data library
Each of a bianry image bianry image aijIt is handled, to obtain the corresponding eigenvectors matrix of the bianry image,
Middle j ∈ (1, k)；
(c) each bianry image in i-th of tables of data is calculated in arc data library according to the corresponding eigenvectors matrix of image
aijThe average value d of Euclidean distance between other all bianry imagesij, and it is taken out the maximum value d of average valueimax, benefit
Dimensionality reduction is carried out with eigenvectors matrix of the PCA method to the corresponding k 1*n size of k bianry images all in i-th of tables of data
Processing, to obtain the eigenvalue matrix of 1*s, wherein s is the integer between 3 to 5；
(d) i=i+1 is set, and judges whether i is equal to n, if yes then enter step (e), otherwise return step (b)；
(e) to k obtained maximum value dimaxBe arranged in maximum value array according to sequence from small to large, by the array two-by-two
Adjacent element is subtracted each other, and new array [Diff is obtained1, Diff2..., Diffk-1], absolute value is selected from new array
Maximum element, and obtain smaller value dist of the element in maximum value array in corresponding two elements；
(f) eigenvalue matrix that step (c) obtains is clustered using density-based algorithms, to obtain n kind electricity consumption
The corresponding cluster result of equipment.
5. according to the method described in claim 4, it is characterized in that,
Convolutional neural networks are 8 layers of structure, in which:
First layer is input layer, and the image of input is the image of p*p*1.
The second layer is convolutional layer, receives the image of the p*p*1 from input layer, and wherein convolution kernel is 3*3*6, which uses complete
0 filling, step-length 1, this layer of output matrix size are p*p*6；
Third layer is pond layer, and convolution kernel is having a size of 2*2, and long and wide step-length is 2, this layer of output matrix is (p/2) * (p/
2)*6；
4th layer is convolutional layer, and convolution kernel is filled having a size of 5*5*16, step-length 1, this layer without using full 0, and the matrix of output is
(p/2-4)*(p/2-4)*16；
Layer 5 is pond layer, and convolution kernel is (p/4-2) * (p/4-2) * having a size of 2*2, step-length 2, the matrix size of output
16；
Layer 6 is full articulamentum, and output node number is 120；
Layer 7 is full articulamentum, and output node number is 64；
8th layer is output layer, and output node number is 10.
6. method according to claim 4 or 5, which is characterized in that loss function used in convolutional neural networks is to hand over
Pitch entropy loss function.
7. the method according to any one of claim 4 to 6, which is characterized in that step (f) specifically, incite somebody to action first
To eigenvalue matrix be mapped in n-dimensional space, then density of setting value minpts be 3 to 5 between integer, cluster core point
Maximum radius is equal to the smaller value dist that above-mentioned steps (e) obtain.
8. method according to any one of claims 4 to 7, which is characterized in that step (4) specifically, will use first
The bianry image of electric equipment inputs electric arc discrimination model, to obtain corresponding eigenvalue matrix, then calculates eigenvalue matrix and exists
Corresponding points in the n-dimensional space of electric arc discrimination model to the distance l of n kind electrical equipment cluster core point, finally compare distance l and
Dist, if l is less than dist, then it represents that the electrical equipment produces electric arc, otherwise indicates that electrical equipment does not generate electric arc.
9. a kind of system that electrical equipment for identification uses route electric arc characterized by comprising
First module, for acquiring the voltage data and current data of electrical equipment under steady state conditions in real time, and to the voltage
It is filtered with current data and sampling processing, with the voltage data and current data for the electrical equipment that respectively obtains that treated；
Second module obtains the voltage-of electrical equipment for the voltage data and current data according to treated electrical equipment
Current locus curve, is normalized the voltage-to-current geometric locus and image binarization is handled, to obtain electrical equipment
Bianry image and backed up；
Third module is used to determine whether that two with the obtained electrical equipment of the second module can be found in arc data library
Value image similarity reaches 80% or more bianry image, if it is shows that the route of electrical equipment in use generates
Electric arc sends warning message to electrical equipment user at this time, and cuts off the circuit of electrical equipment, and then process terminates, otherwise
Into the 4th module；
4th module, for the bianry image of electrical equipment to be input in trained electric arc discrimination model, to obtain electric arc
Recognition result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811342742.7A CN109270384B (en) | 2018-11-13 | 2018-11-13 | A kind of method and system of the electric arc of electrical equipment for identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811342742.7A CN109270384B (en) | 2018-11-13 | 2018-11-13 | A kind of method and system of the electric arc of electrical equipment for identification |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109270384A true CN109270384A (en) | 2019-01-25 |
| CN109270384B CN109270384B (en) | 2019-06-11 |
Family
ID=65193183
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811342742.7A Active CN109270384B (en) | 2018-11-13 | 2018-11-13 | A kind of method and system of the electric arc of electrical equipment for identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109270384B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110084158A (en) * | 2019-04-15 | 2019-08-02 | 杭州拓深科技有限公司 | A kind of electrical equipment recognition methods based on intelligent algorithm |
| CN110954784A (en) * | 2019-12-20 | 2020-04-03 | 吉林省士而德科技有限公司 | Intelligent electricity safety detector |
| CN111103477A (en) * | 2019-10-31 | 2020-05-05 | 武汉旗云高科信息技术有限公司 | Electrical load monitoring method and system based on 3D convolutional neural network |
| CN111834999A (en) * | 2020-05-26 | 2020-10-27 | 浙江华云信息科技有限公司 | Power utilization sensing method based on voltage and current orthogonal curve |
| CN113376473A (en) * | 2021-04-26 | 2021-09-10 | 国网天津市电力公司城南供电分公司 | Fault arc detection method based on support vector machine and image recognition |
| WO2023005976A1 (en) * | 2021-07-27 | 2023-02-02 | 广西电网有限责任公司 | Deep-learning-based identification method and apparatus for electric device |
| TWI830815B (en) * | 2019-10-24 | 2024-02-01 | 鴻海精密工業股份有限公司 | User terminal failure detection method, device, computer device and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101702936A (en) * | 2007-06-08 | 2010-05-05 | 株式会社普来马特 | Arc detecting apparatus and arc monitoring method using the same |
| CN103842836A (en) * | 2011-07-26 | 2014-06-04 | 伊顿工业（奥地利）有限公司 | Method for adapting an arc sensor |
| CN104569684A (en) * | 2015-01-14 | 2015-04-29 | 上海和伍新材料科技有限公司 | Fault electric arc detection method based on electric arc spectrum signals |
| CN104979797A (en) * | 2014-04-02 | 2015-10-14 | 海尔集团公司 | Electric leakage protecting device with arc detection function, and electric power consumption equipment |
| CN105974223A (en) * | 2016-04-27 | 2016-09-28 | 深圳职业技术学院 | Method used for carrying out on-line detection on electric equipment work state and system thereof |
| CN106199131A (en) * | 2016-06-22 | 2016-12-07 | 华为技术有限公司 | The testing circuit of electric arc and microprocessor in a kind of DC current |
| CN106326886A (en) * | 2016-11-07 | 2017-01-11 | 重庆工商大学 | Finger-vein image quality evaluation method and system based on convolutional neural network |
-
2018
- 2018-11-13 CN CN201811342742.7A patent/CN109270384B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101702936A (en) * | 2007-06-08 | 2010-05-05 | 株式会社普来马特 | Arc detecting apparatus and arc monitoring method using the same |
| CN103842836A (en) * | 2011-07-26 | 2014-06-04 | 伊顿工业（奥地利）有限公司 | Method for adapting an arc sensor |
| CN104979797A (en) * | 2014-04-02 | 2015-10-14 | 海尔集团公司 | Electric leakage protecting device with arc detection function, and electric power consumption equipment |
| CN104569684A (en) * | 2015-01-14 | 2015-04-29 | 上海和伍新材料科技有限公司 | Fault electric arc detection method based on electric arc spectrum signals |
| CN105974223A (en) * | 2016-04-27 | 2016-09-28 | 深圳职业技术学院 | Method used for carrying out on-line detection on electric equipment work state and system thereof |
| CN106199131A (en) * | 2016-06-22 | 2016-12-07 | 华为技术有限公司 | The testing circuit of electric arc and microprocessor in a kind of DC current |
| CN106326886A (en) * | 2016-11-07 | 2017-01-11 | 重庆工商大学 | Finger-vein image quality evaluation method and system based on convolutional neural network |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110084158A (en) * | 2019-04-15 | 2019-08-02 | 杭州拓深科技有限公司 | A kind of electrical equipment recognition methods based on intelligent algorithm |
| TWI830815B (en) * | 2019-10-24 | 2024-02-01 | 鴻海精密工業股份有限公司 | User terminal failure detection method, device, computer device and storage medium |
| CN111103477A (en) * | 2019-10-31 | 2020-05-05 | 武汉旗云高科信息技术有限公司 | Electrical load monitoring method and system based on 3D convolutional neural network |
| CN110954784A (en) * | 2019-12-20 | 2020-04-03 | 吉林省士而德科技有限公司 | Intelligent electricity safety detector |
| CN111834999A (en) * | 2020-05-26 | 2020-10-27 | 浙江华云信息科技有限公司 | Power utilization sensing method based on voltage and current orthogonal curve |
| CN113376473A (en) * | 2021-04-26 | 2021-09-10 | 国网天津市电力公司城南供电分公司 | Fault arc detection method based on support vector machine and image recognition |
| WO2023005976A1 (en) * | 2021-07-27 | 2023-02-02 | 广西电网有限责任公司 | Deep-learning-based identification method and apparatus for electric device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109270384B (en) | 2019-06-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109270384B (en) | A kind of method and system of the electric arc of electrical equipment for identification | |
| CN110443302B (en) | Load identification method based on feature fusion and deep learning and application thereof | |
| WO2022067562A1 (en) | Method and device for diagnosing fault arc, and computer-readable storage medium | |
| Sahoo et al. | Trends in partial discharge pattern classification: a survey | |
| Do et al. | Convolutional-neural-network-based partial discharge diagnosis for power transformer using UHF sensor | |
| Xiao et al. | Maximal overlap discrete wavelet transform and deep learning for robust denoising and detection of power quality disturbance | |
| CN106650913A (en) | Deep convolution neural network-based traffic flow density estimation method | |
| KR101433472B1 (en) | Apparatus, method and computer readable recording medium for detecting, recognizing and tracking an object based on a situation recognition | |
| CN103310200B (en) | Face identification method | |
| CN103323749A (en) | Multi-classifier information fusion partial discharge diagnostic method | |
| CN105701470A (en) | Analog circuit fault characteristic extraction method based on optimal wavelet packet decomposition | |
| CN109886242A (en) | A kind of method and system that pedestrian identifies again | |
| Zhang et al. | Assembler: Efficient discovery of spatial co-evolving patterns in massive geo-sensory data | |
| CN104951787A (en) | Power quality disturbance identification method for distinguishing dictionary learning under SRC framework | |
| CN113922361B (en) | Potential safety hazard and fault protection method based on non-invasive distribution and utilization electricity information decomposition | |
| Fahim et al. | An unsupervised protection scheme for overhead transmission line with emphasis on situations during line and source parameter variation | |
| CN108508297A (en) | A kind of fault arc detection method based on change coefficient and SVM | |
| Zheng et al. | Benchmarking unsupervised anomaly detection and localization | |
| CN105425172B (en) | Microgrid inverter adaptive failure diagnostic method based on the analysis of multiband flexure | |
| Lu et al. | Thermal Fault Diagnosis of Electrical Equipment in Substations Based on Image Fusion. | |
| Liao et al. | Recognizing noise-influenced power quality events with integrated feature extraction and neuro-fuzzy network | |
| Rodrigues et al. | Deep learning for power quality event detection and classification based on measured grid data | |
| Ye et al. | Application of bat algorithm for texture image classification | |
| Pan et al. | Study on intelligent anti–electricity stealing early-warning technology based on convolutional neural networks | |
| Li et al. | DC arc fault risk degree evaluation based on back propagation neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |