Power transmission line fault positioning method based on improved YOLOv8 model and double-end TWM

Abstract

The invention relates to a power transmission line fault positioning method based on an improved YOLOv8 model and a double-end TWM, and the method comprises the steps: collecting three-phase current transient signals at M and N ends of a line after a fault occurs, carrying out the phase-mode conversion of the three-phase current transient signals through employing a Karen's transformation method, and selecting a traveling wave current line-mode component as a fault traveling wave research object; using a wavelet analysis method to decompose the traveling wave current line mode component to obtain traveling wave generation pictures with different sampling rates; building a Pix2PixHD model and training the Pix2PixHD model; constructing a fault target detection model, and obtaining a time range of a fault traveling wave head by using the fault target detection model; constructing a traveling wave main body segmentation model, and obtaining a traveling wave head maximum value point by using the traveling wave main body segmentation model; and according to the time range of the fault traveling wave head and the traveling wave head maximum value point information, calculating to obtain a line fault point position. According to the invention, accurate line fault positioning is realized, and the problem of low line fault positioning accuracy in the prior art is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of transmission line fault ranging, and in particular relates to a transmission line fault locating method based on an improved YOLOv8 model and a two-terminal TWM. Background Art

[0002] As the industrial lifeline of a country, electricity has a profound impact on the country's economic development and people's living standards. With the development of society and the needs of production and life, people's expectations for high-quality electricity use are becoming more and more urgent, and their expectations for rapid processing and improved power supply reliability after power grid failures are becoming more and more urgent. With the increasing advancement of modern power grid fault detection technology, in order to meet the high requirements of power supply reliability, the construction of smart grids has become an inevitable trend and important direction for the development of the power industry. Therefore, the power network plays an irreplaceable role in the development of society and the production and life of the people. Ensuring that the power network has high reliability and stability is the requirement of the times and the expectation of mankind.

[0003] When a transmission line fails, it is critical to quickly and promptly detect the fault, accurately locate the fault point, and promptly handle the fault, prevent the fault from expanding, minimize the losses caused by the fault, and ensure the safety of people's lives and property. Since the traveling wave method is almost unaffected by transition resistance, fault type, and line length, the fault distance calculation accuracy is high and it has been widely used. The principle of the traveling wave method is that after a fault occurs, a fault voltage and current traveling wave signal will be generated that propagates to both ends of the line. By identifying the singular point of the fault signal and then determining the time when the traveling wave reaches both ends of the line, the fault distance can be measured.

[0004] The positioning accuracy of the traveling wave method is related to the high-precision sampling rate and fault signal decomposition method. The existing line fault distance measurement technology mainly relies on the measuring device, and the distance measurement accuracy is limited by the sampling rate of the measuring device. The measuring device can only guarantee the distance measurement accuracy at a high sampling rate, and increasing the sampling rate of the measuring device will lead to a significant increase in the manufacturing cost of the measuring device. Summary of the invention

[0005] The purpose of the present invention is to solve the above problems and provide a transmission line fault location method based on an improved YOLOv8 model and a two-terminal TWM, which combines wavelet transform and AI model to construct a fault target detection model and a traveling wave main body segmentation model. The traveling wave main body segmentation model obtains the traveling wave head maximum point according to the traveling wave generated picture, thereby accurately locating the line fault, solving the problem of low line fault location accuracy in the prior art.

[0006] In order to achieve the above-mentioned purpose, the technical solution provided by the present invention is a transmission line fault location method based on an improved YOLOv8 model and a two-terminal TWM. The Pix2PixHD model is used to obtain a traveling wave generation picture at a sampling rate of 2MHz based on an initial traveling wave picture at a sampling rate of 1MHz, and the improved YOLOv8 network is used to construct a fault target detection model and a traveling wave body segmentation model. The fault target detection model is used to detect and determine the fault traveling wave according to the traveling wave generation picture and obtain the time range of the fault traveling wave head. The traveling wave body segmentation model is used to obtain a traveling wave body segmentation map and a traveling wave head maximum point according to the traveling wave generation picture.

[0007] The transmission line fault location method based on the improved YOLOv8 model and the two-terminal TWM includes the following steps: Step 1: Collect the three-phase current transient signals at both ends of the line M and N after the fault occurs, use the Karenberg transformation method to perform phase mode transformation on the three-phase current transient signals, and select the line mode component of the traveling wave current as the fault traveling wave research object; Step 2: Use the wavelet analysis method to decompose the line mode components of the traveling wave current, obtain the traveling wave generation images at 1MHz and 2MHz sampling rates, and construct the traveling wave generation data set and the traveling wave detection data set; Step 3: Build a Pix2PixHD model, and use the traveling wave generation data set obtained in step 2 to train the Pix2PixHD model; Step 4: Build an improved YOLOv8 model, and use the traveling wave detection data set obtained in step 2 to train the improved YOLOv8 model to obtain a fault target detection model and a traveling wave main body segmentation model; Step 5: When a new single-phase fault occurs, phase mode transformation and wavelet transform are performed to obtain a 1MHz initial traveling wave head image, and the 1MHz initial traveling wave image is input into the trained Pix2PixHD model to obtain a 2MHz initial traveling wave generation image; Step 6: calling the fault target detection model, the input of which is the traveling wave generation picture at a sampling rate of 2 MHz output by the Pix2PixHD model, and using the fault target detection model to obtain the time range of the fault traveling wave head; Step 7: calling the traveling wave main body segmentation model, the input of the traveling wave main body segmentation model is the traveling wave generation picture at a sampling rate of 2MHz output by the Pix2PixHD model, and using the traveling wave main body segmentation model to obtain the traveling wave main body segmentation map and the traveling wave head maximum point; Step 8: According to the time range of the fault traveling wave head obtained in step 6 and the maximum point information of the traveling wave head obtained in step 7, the position of the line fault point is calculated.

[0008] Preferably, in step 2, the wavelet analysis method is used to decompose the line mode component of the traveling wave current, specifically: Basic wavelet functions satisfy: ; (1) Where R1 represents the integral range; With a representing the scale factor and b representing the displacement factor, the basic wavelet function is scaled and time-shifted to obtain the wavelet basis function. , ; (2) Given a continuous signal , and its corresponding continuous wavelet transform is defined as: ; (3) In the formula is the continuous wavelet transform function; Discretize the scale factor a and displacement factor b, and set , , , R represents the real number space, j and k are coefficients, a0 represents the base of the power-order discretization of the scale factor a, and b0 represents the normalization coefficient of the uniform discrete value of the scale factor b, and the discrete wavelet transform is obtained: ; (4) In the formula Represents the discrete wavelet transform function.

[0009] Preferably, in step 3, training the Pix2PixHD model specifically includes: 1) Use matlab / simulink software to simulate a transmission line, set a single-phase grounding fault on the transmission line, and obtain the current transient signals at both ends of the line; 2) Repeat the simulation multiple times with single-phase ground faults at different fault occurrence times, different ground resistances, and different fault occurrence locations; 3) reading the simulation file obtained in step 2), performing phase mode transformation on the three-phase transient current signal, decomposing the line mode component of the traveling wave current by using the wavelet analysis method, and deriving the initial traveling wave head image; 4) Using steps 1) to 3), the transmission line faults at 1MHz and 2MHz sampling rates are simulated respectively, and the generated traveling wave head images at 1MHz and 2MHz sampling rates are used as the input and output of the Pix2PixHD model respectively, to establish a traveling wave generation data set, and divide it into training set, validation set and test set according to the proportion; 5) Establish the operating environment of the Pix2PixHD model, import the traveling wave generation data set, and train the Pix2PixHD model.

[0010] Preferably, the construction process of the improved YOLOv8 network specifically includes: a) Integrate the CAA attention mechanism into the Star module in the StarNet network to obtain the Star_CAA module; b) Use the Star_CAA module to replace the Bottleneck layer of the C2f module of YOLOv8 to obtain the C2f_Star_CAA module; c) Replace all C2f modules in the backbone network Backbone and the neck network Neck in YOLOv8 with C2f_Star_CAA modules to obtain an improved YOLOv8 network.

[0011] Preferably, the C2f_Star_CAA module includes a convolution layer Conv, a segmentation layer Split, multiple Star_CAA modules, a splicing layer Concat and a convolution layer Conv connected in sequence, wherein the Star_CAA module includes a depth-separable convolution layer DWConv, a parallel convolution layer Conv, a CAA attention mechanism unit, a convolution layer Conv and a depth-separable convolution layer DWConv connected in sequence.

[0012] Preferably, the CAA attention mechanism unit includes an average pooling layer AvgPool, a convolution layer Conv, a depthwise separable convolution layer DWConv with a convolution kernel of 1*11, a depthwise separable convolution layer DWConv with a convolution kernel of 11*1, a convolution layer Conv and a Sigmoid activation layer connected in sequence.

[0013] Furthermore, in step 6, the trained Pix2PixHD model is used to generate a traveling wave generation image at a sampling rate of 2 MHz, which is used as the input of the fault target detection model constructed with the improved YOLOv8 network to obtain a traveling wave target detection image with a bounding box, a fault target probability, and a time range of the traveling wave body.

[0014] Furthermore, in step 7, the trained Pix2PixHD model is used to generate a traveling wave generation picture at a sampling rate of 2 MHz, which is used as the input of the traveling wave main body segmentation model constructed with the improved YOLOv8 network to obtain the traveling wave main body segmentation map and the relative coordinates of the traveling wave maximum points therein.

[0015] Furthermore, in step 8, the calculating of the line fault point location includes: 1) Calculate the wave velocity of the zero mode and line mode components corresponding to the fault traveling wave of the line according to the parameters of the transmission line , the calculation formula is: ; (5) In the formula , They are the inductance and capacitance of the line per unit length under the same phase sequence respectively; 2) According to the principle of double-terminal traveling wave ranging, the location of the fault on the line is assumed to be point F, and the time of the fault is recorded as , the fault traveling wave propagates from point F to points M and N at the left and right ends of the line, where M is the left end point of the line and N is the right end point of the line; The time when the initial wave head of the fault traveling wave reaches point M is recorded as The time when it reaches point N is recorded as , then: ; (6) ; (7) ; (8) In the formula Indicates the distance from point F to point M; 3) For the mixed lines of overhead lines and cable lines, normalization is performed, and the cable lines are equivalent to overhead lines of corresponding lengths. The wave velocities of traveling waves in overhead lines and cable lines are expressed as , It means that the calculation formula for the wave speed conversion of different lines is: ; (9) In the formula is the normalized equivalent overhead line length of the cable, Indicates the length of cables in the distribution network.

[0016] The above-mentioned fault location system of the transmission line fault location method based on the improved YOLOv8 model and the two-terminal TWM includes: Traveling wave generation module: used to simulate and obtain the three-phase fault transient current signal at a sampling rate of 1MHz after a line fault occurs; Phase mode conversion module: used to convert the three-phase fault transient current signal into the corresponding line mode component through the Karenberg transformation matrix; Pix2PixHD model module: used to record the main body of the fault traveling wave at a sampling rate of 1MHz in the form of a picture, and generate a traveling wave generation picture at a sampling rate of 2MHz; Fault target detection module: used to process the traveling wave generated image at a sampling rate of 2MHz to obtain the traveling wave detection image with a bounding box, the probability of the fault target, and the time range of the traveling wave body; Traveling wave main body segmentation module: used to generate image processing of traveling waves at a sampling rate of 2MHz, the traveling wave main body segmentation map and the relative coordinates of the traveling wave maximum points; Line fault point calculation module: It is used to integrate the output information of the fault target detection module and the traveling wave main body segmentation module to obtain the time when the traveling wave reaches both ends of the line, that is, the time corresponding to the maximum point of the traveling wave, so as to calculate the accurate fault point position of the line.

[0017] Compared with the prior art, the beneficial effects of the present invention include: 1) The present invention combines wavelet transform and improved YOLOv8 network to construct a fault target detection model and a traveling wave main body segmentation model. The traveling wave main body segmentation model is used to generate pictures according to the traveling waves to obtain the maximum points of the traveling wave head, thereby realizing accurate positioning of the line fault and solving the problem of low line fault positioning accuracy in the prior art.

[0018] 2) The present invention uses wavelet transform to decompose the fault traveling wave current to obtain time domain information, converts the data transmission medium from digital to image, and relies on the Pix2PixHD algorithm to improve the fault signal sampling rate.

[0019] 3) The present invention uses an improved YOLOv8 model to evaluate the similarity of fault traveling wave images generated by Pix2PixHD at a high sampling rate, and automatically reads the arrival time of the traveling wave, thereby solving the problem of inconvenience in reading time caused by converting the traveling wave from data to images, and increasing the practicality of the method.

[0020] 4) The transmission line fault location method based on the improved YOLOv8 model and the two-terminal TWM proposed in the present invention is applicable to pure overhead lines, pure cable lines and hybrid lines, and is not affected by the fault time and the size of the transition resistance. It has strong stability and anti-interference performance, and the positioning error is small. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0022] Figure 1 Schematic diagram of a transmission line fault location method according to an embodiment of the present invention.

[0023] Figure 2 This is a simulation circuit diagram of an embodiment of the present invention.

[0024] Figure 3 This is a traveling wave target detection diagram according to an embodiment of the present invention.

[0025] Figure 4 A traveling wave main body segmentation diagram is generated for the fault of an embodiment of the present invention.

[0026] Figure 5 Schematic diagram of the double-terminal traveling wave method according to an embodiment of the present invention.

[0027] Figure 6 Schematic diagram of the structure of the improved YOLOv8 according to an embodiment of the present invention.

[0028] Figure 7 A schematic diagram of the structure of a C2f_Star_CAA module according to an embodiment of the present invention.

[0029] Figure 8 A schematic diagram of the structure of a CAA attention mechanism unit according to an embodiment of the present invention.

[0030] Fig. 9 A schematic diagram of the structure of the Pix2PixHD model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0032] like Figure 1 As shown, the transmission line fault location method based on the improved YOLOv8 model and the two-terminal TWM of the embodiment includes the following steps: Step 1: Collect a large number of three-phase current transient signals at both ends of the line M and N after the fault, use Karen Bell transformation to perform phase mode transformation on the three-phase current transient signals, obtain the line mode component and ground mode component of the current, and select the line mode component as the fault traveling wave research object.

[0033] Use matlab / simulink to simulate five transmission lines, such as Figure 2 As shown, it includes pure overhead lines, pure cable lines and hybrid lines. Set one of the lines to have a single-phase grounding fault at 1MHz and 2MHz sampling rates, repeat the simulation multiple times with different fault occurrence times, different grounding resistances, different fault occurrence locations, etc., and obtain the current transient signal at both ends of the line each time and save the file.

[0034] Step 2: Use wavelet analysis to decompose the line mode components of the traveling wave current, output the initial traveling wave head images at 1MHz and 2MHz sampling rates, and establish the traveling wave generation data set and the traveling wave detection data set; Read the simulation file obtained each time, perform phase mode transformation on its three-phase transient current signal, and use wavelet analysis to decompose the line mode component of the traveling wave current. The specific method is as follows: Basic wavelet functions satisfy: (1) Where R1 represents the integral range; With a representing the scale factor and b representing the displacement factor, the basic wavelet is scaled and time-shifted to obtain the wavelet basis function. : (2) Given a continuous signal , the corresponding continuous wavelet transform is defined as: (3) Discretize the scale factor a and displacement factor b, and set , , , R represents the real number space, j and k are coefficients, a0 represents the base of the power-order discretization of the scale factor a, and b0 represents the normalization coefficient of the uniform discrete value of the scale factor b, and the discrete wavelet transform is obtained: (4) In the formula Represents the discrete wavelet transform function.

[0035] A total of one thousand initial traveling wave head images at 1MHz and 2MHz sampling rates were obtained to form the traveling wave generation data set.

[0036] A traveling wave generation data set at a sampling rate of 2MHz was randomly selected, totaling 500 images, to form the traveling wave test data set.

[0037] Step 3: Build the Pix2PixHD model and use the traveling wave to generate a dataset to train the model; The Pix2PixHD network is used for image-to-image translation, and its structure is as follows Fig. 9 As shown in the figure, the data sets required for training the model require data set A and data set B, and the relationship between the two is one-to-one. The traveling wave generation data set includes the initial traveling wave head images at 1MHz and 2MHz sampling rates, which meet the training requirements. The training, validation and test sets are divided into training, validation and test sets in a ratio of 7:2:1.

[0038] Establish the operating environment, set the number of training rounds to 300, and obtain the corresponding trained Pix2PixHD model.

[0039] Step 4: Build an improved YOLOv8 model and train it using the traveling wave detection dataset to obtain a fault target detection model and a traveling wave subject segmentation model; For improving the structure of the YOLOv8 model Figure 6 As shown in the figure, the construction process specifically includes: a) Integrate the CAA attention mechanism into the Star module in the StarNet network to obtain the Star_CAA module; b) Use the Star_CAA module to replace the Bottleneck layer of the C2f module of YOLOv8 to obtain the C2f_Star_CAA module; c) Replace all C2f modules in the backbone network Backbone and the neck network Neck in YOLOv8 with C2f_Star_CAA modules to obtain an improved YOLOv8 network.

[0040] For the C2f_Star_CAA module, its structure is as follows Figure 7 As shown, it includes sequentially connected convolutional layers Conv, segmentation layers Split, multiple Star_CAA modules, splicing layers Concat and convolutional layers Conv, wherein the Star_CAA module includes sequentially connected depthwise separable convolutional layers DWConv, parallel convolutional layers Conv, CAA attention mechanism units, convolutional layers Conv and depthwise separable convolutional layers DWConv.

[0041] For the CAA attention mechanism unit, its structure is as follows Figure 8 As shown, it includes an average pooling layer AvgPool, a convolution layer Conv, a depthwise separable convolution layer DWConv with a convolution kernel of 1*11, a depthwise separable convolution layer DWConv with a convolution kernel of 11*1, a convolution layer Conv and a Sigmoid activation layer connected in sequence.

[0042] To train and improve the YOLOv8 model, it is necessary to annotate specific targets in the traveling wave detection dataset images, use the Labelimg tool to annotate the traveling wave categories and digital categories, obtain the annotated traveling wave test dataset, and divide it into training, validation, and test sets in a ratio of 7:2:1.

[0043] Establish the operating environment, import the traveling wave inspection dataset, set the corresponding training parameters, and train the improved YOLOv8 network for traveling wave detection and traveling wave segmentation functions respectively to obtain the fault target detection model for traveling wave detection and the traveling wave subject segmentation model for traveling wave segmentation.

[0044] Step 5: When a new single-phase fault occurs, phase mode transform and wavelet transform are performed to obtain a 1MHz initial traveling wave head image. The 1MHz initial traveling wave image is input into the trained Pix2PixHD model to obtain a 2MHz initial traveling wave generation image.

[0045] Use matlab / simulink to simulate five transmission lines, such as Figure 2As shown, it includes pure overhead lines, pure cable lines and hybrid lines. Set one of the lines to have a single-phase grounding fault at a sampling rate of 1MHz. The fault occurrence time, grounding resistance, and fault location are different from the fault conditions of the traveling wave generation data set. Obtain the current transient signals at both ends of the line and save the files.

[0046] Read the file, use Karen Bell transform to perform phase mode transformation on the three-phase current transient signal to obtain the line mode component of the current. Then use wavelet analysis to decompose the line mode component of the traveling wave current to obtain the initial traveling wave head image at both ends of the line at a sampling rate of 1MHz.

[0047] The initial traveling wave images at both ends of the line at a sampling rate of 1 MHz are input into the trained Pix2PixHD model, and the initial traveling wave generation images at a sampling rate of 2 MHz are output. In this way, there is no need to simulate at a sampling rate of 2 MHz to obtain the corresponding initial traveling wave head images.

[0048] Step 6: calling the fault target detection model, the input of which is the traveling wave generation picture at a sampling rate of 2 MHz output by the Pix2PixHD model, and using the fault target detection model to obtain the time range of the fault traveling wave head; For the fault target detection model, the input is the initial traveling wave generation image at a sampling rate of 2MHz, and the output is the traveling wave target detection image at a sampling rate of 2MHz, such as Figure 3 As shown in the figure, as well as the probability of detecting the target and the position and size of the detection box of the selected detection target, the time range of the fault traveling wave head is obtained by integrating the information.

[0049] Step 7: calling the traveling wave main body segmentation model, the input of the traveling wave main body segmentation model is the traveling wave generation picture at a sampling rate of 2MHz output by the Pix2PixHD model, and using the traveling wave main body segmentation model to obtain the traveling wave main body segmentation map and the traveling wave head maximum point; For the traveling wave main segmentation model, the input is the initial traveling wave generation image at a sampling rate of 2MHz, and the output is the traveling wave main segmentation map at a sampling rate of 2MHz, such as Figure 4 As shown, as well as the probability of segmenting the target and the position of the segmentation points that depict the segmented target, the relative coordinates of the maximum point of the traveling wave head are obtained by integrating the information.

[0050] Step 8: Integrate the wave head time range and the relative coordinate information of the maximum point to obtain the arrival time of the fault traveling wave at both ends of the line, and calculate the position of the line fault point.

[0051] The principle of the double-ended traveling wave method is as follows Figure 5 As shown in the figure, after obtaining the accurate arrival time of the initial traveling wave head at both ends of the line, the fault point location is calculated according to the principle. The calculation process is as follows: 1) Calculate the wave velocity of the zero mode and line mode components corresponding to the fault traveling wave of the line according to the parameters of the transmission line , the calculation formula is: (5) In the formula , They are the inductance and capacitance of the line per unit length under the same phase sequence respectively; 2) According to the principle of double-terminal traveling wave ranging, the location of the fault on the line is assumed to be point F, and the time of the fault is recorded as , the fault traveling wave propagates from point F to points M and N at the left and right ends of the line, where M is the left end point of the line and N is the right end point of the line; The time when the initial wave head of the fault traveling wave reaches point M is recorded as The time when it reaches point N is recorded as , then: (6) (7) (8) 3) For the mixed lines of overhead lines and cable lines, normalization is performed, and the cable lines are equivalent to overhead lines of corresponding lengths. The wave velocities of traveling waves in overhead lines and cable lines are expressed as , It means that the calculation formula for the wave speed conversion of different lines is: (9) In the formula is the normalized equivalent overhead line length of the cable, Indicates the length of cables in the distribution network.

Claims

1. A transmission line fault location method based on an improved YOLOv8 model and two-terminal TWM, characterized in that: The Pix2PixHD model is used to obtain the traveling wave generation picture at a sampling rate of 2MHz based on the initial traveling wave picture at a sampling rate of 1MHz, and the improved YOLOv8 network is used to construct a fault target detection model and a traveling wave main body segmentation model. The fault target detection model is used to detect and determine the fault traveling wave based on the traveling wave generation picture and obtain the time range of the fault traveling wave head. The traveling wave main body segmentation model is used to obtain the traveling wave main body segmentation map and the traveling wave head maximum point based on the traveling wave generation picture. The method comprises the following steps: Step 1: Collect the three-phase current transient signals at both ends of the line M and N after the fault occurs, use the Karenberg transformation method to perform phase mode transformation on the three-phase current transient signals, and select the line mode component of the traveling wave current as the fault traveling wave research object; Step 2: Use the wavelet analysis method to decompose the line mode components of the traveling wave current, obtain the traveling wave generation images at 1MHz and 2MHz sampling rates, and construct the traveling wave generation data set and the traveling wave detection data set; Step 3: Build a Pix2PixHD model, and use the traveling wave generation data set obtained in step 2 to train the Pix2PixHD model; Step 4: Build an improved YOLOv8 model, and use the traveling wave detection data set obtained in step 2 to train the improved YOLOv8 model to obtain a fault target detection model and a traveling wave main body segmentation model; Step 5: When a new single-phase fault occurs, phase mode transformation and wavelet transform are performed to obtain a 1MHz initial traveling wave head image, and the 1MHz initial traveling wave image is input into the trained Pix2PixHD model to obtain a 2MHz initial traveling wave generation image; Step 6: calling the fault target detection model, the input of which is the traveling wave generation picture at a sampling rate of 2 MHz output by the Pix2PixHD model, and using the fault target detection model to obtain the time range of the fault traveling wave head; Step 7: calling the traveling wave main body segmentation model, the input of the traveling wave main body segmentation model is the traveling wave generation picture at a sampling rate of 2MHz output by the Pix2PixHD model, and using the traveling wave main body segmentation model to obtain the traveling wave main body segmentation map and the traveling wave head maximum point; Step 8: According to the time range of the fault traveling wave head obtained in step 6 and the maximum point information of the traveling wave head obtained in step 7, the position of the line fault point is calculated.

2. According to claim 1, the method for locating power line faults based on the improved YOLOv8 model and the two-terminal TWM is characterized in that: In step 2, the wavelet analysis method is used to decompose the line mode component of the traveling wave current, specifically: Basic wavelet functions satisfy: ； (1) Where R1 represents the integral range; With a representing the scale factor and b representing the displacement factor, the basic wavelet function is scaled and time-shifted to obtain the wavelet basis function. , ； (2) Given a continuous signal , and its corresponding continuous wavelet transform is defined as: ；(3) In the formula is the continuous wavelet transform function; Discretize the scale factor a and displacement factor b, and set , , , R represents the real number space, j and k are coefficients, a0 represents the base of the power-order discretization of the scale factor a, and b0 represents the normalization coefficient of the uniform discrete value of the scale factor b, and the discrete wavelet transform is obtained: ； (4) In the formula Represents the discrete wavelet transform function.

3. The method for locating a power transmission line fault based on the improved YOLOv8 model and the two-terminal TWM according to claim 1 or 2, characterized in that: In step 3, training the Pix2PixHD model specifically includes: 1) Use matlab / simulink software to simulate a transmission line, set a single-phase grounding fault on the transmission line, and obtain the current transient signals at both ends of the line; 2) Repeat the simulation multiple times with single-phase ground faults at different fault occurrence times, different ground resistances, and different fault occurrence locations; 3) reading the simulation file obtained in step 2), performing phase mode transformation on the three-phase transient current signal, decomposing the line mode component of the traveling wave current by using the wavelet analysis method, and deriving the initial traveling wave head image; 4) Using steps 1) to 3), the transmission line faults at 1MHz and 2MHz sampling rates are simulated respectively, and the generated traveling wave head images at 1MHz and 2MHz sampling rates are used as the input and output of the Pix2PixHD model respectively, to establish a traveling wave generation data set, and divide it into training set, validation set and test set according to the proportion; 5) Establish the operating environment of the Pix2PixHD model, import the traveling wave generation data set, and train the Pix2PixHD model.

4. The method for locating a power transmission line fault based on the improved YOLOv8 model and the two-terminal TWM according to claim 3 is characterized in that: The construction process of the improved YOLOv8 network specifically includes: a) Integrate the CAA attention mechanism into the Star module in the StarNet network to obtain the Star_CAA module; b) Use the Star_CAA module to replace the Bottleneck layer of the C2f module of YOLOv8 to obtain the C2f_Star_CAA module; c) Replace all C2f modules in the backbone network Backbone and the neck network Neck in YOLOv8 with C2f_Star_CAA modules to obtain an improved YOLOv8 network.

5. According to claim 4, the method for locating power line faults based on the improved YOLOv8 model and the two-terminal TWM is characterized in that: The C2f_Star_CAA module includes a sequentially connected convolution layer Conv, a segmentation layer Split, multiple Star_CAA modules, a splicing layer Concat and a convolution layer Conv, wherein the Star_CAA module includes a sequentially connected depth-separable convolution layer DWConv, a parallel convolution layer Conv, a CAA attention mechanism unit, a convolution layer Conv and a depth-separable convolution layer DWConv.

6. The method for locating a power transmission line fault based on the improved YOLOv8 model and the two-terminal TWM according to claim 4 is characterized in that: The CAA attention mechanism unit includes an average pooling layer AvgPool, a convolution layer Conv, a depthwise separable convolution layer DWConv with a convolution kernel of 1*11, a depthwise separable convolution layer DWConv with a convolution kernel of 11*1, a convolution layer Conv and a Sigmoid activation layer, which are connected in sequence.

7. The method for locating a power transmission line fault based on the improved YOLOv8 model and the two-terminal TWM according to claim 5 or 6, characterized in that: In step 6, the trained Pix2PixHD model is used to generate a traveling wave generation image at a sampling rate of 2 MHz as the input of the fault target detection model built with the improved YOLOv8 network, and a traveling wave target detection image with a bounding box, a fault target probability, and a time range of the traveling wave body are obtained.

8. The method for locating a power transmission line fault based on the improved YOLOv8 model and the two-terminal TWM according to claim 7 is characterized in that: In step 7, the trained Pix2PixHD model is used to generate a traveling wave generation picture at a sampling rate of 2 MHz, which is used as the input of the traveling wave main body segmentation model constructed with the improved YOLOv8 network to obtain the traveling wave main body segmentation map and the relative coordinates of the traveling wave maximum points therein.

9. The method for locating a power transmission line fault based on the improved YOLOv8 model and the two-terminal TWM according to claim 8, characterized in that: In step 8, the calculation to obtain the position of the line fault point includes: 1) Calculate the wave velocity of the zero mode and line mode components corresponding to the fault traveling wave of the line according to the parameters of the transmission line , the calculation formula is: ；(5) In the formula , They are the inductance and capacitance of the line per unit length under the same phase sequence respectively; 2) According to the principle of double-terminal traveling wave ranging, the location of the fault on the line is assumed to be point F, and the time of the fault is recorded as , the fault traveling wave propagates from point F to points M and N at the left and right ends of the line, where M is the left end point of the line and N is the right end point of the line; The time when the initial wave head of the fault traveling wave reaches point M is recorded as The time when it reaches point N is recorded as , then: ；(6) ；(7) ；(8) In the formula Indicates the distance from point F to point M; 3) For the mixed lines of overhead lines and cable lines, normalization is performed, and the cable lines are equivalent to overhead lines of corresponding lengths. The wave velocities of traveling waves in overhead lines and cable lines are expressed as , It means that the calculation formula for the wave speed conversion of different lines is: ；(9) In the formula is the normalized equivalent overhead line length of the cable, Indicates the length of cables in the distribution network.

10. A system for locating a power transmission line fault based on an improved YOLOv8 model and a two-terminal TWM as claimed in claim 1 or 2 or 4 or 5 or 6 or 8 or 9, characterized in that: The system comprises: Traveling wave generation module: used to simulate and obtain the three-phase fault transient current signal at a sampling rate of 1MHz after a line fault occurs; Phase mode conversion module: used to convert the three-phase fault transient current signal into the corresponding line mode component through the Karenberg transformation matrix; Pix2PixHD model module: used to record the main body of the fault traveling wave at a sampling rate of 1MHz in the form of a picture, and generate a traveling wave generation picture at a sampling rate of 2MHz; Fault target detection module: used to process the traveling wave generated image at a sampling rate of 2MHz to obtain the traveling wave detection image with a bounding box, the probability of the fault target, and the time range of the traveling wave body; Traveling wave main body segmentation module: used to generate image processing of traveling waves at a sampling rate of 2MHz, the traveling wave main body segmentation map and the relative coordinates of the traveling wave maximum points; Line fault point calculation module: It is used to integrate the output information of the fault target detection module and the traveling wave main body segmentation module to obtain the time when the traveling wave reaches both ends of the line, that is, the time corresponding to the maximum point of the traveling wave, so as to calculate the accurate fault point position of the line.

Images