CN-122016881-A - Overhead insulated cable defect diagnosis method

CN122016881ACN 122016881 ACN122016881 ACN 122016881ACN-122016881-A

Abstract

The invention discloses a defect diagnosis method for an overhead insulated cable, and relates to the technical field of intelligent operation and maintenance and state monitoring of overhead power lines. The method comprises the steps of controlling a first unmanned aerial vehicle and a second unmanned aerial vehicle to cooperatively fly and accurately position, using an X-ray emitter and a detector to scan an aerial insulated cable to obtain an original X-ray image, debluring the original X-ray image through an image deblurring module to obtain a clear image, inputting the clear image into an improved YOLOX-Tiny target detection module to obtain a boundary frame of the cable, a state classification label of the cable and a direction label of the cable, and realizing defect diagnosis of the cable, and generating a course guiding instruction flying along the extending direction of the cable through the first unmanned aerial vehicle and the second unmanned aerial vehicle based on the direction label to control the unmanned aerial vehicle to fly along the extending direction of the cable until continuous scanning and diagnosis of the whole aerial insulated cable are completed. The invention can realize the internal defect diagnosis of the overhead insulated cable by an automatic means.

Inventors

Shen Zeyang
YU YING
LI XINRAN
ZHANG ZEHUA

Assignees

齐齐哈尔大学

Dates

Publication Date: 20260512
Application Date: 20260213

Claims (9)

1. A method for diagnosing defects of an overhead insulated cable, comprising: acquiring an original X-ray image of a target overhead insulated cable; Constructing a pre-trained improved YOLOX-Tiny target detection model, wherein the improved YOLOX-Tiny target detection model comprises a main network, a characteristic pyramid network and a decoupling head which are sequentially connected, and a direction classification branch network parallel to the decoupling head; The method comprises the steps of obtaining an original X-ray image, carrying out downsampling and feature extraction through a backbone network to obtain multi-scale features, merging the multi-scale features through a feature pyramid network to obtain three enhancement feature images with different scales, carrying out convolution on the enhancement feature images with each scale through a decoupling head to obtain bounding box information and state classification labels of a target overhead insulated cable, extracting the directional features of the enhancement feature images through a direction classification branch network, and carrying out feature merging and classification according to the directional features to obtain the direction labels of the cable; and diagnosing the defects of the target aerial insulated cable based on the boundary frame coordinates, the state classification labels and the direction labels of the cable.
2. The method for diagnosing defects of an overhead insulated cable according to claim 1, wherein the acquiring of the original X-ray image specifically comprises: Based on a real-time dynamic positioning system, carrying out cooperative flight and positioning control on a first unmanned aerial vehicle carrying an X-ray emitter and a second unmanned aerial vehicle carrying an X-ray detector, so that the two unmanned aerial vehicles are kept synchronous; And the X-ray emitter, the detected part of the target overhead insulated cable and the X-ray detector are positioned in the same straight line, so that the detected part of the target overhead insulated cable is subjected to X-ray scanning, and an original X-ray image is obtained.
3. A method of diagnosing an overhead insulated cable defect according to claim 1, wherein the original X-ray image is deblurred by generating an countermeasure network model to obtain a clear image suitable for a modified YOLOX-Tiny target detection model.
4. A method of diagnosing a fault in an insulated overhead cable according to claim 3, wherein said generating an countermeasure network model is comprised of a generator network G and a discriminator network D; The generator network G adopts a U-Net structure and is composed of an encoder and a decoder, wherein the encoder extracts characteristics through multi-layer convolution and downsampling, and the decoder reconstructs an image through multi-layer upsampling and convolution fusion characteristics; the discriminator network D adopts PatchGAN structure and is used for discriminating the authenticity of the input clear image.
5. A method for diagnosing a defect in an insulated overhead cable according to claim 3, wherein said training process for generating an countermeasure network model comprises: constructing an image dataset of a 'fuzzy-clear' pairing; training the generated countermeasure network model by configuring a composite loss function based on the image data set to obtain a trained countermeasure network model; The composite loss function includes a counter loss L adv , a pixel level L1 loss L L1 , and a perceived loss L perc of the generator network G; The counterloss L adv adopts Wasserstein distance with gradient penalty, and the expression is: ; Wherein, I sharp is a clear reference image, I blurry is an input blurred image; Lambda gp is a preset weight coefficient of a gradient penalty term, E [ ] represents a mathematical expected value of variables in brackets, G ([ lambda ]) represents forward propagation operation of the generator network G, D ([ lambda ]) represents forward propagation operation of the discriminator network D; Representation pair function Solving for input I, 2 denotes solving the vector in brackets for its L2 norm, i.e. euclidean length; The pixel level L1 loss L L1 has the expression: ; Wherein, || 1 represents the sum of the absolute values of all elements in the bracketed image, i.e., the L1 norm; the perceptual loss L perc has the expression: ; Wherein phi (,) represents a feature extraction function of a pre-trained deep convolutional neural network, the input of which is an image and the output of which is a feature map extracted at a specific network layer; All elements representing the difference between two feature graphs in a bracket are squared and then squared to obtain an L2 norm, and then square operation is carried out on the norm value; The composite loss function is formed by weighted summation of the counterloss L adv , the pixel level L1 loss L L1 and the perception loss L perc of the generator network G, and the expression is as follows: ; Wherein lambda L1 and lambda perc are preset weight coefficients.
6. The method for diagnosing defects of an overhead insulated cable according to claim 1, wherein the obtaining of the direction label of the cable specifically comprises: Based on three feature graphs with different scales output by a feature pyramid network, performing dimension reduction convolution on the feature graph with each scale through a 1X 1 convolution layer respectively, and uniformly reducing the number of channels to 256; Based on the feature map after dimension reduction, the feature characterization capability is enhanced by sequentially passing through two 3×3 convolution layers; mapping the enhanced features into a bottom layer representation for direction classification through a 1X 1 convolution layer to obtain three-scale feature graphs for direction classification representation; Feature fusion is carried out on the feature graphs of the three scales for direction classification representation, and a result containing 8 predefined direction category scores is obtained through classification; The result is converted to a probability distribution by a Softmax function and the highest probability class is used as the predicted cable direction label.
7. The method for diagnosing defects of an overhead insulated cable according to claim 1, wherein the improved YOLOX-Tiny target detection model has a total loss function in the training process, and specifically comprises the following steps: The original loss L yolox of the main network and the loss L direction of the direction classification branch form a total loss function L total through weighted summation, and the expression is as follows: ; Wherein, gamma is a preset weight coefficient for balancing the two losses; The direction classification loss L direction is calculated using a cross entropy loss function.
8. The method for diagnosing a defect in an insulated overhead cable according to claim 1, wherein said status classification labels include a health label, a fracture label, and a fusion label.
9. The method for diagnosing defects of an aerial insulated cable according to claim 1, wherein the direction tag is used for generating a guiding instruction of a flying heading of the unmanned aerial vehicle when the unmanned aerial vehicle scans the target aerial insulated cable.

Description

Overhead insulated cable defect diagnosis method Technical Field The application relates to the technical field of intelligent operation and maintenance and state monitoring of overhead power lines, in particular to a defect diagnosis method for overhead insulated cables. Background The overhead insulated cable is used as a key carrier of a power distribution network and is widely applied to power supply lines of cities, villages and industrial parks. The solar energy battery is exposed to outdoor severe environments for a long time, is not only subjected to natural aging such as sun exposure, rain exposure and wind vibration, but also directly bears the impact such as lightning stroke, overload and external force hanging collision. The insulating layer and the sheath of the cable are barriers for ensuring the electrical safety and the mechanical strength of the cable, and once the internal wire core or the connecting point is broken, melted or insulated and deteriorated due to lightning erosion, electrochemical corrosion, mechanical fatigue and the like, serious faults such as discharge, short circuit, even fire and wire breakage and the like are directly initiated, so that large-scale power failure is caused, and the reliability of power supply is greatly threatened. Therefore, the method has important engineering value and social and economic significance in regular, efficient and accurate internal state detection and defect diagnosis of the running overhead insulated cable. The unmanned aerial vehicle is carried with the visible light camera and carries out automatic line inspection with the thermal infrared imager and becomes the common practice. The infrared thermal imager can automatically identify obvious damage or foreign matters on the surface of the cable by utilizing the visible light image of the unmanned plane through a machine learning algorithm, and can locate defects such as overheat of a connecting point and the like by detecting abnormal temperature rise of a circuit. However, these mainstream technologies have a fundamental limitation in that they can only sense the external state of the line surface or indirectly inferred based on the temperature field, cannot penetrate the insulating sheath of the cable, and directly detect the physical defects of the internal conductor or early deterioration of the insulating layer thereof, which have extremely strong concealment but are core causes of faults. For overhead insulated cables, how to utilize the available inspection data to indirectly and accurately diagnose the potential internal defects of the overhead insulated cables through the innovation of the model and the method is still a difficulty and a blank in the current technical field. In summary, in the prior art, in the indirect characterization and identification of the cable internal defects with a complex background, small targets and weak features, the requirements of high-reliability diagnosis on the concealed internal defects cannot be met due to the problems of insufficient precision, high false alarm rate, weak generalization capability and the like. Disclosure of Invention In view of the foregoing, it is desirable to provide a method for diagnosing defects of an overhead insulated cable. The technical scheme adopted in the specification is as follows: the specification provides a fault diagnosis method for an overhead insulated cable, which comprises the following steps: acquiring an original X-ray image of a target overhead insulated cable; Constructing a pre-trained improved YOLOX-Tiny target detection model, wherein the improved YOLOX-Tiny target detection model comprises a main network, a characteristic pyramid network and a decoupling head which are sequentially connected, and a direction classification branch network parallel to the decoupling head; The method comprises the steps of obtaining an original X-ray image, carrying out downsampling and feature extraction through a backbone network to obtain multi-scale features, merging the multi-scale features through a feature pyramid network to obtain three enhancement feature images with different scales, carrying out convolution on the enhancement feature images with each scale through a decoupling head to obtain bounding box information and state classification labels of a target overhead insulated cable, extracting the directional features of the enhancement feature images through a direction classification branch network, and carrying out feature merging and classification according to the directional features to obtain the direction labels of the cable; and diagnosing the defects of the target aerial insulated cable based on the boundary frame coordinates, the state classification labels and the direction labels of the cable. Further, the acquiring of the original X-ray image specifically includes: Based on a real-time dynamic positioning system, carrying out cooperative flight and positioning control on a first unman