CN-122020320-A - Distribution line fault type identification method based on data enhancement algorithm

CN122020320ACN 122020320 ACN122020320 ACN 122020320ACN-122020320-A

Abstract

The application relates to the technical field of power distribution network protection control and discloses a power distribution line fault type identification method based on a data enhancement algorithm, which comprises the steps of obtaining fault time sequence data corresponding to each fault type of a power distribution line, and processing to obtain three-phase voltage and three-phase current transient fault data with long time scale; the method comprises the steps of carrying out data enhancement on three-phase voltage and three-phase current transient fault data in a long time scale to obtain a training data set, constructing a first power distribution network fault identification model, inputting the training data set into the first power distribution network fault identification model to carry out pre-training to obtain a second power distribution network fault identification model, identifying to-be-identified fault data of a power distribution line based on the second power distribution network fault identification model, and outputting the fault type of the power distribution line. The application has the effects of rapidly judging the specific reason of the power distribution line fault in the power distribution network fault repair link and meeting the requirement of high power supply reliability.

Inventors

ZHOU TAIBIN
YING HAIJUE
YI YONGLI
KONG FANFANG
WANG FAN
WANG YU
YU YUANLI
LI DANYAN
LI HAIPENG
XU BIN
WANG XINGJIE
GU Yu
TANG YAOJING
WU XUGUANG
CAO WANGSHU
WANG ZHANG
WANG LINHAI
LIN ZHENXING

Assignees

国网浙江省电力有限公司温州供电公司

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (10)

1. A distribution line fault type identification method based on a data enhancement algorithm is characterized by comprising the following steps, Acquiring fault time sequence data corresponding to each fault type of a distribution line, and processing to obtain three-phase voltage and three-phase current transient fault data with long time scale; Performing data enhancement on the transient fault data of the three-phase voltage and the three-phase current in a long time scale to obtain a training data set; a first power distribution network fault identification model is constructed, including, An input layer comprising convolution kernels of different sizes, the input layer for parallel processing of fault data; The residual error layer comprises a plurality of cascaded residual error modules, and the residual error layer is used for reducing the sequence length output by the input layer; Introducing a channel attention mechanism for identifying a characteristic channel important program and a time attention mechanism for locating a fault occurrence period into the residual layer; the self-adaptive average pooling layer is used for pooling the input sequence input into the residual error layer into a preset dimension, reserving global statistical information and providing a unified input dimension for the classifier; Replacing a second convolution layer in the residual layer by using an SKN model, focusing on the characteristic scale of the three-phase voltage and the three-phase current related to the fault type, and obtaining a first power distribution network fault identification model; Inputting the training data set into the first power distribution network fault identification model for pre-training to obtain a second power distribution network fault identification model; And identifying the to-be-identified fault data of the distribution line based on the second distribution network fault identification model, and outputting the fault type of the distribution line.
2. The method for identifying the fault type of a distribution line based on a data enhancement algorithm according to claim 1, further comprising the steps of, Acquiring fault data of a current distribution line for model migration and fine adjustment of the second distribution network fault identification model to obtain a target distribution network fault identification model; and replacing the second power distribution network fault identification model by using the target power distribution network fault identification model.
3. The method for identifying the fault type of the distribution line based on the data enhancement algorithm according to claim 2, wherein the step of obtaining the fault data of the current distribution line for model migration and fine tuning of the second distribution network fault identification model to obtain the target distribution network fault identification model comprises, Carrying out standardized processing on the existing fault data in the power distribution network to obtain real fault data; Freezing the weight parameters of the second power distribution network fault identification model; Thawing SKN module parameters of the second power distribution network fault identification model, and replacing the classifier module of the second power distribution network fault identification model by using a new classifier module which is initialized randomly and has an output dimension of 1; and inputting the real fault data into the second power distribution network fault recognition model for training, and finishing the fine tuning of the SKN module and the training of the classifier module to obtain a target power distribution network fault recognition model.
4. The method of claim 1, wherein the channel attention mechanism comprises a plurality of SE modules connected in series, wherein the SE module at the end is connected to the output of the last convolutional layer of the residual layer.
5. The method for identifying the fault type of the distribution line based on the data enhancement algorithm according to claim 1, wherein the time attention mechanism sends out a query to the characteristic representation of the other time step based on the characteristic representation of the current time step, and obtains the correlation degree between the characteristic representation of the other time step and the characteristic representation of the current time step; Respectively taking the characteristic representation of the current time step and the characteristic representation of other time steps as index keys of inquiry, and carrying out matching calculation with a preset inquiry vector to obtain a matching value; and representing the characteristic of the time step corresponding to the largest matching value as a key fault occurrence period in the query vector.
6. The method for identifying fault types of distribution lines based on data enhancement algorithm according to any one of claims 1 to 5, wherein the step of data enhancing three-phase voltage and three-phase current transient fault data on a long time scale to obtain a training data set comprises, Initializing 24 models TimeGAN, wherein the 1 st model TimeGAN is used for enhancing the A-phase voltage sequence fault data collected by the line lightning breakage non-dropping fault type, the 2 nd model TimeGAN is used for enhancing the B-phase voltage sequence fault data collected by the line lightning breakage non-dropping fault type, the 3 rd model TimeGAN is used for enhancing the C-phase voltage sequence fault data collected by the line lightning breakage non-dropping fault type, the 4 th model TimeGAN is used for enhancing the A-phase current sequence fault data collected by the line lightning breakage non-dropping fault type, the 5 th model TimeGAN is used for enhancing the B-phase current sequence fault data collected by the line lightning breakage non-dropping fault type, the 6 th model TimeGAN is used for enhancing the C-phase current sequence fault data collected by the line lightning breakage non-dropping fault type, the 7 th TimeGAN model is used for enhancing the A-phase voltage sequence fault data adopted by the line lightning disconnection and grounding fault type, the 8 th TimeGAN model is used for enhancing the B-phase voltage sequence fault data adopted by the line lightning disconnection and grounding fault type, the 9 th TimeGAN model is used for enhancing the C-phase voltage sequence fault data adopted by the line lightning disconnection and grounding fault type, the 10 th TimeGAN model is used for enhancing the A-phase current sequence fault data adopted by the line lightning disconnection and grounding fault type, the 11 th TimeGAN model is used for enhancing the B-phase current sequence fault data adopted by the line lightning disconnection and grounding fault type, the 12 th TimeGAN model is used for enhancing the C-phase current sequence fault data adopted by the line lightning disconnection and grounding fault type, the 13 th TimeGAN model is used for enhancing the A-phase voltage sequence fault data adopted by the line insulator flashover fault type, the 14 th TimeGAN model is used for enhancing the B-phase voltage sequence fault data adopted by the line insulator flashover fault type, the 15 th TimeGAN model is used for enhancing the C-phase voltage sequence fault data adopted by the line insulator flashover fault type, the 16 th TimeGAN model is used for enhancing the A-phase current sequence fault data adopted by the line insulator flashover fault type, the 17 th TimeGAN model is used for enhancing the B-phase current sequence fault data adopted by the line insulator flashover fault type, the 18 th TimeGAN model is used for enhancing the C-phase current sequence fault data adopted by the line insulator flashover fault type, the No. 19 TimeGAN model is used for enhancing the A-phase voltage sequence fault data adopted by the line tree antenna fault type, the No. 20 TimeGAN model is used for enhancing the B-phase voltage sequence fault data adopted by the line tree antenna fault type, the No. 21 TimeGAN model is used for enhancing the C-phase voltage sequence fault data adopted by the line tree antenna fault type, the No. 22 TimeGAN model is used for enhancing the A-phase current sequence fault data adopted by the line tree antenna fault type, the No. 23 TimeGAN model is used for enhancing the B-phase current sequence fault data adopted by the line tree antenna fault type, and the No. 24 TimeGAN model is used for enhancing the C-phase current sequence fault data adopted by the line tree antenna fault type; the method comprises the steps of presetting a condition judgment function and a loss function, respectively inputting the three-phase voltage and three-phase current transient fault data of a long time scale into 24 TimeGAN models for training, enabling the value of the loss function to be minimum, and obtaining 24 TimeGAN models for carrying out data enhancement on corresponding phase voltage sequences or current sequences under each fault type of a distribution line; and inputting the three-phase voltage and three-phase current transient fault data with long time scale into 24 TimeGAN models for data enhancement to obtain the training data set.
7. The method for identifying the fault type of the distribution line based on the data enhancement algorithm according to claim 6, wherein the step of inputting the transient fault data of the three-phase voltage and the three-phase current with a long time scale into 24 TimeGAN models for data enhancement comprises the steps of, Inputting a voltage sequence or a current sequence corresponding to the ith TimeGAN model into an embedded component in a corresponding TimeGAN model, and reserving characteristic information of a fault waveform of the distribution line by adopting a condition judgment function Sigmoid function to obtain a hidden state sequence; Inputting the hidden state sequence into a recovery component in a corresponding TimeGAN model, recovering the characteristic information of the fault waveform of the distribution line, reconstructing to obtain a fitted voltage sequence or current sequence, and outputting the fitted sequence; the values of the corresponding voltage sequence or current sequence and the fitting sequence are minimized.
8. The method for identifying the fault type of the distribution line based on the data enhancement algorithm according to claim 7, wherein the step of inputting the transient fault data of the three-phase voltage and the three-phase current with a long time scale into 24 TimeGAN models for data enhancement comprises the steps of, Inputting the hidden state sequence into a generator in a corresponding TimeGAN model, and simulating the real fluctuation characteristic of the fault waveform data of the distribution line to obtain a generated data sequence; Minimizing errors of the generated data sequence and the hidden state sequence.
9. The method for identifying the fault type of the distribution line based on the data enhancement algorithm according to claim 8, wherein the step of inputting the transient fault data of the three-phase voltage and the three-phase current with a long time scale into 24 TimeGAN models for data enhancement comprises the steps of, Inputting the hidden state sequence and the generated data sequence into a corresponding TimeGAN model; Training the discriminant using a bi-classification cross entropy loss function.
10. The method for identifying the fault type of the distribution line based on the data enhancement algorithm according to claim 9, wherein the step of inputting the transient fault data of the three-phase voltage and the three-phase current with a long time scale into 24 TimeGAN models for data enhancement comprises the steps of, And jointly optimizing the embedded component, the recovery component, the generator and the discriminator by adopting a loss function for calculating the Euclidean distance, the mean value and the variance sum of the real data sequence and the generated data sequence until the Euclidean distance, the mean value and the variance sum of the real data sequence and the generated data sequence are minimum.

Description

Distribution line fault type identification method based on data enhancement algorithm Technical Field The application relates to the technical field of protection control of power distribution networks, in particular to a power distribution line fault type identification method based on a data enhancement algorithm. Background At present, the intelligent power distribution acquisition terminal and the power distribution automation master station not only improve the operation efficiency and the management level of a power grid, but also basically realize fault positioning, fault isolation and power supply recovery of a non-fault area of the power distribution network. According to analysis of national power grid construction and transformation engineering, high power supply reliability is a core index of a future power distribution network, and an existing power distribution network fault positioning system can only realize positioning and isolation of a power distribution line short circuit fault and a ground fault, but the efficiency of a fault repair link is lower, the prior art cannot judge whether a line is broken by lightning, insulator pollution flashover, tree contact line or other types of faults, the improvement of the efficiency of a field repair team is restricted, on the other hand, the problem that the field repair measures or the preparation of a repair tool are insufficient to cause the overlong repair time occurs is solved, and on the other hand, the problem that the reinforcement engineering of the power distribution line lacks quantitative data support under the condition that the fault tracing diagnosis is lost for a long time exists, and the phenomena of investment waste or incomplete transformation exist. The future distribution network is flexible to operate and strong in toughness power supply capacity, but power failure accidents are easy to occur under extreme climate conditions. In case of failure of the distribution line, besides realizing failure positioning, failure identification is also required to be realized, and specific reasons of failure of the distribution line, such as broken line due to lightning, pollution flashover of insulators, contact line of trees and the like, are rapidly judged so as to rapidly form a failure emergency repair scheme and allocate emergency repair materials, guide a subsequent distribution line reinforcement scheme, realize rapid power restoration of power supply load and meet the requirement of high power supply reliability. Aiming at the related technology, the inventor finds that the specific reason for failure of the distribution line cannot be judged in the existing power distribution network failure rush-repair link, and the problem that the requirement of high power supply reliability is difficult to meet is solved. Disclosure of Invention In order to rapidly judge the specific cause of the distribution line fault, the application provides a distribution line fault type identification method based on a data enhancement algorithm. In a first aspect, the present application provides a method for identifying a fault type of a distribution line based on a data enhancement algorithm. The application is realized by the following technical scheme: A distribution line fault type identification method based on a data enhancement algorithm comprises the following steps, Acquiring fault time sequence data corresponding to each fault type of a distribution line, and processing to obtain three-phase voltage and three-phase current transient fault data with long time scale; Performing data enhancement on the transient fault data of the three-phase voltage and the three-phase current in a long time scale to obtain a training data set; a first power distribution network fault identification model is constructed, including, An input layer comprising convolution kernels of different sizes, the input layer for parallel processing of fault data; The residual error layer comprises a plurality of cascaded residual error modules, and the residual error layer is used for reducing the sequence length output by the input layer; Introducing a channel attention mechanism for identifying a characteristic channel important program and a time attention mechanism for locating a fault occurrence period into the residual layer; the self-adaptive average pooling layer is used for pooling the input sequence input into the residual error layer into a preset dimension, reserving global statistical information and providing a unified input dimension for the classifier; Replacing a second convolution layer in the residual layer by using an SKN model, focusing on the characteristic scale of the three-phase voltage and the three-phase current related to the fault type, and obtaining a first power distribution network fault identification model; Inputting the training data set into the first power distribution network fault identification model for pre-training to obtain a second