CN116500391A - Fault arc detection method, system and storage medium based on frequency domain characteristics - Google Patents
Fault arc detection method, system and storage medium based on frequency domain characteristics Download PDFInfo
- Publication number
- CN116500391A CN116500391A CN202310433279.1A CN202310433279A CN116500391A CN 116500391 A CN116500391 A CN 116500391A CN 202310433279 A CN202310433279 A CN 202310433279A CN 116500391 A CN116500391 A CN 116500391A
- Authority
- CN
- China
- Prior art keywords
- fault
- data
- arc
- arc detection
- fault 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 124
- 238000004590 computer program Methods 0.000 claims abstract description 15
- 238000001914 filtration Methods 0.000 claims abstract description 13
- 238000012544 monitoring process Methods 0.000 claims abstract description 9
- 239000013598 vector Substances 0.000 claims description 27
- 238000012549 training Methods 0.000 claims description 21
- 238000000034 method Methods 0.000 claims description 19
- 238000012360 testing method Methods 0.000 claims description 15
- 230000005611 electricity Effects 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 13
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 11
- 230000009466 transformation Effects 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 10
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 230000005540 biological transmission Effects 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 6
- 238000003062 neural network model Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 2
- 238000006243 chemical reaction Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 230000000737 periodic effect Effects 0.000 description 4
- 230000032683 aging Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000002265 prevention Effects 0.000 description 2
- 108010076504 Protein Sorting Signals Proteins 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000003750 conditioning effect Effects 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 230000009849 deactivation Effects 0.000 description 1
- 230000034994 death Effects 0.000 description 1
- 231100000517 death Toxicity 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000009545 invasion Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
- H02J13/00002—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by monitoring
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
- H02J13/00001—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by the display of information or by user interaction, e.g. supervisory control and data acquisition systems [SCADA] or graphical user interfaces [GUI]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
- H02J13/00032—Systems characterised by the controlled or operated power network elements or equipment, the power network elements or equipment not otherwise provided for
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
- H02J13/00032—Systems characterised by the controlled or operated power network elements or equipment, the power network elements or equipment not otherwise provided for
- H02J13/00036—Systems characterised by the controlled or operated power network elements or equipment, the power network elements or equipment not otherwise provided for the elements or equipment being or involving switches, relays or circuit breakers
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/10—Power transmission or distribution systems management focussing at grid-level, e.g. load flow analysis, node profile computation, meshed network optimisation, active network management or spinning reserve management
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2310/00—The network for supplying or distributing electric power characterised by its spatial reach or by the load
- H02J2310/10—The network having a local or delimited stationary reach
- H02J2310/12—The local stationary network supplying a household or a building
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
- Y04S10/52—Outage or fault management, e.g. fault detection or location
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Computer Interaction (AREA)
- Remote Monitoring And Control Of Power-Distribution Networks (AREA)
Abstract
The invention discloses a fault arc detection method, a fault arc detection system and a fault arc detection storage medium based on frequency domain characteristics: acquiring intelligent breaker data; filtering and detecting with low power, wherein if the arc is not generated, if the arc is normal, performing fast Fourier transform, and transmitting the characteristics into an arc detection model 1; a predicted value of 1 indicates an arc; the predicted value of 0 indicates no arc, and a period is directly taken and sent into an arc detection model 2; a predicted value of 1 indicates an arc; a predictive value of 0, indicating no arcing; the fault arc detection system comprises an intelligent circuit breaker, a server and a community monitoring client; the server comprises a memory and a processor, wherein the memory stores a computer program which can run on the processor, and the processor realizes the arc detection method when executing the computer program; the computer readable storage medium stores a computer program thereon, which when executed by a processor, implements the arc detection method described above. The invention has the advantages of stability, strong portability and higher accuracy.
Description
Technical Field
The invention relates to a method and a skill in the fields of machine learning, artificial intelligence and the like, which can be classified into anomaly detection and analysis of time sequences and signal processing of high frequency data, in particular to a fault arc detection method, a fault arc detection system and a storage medium based on frequency domain characteristics.
Background
At present, fire prevention and control in China has various fire phenomena, large management volume, urban risk aggregation, weak infrastructure and other troublesome problems. According to the data of the fire rescue bureau of the emergency management part, along with the perfection of fire rescue work and the enhancement of national fire protection consciousness in recent years, the total number of fire disaster occurrence in China is steadily reduced, but the data still feel tired. In 2021, china receives 74.8 ten thousand fire disasters, deaths 1987, injuries 2225, and direct property loss 67.5 hundred million yuan. The number of resident fires is high, the proportion of resident fires is about 44% and wander, and the proportion of resident fires is largest in the fires classified according to the places of occurrence. For rural fires, the number of residential fires is only 34.5% of the total, but the number of dead people is 73.8% of the total. From the direct cause of the fire, it accounts for 28.4% of the fire caused by electricity, and more than one third of the fire caused by electricity, and it accounts for almost eight times of the total number of electric fires due to the majority of electric circuit faults.
The occurrence of a fault arc is often less predictable, as for example, problems with line aging, rechargeable battery aging, etc. can be predicted in advance, or can be measured. The fault arc generation and the aging of the electrical appliance are not related greatly, and the research of a proper fault arc detection algorithm is very important.
Application number CN202211617734.5, the invention name is: fault arc detection methods, apparatus, devices, and storage media. The invention adopts current information for a load end through a series of specific hardware devices and storage media, then uses wavelet coefficients to fit, and selects two wavelet coefficients by artificial experience to refer to fault arc characteristics. The method is only suitable for specific load detection, adopts an intrusion mode, and the hardware device is matched with the fault detection method and cannot be transplanted.
Application number CN202211272839.1, the invention name is: a dual mode multiple load loop arc fault detection system. The invention takes stm32H7 as a core microprocessor, and forms a dual-mode fault diagnosis system by matching a conditioning circuit, a power supply circuit, a data acquisition circuit and a wireless communication circuit. The method is hardware-level detection as a whole, and cannot be used in a "non-invasive" way.
Disclosure of Invention
The invention aims to overcome the defects of the existing arc detection method and strives to realize the detection of whether an arc is generated in a real household electricity environment in a non-invasive manner. In order to achieve the purpose, the invention provides a fault arc detection method, a fault arc detection system and a fault arc detection storage medium based on frequency domain characteristics, wherein the intelligent circuit breaker is used for collecting real current intensity information in a household electricity environment, and the information is collected on a household electricity main switch (namely the intelligent circuit breaker) which takes a household as a unit, and a software-level algorithm is used for completing detection of the fault arc, so that the purpose of non-invasiveness is achieved.
The aim of the invention can be achieved by the following technical scheme.
The invention discloses a fault arc detection method based on frequency domain characteristics, which comprises the following steps:
the first step: counting the number of intelligent circuit breakers, wherein each intelligent circuit breaker corresponds to one transmission line, each thread is responsible for a certain number of intelligent circuit breakers by using a multithreading algorithm, and data vectors generated by each transmission line are uploaded to a server database through the intelligent circuit breakers to store line information, time information and current intensity information;
and a second step of: extracting a data vector from a server database, and performing low-power filtering detection; if the low-power filtering detection result is abnormal, the existence of no-load and interference is indicated, and then fault-free arc is indicated; if the low-power filtering detection result is normal, turning to the third step;
and a third step of: performing fast Fourier transform on the extracted data vector, mapping the 510-dimensional time domain data into 256-dimensional frequency domain features at the moment, and sending the 256-dimensional frequency domain features into the trained arc detection model 1; if the predicted value of the arc detection model 1 is 1, the arc detection model 1 considers that the detected data contains fault arcs; if the predicted value of the arc detection model 1 is 0, the arc detection model 1 considers that the detected data does not contain fault arcs, and the process goes to a fourth step;
fourth step: directly taking a period of the data vector, and sending the data vector into the trained arc detection model 2; if the predicted value of the arc detection model 2 is 1, a fault arc is indicated; if the predicted value of the arc detection model 2 is 0, it indicates that there is no fault arc.
In the second step, the low power filtering detection means that a power threshold P is set, and if the 510-dimension value of the data vector is smaller than 100, the data is considered to be abnormal data, and no fault arc is contained; if any one dimension data is greater than 100, then it is considered normal data.
In the third step, the arc detection model 1 and the arc detection model 2 in the fourth step both adopt BP neural network models, the network structures of the two models are the same, and the BP neural network adopts a three-layer linear full-connection network mode, wherein the BP neural network comprises 256-node input layers, 200-node hidden layers, a dropout operation layer, an activation function layer and 2-node output layers.
Constructing a data set 1 and a data set 2 for training, testing and evaluating the arc detection model 1 and the arc detection model 2;
there are two types of fault arcs: fault arc type 1 and fault arc type 2; the fault arc type 1 is characterized in that the current intensity suddenly oscillates with the change of time; the fault arc type 2 is characterized in that the current intensity is briefly reset to zero and shows periodicity after the change of time;
acquiring data of fault arc type 1, fault arc type 2 and normal circuit through an intelligent circuit breaker, wherein the data of the fault arc type 1 and the fault arc type 2 are marked with tags [1,0] with fault arcs, and the data of the normal circuit are marked with tags [0,1] with non-fault arcs; the data of the fault arc type 1 and the normal circuit are subjected to fast Fourier transform to form a data set 1; the data of the fault arc type 2 and the normal circuit are directly extracted for one period without fast Fourier transformation to form a data set 2; processing the data set 1 and the data set 2 into npz format files by using python language respectively, and storing the data;
data set 1 and data set 2 were each scaled 8:2 is divided into a training set and a testing set, wherein the training set is used for training an arc detection model, the testing set is used for testing and evaluating the arc detection model, and the trained arc detection model 1 and the trained arc detection model 2 are stored.
The aim of the invention can be achieved by the following technical scheme.
The invention discloses a fault arc detection system based on frequency domain characteristics, which comprises an intelligent circuit breaker, a server and a community monitoring client; the intelligent circuit breaker is used for uploading the data vector generated by each line to the server; the server comprises a memory and a processor, wherein a computer program capable of running on the processor is stored in the memory, and the processor realizes the fault arc detection method based on the frequency domain characteristics when executing the computer program; if the fault arc is found, the phenomenon of unsafe electricity consumption is indicated, the time and user information of the fault arc are accurately sent to a community monitoring client, and community management staff manage the fault arc in a unified mode.
The aim of the invention can be achieved by the following technical scheme.
The invention relates to a computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, and the computer program is characterized in that the fault arc detection method based on the frequency domain characteristics is realized when the computer program is executed by a processor.
Compared with the prior art, the technical scheme of the invention has the following beneficial effects:
(1) The invention does not consider two types of electric arcs, and respectively applies two characteristic extraction modes of a time domain and a frequency domain: firstly, the frequency domain characteristics of current period data are extracted by using a fast Fourier method, the fast Fourier transform can extract the frequency domain characteristics of signals with lower computational complexity, and the nature that fault arcs do not have periodicity is effectively extracted, so that the performance of a model facing a real practical electric environment detection scene is improved; secondly, the time domain characteristics of the current data are directly applied, information is not lost at all, and effective detection is carried out.
(2) The invention relies on a high-quality fault arc labeling data set, and two different neural network models are respectively trained to detect the fault arc by considering two types of the fault arc, thereby improving the detection accuracy rate.
(3) In order to realize the purpose of non-invasion, the invention only needs to install an intelligent breaker for each user, thus the household environment of the user is not invaded, and the total electricity consumption data is only needed to be collected, thereby judging whether fault arc is generated in the electricity consumption process of a certain user, the method and the system based on the frequency domain characteristics can effectively detect whether the fault arc exists in the household environment, so as to achieve the effect of early warning, and promote the development of fire prevention technology
Drawings
Fig. 1 is a flow chart of a fault arc detection method based on frequency domain features of the present invention.
Fig. 2 is a topology diagram of a fault arc detection system device based on frequency domain features of the present invention.
Fig. 3 is a diagram of a BP network architecture of two arc detection models (note: the network architecture of both models is identical).
FIG. 4 is a schematic diagram of a fault arc and normal current;
wherein 4.1 arc fault type 1,4.2 arc fault type 2,4.3 current of the normal circuit.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
As shown in fig. 1, the fault arc detection method based on the frequency domain features of the invention comprises the following steps:
the first step: counting the number of intelligent circuit breakers, wherein each intelligent circuit breaker corresponds to a transmission line, and each thread receives a certain amount of intelligent circuit breaker data in a multi-thread mode. Thereby meeting the real-time requirement. The data vector generated by each transmission line is uploaded to a server database through the intelligent circuit breaker, and line information, time information, current intensity information and the like are stored.
The intelligent circuit breaker collects current intensity information of a formal power utilization environment in real time for 0.05 seconds, and samples 510 data points by a high-frequency sampling method in the 0.05 seconds (two half periods of alternating current) to form a detection data vector, namely, the data vector only comprises the current intensity information at a certain moment. The python language may be used to construct a data receiving program in the server to receive the current intensity information delivered by the intelligent circuit breaker in real time, more than once, and the intelligent circuit breaker delivers the current intensity information for a time of 0.05 seconds to a database in the server through a wireless network, each piece of data being 510 data points. Because the home power environment is a 50HZ ac power environment, the 510 data points contain two half-cycle current intensity information. The detection we do is: based on the 510 data points only, it is determined whether a fault arc exists at a certain time, and one piece of data is detected each time, and the loop is executed.
And a second step of: extracting a data vector from a server database, and performing low-power filtering detection; if the low-power filtering detection result is abnormal, the existence of no-load and interference is indicated, so that no-fault arc is further indicated, and the detection is finished; if the low-power filtering detection result is normal, the method goes to the third step.
Because, influenced by the actual electricity ring, a certain circuit is empty, and the intelligent circuit breaker can still collect weak current, and the current file of this part is filtered in advance, thereby being beneficial to more accurate detection later. The low-power filtering detection means that a power threshold value is set, and is assumed to be P (a natural value, the invention can be set to 100), if the 510-dimension value of a data vector is smaller than 100, the data vector is considered to be abnormal data, no fault arc exists, and the detection is finished; if any one dimension data is greater than 100, the data is considered to be normal data, and detection is continued.
And a third step of: the extracted data vector is subjected to fast fourier transform, at this time, the 510-dimensional time domain data is mapped into 256-dimensional frequency domain features, and the 256-dimensional frequency domain features are sent to the trained arc detection model 1. If the predicted value of the arc detection model 1 is 1, the arc detection model 1 considers that the detected data contains fault arc, and the detection is finished; if the predicted value of the arc detection model 1 is 0, it means that the arc detection model 1 considers that the detected data does not contain a fault arc, and the process proceeds to the fourth step.
Fourth step: the data vector is directly taken for one period (about 220 points) without fast Fourier transformation, and then is sent into the trained arc detection model 2; if the predicted value of the arc detection model 2 is 1, the fault arc is indicated, and the detection is finished; if the predicted value of the arc detection model 2 is 0, no fault arc is detected, and the detection is ended.
In the process, if a fault arc is found, the phenomenon of unsafe electricity consumption is indicated, the time of occurrence of the fault arc, user information and the like are accurately sent to a community monitoring client, and community management staff manage the fault arc in a unified mode.
The following detailed description is directed to several knowledge points involved in the process of the present invention as described above.
1. Fast fourier transform
The operation of the fast fourier transform is completed by applying the fast fourier transform to the first 510-dimensional data vector (510-point data). To this end, the raw data 510-dimensional vector is processed into 256-dimensional vector. The conversion from the time domain signal to the frequency domain signal is completed, and the fault arc characteristics are better distinguished through the space conversion. The formulation of the fast fourier transform is described in detail herein.
For the continuous-time signal x (t), if x (t) is integrable in the time dimension, (actually, it is not necessarily the time t dimension, but may be any dimension, and only needs to be integrable in the corresponding dimension space), that is:
then, the fourier transform of x (t) exists and is calculated as:
the inverse transformation is as follows:
the energy of the function or signal can be expanded by Fourier transformation, X (j omega) obtained by expansion becomes a function of a frequency domain, if a curve is drawn on the frequency, the function is a spectrogram, the inverse transformation is better understood, and if a signal or a function spectral density function is known, the function of a time domain can be correspondingly restored, and a waveform diagram of the time domain can be drawn.
Fourier transforms can be divided into four categories depending on the type of original signal:
1. aperiodic continuous signal: fourier transform (Fourier Transform, abbreviated FT)
2. Periodic continuous signal: fourier Series (Fourier Series, abbreviated as FS)
3. Aperiodic discrete signal: discrete time fourier transform (Discrete Time Fourier Transform, abbreviated as DTFT)
4. Periodic discrete signal: discrete fourier transform (Discrete Fourier Transform, abbreviated DFT)
The invention uses a fast fourier transform (Fast Fourier Transform, abbreviated FFT) which is a calculation method of the discrete fourier transform.
A discrete fourier transform (Discrete Fourier Transform, abbreviated DFT) is a form of fourier transform that is discrete in both the time and frequency domains, transforming the time domain samples of a signal into frequency domain samples of its DTFT.
Formally, the sequences at both ends of the transformation (in the time and frequency domains) are of finite length, and in practice both sets of sequences should be considered as the main sequence of values of the discrete periodic signal. Even if a DFT is performed on a discrete signal of finite length, it should be considered as a transform of its period extension. In practical applications, the DFT is typically calculated using a fast fourier transform.
For the N-point sequence x (N) 0≤n≤N Its Discrete Fourier Transform (DFT) is:
where k=0, 1,.. the above formula is expanded:
the fast fourier transform (Fast Fourier Transform: FFT) is an algorithm that calculates the discrete fourier transform (Discrete Fourier Transform: DFT) or inverse transform (IDFT) of a digital signal sequence. Fourier analysis converts a signal from its original domain (typically time or space) to a representation of the frequency domain and vice versa. The DFT is obtained by decomposing a series of values into components of different frequencies. This operation is useful in many areas, but it is often too slow to be practical to calculate it directly from the definition. The FFT computes this conversion quickly by decomposing the DFT matrix into products of sparse (mostly zero) factors. The essence is an optimization algorithm for realizing discrete Fourier transform, and the time complexity is reduced. When N is very large, this optimization is very significant in the time dimension. Especially in the embedded application field, the effect of the FFT algorithm compared with the DFT is very valuable because the computational power of the chip limited to be adopted is not strong.
The most core idea of the fast Fourier algorithm is a common divide-and-conquer idea in computer science, namely, a complex problem is decomposed into a small similar problem to be solved.
FFTs can be largely divided into two categories, time extraction and frequency extraction, which can only handle lengths n=2 M M is a natural number. The extraction is a process of dividing a long sequence into short sequences, and may be performed in the time domain or the frequency domain. The most commonly used time domain decimation method is to change long sequences into short sequences continuously according to parity, so that the input sequences are in reverse order, and the output sequences are arranged in sequence, which is a kuli-base fast fourier transform algorithm (simply referred to as cool-Tukey algorithm).
Assume that the discrete time series signal length to be converted is n=2 M Grouping x (n) by parity of n:
the above can be transformed into:
and (3) making:
wherein k is 0,1, & gt, N/2-1; a (k) represents n as an odd number group, and B (k) represents n as an even number group; w represents rotationThe weight of the transformation process is determined,representing the weight of the kth point in the N point sequences during the conversion, +.>Representing the weight of the 2rk point in the N point sequence during the conversion,/for>Representing the weight of the (2r+1) th point in the N point sequences during the conversion,/for the (2r+1) th point>Representation->Weight of the rk th point in the sequence of points during conversion, +.>Representation->The weight of the kth point in the sequence of points in the conversion process.
Thereby obtaining the result of the fast fourier transform:
2. arc detection model 1 and arc detection model 2
Analyzing the type of fault arc and deciding to train two BP neural network models respectively. Fig. 4.1 represents one type of fault arc, fig. 4.2 represents another type of fault arc, and fig. 4.3 represents current information of a normal circuit. Looking at the visualized image, we find that: the difference between the fault arc shown in fig. 4.2 and the normal current is that the fault arc of fig. 4.2 does not have periodicity, whereas both the current information of the normal circuit and the fault arc of fig. 4.1 have significant periodicity information. We must therefore be able to separate the fault arc shown in fig. 4.2 from the normal appliance current and the fault arc shown in fig. 4.1 by a fast fourier transform and then using a BP neural network classifier. How does the fault arc shown in fig. 4.1 separate from the normal appliance current? We used a further training of a BP neural network to distinguish between the fault arc and normal appliance current shown in fig. 4.1.
The current intensity data in the time domain range is subjected to fast Fourier transform to the frequency domain data, so that the periodic characteristics of the data can be clearly extracted, the data in the normal power utilization environment are taught to have strong periodicity by experience observation, and the data with electric arcs often have no periodicity, so that the data can be well classified by utilizing the method. There are two types of fault arcs: fault arc type 1 and fault arc type 2. The nature of the fault arc type 1 is such that the current intensity will oscillate suddenly over time, as illustrated in fig. 4.2. The fault arc type 2 is characterized by a temporal change, a brief return of the current intensity to zero, and a periodicity, as illustrated in fig. 4.1.
The type of fault arc is analyzed, and two BP neural network models are respectively trained to serve as an arc detection model 1 and an arc detection model 2. We have to construct data set 1 and data set 2 for training and testing arc detection model 1, arc detection model 2, respectively.
The data of the fault arc type 1, the fault arc type 2 and the normal circuit are obtained through the intelligent circuit breaker, the data of the fault arc type 1 and the fault arc type 2 are marked with tags [1,0] with fault arcs, and the data of the normal circuit are marked with tags [0,1] with non-fault arcs. The data of the fault arc type 1 and the normal circuit are subjected to fast fourier transformation to form a data set 1. The data of the fault arc type 2 and the normal circuit are directly extracted for one period without fast Fourier transformation to form a data set 2. Data set 1 and data set 2 are processed into npz format files by using python language respectively, and data are saved. Data set 1 and data set 2 are each proportional 8:2 is divided into a training set and a testing set, wherein the training set is used for training the arc detection model, and the testing set is used for testing and evaluating the arc detection model.
For example: the data set 1 is constructed as follows: fault arc data file shown in fig. 4.2, fault arc tags [1,0] are attached, data file shown in fig. 4.3, non-fault arc tags [0,1] are attached, and then a data set with a scale of 10K, a training set and a test set are formed through fast fourier transformation, wherein the ratio of the training set to the test set is 8:2. the data set 2 is constructed as follows: the fault arc data file shown in fig. 4.1, labeled with fault arc tag [1,0], the data file shown in fig. 4.3, labeled with non-fault arc tag [0,1], were not subjected to a fast fourier transform, but were directly extracted for one cycle (about 220 data points), forming a 10K scale data set, training set and test set ratio 8:2.
any function can be approximated by applying a suitable BP network structure, which is an approximation theorem. That is, as long as the network structure and the parameter design are reasonable, the fitting capability of the BP neural network is completely available, the task data is not huge, so that the fully connected network is adopted, and then the parameter is optimized and the network structure is adjusted according to the actual situation. Two BP neural network models can be constructed by using the python language and are respectively used as an arc detection model 1 and an arc detection model 2, the network structures of the two models are identical, the BP neural network adopts a three-layer linear full-connection network mode, and the constructed network structure is shown in figure 3. In particular, the first layer is 256-node input layer (input data dimension), the second layer is 200-node hidden layer, the third layer is 2-node output layer (output independent heat vector), and in the practical process, an activation function layer (relu layer) and a dropout operation layer (random deactivation layer, which can be deactivated by 20% of nodes) are arranged between the hidden layer and the output layer for better performance. The neural network learning rate, the dropout percentage, the maximum iteration number and the like can be set, and the optimal parameters are obtained through training and adjustment of long-time real data.
Assuming that the output layer node value of the neural network is [ x1, x2], and the label of this sample is [1,0] (defining [1,0] as a faulty arc data label, [0,1] as no faulty arc data label), we aim to make the similarity between the output of a certain sample [ x1, x2] and the label highest, and this similarity is characterized by a cross entropy function.
Wherein L represents the loss degree, L i Represents the loss degree of the ith sample, p i Indicating the probability that the ith sample is a fault arc, y i Indicating the probability that the real tag is a fault arc, N 1 Representing the number of samples, our optimization goal is to minimize L. Training the model by using a large number of samples, and after the model is optimal, obtaining an optimal solution, namely the optimal weight of the constructed neural network in the task, and storing the optimal weight.
And training the arc detection model 1 and the arc detection model 2 by using the data set 1 and the data set 2 respectively, and storing the optimal arc detection model 1 and the optimal arc detection model 2 after training according to the training principle of the neural network. And detecting user electricity data uploaded in real time by using the trained arc detection model 1 and the arc detection model 2, and giving an alarm to a community monitoring client if abnormality is found.
The detection method of the invention has stable service for 3 months in 3000 users in Shenzhen certain district, the accuracy rate reaches more than 99%, and the time proves that the algorithm is stable and practical.
The invention provides a fault arc detection system based on frequency domain characteristics, which is shown in fig. 2 and comprises an intelligent circuit breaker, a server and a community monitoring client; the intelligent circuit breaker is used for uploading the data vector generated by each line to the server; the server comprises a memory and a processor, wherein a computer program capable of running on the processor is stored in the memory, and the processor realizes the fault arc detection method based on the frequency domain characteristics when executing the computer program; if the fault arc is found, the phenomenon of unsafe electricity consumption is indicated, the time and user information of the fault arc are accurately sent to a community monitoring client, and community management staff manage the fault arc in a unified mode.
The present invention proposes a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the above-described fault arc detection method of the present invention based on frequency domain features.
Although the function and operation of the present invention has been described above with reference to the accompanying drawings, the present invention is not limited to the above-described specific functions and operations, but the above-described specific embodiments are merely illustrative, not restrictive, and many forms can be made by those having ordinary skill in the art without departing from the spirit of the present invention and the scope of the appended claims, which are included in the protection of the present invention.
Claims (6)
1. The fault arc detection method based on the frequency domain features is characterized by comprising the following steps of:
the first step: counting the number of intelligent circuit breakers, wherein each intelligent circuit breaker corresponds to one transmission line, each thread is responsible for a certain number of intelligent circuit breakers by using a multithreading algorithm, and data vectors generated by each transmission line are uploaded to a server database through the intelligent circuit breakers to store line information, time information and current intensity information;
and a second step of: extracting a data vector from a server database, and performing low-power filtering detection; if the low-power filtering detection result is abnormal, the existence of no-load and interference is indicated, and then fault-free arc is indicated; if the low-power filtering detection result is normal, turning to the third step;
and a third step of: performing fast Fourier transform on the extracted data vector, mapping the 510-dimensional time domain data into 256-dimensional frequency domain features at the moment, and sending the 256-dimensional frequency domain features into the trained arc detection model 1; if the predicted value of the arc detection model 1 is 1, the arc detection model 1 considers that the detected data contains fault arcs; if the predicted value of the arc detection model 1 is 0, the arc detection model 1 considers that the detected data does not contain fault arcs, and the process goes to a fourth step;
fourth step: directly taking a period of the data vector, and sending the data vector into the trained arc detection model 2; if the predicted value of the arc detection model 2 is 1, a fault arc is indicated; if the predicted value of the arc detection model 2 is 0, it indicates that there is no fault arc.
2. The fault arc detection method based on frequency domain features as in claim 1, wherein in the second step the low power filtering detection means that a power threshold P is set, and if the 510 dimensions of the data vector are each smaller than 100, it is considered as abnormal data, and no fault arc is contained; if any one dimension data is greater than 100, then it is considered normal data.
3. The fault arc detection method based on the frequency domain features according to claim 1, wherein in the third step, the arc detection model 1 and the arc detection model 2 in the fourth step both adopt BP neural network models, the network structures of the two models are the same, and the BP neural network adopts a three-layer linear full-connection network mode, which comprises an input layer of 256 nodes, an hidden layer of 200 nodes, a dropout operation layer, an activation function layer, and an output layer of 2 nodes.
4. A fault arc detection method based on frequency domain features as in claim 3, wherein for training, testing, evaluating arc detection model 1, arc detection model 2, data set 1 and data set 2 are constructed;
there are two types of fault arcs: fault arc type 1 and fault arc type 2; the fault arc type 1 is characterized in that the current intensity suddenly oscillates with the change of time; the fault arc type 2 is characterized in that the current intensity is briefly reset to zero and shows periodicity after the change of time;
acquiring data of fault arc type 1, fault arc type 2 and normal circuit through an intelligent circuit breaker, wherein the data of the fault arc type 1 and the fault arc type 2 are marked with tags [1,0] with fault arcs, and the data of the normal circuit are marked with tags [0,1] with non-fault arcs; the data of the fault arc type 1 and the normal circuit are subjected to fast Fourier transform to form a data set 1; the data of the fault arc type 2 and the normal circuit are directly extracted for one period without fast Fourier transformation to form a data set 2; processing the data set 1 and the data set 2 into npz format files by using python language respectively, and storing the data;
data set 1 and data set 2 were each scaled 8:2 is divided into a training set and a testing set, wherein the training set is used for training an arc detection model, the testing set is used for testing and evaluating the arc detection model, and the trained arc detection model 1 and the trained arc detection model 2 are stored.
5. The fault arc detection system based on the frequency domain features is characterized by comprising an intelligent circuit breaker, a server and a community monitoring client; the intelligent circuit breaker is used for uploading the data vector generated by each line to the server; the server comprises a memory and a processor, wherein a computer program capable of running on the processor is stored in the memory, and the processor realizes the fault arc detection method based on the frequency domain characteristics when executing the computer program; if the fault arc is found, the phenomenon of unsafe electricity consumption is indicated, the time and user information of the fault arc are accurately sent to a community monitoring client, and community management staff manage the fault arc in a unified mode.
6. A computer readable storage medium having stored thereon a computer program, characterized in that the fault arc detection method based on frequency domain features of any of claims 1 to 4 is implemented when the computer program is executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310433279.1A CN116500391A (en) | 2023-04-21 | 2023-04-21 | Fault arc detection method, system and storage medium based on frequency domain characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310433279.1A CN116500391A (en) | 2023-04-21 | 2023-04-21 | Fault arc detection method, system and storage medium based on frequency domain characteristics |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116500391A true CN116500391A (en) | 2023-07-28 |
Family
ID=87324107
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310433279.1A Pending CN116500391A (en) | 2023-04-21 | 2023-04-21 | Fault arc detection method, system and storage medium based on frequency domain characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116500391A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116933044A (en) * | 2023-09-15 | 2023-10-24 | 深圳高新区信息网有限公司 | Intelligent processing method and system for power supply data |
-
2023
- 2023-04-21 CN CN202310433279.1A patent/CN116500391A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116933044A (en) * | 2023-09-15 | 2023-10-24 | 深圳高新区信息网有限公司 | Intelligent processing method and system for power supply data |
| CN116933044B (en) * | 2023-09-15 | 2023-11-24 | 深圳高新区信息网有限公司 | Intelligent processing method and system for power supply data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | ArcNet: Series AC arc fault detection based on raw current and convolutional neural network | |
| CN108021736B (en) | Load switching action monitoring method based on sliding window residual error model | |
| CN108508320B (en) | Arc grounding fault identification method based on harmonic energy and wave distortion feature | |
| CN110514889A (en) | A kind of method and system of non-intrusion type household electricity remained capacity | |
| CN110133393B (en) | Power consumption monitoring system and method based on non-invasive monitoring technology | |
| KR100918313B1 (en) | Method for diagnosis and analysis of electric power quality using artificial intelligence | |
| CN116500391A (en) | Fault arc detection method, system and storage medium based on frequency domain characteristics | |
| Held et al. | Frequency invariant transformation of periodic signals (FIT-PS) for classification in NILM | |
| CN111579939A (en) | Method for detecting partial discharge phenomenon of high-voltage power cable based on deep learning | |
| CN112748318A (en) | Series fault arc detection method based on improved PSO-BP neural network | |
| Gurbuz et al. | Comprehensive non-intrusive load monitoring process: Device event detection, device feature extraction and device identification using KNN, random forest and decision tree | |
| CN117849691A (en) | Multi-dimensional collaborative operation monitoring and early warning system and method for capacitive voltage transformer | |
| Zhu et al. | Optimization research on abnormal diagnosis of transformer voiceprint recognition based on improved wasserstein GAN | |
| Xiong et al. | Application of multi-kernel relevance vector machine and data pre-processing by complementary ensemble empirical mode decomposition and mutual dimensionless in fault diagnosis | |
| Wilson et al. | Uncertainty Quantification of Capacitor Switching Transient Location Using Machine Learning | |
| Xu et al. | A Wavelet Transform and ResNet Based Method for Sensing the Operational Status of Secondary Equipment in the Distribution Room | |
| CN110888013A (en) | Fault prediction method and device of power equipment and power monitoring system | |
| Wang et al. | Application of cloud model and matter element theory in transformer fault diagnosis | |
| Nguyen et al. | 1-d convolutional graph convolutional networks for fault detection in distributed energy systems | |
| Zhao et al. | Incipient fault detection and condition assessment in DFIGs based on external leakage flux sensing and modified multiscale poincare plots analysis | |
| CN112198361A (en) | Electric singular point monitoring device based on frequency energy ratio | |
| Wang et al. | Lcl: Light contactless low-delay load monitoring via compressive attentional multi-label learning | |
| Quan et al. | The application of Bayesian network theory in transformer condition assessment | |
| CN114740261B (en) | Harmonic sensing method, system, device and storage medium | |
| Wang et al. | Diagnosis for loose switching contact fault of on-load tap-changer in transformer |
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 |