CN110672976B - Fault Diagnosis Method for Multi-terminal HVDC Transmission Line Based on Parallel Convolutional Neural Network - Google Patents
Fault Diagnosis Method for Multi-terminal HVDC Transmission Line Based on Parallel Convolutional Neural Network Download PDFInfo
- Publication number
- CN110672976B CN110672976B CN201910991992.1A CN201910991992A CN110672976B CN 110672976 B CN110672976 B CN 110672976B CN 201910991992 A CN201910991992 A CN 201910991992A CN 110672976 B CN110672976 B CN 110672976B
- Authority
- CN
- China
- Prior art keywords
- fault
- waveform
- electrical signal
- neural network
- current
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
- G01R31/085—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution lines, e.g. overhead
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/088—Aspects of digital computing
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
-
- 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
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Locating Faults (AREA)
Abstract
本发明的基于并联卷积神经网络的多端直流输电线路故障诊断方法，可实现快速准确地识别多端直流输电线路故障类型和故障位置；通过收集多端直流输电系统中各个故障类型发生时的直流线路上的电气信号波形，进行电气信号波形特征预处理，进而给两个单支路卷积神经网络分别做识别故障类型和识别故障区内外的预训练；根据迁移学习的原理，将各个卷积神经网络中间层参数分别赋值给并联卷积神经网络的两个支路上，从而在保持并联卷积神经网络中间层支路参数不变的条件下，训练并联卷积神经网络的输入和输出层，完成并联卷积神经网络预训练；将实时采集的电气信号波形输入到预训练好的并联卷积神经网络实现多端直流输电线路故障诊断。
The multi-terminal direct current transmission line fault diagnosis method based on the parallel convolutional neural network of the present invention can realize fast and accurate identification of the fault type and fault location of the multi-terminal direct current transmission line; According to the principle of transfer learning, each convolutional neural network is divided into two single-branch convolutional neural networks to identify the fault type and identify the pre-training of the fault area. The parameters of the intermediate layer are respectively assigned to the two branches of the parallel convolutional neural network, so that the input and output layers of the parallel convolutional neural network are trained under the condition that the parameters of the intermediate layer branch of the parallel convolutional neural network remain unchanged, and the parallel connection is completed. Convolutional neural network pre-training; input the electrical signal waveform collected in real time into the pre-trained parallel convolutional neural network to realize fault diagnosis of multi-terminal DC transmission lines.
Description
Technical Field
The invention belongs to the technical field of direct current transmission, and relates to a multi-terminal direct current transmission line fault diagnosis method based on a parallel convolution neural network.
Background
With the economic development and global energy shortage, the coastal areas of the east of China need a large amount of electric energy for support urgently, and the electricity generation of a large amount of new energy such as photovoltaic energy, wind power and the like is distributed in the western areas of China and is far away from large urban groups, so that large-capacity remote transmission projects are produced at the same time. Compared with alternating current transmission, high-voltage direct current transmission has the advantages of less faults, low insulation requirement, no requirement for synchronization and the like, and has attracted more and more attention.
At present, fault diagnosis methods based on mechanism models or signal analysis have the problem that fault thresholds are not easy to determine. And the traditional fault diagnosis method based on the artificial intelligence method does not consider the requirement of rapidity.
Disclosure of Invention
In order to solve the technical problems, the invention provides a multi-terminal direct current transmission line fault diagnosis method based on a parallel convolution neural network, which can accurately and quickly diagnose and process faults.
The invention provides a multi-terminal direct current transmission line fault diagnosis method based on a parallel convolution neural network, which comprises the following steps:
step 1: acquiring fault information of a fault line of a multi-terminal direct-current power transmission system and an electric signal waveform corresponding to the fault information;
step 2: combining the fault electrical signal waveform with the normal electrical signal waveform to form a new electrical signal waveform, and carrying out graying processing on the new electrical signal waveform;
and step 3: combining wavelet packet reconstruction coefficients obtained by performing 3-layer wavelet packet decomposition on a fault electrical signal waveform and a normal electrical signal waveform to form a new wavelet packet reconstruction coefficient waveform, and performing graying processing on the new wavelet packet reconstruction coefficient waveform;
and 4, step 4: forming a two-dimensional matrix form by all the grayed signal waveforms obtained in the step 2 and the step 3, and taking the two-dimensional matrix as the characteristics of the electrical signal waveform;
and 5: training the parallel convolution neural network based on the fault type, the fault position and the characteristics of the electrical signal waveform;
step 6: when the multi-terminal direct current transmission line has faults, fault electrical signal waveforms and normal electrical signal waveforms are collected, and fault diagnosis is carried out on the multi-terminal direct current transmission line on the basis of the trained parallel convolution neural network.
In the method for diagnosing the fault of the multi-terminal direct current transmission line based on the parallel convolution neural network, the step 1 specifically comprises the following steps: and acquiring voltage waveforms between two poles, positive voltage waveforms, negative voltage waveforms, positive current waveforms and negative current waveforms when a line fails.
In the method for diagnosing the fault of the multi-terminal direct current transmission line based on the parallel convolution neural network, the step 2 is specifically as follows:
step 2.1: connecting the normal voltage waveform between two poles and the voltage waveform between two poles at fault end to form a new voltage waveform between two poles, connecting the normal voltage waveform between the positive pole and the voltage waveform between the positive pole at fault end to form a new voltage waveform of the positive pole, connecting the normal voltage waveform between the negative pole and the voltage waveform between the negative pole at fault end to form a new voltage waveform of the negative pole, connecting the normal current waveform between the positive pole and the current waveform between the positive pole at fault end to form a new current waveform of the positive pole, and connecting the normal current waveform between the negative pole and the current waveform between the negative pole at fault end to form a new current waveform of the;
step 2.2: and carrying out graying processing on the new voltage waveform between two poles, the new positive voltage waveform, the new negative voltage waveform, the new positive current waveform and the new negative current waveform respectively, and compressing the values in the signal waveforms to values between 0 and 255 in an equal proportion.
In the method for diagnosing the fault of the multi-terminal direct current transmission line based on the parallel convolution neural network, the step 3 is specifically as follows:
step 3.1: respectively carrying out 3-layer wavelet packet decomposition on a normal interpolar voltage waveform, a fault interpolar voltage waveform, a normal anode voltage waveform, a fault anode voltage waveform, a normal cathode voltage waveform, a fault cathode voltage waveform, a normal anode current waveform, a fault anode current waveform, a normal cathode current waveform and a fault cathode current waveform to obtain a wavelet packet reconstruction coefficient of each electrical signal waveform;
step 3.2: respectively connecting the wavelet packet reconstruction coefficients of all electrical signal waveforms when the multi-terminal direct-current power transmission system normally operates and the wavelet packet reconstruction coefficients of electrical signal waveforms when the multi-terminal direct-current power transmission system fails end to form 5 groups of new electrical signal wavelet packet reconstruction coefficient waveforms;
step 3.3: and (5) carrying out graying processing on the wave form of the wavelet packet reconstruction coefficient of the new electrical signal of the 5 groups.
In the method for diagnosing the fault of the multi-terminal direct current transmission line based on the parallel convolution neural network, the step 5 is specifically as follows:
step 5.1: combining the characteristics of the electric signal waveform and the fault type corresponding to the electric signal waveform into a first training set, and training the single-branch convolutional neural network CNN1 through the first training set;
step 5.2: forming the characteristics of the electrical signal waveform and the fault position corresponding to the characteristics into a second training set, and training the single-branch convolutional neural network CNN2 through the second training set;
step 5.3: constructing a parallel convolutional neural network P-CNN with the same intermediate layer structure as the CNN1 and the CNN2, and copying weight parameter values and bias parameter values in the intermediate layers of the CNN1 and the CNN2 to corresponding positions of two branches of the P-CNN;
step 5.4: and fixing the parameter value of the P-CNN branch, and training the input layer parameter and the output layer parameter of the P-CNN by using the characteristics, the fault type and the fault position of the electrical signal waveform.
In the multi-terminal direct-current transmission line fault diagnosis method based on the parallel convolution neural network, the fault types comprise: the direct current power supply system comprises a positive short circuit, a negative short circuit, an inter-pole short circuit, a single-phase grounding fault, a two-phase short circuit fault, a two-phase grounding fault and a three-phase short circuit fault on the alternating current side of a converter station on a direct current line.
The multi-terminal direct-current transmission line fault diagnosis method based on the parallel convolution neural network at least has the following beneficial effects:
1. compared with the traditional method for manually determining the fault type and the internal and external thresholds of the fault area based on a mechanism model and signal analysis, the fault diagnosis method provided by the invention avoids the condition that the accuracy rate is reduced possibly caused by manually determining the thresholds during fault diagnosis, and ensures the accuracy of fault diagnosis of the multi-terminal direct-current transmission line;
2. compared with the traditional fault diagnosis method which only can complete one function by one algorithm, the fault diagnosis method can have a plurality of branches and complete a plurality of fault diagnosis functions simultaneously, and the functions do not interfere with each other, so that the fault diagnosis method has good expansibility;
3. compared with the traditional single-feature fault diagnosis method based on the amplitude feature of the signal waveform or the frequency feature of the signal waveform, the method provided by the invention has the advantages that the amplitude feature of the signal waveform and the frequency feature of the signal waveform are mixed, and the accuracy is improved.
Drawings
FIG. 1 is a flow chart of the fault diagnosis method of the multi-terminal direct current transmission line based on the parallel convolution neural network;
FIG. 2 is a flow chart of the training of the parallel convolutional neural network of the present invention;
FIG. 3 is a four terminal DC power transmission system topology of an embodiment;
FIG. 4 is a graph of a one-branch CNN1 training graph;
FIG. 5 is a one-way CNN2 training graph;
FIG. 6 is a P-CNN input layer parameter and output layer parameter training diagram.
Detailed Description
The invention discloses a fault diagnosis method for a multi-terminal direct-current transmission line based on a parallel convolution neural network, which mainly divides fault diagnosis of the multi-terminal direct-current transmission line into two parts, namely fault type identification and fault position identification, by means of parallel operation capacity of the parallel convolution neural network. The convolutional neural network has a double-branch structure, one branch carries out fault type identification, the other branch carries out fault position external identification, and due to the special structure of the neural network, the two branches can simultaneously work, and finally, a fault diagnosis result is output by the same output layer, so that the number of classifiers is reduced, the fault diagnosis time is shortened, and the requirement on rapidity of the multi-terminal direct-current power transmission system is met.
During specific implementation, the function of the multi-terminal direct-current transmission line fault diagnosis method based on the parallel convolutional neural network is not limited to fault type identification and fault position identification, a plurality of functions can be added under the condition of the same input, a plurality of branches can be added to the convolutional neural network, and the method has expansibility.
As shown in fig. 1, the method for diagnosing the fault of the multi-terminal direct current transmission line based on the parallel convolutional neural network of the present invention includes the following steps:
step 1: acquiring fault information of a fault line of a multi-terminal direct-current power transmission system and an electric signal waveform corresponding to the fault information;
in specific implementation, voltage waveforms between two poles, positive voltage waveforms, negative voltage waveforms, positive current waveforms and negative current waveforms during line faults are collected.
Step 2: combining the fault electrical signal waveform with the normal electrical signal waveform to form a new electrical signal waveform, and performing graying processing on the new electrical signal waveform, wherein the step 2 specifically comprises the following steps of:
step 2.1: will normally generate a voltage waveform U between two electrodes0Voltage between two poles in case of sum faultWave form U1The two ends are connected to form a new voltage waveform between two poles U ═ U0,U1]The normal positive voltage waveform EP0And positive electrode voltage waveform E in case of failureP1The positive voltage waveform E is formed by connecting the head and the tailP＝[EP0,EP1]The normal negative voltage waveform EN0And the negative voltage waveform E in the case of a faultN1The voltage waveform E of the negative electrode is formed by connecting the head and the tailN＝[EN0,EN1]The normal positive electrode current waveform IP0Positive electrode current waveform I in sum faultP1The end to end connection forms a new positive electrode current waveform IP＝[IP0,IP1]The normal negative current waveform IN0And the negative current waveform I in faultN1Make up new negative pole current waveform I end to endN＝[IN0,IN1]；
Step 2.2: for new voltage waveform between two poles U ═ U0,U1]New positive electrode voltage waveform EP＝[EP0,EP1]New negative electrode voltage waveform EN＝[EN0,EN1]New positive electrode current waveform IP＝[IP0,IP1]And a new negative electrode current waveform IN＝[IN0,IN1]Respectively carrying out graying treatment, compressing the numerical values in the signal waveform to the numerical values between 0 and 255 in equal proportion to respectively obtain U' and Ep'、EN'、Ip' and IN'。
And step 3: combining wavelet packet reconstruction coefficients obtained by performing 3-layer wavelet packet decomposition on a fault electrical signal waveform and a normal electrical signal waveform to form a new wavelet packet reconstruction coefficient waveform, and performing graying processing on the new wavelet packet reconstruction coefficient waveform; the step 3 specifically comprises the following steps:
step 3.1: for the normal voltage waveform U between two electrodes0Voltage waveform U between two poles during fault1Normal positive electrode voltage waveform EP0And positive electrode voltage waveform E in failureP1Normal negative electrode voltage waveform EN0And a negative electrode voltage waveform E at the time of failureN1Normal positive electrode current waveform IP0Therefore, it isTime-of-failure positive electrode current waveform IP1Normal negative current waveform IN0And the negative current waveform I in faultN1Respectively carrying out 3 layers of wavelet packet decomposition to obtain wavelet packet reconstruction coefficient U of each electrical signal waveformw0、Uw1、EwP0、EwP1、EwN0、EwN1、IwP0、IwP1、IwN0And IwN1；
Step 3.2: respectively connecting the wavelet packet reconstruction coefficient of each electrical signal waveform of the multi-terminal direct-current power transmission system in normal operation and the wavelet packet reconstruction coefficient of the electrical signal waveform in fault end to form 5 groups of new electrical signal wavelet packet reconstruction coefficient waveforms Uw＝[Uw0,Uw1]、EwP＝[EwP0,EwP1]、EwN＝[EwN0,EwN1]、IwP＝[IwP0,IwP1]And IwN＝[IwN0,IwN1]；
Step 3.3: carrying out gray processing on 5 groups of new electric signal wavelet packet reconstruction coefficient waveforms to obtain Uw'、Ewp'、EwN'、Iwp' and IwN'。
And 4, step 4: combining all the grayed signal waveforms obtained in the step 2 and the step 3 into a two-dimensional matrix form T ═ U', EP',EN',IP',IN',Uw',EwP',EwN',IwP',IwN']TAnd taking the two-dimensional matrix as the characteristics of the waveform of the electrical signal;
and 5: training the parallel convolution neural network based on the fault type, the fault position and the characteristics of the electrical signal waveform, as shown in fig. 2, wherein the step 5 specifically comprises the following steps:
step 5.1: combining the characteristics of the electric signal waveform and the fault type corresponding to the electric signal waveform into a first training set, and training the single-branch convolutional neural network CNN1 through the first training set;
step 5.2: forming the characteristics of the electrical signal waveform and the fault position corresponding to the characteristics into a second training set, and training the single-branch convolutional neural network CNN2 through the second training set;
step 5.3: constructing a parallel convolutional neural network P-CNN with the same intermediate layer structure as the CNN1 and the CNN2, and copying weight parameter values and bias parameter values in the intermediate layers of the CNN1 and the CNN2 to corresponding positions of two branches of the P-CNN;
step 5.4: and fixing the parameter value of the P-CNN branch, and training the input layer parameter and the output layer parameter of the P-CNN by using the characteristics, the fault type and the fault position of the electrical signal waveform.
In specific implementation, the fault types include: the direct current power supply system comprises a positive short circuit, a negative short circuit, an inter-pole short circuit, a single-phase grounding fault, a two-phase short circuit fault, a two-phase grounding fault and a three-phase short circuit fault on the alternating current side of a converter station on a direct current line.
Step 6: when the multi-terminal direct current transmission line has faults, fault electrical signal waveforms and normal electrical signal waveforms are collected, and fault diagnosis is carried out on the multi-terminal direct current transmission line on the basis of the trained parallel convolution neural network.
The four-terminal dc transmission line of fig. 3 is taken as an example to further explain the embodiment of the present invention. As shown in fig. 3, in the figure, terminals B2 and F1 are power transmission terminals, terminals B3 and E1 are power reception terminals, and the dc transmission bus voltage ± 200 kV. Where TLB2E1 and TLE1F1 are 1000km overhead lines, and TLB2B3 and TLB3F1 are 800km overhead lines. And a group of direct current breakers DCB 1-DCB 4 are respectively arranged on the four direct current transmission lines to cut off the fault direct current transmission lines.
The invention relates to a fault diagnosis method for a multi-terminal direct-current transmission line based on a parallel convolution neural network, which can realize the quick and accurate identification of fault types and fault position information of the multi-terminal direct-current transmission line; the method comprises the steps that electrical signal waveform on a line TLB2E1 when each fault type occurs in a multi-terminal direct-current power transmission system is collected, electrical signal waveform characteristic preprocessing of each fault type is conducted, and then pre-training of fault type identification and fault area identification is conducted on a single-branch CNN1 and a single-branch CNN2 respectively; according to the transfer learning principle, intermediate layer parameters in CNN1 and CNN2 are respectively assigned to two branches of P-CNN, so that an input layer and an output layer of the P-CNN are trained under the condition of keeping the branch parameters of the intermediate layer of the P-CNN unchanged; and applying the pre-trained P-CNN to fault diagnosis, and when the multi-end direct-current transmission line has a fault, acquiring an electrical signal waveform on the TLB2E1, preprocessing the electrical signal waveform characteristics, and inputting the electrical signal waveform characteristics into the P-CNN to realize the multi-end direct-current transmission line fault diagnosis.
Step 1: collecting fault information of a multi-terminal direct current transmission line and an electric signal waveform corresponding to the fault information; specifically, a voltage waveform between two poles, a positive voltage waveform, a negative voltage waveform, a positive current waveform and a negative current waveform during line fault are collected.
Step 2.1: respectively connecting the electric signal waveforms of the multi-end direct current transmission line in normal operation and the electric signal waveforms of the multi-end direct current transmission line in a fault occurrence end to end manner to form the following 5 groups of new electric signal waveforms:
U＝[U0,U1]＝[U01,U02，,...,U0n,U11,U12,...,U1n]
EP＝[EP0,EP1]＝[EP01,EP02,...,EP0n,EP11,EP12,...,EP1n]
EN＝[EN0,EN1]＝[EN01,EN02,...,EN0n,EN11,EN12,...,EN1n]
IP＝[IP0,IP1]＝[IP01,IP02,...,IP0n,IP11,IP12,...,IP1n]
IN＝[IN0,IN1]＝[IN01,IN02,...,IN0n,IN11,IN12,...,IN1n]
wherein, U1＝[U11,U12,,...,U1n]Is the voltage waveform between two poles, U, on the line TLB2E1 at fault1nIs the nth point in the waveform, U0＝[U01,U02,,...,U0n]Is the voltage waveform between two poles, U, during normal operation before a fault0nIs in the waveformThe nth point. EP1＝[EP11,EP12,...,EP1n]Is the positive voltage waveform on line TLB2E1, EP1nIs the nth point in the waveform, EP0＝[EP01,EP02,...,EP0n]Is the positive electrode voltage waveform in normal operation before failure, EP0nIs the nth point in the waveform. EN1＝[EN11,EN12,...,EN1n]Is the negative voltage waveform on line TLB2E1, EN1nIs the nth point in the waveform. EN0＝[EN01,EN02,...,EN0n]Is the negative electrode voltage waveform in normal operation before failure, EN0nIs the nth point in the waveform. I isP1＝[IP11,IP12,...,IP1n]Is the positive current waveform, I, on line TLB2E1P1nIs the nth point in the waveform. I isP0＝[IP01,IP02,...,IP0n]Is the positive current waveform before failure in normal operation, IP0nIs the nth point in the waveform. I isN1＝[IN11,IN12,...,IN1n]Is the negative current waveform, I, on line TLB2E1N1nIs the nth point in the waveform, IN0＝[IN01,IN02,...,IN0n]Is the negative current wave, I, in normal operation before a faultN0nIs the nth point in the waveform. Wherein each electrical signal waveform is in a row vector form;
step 2.2: combine 5 new sets of electrical signal waveforms U, EP、EN、IPAnd INRespectively graying to obtain U' and Ep'、EN'、Ip' and IN', where the graying formula is:
where x' is the electrical signal waveform after graying, and x (n) is the original electrical signal waveform.
Step 3.1: waveform U of electric signal0、U1、EP0、EP1、EN0、EN1、IP0、IP1、IN0And IN1Respectively carrying out 3 layers of wavelet packet decomposition to obtain wavelet packet reconstruction coefficient U of each electrical signal waveformw0、Uw1、EwP0、EwP1、EwN0、EwN1、IwP0、IwP1、IwN0And IwN1Wherein the wavelet packet reconstruction coefficient of each electrical signal waveform is in a matrix form of 8 multiplied by n;
step 3.2: respectively connecting the wavelet packet reconstruction coefficient of each electrical signal waveform when the multi-terminal direct-current power transmission system normally operates and the wavelet packet reconstruction coefficient of the electrical signal waveform when a fault occurs end to form 5 groups of new electrical signal wavelet packet reconstruction coefficient waveforms Uw＝[Uw0,Uw1]、EwP＝[EwP0,EwP1]、EwN＝[EwN0,EwN1]、IwP＝[IwP0,IwP1]And IwN＝[IwN0,IwN1]；
Step 3.3: 5 groups of new electric signal wavelet packet reconstruction coefficient waveforms Uw＝[Uw0,Uw1]、EwP＝[EwP0,EwP1]、EwN＝[EwN0,EwN1]、IwP＝[IwP0,IwP1]And IwN＝[IwN0,IwN1]Respectively graying to obtain Uw'、Ewp'、EwN'、Iwp' and IwN'；
And 4, step 4: putting all the grayed signal waveforms into a 45 x 2n matrix to form a two-dimensional matrix form T ═ U', EP',EN',IP',IN',Uw',EwP',EwN',IwP',IwN']THere, T is a characteristic of an electrical signal waveform on the dc transmission line TLB2E1 corresponding to each fault of the multi-terminal dc transmission system.
Step 5.1: the collected fault types in the M four-terminal direct-current power transmission systems and M fault characteristics T corresponding to the fault types are submitted to a single branch CNN1 for training, and a training process curve is shown in FIG. 4;
step 5.2: the collected fault position information in the M four-terminal direct-current power transmission systems and M fault characteristics T corresponding to the fault position information are submitted to a single branch CNN2 for training, and a training process curve is shown in FIG. 5;
step 5.3: constructing a parallel convolutional neural network P-CNN with the same intermediate layer structure as the CNN1 and the CNN2, and respectively assigning a weight parameter value omega and a bias parameter value b in the intermediate layers of the CNN1 and the CNN2 to corresponding positions of two branches of the P-CNN;
step 5.4: and fixing the parameter values of the P-CNN branch circuits unchanged, and training the input layer parameters and the output layer parameters of the P-CNN by using the collected fault types and position information in the M four-terminal direct-current power transmission systems and the M fault characteristics T corresponding to the fault types and position information.
Step 6: when the relay protection devices DCB 1-DCB 4 of the multi-terminal direct current transmission line identify faults in the system, starting the multi-terminal direct current transmission line fault diagnosis method based on the parallel convolution neural network; collecting fault electrical signal waveforms and normal electrical signal waveforms on the line TLB2E1 during fault; repeating the step 2 to the step 4 to obtain the current fault characteristic t by the method for preprocessing the fault electrical signal waveform and the normal electrical signal waveform; and inputting the current fault feature t into the trained P-CNN to obtain the current fault type and fault position information.
The method comprises two functions of fault type identification and fault position identification, wherein the fault position identification is to determine whether a fault occurs on a line where a detection device is located according to the characteristics of a fault signal. The invention realizes the fault diagnosis method for determining the fault type and the occurrence position through the characteristics of the fault signal, and the fault diagnosis result can help the relay protection device to isolate the fault and can also help operation and dispatching personnel to know and process the fault. The fault location identification function of the invention is to determine which line of the multi-terminal direct current transmission system the fault occurs on according to the characteristics of the fault signal, namely the fault location method is to judge whether the fault point occurs on the line of the detection device.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the scope of the present invention, which is defined by the appended claims.
Claims (5)
1. The method for diagnosing the fault of the multi-terminal direct-current transmission line based on the parallel convolutional neural network is characterized by comprising the following steps of:
step 1: collecting fault information of a multi-terminal direct current transmission line and an electric signal waveform corresponding to the fault information;
step 2: combining the fault electrical signal waveform with the normal electrical signal waveform to form a new electrical signal waveform, and performing graying processing on the new electrical signal waveform, wherein the step 2 specifically comprises the following steps of:
step 2.1: connecting the normal voltage waveform between two poles and the voltage waveform between two poles at fault end to form a new voltage waveform between two poles, connecting the normal voltage waveform between the positive pole and the voltage waveform between the positive pole at fault end to form a new voltage waveform of the positive pole, connecting the normal voltage waveform between the negative pole and the voltage waveform between the negative pole at fault end to form a new voltage waveform of the negative pole, connecting the normal current waveform between the positive pole and the current waveform between the positive pole at fault end to form a new current waveform of the positive pole, and connecting the normal current waveform between the negative pole and the current waveform between the negative pole at fault end to form a new current waveform of the;
step 2.2: carrying out graying processing on the new voltage waveform between two poles, the new positive voltage waveform, the new negative voltage waveform, the new positive current waveform and the new negative current waveform respectively, and compressing the numerical values in the signal waveforms to numerical values between 0 and 255 in an equal proportion;
and step 3: combining wavelet packet reconstruction coefficients obtained by performing 3-layer wavelet packet decomposition on a fault electrical signal waveform and a normal electrical signal waveform to form a new wavelet packet reconstruction coefficient waveform, and performing graying processing on the new wavelet packet reconstruction coefficient waveform;
and 4, step 4: forming a two-dimensional matrix form by all the grayed signal waveforms obtained in the step 2 and the step 3, and taking the two-dimensional matrix as the characteristics of the electrical signal waveform;
and 5: training the parallel convolution neural network based on the fault type, the fault position and the characteristics of the electrical signal waveform;
step 6: when the multi-terminal direct current transmission line has faults, fault electrical signal waveforms and normal electrical signal waveforms are collected, and fault diagnosis is carried out on the multi-terminal direct current transmission line on the basis of the trained parallel convolution neural network.
2. The method for diagnosing the fault of the multi-terminal direct-current transmission line based on the parallel convolutional neural network as claimed in claim 1, wherein the step 1 specifically comprises: and acquiring voltage waveforms between two poles, positive voltage waveforms, negative voltage waveforms, positive current waveforms and negative current waveforms when a line fails.
3. The method for diagnosing the fault of the multi-terminal direct-current transmission line based on the parallel convolutional neural network as claimed in claim 2, wherein the step 3 specifically comprises:
step 3.1: respectively carrying out 3-layer wavelet packet decomposition on a normal interpolar voltage waveform, a fault interpolar voltage waveform, a normal anode voltage waveform, a fault anode voltage waveform, a normal cathode voltage waveform, a fault cathode voltage waveform, a normal anode current waveform, a fault anode current waveform, a normal cathode current waveform and a fault cathode current waveform to obtain a wavelet packet reconstruction coefficient of each electrical signal waveform;
step 3.2: respectively connecting the wavelet packet reconstruction coefficients of all electrical signal waveforms when the multi-terminal direct-current power transmission system normally operates and the wavelet packet reconstruction coefficients of electrical signal waveforms when the multi-terminal direct-current power transmission system fails end to form 5 groups of new electrical signal wavelet packet reconstruction coefficient waveforms;
step 3.3: and (5) carrying out graying processing on the wave form of the wavelet packet reconstruction coefficient of the new electrical signal of the 5 groups.
4. The method for diagnosing the fault of the multi-terminal direct-current transmission line based on the parallel convolutional neural network as claimed in claim 1, wherein the step 5 specifically comprises:
step 5.1: combining the characteristics of the electric signal waveform and the fault type corresponding to the electric signal waveform into a first training set, and training the single-branch convolutional neural network CNN1 through the first training set;
step 5.2: forming the characteristics of the electrical signal waveform and the fault position corresponding to the characteristics into a second training set, and training the single-branch convolutional neural network CNN2 through the second training set;
step 5.3: constructing a parallel convolutional neural network P-CNN with the same intermediate layer structure as the CNN1 and the CNN2, and respectively copying the weight parameter values and the bias parameter values in the intermediate layers of the CNN1 and the CNN2 to the corresponding positions of two branches of the P-CNN;
step 5.4: and fixing the parameter value of the P-CNN branch, and training the input layer parameter and the output layer parameter of the P-CNN by using the characteristics, the fault type and the fault position of the electrical signal waveform.
5. The method for diagnosing the fault of the multi-terminal direct-current transmission line based on the parallel convolutional neural network as claimed in claim 1 or 4, wherein the fault type comprises: the direct current power supply system comprises a positive short circuit, a negative short circuit, an inter-pole short circuit, a single-phase grounding fault, a two-phase short circuit fault, a two-phase grounding fault and a three-phase short circuit fault on the alternating current side of a converter station on a direct current line.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910991992.1A CN110672976B (en) | 2019-10-18 | 2019-10-18 | Fault Diagnosis Method for Multi-terminal HVDC Transmission Line Based on Parallel Convolutional Neural Network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910991992.1A CN110672976B (en) | 2019-10-18 | 2019-10-18 | Fault Diagnosis Method for Multi-terminal HVDC Transmission Line Based on Parallel Convolutional Neural Network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110672976A CN110672976A (en) | 2020-01-10 |
| CN110672976B true CN110672976B (en) | 2021-04-16 |
Family
ID=69083421
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910991992.1A Expired - Fee Related CN110672976B (en) | 2019-10-18 | 2019-10-18 | Fault Diagnosis Method for Multi-terminal HVDC Transmission Line Based on Parallel Convolutional Neural Network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110672976B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111441783B (en) * | 2020-04-17 | 2021-06-15 | 中铁隧道局集团有限公司 | Centralized detection method for performance of low-voltage apparatus of shield machine |
| CN112098889B (en) * | 2020-09-09 | 2022-03-04 | 青岛鼎信通讯股份有限公司 | Single-phase earth fault positioning method based on neural network and feature matrix |
| CN113049914B (en) * | 2021-03-15 | 2023-02-07 | 广东电网有限责任公司清远供电局 | Power transmission line fault diagnosis method and system, electronic equipment and storage medium |
| CN113189448A (en) * | 2021-04-29 | 2021-07-30 | 广东电网有限责任公司清远供电局 | Method and device for detecting fault type of power transmission line |
| CN116049646A (en) * | 2022-09-29 | 2023-05-02 | 佳源科技股份有限公司 | Transmission line fault detection method, device, computer equipment and storage medium |
| CN115825648A (en) * | 2022-11-16 | 2023-03-21 | 昆明理工大学 | Extra-high voltage direct current transmission line fault identification method based on multi-decomposition and parallel residual error network |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5537327A (en) * | 1993-10-22 | 1996-07-16 | New York State Electric & Gas Corporation | Method and apparatus for detecting high-impedance faults in electrical power systems |
| JP2006162541A (en) * | 2004-12-10 | 2006-06-22 | Meiji Univ | Accident point location method, system and program |
| CN104808107A (en) * | 2015-04-16 | 2015-07-29 | 国家电网公司 | XLPE cable partial discharge defect type identification method |
| CN107271925B (en) * | 2017-06-26 | 2019-11-05 | 湘潭大学 | Five level converter Fault Locating Method of modularization based on depth convolutional network |
-
2019
- 2019-10-18 CN CN201910991992.1A patent/CN110672976B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN110672976A (en) | 2020-01-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110672976B (en) | Fault Diagnosis Method for Multi-terminal HVDC Transmission Line Based on Parallel Convolutional Neural Network | |
| CN110398663B (en) | Flexible direct current power grid fault identification method based on convolutional neural network | |
| CN105116208B (en) | A kind of extra-high voltage DC transmission system commutation failure method for diagnosing faults | |
| CN109274079B (en) | A single-ended protection method for ring-shaped flexible direct grid lines | |
| CN110609213B (en) | High-resistance grounding fault location method for MMC-HVDC transmission lines based on optimal features | |
| CN113791294B (en) | An edge-cloud collaborative commutation failure source tracing method for UHV DC transmission systems | |
| CN107329044A (en) | A kind of wire selection method for power distribution network single phase earthing failure based on electric arc transient state component | |
| CN111429043B (en) | Voltage sag characteristic quantity random evaluation method based on scene construction | |
| CN117937368A (en) | Flexible direct-current power grid pilot protection method based on frequency domain energy matrix similarity | |
| CN115310604A (en) | UPS system online fault diagnosis method based on neural network | |
| CN114325236B (en) | Power distribution network fault identification method and system based on frequency spectrum entropy and random forest | |
| CN114236308B (en) | Distribution network fault detection positioning method based on feeder information | |
| CN110703134B (en) | Small current grounding line selection and phase selection method based on fault sequence component | |
| CN112595932A (en) | Single-pole fault line selection method suitable for medium-voltage direct-current power distribution network | |
| CN104537196B (en) | A kind of method for obtaining the MTDC transmission system reliability based on VSC | |
| CN110601153A (en) | Relay protection method for single-end quantity of direct-current power distribution network | |
| CN115906350A (en) | A Microgrid Fault Location and Diagnosis Method | |
| CN117269676A (en) | A lightning strike fault location method and device for distribution network line equipment | |
| Wang et al. | Fault Diagnosis of UHVDC Transmission Line Based on Deep Neural Network | |
| Yang et al. | Multi-criteria fusion method for line selection of single-phase-to-ground fault with small current based on GRU neural network | |
| CN113125902A (en) | Direct current power distribution system fault detection method based on IEWT and IMDMF | |
| Likhitha et al. | Setting-free Smart Recloser for Three Terminal Mixed Lines | |
| KR20050020841A (en) | Determination method for equivalence impedance of power line | |
| CN118641880B (en) | Method and system for identifying internal and external faults of metal reflux direct current power grid area | |
| Mao et al. | Fast Detection Method of DC Short Circuit Fault of MMC-HVDC Grid |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210416 |