CN112834959A - Direct-current power distribution system fault detection method based on high-frequency feature extraction - Google Patents

Direct-current power distribution system fault detection method based on high-frequency feature extraction Download PDF

Info

Publication number
CN112834959A
CN112834959A CN202110001588.2A CN202110001588A CN112834959A CN 112834959 A CN112834959 A CN 112834959A CN 202110001588 A CN202110001588 A CN 202110001588A CN 112834959 A CN112834959 A CN 112834959A
Authority
CN
China
Prior art keywords
fault
direct current
power distribution
feature extraction
current power
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110001588.2A
Other languages
Chinese (zh)
Inventor
洪翠
连淑婷
郭谋发
高伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fuzhou University
Original Assignee
Fuzhou University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Fuzhou University filed Critical Fuzhou University
Priority to CN202110001588.2A priority Critical patent/CN112834959A/en
Publication of CN112834959A publication Critical patent/CN112834959A/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01RMEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/50Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
    • G01R31/56Testing of electric apparatus

Abstract

The invention relates to a direct current power distribution system fault detection method based on high-frequency feature extraction, which comprises the following steps of S1: constructing a fault classification model combining an automatic encoder and a support vector machine; step S2, mounting the protection device on the direct current line, and collecting the current and voltage waveform of the direct current line to obtain the positive voltage upStep S3, self-adaptively decomposing the current waveform by adopting an empirical wavelet algorithm to obtain an empirical wavelet modal componentf 0~2(ii) a Step S4, calculatingf 2The maximum amplitude of the component and comparing with a set threshold to detect a fault; step S5, according to the obtained empirical wavelet modal componentf 0~2And positive electrode voltage upAnd fusing the constructed characteristic vectors, inputting the constructed characteristic vector matrix into a fault classification model, and dividing the fault types. The invention effectively improves the fault detection efficiency and accuracy of the direct current power distribution system.

Description

Direct-current power distribution system fault detection method based on high-frequency feature extraction
Technical Field
The invention relates to the field of power distribution system fault detection, in particular to a direct-current power distribution system fault detection method based on high-frequency feature extraction.
Background
The direct-current power distribution network has a series of advantages of improving the transmission capacity of a line, facilitating the access of distributed new energy and energy storage equipment, improving the quality of electric energy, reducing the loss of the line and the like, and provides a new solution for the development of urban power distribution networks. However, the rising speed of the fault current of the direct-current power distribution network is high, the characteristic of a natural zero crossing point is not provided, the current value can be kept at a high level after the fault enters a steady state, high requirements on the aspects of quickness, sensitivity and selectivity of direct-current fault treatment are provided, and the quick and reliable fault detection is more significant.
At present, methods for detecting faults of a direct-current power distribution network can be roughly divided into two main categories: one is fault detection based on electrical quantities, which are usually current voltages on dc lines. The methods directly taking the amplitude of the electrical quantity as the fault detection criterion include overcurrent and overvoltage detection and current reverse zero crossing point detection, the methods are simple and easy to implement, but the identification speed is difficult to meet the requirement of quick action, and the method is greatly influenced by fault transition resistance, is easy to generate false detection and missing detection, and generally needs to be matched with other detection means for use. More commonly, some processing means is applied to the electrical quantity, for example, fourier transform, wavelet transform, empirical mode decomposition, etc. are used as signal processing methods, and intelligent processing methods include automatic encoders, neural networks, random forests, etc., and mathematical differential processing. The signal processing method can accurately capture singular points of the electrical quantity, extract fault characteristics, and is high in reliability and less affected by fault transition resistance. The intelligent processing method has the advantages that the advantages are gradually revealed, the fault characteristics do not need to be manually selected, the reliability is good, and the fault detection model of the corresponding power distribution network needs to be trained in advance. The method of electrical quantity differentiation processing increases the speed with respect to overcurrent overvoltage detection, but still fails to overcome the disadvantage of being greatly affected by the fault transition resistance. Secondly, communication-based fault detection needs to be carried out, communication equipment is arranged at two ends of a direct-current line, an electric quantity analog signal is converted into a digital signal, and a fault is detected through comparison of difference values of the digital signals at the two ends, so that the fault identification speed is high due to the fact that analog-to-digital conversion is fast, but the problem of communication synchronism exists.
Disclosure of Invention
In view of this, the present invention provides a method for detecting a fault of a dc power distribution system based on high-frequency feature extraction, so as to effectively improve detection efficiency and accuracy.
In order to achieve the purpose, the invention adopts the following technical scheme:
a direct current power distribution system fault detection method based on high-frequency feature extraction comprises the following steps
Step S1: constructing a fault classification model combining an automatic encoder and a support vector machine;
step S2, mounting the protection device on the direct current line, and collecting the current and voltage waveform of the direct current line to obtain the positive voltage up
Step S3, self-adaptively decomposing the current waveform by adopting an empirical wavelet algorithm to obtain an empirical wavelet modal component f0~2
Step S4 calculating f2The maximum amplitude of the component and comparing with a set threshold to detect a fault;
step S5, according to the obtained empirical wavelet modal component f0~2And positive electrode voltage upAnd fusing the constructed characteristic vectors, inputting the constructed characteristic vector matrix into a fault classification model, and dividing the fault types.
Further, the step S1 is specifically:
s11: acquiring various fault data and constructing a training set database;
s12: according to the parameters of the automatic encoder trained by the training set, obtaining a fault characteristic h after dimensionality reduction;
s13: training a support vector machine classifier by using the characteristic quantity h obtained in the step S12;
s14: and (4) saving parameters of the automatic encoder and the support vector machine, and establishing a direct-current line fault classification model.
Further, the step S12 is specifically:
a: encoding a signal according to equation (1);
h=f(Wx+b) (1)
wherein W is weight, b is offset, and x is f0~2、upA signal;
b: decoding the fault feature h extracted in the step a according to a formula (2);
wherein W' is weight and d is offset;
c: calculating the signal x and the reconstructed signal according to equation (3)A loss function of (d);
d: and c, judging whether the error calculated in the step c is in accordance with the expectation, if so, saving the weight and the bias parameter of the automatic encoder, and if not, updating the weight and the bias parameter and returning to the step a.
Further, the step S3 is specifically:
s31: dividing a Fourier amplitude spectrum by adopting self-adaption division to obtain N areas;
s32: boundary points ω according to the divided N regionsnConstructing an empirical wavelet function and an empirical scale function;
s33: calculating an empirical wavelet detail coefficient and an approximate coefficient according to the obtained empirical wavelet function and the empirical scale function;
s34: calculating to obtain an empirical wavelet modal component f according to the calculated empirical wavelet detail coefficient and approximation coefficient0~2
Further, the step S31 is specifically:
a: the collected current waveform i is processed according to the formula (4)cPerforming discrete Fourier transform, and obtaining amplitude frequency spectrum by using formula (5)
Where x (M) is the sampling signal, M is the number of data
Where Re (k) denotes the real part and im (k) denotes the imaginary part;
b: finding local maxima in the frequency range, the middle point of the two maxima being set as the boundary omeganFinding out N-1 maximum value points and dividing N regions Lambdan
Further, the empirical wavelet function and the empirical scale function are respectively:
wherein β (x) is x4(35-84x+70x2-20x3),
Further, the step S33 is specifically: calculating empirical wavelet detail coefficients and approximation coefficients according to equation (8) and equation (9):
wherein, F-1Representing an inverse FourierThe result of the transformation is a transformation,is the fourier transform of the signal f (τ).
Further, the step S34 is specifically: calculating empirical wavelet modal component f according to formula (10)n
Further, the step S4 is specifically:
step S41, adopting Hilbert transform to identify f2And the maximum value f of the amplitude of the component is obtained2_Amp
S42: setting a threshold value delta when f is satisfied2_Amp>Delta, the fault is determined to occur.
Further, the step S41 is specifically:
calculating a Hilbert transform of the signal according to formula (11);
calculating an analytic signal according to formula (12);
calculating the instantaneous amplitude of the signal according to equation (13) and obtaining the maximum amplitude value f2_Amp
Compared with the prior art, the invention has the following beneficial effects:
1. the invention adopts the empirical wavelet algorithm to extract the high-frequency component of the current in a self-adaptive manner, thereby effectively reducing the detection time consumption and improving the detection efficiency.
2. The invention has stronger fault transition resistance tolerance capability, further extracts the depth characteristic of the fault by using the automatic encoder, avoids the complexity of manual selection and fault characteristic calculation, and can automatically extract the characteristic.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
fig. 2 is a schematic diagram of an embodiment of a dc distribution network fault line detection.
Detailed Description
The invention is further explained below with reference to the drawings and the embodiments.
Referring to fig. 1, the present invention provides a method for detecting a fault of a dc power distribution system based on high frequency feature extraction, which includes the following steps
Step S1: constructing a fault classification model combining an automatic encoder and a support vector machine;
s11: acquiring various fault data and constructing a training set database;
s12: according to the parameters of the automatic encoder trained by the training set, obtaining a fault characteristic h after dimensionality reduction;
a: encoding a signal according to equation (1);
h=f(Wx+b) (1)
wherein W is weight, b is offset, and x is f0~2、upA signal;
b: decoding the fault feature h extracted in the step a according to a formula (2);
wherein W' is weight and d is offset;
c: calculating the signal x and the reconstructed signal according to equation (3)A loss function of (d);
d: and c, judging whether the error calculated in the step c is in accordance with the expectation, if so, saving the weight and the bias parameter of the automatic encoder, and if not, updating the weight and the bias parameter and returning to the step a.
S13: training a support vector machine classifier by using the characteristic quantity h obtained in the step S12;
s14: and (4) saving parameters of the automatic encoder and the support vector machine, and establishing a direct-current line fault classification model.
Step S2, mounting the protection device on the direct current line, and collecting the current and voltage waveform of the direct current line to obtain the positive voltage up
Step S3, self-adaptively decomposing the current waveform by adopting an empirical wavelet algorithm to obtain an empirical wavelet modal component f0~2
S31: dividing a Fourier amplitude spectrum by adopting self-adaption division to obtain N areas;
a: the collected current waveform i is processed according to the formula (4)cPerforming discrete Fourier transform, and obtaining amplitude frequency spectrum by using formula (5)
Where x (M) is the sampling signal, M is the number of data
Where Re (k) denotes the real part and im (k) denotes the imaginary part;
b: finding local maxima in the frequency range, the middle point of the two maxima being set as the boundary omeganFinding out N-1 maximum value points and dividing N regions Lambdan
S32: boundary points ω according to the divided N regionsnConstructing an empirical wavelet function and an empirical scale function;
wherein β (x) is x4(35-84x+70x2-20x3),
S33: calculating empirical wavelet detail coefficients and approximation coefficients according to equation (8) and equation (9):
wherein, F-1Which represents the inverse fourier transform of the signal,is the fourier transform of the signal f (τ).
S34: calculating empirical wavelet modal component f according to formula (10)n
Step S4 calculating f2The maximum amplitude of the component and comparing with a set threshold to detect a fault;
step S41, adopting Hilbert transform to identify f2And the maximum value f of the amplitude of the component is obtained2_Amp
Calculating a Hilbert transform of the signal according to formula (11);
calculating an analytic signal according to formula (12);
calculating the instantaneous amplitude of the signal according to equation (13) and obtaining the maximum amplitude value f2_Amp
S42: setting a threshold value delta when f is satisfied2_Amp>Delta, the fault is determined to occur.
Step S5, according to the obtained empirical wavelet modal component fnAnd positive electrode voltage upAnd fusing the constructed characteristic vectors, inputting the constructed characteristic vector matrix into a fault classification model, and dividing the fault types.
The above description is only a preferred embodiment of the present invention, and all equivalent changes and modifications made in accordance with the claims of the present invention should be covered by the present invention.

Claims (10)

1. A direct current power distribution system fault detection method based on high-frequency feature extraction is characterized by comprising the following steps
Step S1: constructing a fault classification model combining an automatic encoder and a support vector machine;
step S2, mounting the protection device on the direct current line, and collecting the current and voltage waveform of the direct current line to obtain the positive voltage up
Step S3, self-adaptively decomposing the current waveform by adopting an empirical wavelet algorithm to obtain an empirical wavelet modal component f0~2
Step S4 calculating f2The maximum amplitude of the component and comparing it with a set threshold to detect a faultA barrier;
step S5, according to the obtained empirical wavelet modal component f0~2And positive electrode voltage upAnd fusing the constructed characteristic vectors, inputting the constructed characteristic vector matrix into a fault classification model, and dividing the fault types.
2. The method for detecting faults of a direct current power distribution system based on high-frequency feature extraction as claimed in claim 1, wherein the step S1 specifically includes:
s11: acquiring various fault data and constructing a training set database;
s12: according to the parameters of the automatic encoder trained by the training set, obtaining a fault characteristic h after dimensionality reduction;
s13: training a support vector machine classifier by using the characteristic quantity h obtained in the step S12;
s14: and (4) saving parameters of the automatic encoder and the support vector machine, and establishing a direct-current line fault classification model.
3. The method for detecting faults of a direct current power distribution system based on high-frequency feature extraction as claimed in claim 2, wherein the step S12 specifically includes:
a: encoding a signal according to equation (1);
h=f(Wx+b) (1)
wherein W is weight, b is offset, and x is f0~2、upA signal;
b: decoding the fault feature h extracted in the step a according to a formula (2);
wherein W' is weight and d is offset;
c: calculating the signal x and the reconstructed signal according to equation (3)A loss function of (d);
d: and c, judging whether the error calculated in the step c is in accordance with the expectation, if so, saving the weight and the bias parameter of the automatic encoder, and if not, updating the weight and the bias parameter and returning to the step a.
4. The method for detecting faults of a direct current power distribution system based on high-frequency feature extraction as claimed in claim 1, wherein the step S3 specifically includes:
s31: dividing a Fourier amplitude spectrum by adopting self-adaption division to obtain N areas;
s32: boundary points ω according to the divided N regionsnConstructing an empirical wavelet function and an empirical scale function;
s33: calculating an empirical wavelet detail coefficient and an approximate coefficient according to the obtained empirical wavelet function and the empirical scale function;
s34: calculating to obtain an empirical wavelet modal component f according to the calculated empirical wavelet detail coefficient and approximation coefficient0~2
5. The method for detecting faults of a direct current power distribution system based on high-frequency feature extraction as claimed in claim 4, wherein the step S31 specifically comprises:
a: the collected current waveform i is processed according to the formula (4)cPerforming discrete Fourier transform, and obtaining amplitude frequency spectrum by using formula (5)
Where x (M) is the sampling signal, M is the number of data
Where Re (k) denotes the real part and im (k) denotes the imaginary part;
b: finding local maxima in the frequency range, the middle point of the two maxima being set as the boundary omeganFinding out N-1 maximum value points and dividing N regions Lambdan
6. The method for detecting the fault of the direct current power distribution system based on the high-frequency feature extraction as claimed in claim 5, wherein the empirical wavelet function and the empirical scale function are respectively as follows:
wherein β (x) is x4(35-84x+70x2-20x3),
7. The method for detecting faults of a direct current power distribution system based on high-frequency feature extraction as claimed in claim 6, wherein the step S33 is specifically as follows: calculating empirical wavelet detail coefficients and approximation coefficients according to equation (8) and equation (9):
wherein, F-1Which represents the inverse fourier transform of the signal,is the fourier transform of the signal f (τ).
8. The method for detecting faults of a direct current power distribution system based on high-frequency feature extraction as claimed in claim 6, wherein the step S34 is specifically as follows: calculating empirical wavelet modal component f according to formula (10)n
9. The method for detecting faults of a direct current power distribution system based on high-frequency feature extraction as claimed in claim 1, wherein the step S4 specifically includes:
step S41, adopting Hilbert transform to identify f2And the maximum value f of the amplitude of the component is obtained2_Amp
S42: setting a threshold value delta when f is satisfied2_Amp>Delta, the fault is determined to occur.
10. The method for detecting faults of a direct current power distribution system based on high-frequency feature extraction as claimed in claim 1, wherein the step S41 specifically includes:
calculating a Hilbert transform of the signal according to formula (11);
calculating an analytic signal according to formula (12);
calculating the instantaneous amplitude of the signal according to equation (13) and obtaining the maximum amplitude value f2_Amp
CN202110001588.2A 2021-01-04 2021-01-04 Direct-current power distribution system fault detection method based on high-frequency feature extraction Pending CN112834959A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110001588.2A CN112834959A (en) 2021-01-04 2021-01-04 Direct-current power distribution system fault detection method based on high-frequency feature extraction

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110001588.2A CN112834959A (en) 2021-01-04 2021-01-04 Direct-current power distribution system fault detection method based on high-frequency feature extraction

Publications (1)

Publication Number Publication Date
CN112834959A true CN112834959A (en) 2021-05-25

Family

ID=75927267

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110001588.2A Pending CN112834959A (en) 2021-01-04 2021-01-04 Direct-current power distribution system fault detection method based on high-frequency feature extraction

Country Status (1)

Country Link
CN (1) CN112834959A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113419141A (en) * 2021-08-26 2021-09-21 中国南方电网有限责任公司超高压输电公司广州局 Direct-current line fault positioning method and device

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107193271A (en) * 2017-06-13 2017-09-22 青岛科技大学 Preposition service adapter and trouble-shooter for industrial information physical system
CN107451557A (en) * 2017-07-29 2017-12-08 吉林化工学院 Transmission line short-circuit fault diagnostic method based on experience wavelet transformation and local energy
CN109766927A (en) * 2018-12-10 2019-05-17 清华大学 High-speed railway track switch intelligent trouble detection method based on interacting depth study
CN110286279A (en) * 2019-06-05 2019-09-27 武汉大学 Based on extreme random forest and the sparse Fault Diagnosis of Power Electronic Circuits method from encryption algorithm of stacking-type
CN110376522A (en) * 2019-09-03 2019-10-25 宁夏西北骏马电机制造股份有限公司 A kind of Method of Motor Fault Diagnosis of the deep learning network of data fusion
CN110514924A (en) * 2019-08-12 2019-11-29 武汉大学 The Winding in Power Transformer Fault Locating Method of depth convolutional neural networks fusion visual recognition
CN110672905A (en) * 2019-09-16 2020-01-10 东南大学 CNN-based self-supervision voltage sag source identification method
CN111142001A (en) * 2020-01-10 2020-05-12 三峡大学 Transformer multi-source partial discharge mode identification method based on parallel characteristic domain
CN111259912A (en) * 2020-01-13 2020-06-09 金陵科技学院 Instrument image recognition method based on AE-SVM substation inspection robot
CN111413646A (en) * 2020-04-10 2020-07-14 华中科技大学 Real-time detection method for open-circuit fault of bridge arm of three-phase voltage source inverter

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107193271A (en) * 2017-06-13 2017-09-22 青岛科技大学 Preposition service adapter and trouble-shooter for industrial information physical system
CN107451557A (en) * 2017-07-29 2017-12-08 吉林化工学院 Transmission line short-circuit fault diagnostic method based on experience wavelet transformation and local energy
CN109766927A (en) * 2018-12-10 2019-05-17 清华大学 High-speed railway track switch intelligent trouble detection method based on interacting depth study
CN110286279A (en) * 2019-06-05 2019-09-27 武汉大学 Based on extreme random forest and the sparse Fault Diagnosis of Power Electronic Circuits method from encryption algorithm of stacking-type
CN110514924A (en) * 2019-08-12 2019-11-29 武汉大学 The Winding in Power Transformer Fault Locating Method of depth convolutional neural networks fusion visual recognition
CN110376522A (en) * 2019-09-03 2019-10-25 宁夏西北骏马电机制造股份有限公司 A kind of Method of Motor Fault Diagnosis of the deep learning network of data fusion
CN110672905A (en) * 2019-09-16 2020-01-10 东南大学 CNN-based self-supervision voltage sag source identification method
CN111142001A (en) * 2020-01-10 2020-05-12 三峡大学 Transformer multi-source partial discharge mode identification method based on parallel characteristic domain
CN111259912A (en) * 2020-01-13 2020-06-09 金陵科技学院 Instrument image recognition method based on AE-SVM substation inspection robot
CN111413646A (en) * 2020-04-10 2020-07-14 华中科技大学 Real-time detection method for open-circuit fault of bridge arm of three-phase voltage source inverter

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113419141A (en) * 2021-08-26 2021-09-21 中国南方电网有限责任公司超高压输电公司广州局 Direct-current line fault positioning method and device

Similar Documents

Publication Publication Date Title
CN111007364B (en) Method for identifying early self-recovery fault of cable
CN103941163A (en) Resonant earthed system fault line selection method utilizing fuzzy K-means clustering
CN108896875B (en) Single-phase earth fault line selection method and device
CN106990324B (en) Power distribution network ground fault detection and positioning method
CN112834959A (en) Direct-current power distribution system fault detection method based on high-frequency feature extraction
CN112098889B (en) Single-phase earth fault positioning method based on neural network and feature matrix
CN108896805B (en) Chopped wave signal verification method and system and electric energy detection device
CN109406949B (en) Power distribution network early fault detection method and device based on support vector machine
CN110726898A (en) Power distribution network fault type identification method
CN108108659B (en) Island detection key feature extraction method based on empirical mode decomposition
CN108594156B (en) Improved current transformer saturation characteristic identification method
CN109145762A (en) A kind of cable accessory Recognition of Partial Discharge based on mathematical morphology and fractal theory
CN108493909A (en) The detection method of Distribution Network Failure based on Voltage Drop
CN107632237A (en) A kind of fault line selection method for single-phase-to-ground fault based on wavelet character amount correlation after failure
CN109375041B (en) Single-phase grounding fault judgment method for small-current grounding system
CN111080105A (en) Transformer area user-to-user relationship identification method and system based on voltage time sequence data
CN111753968A (en) Non-invasive load monitoring intelligent ammeter and electricity decomposition method
Wang et al. The hydrophobic detection of transformer composite insulator bushing based on digital image processing technique
CN111426905B (en) Power distribution network common bus transformation relation abnormity diagnosis method, device and system
CN113917364A (en) High-resistance grounding identification method and device for power distribution network
CN112444758B (en) Intelligent power distribution network line fault diagnosis and classification evaluation method
CN106019043B (en) Power grid fault diagnosis method based on fuzzy integral multi-source information fusion theory
CN110212503B (en) High-resistance grounding protection method suitable for small-resistance grounding system
CN113125902A (en) Direct current power distribution system fault detection method based on IEWT and IMDMF
CN113608066A (en) Distribution network overhead line single-phase earth fault positioning method based on magnetic field detection

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