CN114414942A - Power transmission line fault identification classifier, identification method and system based on transient waveform image identification - Google Patents
Power transmission line fault identification classifier, identification method and system based on transient waveform image identification Download PDFInfo
- Publication number
- CN114414942A CN114414942A CN202210041191.0A CN202210041191A CN114414942A CN 114414942 A CN114414942 A CN 114414942A CN 202210041191 A CN202210041191 A CN 202210041191A CN 114414942 A CN114414942 A CN 114414942A
- Authority
- CN
- China
- Prior art keywords
- fault
- transmission line
- power transmission
- transient waveform
- phase
- 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
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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Locating Faults (AREA)
Abstract
The invention discloses a power transmission line fault identification classifier based on transient waveform image identification, which comprises the following components: the AlexNet network is constructed in the following way, and the transient waveform data of the voltage/current signal at the head end, the voltage/current signal at the tail end and the zero sequence component at the head end and the tail end of the power transmission line corresponding to the power transmission line fault are normalized; generating a two-dimensional transient waveform image as a fault sample by using the transient waveform data after normalization processing corresponding to the same fault; obtaining a fault sample set formed by a plurality of fault samples; and (3) taking the AlexNet network which is subjected to pre-training and the first full connection layer, the second full connection layer and the classification output layer are reset to be in a state to be trained, and inputting the fault sample set to train until convergence. The invention also discloses a fault identification method and a fault identification system based on the classifier, and the invention utilizes image feature extraction to avoid the difficulty of feature quantity selection and can realize accurate fault identification of the power transmission line.
Description
Technical Field
The invention relates to the technical field of power transmission line fault identification, in particular to a power transmission line fault identification classifier, a power transmission line fault identification method and a power transmission line fault identification system based on transient waveform image identification.
Background
Overhead transmission line region distributes extensively, the structure is more complicated, the operational environment is changeable, leads to electric power system safe operation easily to be influenced. The fault identification of the power transmission line can realize timely and accurate identification of fault reasons and fault types, and has important significance for guiding self-adaptive reclosing and recovering power transmission of the line, reducing line outage time and ensuring safe and stable operation of a power system.
The existing power transmission line fault identification idea is to utilize current and voltage transient waveform data recorded by a fault recording device and carry out fault classification identification through characteristics such as peak values and duration of time domain signals and energy spectra of frequency domain signals. The main existing methods for on-line monitoring of power transmission lines include: selecting waveform characteristics, meteorological factors, seasonal time period characteristics, background information and the like as characteristic quantities, establishing a classifier model, and training and testing the model by using actual fault recording data; selecting waveform characteristics, positive sequence current fault components at two ends of a line, initial polarity of fault current high-frequency traveling wave mode components and the like as characteristic quantities, establishing a classifier model, and training and testing the model by using a balanced fault simulation sample; and thirdly, extracting characteristic quantity by using a mathematical morphology method.
The above methods have the following disadvantages: firstly, in practical application, the most effective characteristic quantity is selected as a key point and a difficulty point influencing the fault identification effect, and the evaluation index of the reasonability of the characteristic quantity selection has subjectivity; secondly, the selected fault characteristics are easily influenced by factors such as voltage and current waveforms, fault distance, transition resistance and the like, and the characteristic selection process is complex; the fault labels of the actual wave recording data are often lost, fuzzy and the like, the fault sample of the same line is limited, and the effect on model training and testing is poor; fourthly, the unbalance of factors such as fault types, fault phases, transition resistances and the like is not considered in the construction of the balanced fault sample, the difference between the sample and the actual is large, and the result of training and testing the classification model is questioned.
Therefore, it is highly desirable to invent a transmission line fault identification method that does not need to select feature quantities independently and utilizes a large number of labeled fault samples to train and test a classifier model.
Disclosure of Invention
Aiming at the defects of the existing method, the invention aims to utilize the characteristic that an AlexNet network has strong image classification learning capability, construct a power transmission line fault identification classifier based on transient waveform image identification by a transfer learning method, and construct a power transmission line fault identification method and system based on the classifier.
One of the objectives of the present invention is to provide a power transmission line fault identification classifier based on transient waveform image identification, which includes:
an AlexNet network;
the AlexNet network is constructed in the following way:
s1, normalizing the transient waveform data of the voltage/current signals at the head end of the power transmission line, the voltage/current signals at the tail end and the zero-sequence components at the head end and the tail end corresponding to the power transmission line fault;
generating a two-dimensional transient waveform image as a fault sample according to the transient waveform data after normalization processing corresponding to the same fault;
obtaining a fault sample set formed by a plurality of fault samples;
and S2, taking the AlexNet network which is subjected to pre-training and the first full connection layer, the second full connection layer and the Softmax layer are reset to be in a state to be trained, and inputting the fault sample set for training until convergence.
Preferably, the classification label of the fault sample comprises a fault type, a fault phase and a fault reason.
Preferably, in the normalization process, the sampling time of each of the transient waveform data is the same.
Preferably, in the normalization processing, the ordinate scale of each waveform on the two-dimensional transient waveform image is set according to the voltage class and the power transmission capacity of the power transmission line, and the following principle is followed:
a. the same ordinate scale is adopted for the voltage signals at the head end/the tail end of the power transmission line;
b. the same longitudinal coordinate scale is adopted for the head end/tail end current signals of the power transmission line;
c. the zero-sequence components of the head end and the tail end adopt the same longitudinal coordinate scale;
d. each ordinate scale is as small as possible on the premise that any transient waveform data is not truncated and displayed.
Preferably, in the two-dimensional transient waveform image, transient waveform data of each phase of the head end/tail end voltage/current signal of the power transmission line are displayed in an overlapping manner.
Preferably, A, B, C each phase transient waveform data is differentiated by a different color.
Preferably, in the two-dimensional transient waveform image, the transient waveform data of the first/last zero sequence component are displayed in an overlapping manner.
Preferably, transient waveform data of zero sequence voltage and current are distinguished by different colors.
Preferably, the fault sample set is an unbalance-like fault sample set generated according to a simulation model constructed from actual transmission line segments.
Preferably, the step of generating an imbalance-like fault sample set by the simulation model comprises:
(1) according to the length L of the power transmission line MN, a fault distance Lm from an M-side bus is generated in a uniform sampling mode in the range of the power transmission line MN, and the fault distance Ln from an N-side bus is equal to L-Lm;
(2) uniformly sampling within the range of 0-90 degrees to generate fault phase angles;
(3) generating fault types according to probability sampling corresponding to fault types such as single-phase earth faults, two-phase short-circuit faults, two-phase earth faults or three-phase short-circuit faults;
(4) judging the fault type, if the fault type is a single-phase fault, firstly generating a phase class according to the fault probability of different phase classes, then sampling according to the probability corresponding to the specific single-phase fault reason to generate a single-phase fault reason and matching the single-phase fault reason with corresponding transition resistance;
other fault types are uniformly sampled to generate fault phases;
(5) for single-phase earth faults, uniformly sampling and generating transition resistors according to the probability of the corresponding transition resistors of the generated fault reasons;
(6) setting fixed setting for two-phase short-circuit fault and three-phase short-circuit fault transition resistance according to corresponding values;
(7) simulating to generate a fault sample, namely a two-dimensional transient waveform image, and taking the fault type, the fault phase and the fault reason determined in the previous step as labels of the fault sample;
(8) and repeating the steps until the number of the simulation samples reaches the required number.
Another objective of the present invention is to provide a method for identifying a power transmission line fault based on transient waveform image recognition, which includes:
recording transient waveform data of a voltage/current signal at the head end of the power transmission line, a voltage/current signal at the tail end of the power transmission line and zero sequence components of the head end and the tail end of the power transmission line when the power transmission line fails, and generating a two-dimensional transient waveform image;
and inputting the obtained two-dimensional transient waveform image into the classifier to identify the fault of the faulty power transmission line.
Further, the power transmission line fault identification classifier is deployed at transformer substations at two ends of the power transmission line.
Still another object of the present invention is to provide a power transmission line fault identification system based on transient waveform image recognition, which comprises,
the two-dimensional transient waveform image generation module is used for acquiring transient waveform data of a head end voltage/current signal, a tail end voltage/current signal and a head-tail end zero sequence component of the power transmission line corresponding to the power transmission line fault, and generating and outputting a two-dimensional transient waveform image;
the power transmission line fault identification classifier is connected with the two-dimensional transient waveform image generation module and is used for identifying a fault of a faulty power transmission line according to the input two-dimensional transient waveform image.
Compared with the prior art, the power transmission line fault identification classifier, the identification method and the system based on transient waveform image identification have the following beneficial effects:
(1) the fault transient waveform image monitored at the double-end bus of the power transmission line is used as an input quantity, and the transient waveform image is classified and identified by using a deep learning method, so that the complexity of fault feature extraction is reduced.
(2) The existing method for carrying out fault identification based on a small-scale balance fault sample set has optimistic classification accuracy; when the actual fault sample set is seriously unbalanced, even if training modes such as a sampling method and the like are adopted, the obtained classification accuracy rate is still to be improved; and the fault classifier based on transfer learning-AlexNet can well cope with the influence of class unbalance fault samples, and has higher accuracy rate for identifying fault types and fault reasons.
(3) According to the fault characteristics and probability distribution of the power transmission line, a sample set which accords with the actual fault position, the initial phase angle, the fault type, the fault phase, the fault reason and the transition resistance value distribution characteristics of the fault is generated in batches, so that the training of a transfer learning model is facilitated, and the accuracy of the transfer-Alexnet network is improved.
Drawings
Fig. 1 is a schematic view of a construction process of an AlexNet network in the electric transmission line fault identification classifier in the embodiment of the present invention.
Fig. 2 is an exemplary diagram of a two-dimensional transient waveform image generated by normalizing transient waveform data in an embodiment of the invention.
Fig. 3 is a schematic structural diagram of an AlexNet network in the embodiment of the present invention.
Fig. 4 is a schematic logic block diagram of a power transmission line simulation model in the embodiment of the present invention.
FIG. 5 is a flowchart of constructing an imbalance-like fault sample set according to an embodiment of the present invention.
Fig. 6 is a schematic diagram of an AlexNet network test result based on a class imbalance fault sample in an embodiment of the present invention.
Fig. 7 is a schematic block diagram of a power transmission line fault identification system in an embodiment of the present invention.
Fig. 8 is a schematic block diagram of another transmission line fault identification system in the embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
The embodiment discloses transmission line fault identification classifier based on transient waveform image recognition, including: an AlexNet network obtained by means of transfer learning, the specific construction process of which is basically shown in fig. 1, includes:
on one hand, the transient waveform data of the voltage/current signal at the head end of the power transmission line, the voltage/current signal at the tail end and the zero sequence component at the head end and the tail end corresponding to the power transmission line fault are normalized; generating a two-dimensional transient waveform image as a fault sample according to the transient waveform data after normalization processing corresponding to the same fault; obtaining a fault sample set formed by a plurality of fault samples; the normalization processing mainly aims at unifying the forms of fault samples, including the size, resolution, sampling time span, coordinate scale of waveform and the like of the picture;
fig. 2 shows an example of a two-dimensional transient waveform image obtained by normalizing transient waveform data, where the upper half of the image includes transient waveform data of a voltage/current signal at the head end, a voltage/current signal at the tail end, and a zero-sequence component at the head end and the tail end of a power transmission line; the lower half of the figure is a two-dimensional transient waveform image composed of normalized waveform data, i.e., a standard form of a fault sample used in this example.
As can be seen in the lower half of fig. 2, the two-dimensional transient waveform image sample is basically divided into six subgraphs, each subgraph occupying basically the same area,
wherein, each phase voltage waveform of the head end of the transmission line (V in the head end signal part in the upper half part of figure 2)a、VbAnd Vc) Overlapping the voltage waves of the phases shown in the same sub-diagram at the end of the transmission line (V in the end signal part in the upper half of FIG. 2)a、VbAnd Vc) The display of the shape superposition is in the same subgraph, and in the two subgraphs, the waveform data book is scaled to the same coordinate scale, namely-4 x 105V to 4X 105V；
Phase current waveforms of head end of transmission line (I in head end signal section in upper half of FIG. 2)a、IbAnd Ic) The overlapping display is in the same sub-diagram, and the display in which the waveforms of the voltages of the phases at the end of the transmission line are overlapped is in the same sub-diagram (I in the end signal part in the upper half of fig. 2)a、IbAnd Ic) In both sub-graphs, the waveform data is scaled to the same coordinate scale, i.e., -1 × 104V to 1X 104V；
The overlapping display of the zero sequence voltage waveforms of the head end and the tail end of the power transmission line is carried out in the same subgraph, the overlapping display of the zero sequence current waveforms of the head end and the tail end of the power transmission line is carried out in the same subgraph, and the waveform data is scaled to the same coordinate scale, namely-2 multiplied by 104V or A to 2X 104V or A;
the sampling time of all waveform data was 0.04 s.
In addition, it cannot be shown in the figure that the waveforms displayed in the same sub-diagram in a superimposed manner are distinguished by different colors, for example, the three colors of red, yellow and green are distinguished among the voltage/current waveforms of each phase, the blue and black are distinguished among the zero-sequence voltage waveforms at the head end and the tail end, and the blue and black are distinguished among the zero-sequence current waveforms at the head end and the tail end.
In this embodiment, the size of the two-dimensional transient waveform image is 227 × 227 pixels, and since a color image is adopted, each pixel has R, G, B pixel values, and the two-dimensional transient waveform image can be divided into three monochromatic layers (R, G, B layers).
In order to provide characteristic information as much as possible in limited pixels, it is preferable that, in addition to the overlapping display of waveforms, the use of the phase and component of the color region as in this embodiment, the setting of the transmission line in the ordinate scale should be based on the voltage class and the transmission capacity of the current transmission line, and the voltage signals at the head end/the tail end of the transmission line use the same ordinate scale, the current signals at the head end/the tail end of the transmission line use the same ordinate scale, and the zero sequence component at the head end and the tail end use the same ordinate scale; and each ordinate scale is as small as possible on the premise that any transient waveform data is not cut off and displayed, so that the transient waveform is ensured to be of sufficient size in a two-dimensional transient waveform image without transient waveform data loss, and extraction and improvement of characteristic information by an AlexNet network are facilitated.
On the other hand, in the present embodiment, an AlexNet network pre-trained on ImageNet is introduced.
As shown in fig. 2, the pre-trained AlexNet network in this embodiment has 8 layers, and the input layers are also shown in the figure, so that the layers in the figure from left to right are:
an input layer having a size of 227 × 227 × 3 corresponding to an input image format;
the pre-trained AlexNet network includes:
a first convolution layer with convolution kernel size of 55 × 55 × 96, after convolution, a local response normalization operation and a pooling layer are added, and the pooling kernel size is 27 × 27 × 96 (all pooling layers in the network structure are realized by a maximum pooling method);
the convolution kernel size of the second convolution layer is 27 multiplied by 256, a local response normalization process is still added after the convolution of the second convolution layer, and then the second convolution layer enters a pooling layer, wherein the kernel size is 13 multiplied by 256;
a third convolution layer having a convolution kernel size of 13 × 13 × 384;
a fourth convolution layer having a convolution kernel size of 13 × 13 × 384;
a fifth convolution layer with convolution kernel size of 13 × 13 × 256 and then entering the pooling layer, where the pooling kernel size is 6 × 6 × 256;
4096 nodes in total on the first full-link layer;
second full connection layer of 4096 nodes in total
The Softmax layer, also called the prediction layer, has 1000 nodes.
Then, reserving parameters and weights of the pre-trained AlexNet network, and only replacing the last three layers with a new first full connection layer, a new second full connection layer and a new Softmax layer; the replaced pretrained AlexNet network is the migrated AlexNet network, the first to fifth convolutional layers of the pretrained AlexNet network reserve the parameters of the pretrained AlexNet network, and the initial parameters of the first full connection layer, the second full connection layer and the Softmax layer are generated randomly; the migration AlexNet network saves the capability of extracting feature quantities from images and also provides a space for adapting to new feature classification.
After the above two aspects are ready, the training of the model can be started, in this embodiment, the training environment is first set, i.e., the minimum training sample number and the iteration number are set; in the training environment, a leave-out method is adopted, 70% of data of each type of fault of a fault sample set is used as a training set and is input into a migration AlexNet network to achieve network fine adjustment, the migration AlexNet network after fine adjustment can adapt to classification of transient waveform images, and the classification effect of a model is tested by using the remaining 30% of samples.
In the practical work, under the condition of lacking a large number of effective Fault category labels, a simulation model can be constructed according to an actual transmission line section, a Fault sample set with unbalanced Fault types is generated for model training, in the embodiment, the used transmission line simulation model is shown in fig. 4, software implementation comprises EM (effective energy management) as a head-end power supply, EN (end energy management) as an end power supply, ZM (zero internal resistance) as a head-end power supply, ZN (zero internal resistance) as an end power supply, the resistance value of an adjustable resistor L1 is in direct proportion to the length Lm of the transmission line from a Fault point to a head-end bus, the resistance value of an adjustable resistor L2 is in direct proportion to the length Ln of the transmission line from the Fault point to the end bus, the sum of the resistance values of L1 and L2 is in direct proportion to the length L of the transmission line, so that different Fault point positions on the transmission line can be simulated by changing the resistance values of L1 and L2, the Fault module Fault integrates an ideal breaker and a transition resistor to realize the connection between the Fault points through a transition resistor, for example, the phase A grounding fault point is phase A and phase ground, and the AC phase short-circuit fault point is phase A and phase C. The simulation may employ, but is not limited to, MATLAB/Simulink software, and the steps of generating the fault sample set are substantially as shown in fig. 5, including:
(1) according to the length L of the power transmission line MN, a fault distance Lm from an M-side bus is generated in a uniform sampling mode in the range of the power transmission line MN, and the fault distance Ln from an N-side bus is equal to L-Lm;
(2) uniformly sampling within the range of 0-90 degrees to generate fault phase angles;
(3) generating fault types according to probability sampling corresponding to fault types such as single-phase earth faults, two-phase short-circuit faults, two-phase earth faults or three-phase short-circuit faults;
(4) judging the fault type, if the fault type is a single-phase fault, firstly generating a phase difference according to the fault probability of different phase differences, and then sampling according to the probability corresponding to the single-phase fault reasons such as lightning stroke, mountain fire and the like to generate a single-phase fault reason and matching corresponding transition resistance;
other fault types are uniformly sampled to generate fault phases, namely the fault phases of the two-phase short circuit/ground fault include the possibility of equal probability of AC, AB and BC;
(5) for a single-phase earth fault, transition resistors are generated by uniformly sampling according to the probability of the corresponding transition resistor of the generated fault reason, for a two-phase short-circuit fault and a three-phase short-circuit fault, the transition resistors are set to be as small as possible in consideration of the fault characteristic of metallic earth, the transition resistors corresponding to the two-phase earth short-circuit fault are set according to corresponding values, in the embodiment, the transition resistors of the two-phase and three-phase short-circuit faults are set to be 0.01 omega, and the transition resistors of the two-phase earth are set to be 5 omega.
(6) And (4) generating a fault sample, namely a two-dimensional transient waveform image, by simulation, and taking the fault type, the fault phase and the fault reason determined in the previous step as labels of the fault sample.
(7) And when the number n of the existing simulation samples is less than the number D of the target simulation samples, repeating the steps.
In order to be closer to the actual occurrence of the fault, the above method and model can be used to generate an imbalance-like fault simulation sample set, and the following settings are used in the embodiment, but not limited to:
the length L of the power transmission line MN is 100km, the single-phase ground fault proportion is 90%, the two-phase ground fault proportion is 6%, the two-phase short circuit fault proportion is 3%, and the three-phase short circuit fault proportion is 1%;
wherein, the A phase fault, the B phase fault and the C phase fault in the fault phase of the single-phase fault type are 40%, 20% and 40%;
in the single-phase fault type, the lightning stroke fault accounts for 56.7%, the transition resistance takes 6 omega as an average value, 1.8 omega is extracted and generated as a standard deviation, the mountain fire fault accounts for 27%, the transition resistance takes 600 omega as an average value, 20 omega is extracted and generated as a standard deviation, the foreign matter fault accounts for 3.6%, the transition resistance takes 20 omega as an average value, 2.5 omega is extracted and generated as a standard deviation, the tree flash fault accounts for 2.7%, the transition resistance takes 300 omega as an average value, and 10 omega is extracted and generated as a standard deviation.
In this embodiment, after 5000 sets of fault samples are generated with the above settings, 70% of samples are randomly extracted from the faults of each fault type as a training set, and in combination with a pre-trained transfer learning-AlexNet network, the minimum number of training samples per round is set to 128, the round number is set to 30, and the weight of the network is trained by using a self-adaptive moment estimation solver. In this embodiment, the accuracy of training is already 90% after 50 iterations, and the variation trends of the training result and the test result are both stable during the 250 th iteration, which basically completes the learning of extracting and classifying the transient waveform image features. Fig. 6 shows the test results of the test set, and the fault identification accuracy can be obtained by calculating the test results by using the classification result accuracy calculation method.
The results show that the fault identification method based on the transfer learning-AlexNet network can accurately distinguish the fault types, and can accurately distinguish fault phases for single-phase grounding short circuit, two-phase grounding short circuit and three-phase short circuit, and the accuracy rate is 100%. The method provided by the invention has high identification capability for fault reasons, as shown in fig. 6, only 1 foreign matter fault has identification errors, 98 foreign matter fault identification accuracy is 99.71% (349/350), and the identification accuracy of the rest fault reasons is 100%, so that the power transmission line fault identification classifier based on transient waveform image identification provided by the invention can realize expected classification effect.
The power transmission line fault identification classifier can be deployed at substations at two ends of a power transmission line, fault identification is carried out through recorded transient waveform images when the power transmission line fails, and the identified power transmission line fault type and reason are used as bases for adaptive reclosing and carrying out rapid recovery handling decision of the power transmission line fault. Therefore, an embodiment further provides a transmission line fault identification system based on transient waveform image recognition, which is used for being deployed in a substation, and the system is basically as shown in fig. 7, and includes a two-dimensional transient waveform image generation module, configured to obtain a voltage/current signal at a head end of a transmission line, a voltage/current signal at a tail end of the transmission line, and transient waveform data of zero-sequence components at the head end and the tail end of the transmission line, which correspond to a transmission line fault, and generate and output a two-dimensional transient waveform image; the power transmission line fault identification classifier is connected with the two-dimensional transient waveform image generation module and is used for identifying a fault of a faulty power transmission line according to the input two-dimensional transient waveform image.
In addition, the present embodiment further provides a locally learnable transmission line fault identification system based on transient waveform image recognition, the system is basically as shown in fig. 8, and the system is different from the system in fig. 7 in that the system further includes an imbalance-like fault sample generation module, configured to generate a required fault sample set according to a specific imbalance probability configuration according to the foregoing fault sample set generation step; and initially, the pre-trained AlexNet network is not trained as a power transmission line fault identification classifier;
the training module is used for carrying out power transmission line fault classification training on the pre-trained AlexNet network by using the obtained fault sample set.
According to the system, the unbalance-like fault sample generation module can be configured locally as required, so that the generated unbalance-like fault sample set is more consistent with the actual condition of the current power transmission line. And in some examples, the system can also cope with changes at any time by resetting and retraining the last three layers of the AlexNet network.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that, while the invention has been described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Meanwhile, the detailed structures, characteristics and the like of the common general knowledge in the embodiments are not described in detail herein. Finally, the scope of the claims should be determined by the content of the claims, and the description of the embodiments and the like in the specification should be used for interpreting the content of the claims.
Claims (10)
1. Transmission line fault identification classifier based on transient state waveform image recognition, its characterized in that includes:
an AlexNet network;
the AlexNet network is constructed in the following way:
s1, normalizing the transient waveform data of the voltage/current signals at the head end of the power transmission line, the voltage/current signals at the tail end and the zero-sequence components at the head end and the tail end corresponding to the power transmission line fault;
generating a two-dimensional transient waveform image as a fault sample according to the transient waveform data after normalization processing corresponding to the same fault;
obtaining a fault sample set formed by a plurality of fault samples;
and S2, taking the AlexNet network which is subjected to pre-training and the first full connection layer, the second full connection layer and the classification output layer are reset to be in a state to be trained, and inputting the fault sample set for training until convergence.
2. The electric transmission line fault identification classifier according to claim 1, wherein the classification labels of the fault samples include fault type, fault phase and fault reason.
3. The power transmission line fault identification classifier according to claim 1, wherein in the normalization process, the ordinate scale of each waveform on the two-dimensional transient waveform image is set according to the voltage class and the power transmission capacity of the power transmission line, and the following principle is followed:
a. the same ordinate scale is adopted for the voltage signals at the head end/the tail end of the power transmission line;
b. the same longitudinal coordinate scale is adopted for the head end/tail end current signals of the power transmission line;
c. the zero-sequence components of the head end and the tail end adopt the same longitudinal coordinate scale;
d. each ordinate scale is as small as possible on the premise that any transient waveform data is not truncated and displayed.
4. The power transmission line fault identification classifier according to claim 1, wherein in the two-dimensional transient waveform image, transient waveform data of each phase of the power transmission line head end/tail end voltage/current signal are displayed in an overlapping manner.
5. The power transmission line fault identification classifier of claim 4, wherein A, B, C the transient waveform data of each phase are differentiated by different colors.
6. The power transmission line fault identification classifier according to claim 1, wherein in the two-dimensional transient waveform image, transient waveform data of the first/last zero sequence components are displayed in an overlapping manner.
7. The electric transmission line fault identification classifier according to claim 1, wherein the fault sample set is an unbalance-like fault sample set generated from a simulation model constructed from actual electric transmission line segments.
8. The electric transmission line fault identification classifier of claim 7, wherein the step of generating the imbalance-like fault sample set by the simulation model comprises:
(1) according to the length L of the power transmission line MN, a fault distance Lm from an M-side bus is generated in a uniform sampling mode in the range of the power transmission line MN, and the fault distance Ln from an N-side bus is equal to L-Lm;
(2) uniformly sampling within the range of 0-90 degrees to generate fault phase angles;
(3) generating fault types according to probability sampling corresponding to fault types such as single-phase earth faults, two-phase short-circuit faults, two-phase earth faults or three-phase short-circuit faults;
(4) judging the fault type, if the fault type is a single-phase fault, firstly generating a phase class according to the fault probability of different phase classes, then sampling according to the probability corresponding to the specific single-phase fault reason to generate a single-phase fault reason and matching the single-phase fault reason with corresponding transition resistance;
other fault types are uniformly sampled to generate fault phases;
(5) for single-phase earth faults, uniformly sampling and generating transition resistors according to the probability of the corresponding transition resistors of the generated fault reasons;
(6) setting fixed setting for two-phase short-circuit fault and three-phase short-circuit fault transition resistance according to corresponding values;
(7) simulating to generate a fault sample, namely a two-dimensional transient waveform image, and taking the fault type, the fault phase and the fault reason determined in the previous step as labels of the fault sample;
(8) and repeating the steps until the number of the simulation samples reaches the required number.
9. The method for identifying the faults of the power transmission line based on the transient waveform image recognition is characterized by comprising the following steps of:
recording transient waveform data of a voltage/current signal at the head end of the power transmission line, a voltage/current signal at the tail end of the power transmission line and zero sequence components of the head end and the tail end of the power transmission line when the power transmission line fails, and generating a two-dimensional transient waveform image;
inputting the obtained two-dimensional transient waveform image into the power transmission line fault identification classifier according to any one of claims 1 to 8 for identifying fault of the faulty power transmission line.
10. The power transmission line fault identification system based on transient waveform image identification is characterized by comprising,
the two-dimensional transient waveform image generation module is used for acquiring transient waveform data of a head end voltage/current signal, a tail end voltage/current signal and a head-tail end zero sequence component of the power transmission line corresponding to the power transmission line fault, and generating and outputting a two-dimensional transient waveform image;
the transmission line fault identification classifier according to any one of claims 1 to 8, connected to the two-dimensional transient waveform image generation module, for identifying a fault of a faulty transmission line according to the input two-dimensional transient waveform image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210041191.0A CN114414942A (en) | 2022-01-14 | 2022-01-14 | Power transmission line fault identification classifier, identification method and system based on transient waveform image identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210041191.0A CN114414942A (en) | 2022-01-14 | 2022-01-14 | Power transmission line fault identification classifier, identification method and system based on transient waveform image identification |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114414942A true CN114414942A (en) | 2022-04-29 |
Family
ID=81273867
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210041191.0A Pending CN114414942A (en) | 2022-01-14 | 2022-01-14 | Power transmission line fault identification classifier, identification method and system based on transient waveform image identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114414942A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115932484A (en) * | 2023-02-15 | 2023-04-07 | 重庆大学 | Method and device for identifying and ranging faults of power transmission line and electronic equipment |
| CN116008731A (en) * | 2023-02-15 | 2023-04-25 | 重庆大学 | Power distribution network high-resistance fault identification method and device and electronic equipment |
| CN116070151A (en) * | 2023-03-17 | 2023-05-05 | 国网安徽省电力有限公司超高压分公司 | Ultra-high voltage direct current transmission line fault detection method based on generalized regression neural network |
| CN117290756A (en) * | 2023-09-25 | 2023-12-26 | 重庆大学 | Power transmission line fault identification method and device based on federal learning and electronic equipment |
| CN117314883A (en) * | 2023-10-27 | 2023-12-29 | 国网四川省电力公司电力科学研究院 | Power distribution network fault line selection method and system based on EWT and VGGNet |
| CN118040679A (en) * | 2024-04-11 | 2024-05-14 | 西南交通大学 | Disturbance and fault identification method for in-phase power supply system |
-
2022
- 2022-01-14 CN CN202210041191.0A patent/CN114414942A/en active Pending
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115932484A (en) * | 2023-02-15 | 2023-04-07 | 重庆大学 | Method and device for identifying and ranging faults of power transmission line and electronic equipment |
| CN116008731A (en) * | 2023-02-15 | 2023-04-25 | 重庆大学 | Power distribution network high-resistance fault identification method and device and electronic equipment |
| CN115932484B (en) * | 2023-02-15 | 2023-07-18 | 重庆大学 | Power transmission line fault identification and fault location method and device and electronic equipment |
| CN116008731B (en) * | 2023-02-15 | 2023-08-25 | 重庆大学 | Power distribution network high-resistance fault identification method and device and electronic equipment |
| CN116070151A (en) * | 2023-03-17 | 2023-05-05 | 国网安徽省电力有限公司超高压分公司 | Ultra-high voltage direct current transmission line fault detection method based on generalized regression neural network |
| CN116070151B (en) * | 2023-03-17 | 2023-06-20 | 国网安徽省电力有限公司超高压分公司 | Ultra-high voltage direct current transmission line fault detection method based on generalized regression neural network |
| CN117290756A (en) * | 2023-09-25 | 2023-12-26 | 重庆大学 | Power transmission line fault identification method and device based on federal learning and electronic equipment |
| CN117290756B (en) * | 2023-09-25 | 2024-04-16 | 重庆大学 | Power transmission line fault identification method and device based on federal learning and electronic equipment |
| CN117314883A (en) * | 2023-10-27 | 2023-12-29 | 国网四川省电力公司电力科学研究院 | Power distribution network fault line selection method and system based on EWT and VGGNet |
| CN117314883B (en) * | 2023-10-27 | 2024-04-16 | 国网四川省电力公司电力科学研究院 | Power distribution network fault line selection method and system based on EWT and VGGNet |
| CN118040679A (en) * | 2024-04-11 | 2024-05-14 | 西南交通大学 | Disturbance and fault identification method for in-phase power supply system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114414942A (en) | Power transmission line fault identification classifier, identification method and system based on transient waveform image identification | |
| CN110082640B (en) | Distribution network single-phase earth fault identification method based on long-time memory network | |
| Du et al. | Single line-to-ground faulted line detection of distribution systems with resonant grounding based on feature fusion framework | |
| CN108562821B (en) | Method and system for determining single-phase earth fault line selection of power distribution network based on Softmax | |
| CN110247420B (en) | Intelligent fault identification method for HVDC transmission line | |
| CN116008731B (en) | Power distribution network high-resistance fault identification method and device and electronic equipment | |
| CN108805107B (en) | Method for identifying partial discharge defects in GIS based on PRPS signal | |
| CN106841905A (en) | A kind of recognition methods of transformer short circuit fault and device | |
| CN112098889B (en) | Single-phase earth fault positioning method based on neural network and feature matrix | |
| CN109828181A (en) | A kind of transformer winding minor failure detection method based on MODWT | |
| CN115932484B (en) | Power transmission line fault identification and fault location method and device and electronic equipment | |
| CN113850330A (en) | Power distribution network fault cause detection method based on short-time Fourier transform and convolutional neural network | |
| Du et al. | Detection of single line-to-ground fault using convolutional neural network and task decomposition framework in distribution systems | |
| CN111008641A (en) | Power transmission line tower external force damage detection method based on convolutional neural network | |
| CN110261746A (en) | Electric cable stoppage detection method based on oscillating wave voltage periodic attenuation characteristic | |
| CN114021433A (en) | Construction method and application of dominant instability mode recognition model of power system | |
| CN116540025A (en) | Fault detection method based on transfer learning and residual error network | |
| CN113610119B (en) | Method for identifying power transmission line development faults based on convolutional neural network | |
| CN114169249A (en) | Artificial intelligence identification method for high-resistance grounding fault of power distribution network | |
| CN111999591A (en) | Method for identifying abnormal state of primary equipment of power distribution network | |
| CN112946425A (en) | Fault positioning method for mining travelling wave time-frequency domain characteristics by utilizing deep learning | |
| CN111953657B (en) | Sequence-data joint driven CPS network attack identification method for power distribution network | |
| CN113406437B (en) | Power transmission line fault detection method for generating countermeasure network based on auxiliary classification | |
| CN106646138A (en) | Method for locating grounding fault of power distribution network based on multi-sample frequency wavelet character energy conversion | |
| Asbery et al. | Electric transmission system fault identification using modular artificial neural networks for single transmission lines |
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 |