A method for generating maintenance strategies for distribution networks using a heuristic algorithm

Abstract

The present invention discloses a method for generating a maintenance strategy for a distribution network using a heuristic algorithm, including multi-dimensional health status indicators such as the aging degree of the insulator surface in the distribution network, the glaze layer crystal structure, geometric dimension deviation, surface cracks, and temperature distribution, evaluating the aging degree of the coating and determining whether to replace the coating or repair the glaze layer, dividing the maintenance area and determining the optimal maintenance path by analyzing the geographical location of the pole where the insulator is located, the surrounding vegetation coverage, and the electromagnetic interference level; during the ground inspection process, obtaining the three-dimensional point cloud data of the transmission line to determine whether there are damages or dirt on the insulator; measuring the width and length of the surface cracks, predicting the crack propagation trend and remaining life, formulating the maintenance time window and scheduling maintenance resources; after the insulator replacement or repair is completed, detecting its temperature distribution to judge the installation quality; finally, generating the optimal maintenance decision sequence for the entire life cycle of the insulator through multi-dimensional data analysis.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for generating a maintenance strategy for a distribution network using a heuristic algorithm. Background Art

[0002] There is a technical contradiction in the maintenance process of distribution network insulators, that is, how to minimize the negative impact of the maintenance process on the environment and human health while ensuring the maintenance quality and efficiency. In order to accurately identify the overheating points, potential fault locations, and corona discharge conditions of insulators, it is necessary to use high-precision temperature sensors and corona discharge detection equipment to monitor the status of insulators in real time, and conduct a fine-grained analysis and evaluation of the maintenance requirements of each insulator based on the monitoring data. When implementing the maintenance strategy of power outage maintenance, disassembling old insulators, and replacing new insulators, it will inevitably generate noise and electromagnetic radiation emissions, which will have a certain impact on the surrounding environment and human health. When formulating a maintenance plan, it is also necessary to comprehensively consider the dependency relationships between various maintenance strategies, the urgency of maintenance tasks, and the availability factors of maintenance resources, and reasonably arrange the execution order of maintenance tasks and resource allocation to achieve a balance between maintenance efficiency and cost. Summary of the Invention

[0003] In order to solve the above-mentioned existing technical problems, the present invention provides a method for generating a maintenance strategy for a distribution network using a heuristic algorithm.

[0004] The technical solution of the present invention is realized as follows: A method for generating a maintenance strategy for a distribution network using a heuristic algorithm, including:

[0005] Analyze the aging degree of the coating on the surface of the insulator in the distribution network, including obtaining the chemical bond vibration characteristic data of the coating material by using Fourier transform infrared spectroscopy analysis technology, constructing an evaluation model of the aging degree, judging whether the aging degree of the insulator is greater than a threshold value, if so, identifying that the coating needs to be replaced, entering the insulator number and position information into the fault maintenance database, and determining the maintenance priority according to the aging degree of the coating;

[0006] Use an X-ray diffractometer to analyze the phase composition of the glaze layer of the insulator, obtain the diffraction spectrum data of the crystal structure of the glaze layer, perform feature extraction and classification on the diffraction spectrum data through a convolutional neural network model, judge whether the glaze layer is distorted, if it is identified as distorted, enter the insulator number and position information into the fault maintenance database, and determine the glaze layer repair or replacement plan according to the type of distortion;

[0007] For each insulator to be repaired in the fault repair database, by analyzing the geographical location and electromagnetic interference level of the tower where it is located, the fuzzy C-means clustering module divides the insulators to be repaired into multiple repair areas, determines the optimal repair path for each repair area, sends the optimal repair path to the terminal of the repair personnel, and real-time locates and tracks the repair personnel through the position perception module to ensure repair along the optimal repair path;

[0008] During the repair process, the photogrammetry module is used to obtain the image data of the transmission line, extract the geometric model of the insulator, and judge whether there are defects in the insulator, including breakage or contamination. If there are defects, the defect type and position coordinates are uploaded to the fault repair database, and the local cleaning or replacement process is triggered;

[0009] For the insulators with defects, measure the width and length of the cracks on the surface of the insulators, determine the crack propagation path through the ant colony search module, predict the crack propagation trend and the remaining life of the insulators. If the remaining life is lower than the threshold, the insulator number and position information are entered into the fault repair database, and the repair time window is determined according to the crack propagation rate. The heuristic rule module is used to schedule the repair resources to minimize the risk brought by crack propagation;

[0010] After the insulator is replaced or repaired, the temperature distribution on the surface of the insulator is detected through the thermal image data on the surface of the insulator, the frequency domain characteristics of the thermal image data are extracted, and the convolutional autoencoder network is used to judge the current quality of the insulator. If there is an abnormal temperature distribution, the thermal image data is uploaded to the fault repair database, and the repair process is triggered to dynamically adjust the repair strategy;

[0011] Based on the multi-dimensional data of the insulator, including coating aging degree, glaze distortion, surface cracks, and temperature anomalies, the comprehensive health index of the insulator is obtained, and the Markov chain prediction model is used to estimate the state transition probability under different repair strategies. The multi-objective particle swarm optimization module is used to generate the optimal repair decision sequence for the whole life cycle of the insulator. Beneficial effects

[0012] A method for generating a maintenance strategy for a distribution network using a heuristic algorithm provided by the present invention can non-destructively evaluate the aging degree of insulator coatings and glaze layer distortion through Fourier transform infrared spectroscopy analysis and X-ray diffraction technology, realizing early identification of potential problems and reducing power outages caused by insulator failures; the application of fuzzy C-means clustering and location awareness technology can optimize the maintenance path planning, ensure the effective allocation of maintenance resources, and at the same time monitor the activities of maintenance personnel in real time, improving operation efficiency and safety; combining photogrammetry and deep learning technology to automatically detect insulator defects, reducing omissions in manual inspections, shortening response times, and the use of ant colony search and multi-objective optimization models further enhances the intelligent level of dealing with complex cracks, ensuring the scientificity and economy of maintenance decisions; through the simulation of crack propagation and the prediction of remaining life, combined with the Markov chain model, it realizes the transformation from passive fault response to active preventive maintenance, effectively extending the service life of insulators and reducing maintenance costs; the application of multi-objective particle swarm optimization technology comprehensively considers various factors, formulates the optimal maintenance decision sequence, ensures the optimal execution of maintenance work under limited resources, and at the same time reduces the system risk brought by insulator failures; integrating multi-dimensional data of insulators to construct a comprehensive health index, providing a full-cycle health management plan for each insulator from installation to retirement, and enhancing the comprehensive management level of power grid assets. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a structural block diagram of a method for generating a maintenance strategy for a distribution network using a heuristic algorithm in an embodiment of the present invention;

[0014] Figure 2 It is a step block diagram of a method for generating a maintenance strategy for a distribution network using a heuristic algorithm in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] In order to make the objectives and advantages of the present invention clearer, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0016] The preferred implementation methods of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these implementation methods are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0017] It should be noted that when an element is considered to be "connected" to another element, it can be directly connected to the other element or connected to the other element through an intermediate element. In addition, "connection" in the following embodiments should be understood as "electrical connection", "communication connection", etc. if there is an electrical signal or data transmission between the connected objects.

[0018] As used herein, the singular forms "a", "an" and "the" may also include the plural forms unless the context clearly dictates otherwise. It should also be understood that the terms "comprises / comprising", "has / including", etc. specify the presence of the stated features, integers, steps, operations, components, parts, or combinations thereof, but do not preclude the presence or addition of one or more other features, integers, steps, operations, components, parts, or combinations thereof. At the same time, the term "and / or" used in this specification includes any and all combinations of the related listed items.

[0019] Please refer to Figure 1-2 As shown, in this embodiment, a method for generating a maintenance strategy for a distribution network using a heuristic algorithm specifically includes:

[0020] In step S101, the aging degree of the coating on the surface of the insulator in the distribution network is analyzed, including obtaining the chemical bond vibration characteristic data of the coating material by using Fourier transform infrared spectroscopy analysis technology, constructing an evaluation model for the aging degree, judging whether the aging degree of the insulator is greater than a threshold value. If so, it is identified that the coating needs to be replaced, the number and position information of the insulator are entered into the fault maintenance database, and the maintenance priority is determined according to the aging degree of the coating.

[0021] Specifically, Fourier transform infrared spectroscopy analysis technology is used to analyze the chemical composition of the coating material on the surface of the insulator, and the chemical bond vibration characteristic data of the coating material are obtained. A small amount of sample is taken from the coating on the surface of the insulator and ground into powder with a mortar. The powder sample is pressed into a tablet and put into an FTIR instrument for testing. The test data are preprocessed by baseline correction and smoothing using FTIR spectrum analysis software, and then compared with the data in the standard spectrum library to calculate the similarity. The similarity is measured by Euclidean distance and correlation coefficient indicators. According to the size of the similarity, the aging degree of the coating material is judged. The smaller the similarity, the greater the change in the chemical composition of the material and the higher the aging degree. A threshold value is set. When the similarity is lower than this threshold value, it is considered that the coating has been severely aged and needs to be replaced. By analyzing the chemical composition and aging characteristics of different materials, a performance prediction model for the coating material is established, so as to select the optimal coating material, construct an evaluation model for the aging degree of the coating based on a convolutional neural network, splice the chemical bond vibration characteristics obtained by FTIR analysis with the image characteristics of the insulator coating surface to form a complete sample feature input, use the aging degree category as the output, and use the cross-entropy loss function and Adam optimizer for training. A large number of insulator coating surface images and corresponding aging degree annotation data are collected and divided into a training set, a validation set and a test set. The pre-trained CNN model is used to extract features from the images to obtain high-dimensional feature vectors;

[0022] Build a CNN model, continuously optimize the model performance by adjusting hyperparameters, evaluate the precision, recall rate, and F1 value metrics of the model on the validation set, conduct error analysis to identify the deficiencies of the model, adopt the early stopping method to prevent overfitting, apply the trained model to the test set, output the predicted results of the aging degree of each sample, compare with the manually labeled results, calculate the accuracy of the model, set the threshold for the aging degree of the coating, compare the aging degree output by the evaluation model with the threshold. If the aging degree exceeds the threshold, enter the number and location information of the insulators that need to replace the coating into the fault repair database, determine the maintenance priority according to the aging degree of the coating, and the higher the aging degree of the insulator, the higher the priority. Adopt the earliest deadline first (EDF) algorithm based on deadline to sort and schedule the priorities of maintenance tasks. Calculate the priority of each maintenance task according to the aging degree of the coating, set a mapping function to map the aging degree to a priority value, calculate its deadline according to the priority and duration of the maintenance task, deadline = current time + duration * (1 + priority factor), and the priority factor is set according to the specific situation, such as taking 0.2. Sort all maintenance tasks in ascending order according to the deadline, and sequentially assign the maintenance tasks to the maintenance personnel in the sorted order. After the maintenance personnel complete one task, they request to assign the next task until all tasks are completed;

[0023] After determining the maintenance priority, use robotics to automatically complete the insulator coating replacement operation. Achieve precise cleaning and re-spraying of the coating through vision guidance and force control technology, configure the new coating material through the automatic proportioning system of the coating material, inspect the quality of the new coating through the quality inspection system, and establish an insulator full-life cycle management system to record the whole-process data of the insulator from installation, operation, maintenance to scrapping.

[0024] In one embodiment, when performing chemical composition analysis on the surface coating material of an insulator, a Fourier transform infrared spectroscopy instrument Nicoleti S50 can be used. The scanning wavenumber range is set to 400 - 4000 cm-1, the resolution is 4 cm-1, and the number of scans is 32 times. The OMNIC software is used to process the test data. By comparing with the built-in polymer additive spectral library, the similarity index is calculated. When the similarity is lower than 0.8, it is determined that the coating is severely aged and needs to be replaced. At the same time, a performance prediction model for the coating material is established using the principal component analysis (PCA) and support vector machine (SVM) algorithms. By analyzing the infrared spectral characteristics and aging test data of different materials, the epoxy silicone coating material with the best comprehensive performance is selected. When constructing the convolutional neural network model, ResNet-50 is used as the backbone network and pre-trained on the ImageNet dataset, and then transfer learning is performed on the coating surface image dataset. By setting the learning rate to 0.001, the batch size to 32, and training for 100 epochs, an accuracy of 98% is achieved on the validation set. The model is deployed to the embedded device Jetson Xavier NX to monitor the coating aging situation in real time. When the aging degree exceeds 0.7, a maintenance task is automatically triggered, and the task information is written into the MySQL database. When scheduling tasks, according to the coating aging rate and the estimated remaining life, the priority and deadline of the maintenance task are calculated, and the improved EDF algorithm is used for task sorting and allocation. Through the dynamic priority adjustment mechanism, the reasonable configuration of maintenance resources is realized. For the coating replacement operation, an ABB IRB1200 robot is used, equipped with a high-precision laser scanner and a spraying end effector. Automatic positioning and trajectory planning are realized through machine vision guidance, and the closed-loop control algorithm is used to achieve precise control of the coating thickness and uniformity, with a spraying accuracy of ±0.1 mm. In the insulator full-life cycle management system, distributed data acquisition and edge computing technologies are adopted to realize real-time monitoring of the insulator status and fault warning.

[0025] In step S102, an X-ray diffractometer is used to analyze the phase composition of the glaze layer of the insulator, and diffraction spectrum data on the crystal structure of the glaze layer is obtained. The convolutional neural network model is used to extract features and classify the diffraction spectrum data to determine whether the glaze layer is distorted. If it is identified as distorted, the insulator number and location information are entered into the fault repair database, and the glaze layer repair or replacement plan is determined according to the type of distortion.

[0026] Specifically, an X-ray diffractometer is used to analyze the phase composition of the glaze layer of the insulator to obtain the diffraction spectrum data of the crystal structure of the glaze layer. The setting parameters of the X-ray diffractometer include the X-ray wavelength, the incident angle, the scanning step size, and the scanning speed. Through data preprocessing and feature engineering, the key feature parameters of the diffraction spectrum are extracted, such as the crystal plane spacing corresponding to the diffraction peak position, the crystal plane orientation corresponding to the peak intensity, the grain size corresponding to the peak width, and the amorphous phase content corresponding to the background intensity. A convolutional neural network model is used to perform deep learning and classification on the extracted features. By training the diffraction spectrum samples of the normal glaze layer, a multi-class distortion recognition model is established to judge whether the glaze layer has undergone distortion. The convolutional neural network model adopts the LeNet-5 or AlexNet structure. The input layer is the diffraction spectrum image. The convolutional layer and the pooling layer are used to extract the local features of the image. The fully connected layer is used for feature integration and classification judgment. For the main distortion types of the glaze layer, such as mullite crystal phase, quartz crystal phase, crystallization, and residual stress, corresponding labels are set for multi-class training. The hyperparameters of the model, including the convolutional kernel size, the number of convolutional layers, and the number of nodes, are optimized by the gridsearch method. When it is recognized that the glaze layer has undergone distortion, the insulator number and position information are automatically obtained and entered into the fault repair database. At the same time, according to the type and severity of the distortion, a scheme for glaze layer repair or replacement is intelligently generated. The Apriori algorithm for association rule mining, the decision tree ID3, C4.5 Algorithm data mining method, analyze historical maintenance data, find out the corresponding relationship between different distortion types and maintenance measures, form a rule base in the form of IF-THEN. When generating a repair and replacement plan, through rule matching and reasoning, find the historical case most similar to the current distortion type and associate its maintenance measures. For the glaze repair plan, adopt image segmentation algorithms including threshold segmentation, edge detection, and region growing to automatically detect and locate the surface defects of the glaze, separate the defect area from the normal area. For the segmented defect area, extract its geometric features such as shape, size, and direction. Through decision tree or support vector machine classification algorithms, judge the types of defects including cracks, bubbles, and spalling. According to the defect type and features, automatically generate the process parameters of laser cladding or plasma spraying including power, speed, and powder feeding rate, and use a robot to precisely cladding or spray the defect area. For the glaze replacement plan, adopt big data analysis technology, comprehensively consider the operating conditions and environmental condition factors of the insulator, optimize the best glaze material formula, and use an intelligent glaze configuration system to automatically batch, mix, and prepare the glaze according to the optimized formula to ensure the consistency of glaze performance. Adopt laser cleaning and water jet cleaning methods to remove the old glaze layer, and then use robot automatic spraying technology to coat a new glaze layer on the insulator. Through computer control of the movement trajectory and spraying parameters of the spraying gun, achieve precise control of the glaze thickness and uniformity. The sprayed insulator needs to be sintered in a tunnel kiln. Through the control of temperature and atmosphere parameters, make the glaze fully combine with the porcelain body. During the glaze repair and replacement process, adopt on-line monitoring and quality inspection technology to real-time detect the key indicators such as the thickness, uniformity, and insulation resistance of the glaze. Once a deviation is found, immediately optimize and adjust the process parameters to ensure that the repair and replacement quality meets the standard requirements. Establish an on-line monitoring system for insulators based on the Internet of Things. By installing temperature, vibration, and strain wireless sensors on the insulators, real-time collect the state parameters of the glaze. Through edge computing nodes, real-time analyze the monitoring data, establish a correlation model between temperature, vibration, strain and glaze distortion. When the monitoring parameters exceed the normal range or show abnormal fluctuations, timely warn of the risk of glaze distortion and trigger a maintenance task.

[0027] In one embodiment, the setting parameters of the X-ray diffractometer include an X-ray wavelength of CuKα, λ = 0.15406 nm, an incident angle of 10° - 90°, a scanning step of 0.02°, and a scanning speed of 5° / min. Relevant information of the insulator to be repaired is extracted from the fault repair database, including the geographical location coordinates of the pole tower where it is located, such as a longitude of 1123 degrees, a latitude of 312 degrees, and an elevation of 25 meters, as well as the power frequency electric field intensity of the surrounding environment being 35 V / m, the radio electric field intensity being 2 V / m, and the magnetic field intensity being 8 A / m. The spatial visualization analysis of the pole tower position is carried out using the ArcGIS system, and the distance matrix between the pole towers is calculated through the Geodesic distance calculation function. For example, the distance between pole tower A and pole tower B is 153 meters. The fuzzy C-means clustering algorithm is used to divide the repair area of the insulator to be repaired. The geographical location of the pole tower and the electromagnetic interference intensity are selected as clustering features, and the Mahalanobis distance is used to calculate the similarity between the insulators. The Mahalanobis distance takes into account the correlation between the features and can better reflect the true differences between the insulators. The Gaussian membership function is used, and the membership value ranges from 0 to 1. The larger the membership value, the more likely the insulator belongs to this category. The membership matrix is calculated based on the Mahalanobis distance between the insulator and the clustering center. The belonging area of each insulator is determined through the maximum membership principle, and the central coordinates of each area are calculated. The FHI index is used to evaluate the goodness of the clustering result. The FHI index comprehensively considers the clustering compactness and the clustering separation degree. The larger the value, the better the clustering effect. The results of different clustering numbers are compared through the FHI index. When the clustering number is 5, the FHI index reaches the maximum value of 92. Therefore, the insulators to be repaired are divided into 5 repair areas. In each repair area, the improved ant colony optimization algorithm is used to search for the best repair path. The repair area is abstracted as an undirected weighted graph, with the insulators as nodes and the distance between the insulators as the edge weights. At the same time, the electromagnetic interference intensity between the insulators is used as heuristic information. The greater the interference intensity, the greater the edge weight value, which attracts the ants to search. The global pheromone and local pheromone are introduced. The initial value of the global pheromone concentration is set to 2, the initial value of the local pheromone concentration is set to 5, and the pheromone evaporation coefficient is set to 1.The local pheromone attenuation coefficient is set to 7. An evaluation index for the diversity of the ant population is introduced, and the Simpson index is used to quantify the population diversity. When the Simpson index is lower than 6, 20% of the ant individuals are regenerated through mutation operations. After the ant colony algorithm iterates 500 times, it converges to the global optimal solution, obtaining the best inspection path for each inspection area. The best inspection path is converted into JSON format data, including inspection area ID, longitude and latitude of path nodes, distance between nodes, and estimated time-consuming attributes. The JSON data is pushed to the / api / inspection / tasks interface of the mobile terminal APP through an HTTP POST request, and different inspection areas and paths are marked with different colors on the Amap API. Maintenance personnel can query task details through the voice interaction function, such as "Query today's maintenance tasks" and "Navigate to the next inspection point". The mobile terminal APP integrates a position perception module, using GPS / Beidou dual-mode positioning, combined with the Chinese Precision CORS system for RTK differential positioning. The horizontal positioning accuracy is better than 5 cm, and the vertical positioning accuracy is better than 10 cm. At the same time, the extended Kalman filter algorithm is used to fuse GPS positioning data and mobile phone IMU inertial navigation data, and the position update frequency reaches 20 Hz. The position perception module captures on-site images through the mobile phone camera, and uses the YOLOv5s object detection algorithm based on MobileNet-SSD to perform real-time detection and positioning of insulators. The average detection accuracy mAP reaches 95%, and the detection speed reaches 25 FPS. The front image of the insulator is extracted through affine transformation and perspective transformation and sent into the Tesseract 0 OCR model based on CNN for nameplate information recognition. The average recognition accuracy reaches 95%. Finally, the recognition result is compared with the data in the asset management system to verify the correctness of the maintenance object. The system matches the trajectory data with the best inspection path, and uses the dynamic time warping (DTW) algorithm to calculate the trajectory deviation. Once the deviation exceeds 10 meters, a warning prompt is sent to the maintenance personnel in a timely manner to prompt the maintenance personnel to adjust the inspection path until all maintenance tasks are completed. The verification result will also be fed back to the maintenance personnel in real time to avoid maintenance omissions and errors. When performing X-ray diffraction analysis on the glaze layer of the insulator, a Bruker D8 Advance diffractometer is used, with the wavelength of the CuKα radiation source set to 0.15406 nm, the scanning range of 10° - 90°, the scanning step size of 0.02°, and the scanning speed of 5° / min. The JADE 6.5 software is used to perform background subtraction, peak search, and crystal phase calibration on the diffraction spectrum. By comparing with the ICDD PDF-2 database, the phase composition of the glaze layer is determined. The Scherrer formula is used to calculate the full width at half maximum of the diffraction peak, and then the grain size of the glaze layer is estimated. The diffraction spectrum data is converted into a grayscale image of 1024×1024 pixels and sent into the pre-trained AlexNet network for feature extraction and classification judgment. The input layer of the network is 227×227 pixels,The convolution kernel sizes of the 5 convolutional layers are successively 11×11×3, 5×5×64, 3×3×192, 3×3×384, and 3×3×256. The number of nodes in the 3 fully connected layers are successively 4096, 4096, and 8, corresponding to 8 typical types of glaze layer distortion. The Adam optimizer and cross-entropy loss function are used to train the network. The batch size is 128, the initial learning rate is 0.001, the momentum factor is 0.9, and the regularization coefficient is 0.0005. When the accuracy on the validation set has not improved for 5 consecutive epochs, the early stopping mechanism is triggered, and the current optimal model parameters are saved. The SEM image of the glaze layer surface is subjected to Ostu threshold segmentation, and the morphological features of the defect area, such as area, perimeter, roundness, and aspect ratio, are extracted. The CART decision tree algorithm is used to perform five-classification of the defect types, including cracks, bubbles, spalling, pinholes, and impurities, and the processing parameters of laser cladding, such as laser power of 300W, spot diameter of 0.5mm, scanning speed of 800mm / min, and powder feeding rate of 12g / min, are automatically generated according to the defect size and quantity. The AdaBoost algorithm is used to evaluate the comprehensive performance of different material formulations, and the optimal glaze composition is automatically selected to achieve the intelligent optimization of the glaze formulation. The time-domain signal collected by the on-line monitoring system is transformed to the frequency domain through fast Fourier transform, and the characteristic parameters of the spectrum, such as root mean square value, kurtosis, and skewness, are extracted. A wavelet neural network is used to establish the mapping relationship between the characteristic parameters and the damage degree of the glaze layer to realize the real-time evaluation of the health state of the glaze layer and the prediction of the remaining life.,

[0028] S103. For each insulator to be repaired in the fault repair database, by analyzing the geographical location and electromagnetic interference level of the pole tower where it is located, it is divided into multiple repair areas on the insulator to be repaired through the fuzzy C-means clustering module, and the best repair path for each repair area is determined. The best repair path is sent to the terminal of the repair personnel, and the repair personnel are real-time located and trajectory tracked through the position perception module to ensure that the repair is carried out along the best repair path.

[0029] Specifically, relevant information of the insulators to be repaired is extracted from the fault repair database, including the geographical location coordinates of the poles and towers where they are located, including longitude, latitude, and elevation, and the electromagnetic interference intensity of the surrounding environment, including power frequency electric field, radio electric field, and magnetic field. The GIS system is used to perform spatial visualization analysis on the positions of the poles and towers, calculate the distance matrix between the poles and towers, which is used as the input for fuzzy C-means clustering. The fuzzy C-means clustering algorithm is used to divide the repair areas of the insulators to be repaired. The geographical location of the poles and towers and the electromagnetic interference intensity are selected as clustering features. The Euclidean distance or Mahalanobis distance is used to calculate the similarity between the insulators. The exponential membership function is used to calculate the membership matrix according to the distance between the insulators and the cluster centers. The belonging area of each insulator is determined through the maximum membership principle, and the central coordinates of each area are calculated. The Xie-Beni index or FHI fuzzy hypersphere index is used to evaluate the goodness of the clustering results, and the optimal number of clusters is selected. In each repair area, the improved ant colony optimization algorithm is used to search for the best repair path. The repair area is abstracted as an undirected weighted graph, with the insulators as nodes and the distances between the insulators as edge weights. At the same time, the electromagnetic interference intensity between the insulators is used as heuristic information. The greater the interference intensity, the greater the edge weight value, which attracts ants to search. The global pheromone and local pheromone are introduced to enhance the global search ability and local development ability of the algorithm. An ant population diversity evaluation index is introduced. When the population tends to be homogenized, a part of the ant individuals are regenerated through mutation operations. The ant colony algorithm continuously optimizes the search path through the pheromone update mechanism until it converges to the global optimal solution. The best repair path is converted into JSON format data, including the repair area ID, the longitude and latitude of the path nodes, the distance between the nodes, and the estimated time-consuming attributes. The JSON data is pushed to the specified API interface of the mobile terminal APP through an HTTP POST request, and different repair areas and paths are marked with different colors on the electronic map. The repair personnel can query the task details through the voice interaction function and perform on-site operations according to the navigation prompts. The mobile terminal APP integrates a position perception module, which uses GPS / Beidou dual-mode positioning and combines with base station differential positioning (RTK) technology to achieve centimeter-level high-precision positioning. At the same time, the Kalman filter algorithm is used to fuse the GPS positioning data and inertial navigation data to achieve smooth trajectory tracking. The position perception module captures the on-site images through the mobile phone camera, uses the YOLOv5 object detection algorithm based on deep learning to perform real-time detection and positioning of the insulators, extracts the front image of the insulators through affine transformation and perspective transformation, sends it to the OCR model for nameplate information recognition, and finally compares the recognition result with the data in the asset management system to verify the correctness of the repair object. The system matches the trajectory data with the best repair path, calculates the trajectory deviation degree. Once the deviation degree exceeds the threshold, a warning prompt is sent to the repair personnel in a timely manner until all repair tasks are completed. The verification result will also be fed back to the repair personnel in real time to avoid repair omissions and errors.

[0030] In one embodiment, relevant information of the insulator to be inspected and repaired is extracted from the fault repair database, including the geographical location coordinates of the pole tower where it is located, such as longitude of 1123 degrees, latitude of 312 degrees, and elevation of 25 meters, as well as the power frequency electric field intensity of the surrounding environment of 35 V / m, the radio electric field intensity of 2 V / m, and the magnetic field intensity of 8 A / m. The spatial visualization analysis of the pole tower position is carried out using the ArcGIS system, and the distance matrix between the pole towers is calculated through the Geodesic distance calculation function. For example, the distance between pole tower A and pole tower B is 153 meters. The fuzzy C-means clustering algorithm is used to divide the inspection area of the insulator to be inspected and repaired. The geographical location of the pole tower and the electromagnetic interference intensity are selected as the clustering features. The Mahalanobis distance is used to calculate the similarity between the insulators. The Mahalanobis distance takes into account the correlation between the features and can better reflect the true differences between the insulators. The Gaussian membership function is adopted, and the membership value ranges from 0 to 1. The larger the membership value, the more likely the insulator belongs to this category. The membership matrix is calculated according to the Mahalanobis distance between the insulator and the cluster center. The belonging area of each insulator is determined by the maximum membership principle, and the central coordinates of each area are calculated. The FHI index is used to evaluate the goodness of the clustering result. The FHI index comprehensively considers the clustering compactness and the clustering separation degree. The larger the value, the better the clustering effect. The results of different clustering numbers are compared through the FHI index. When the clustering number is 5, the FHI index reaches the maximum value of 92. Therefore, the insulators to be inspected and repaired are divided into 5 inspection areas. In each inspection area, the improved ant colony optimization algorithm is used to search for the best inspection path. The inspection area is abstracted as an undirected weighted graph, with the insulators as nodes and the distance between the insulators as the edge weights. At the same time, the electromagnetic interference intensity between the insulators is used as heuristic information. The greater the interference intensity, the greater the edge weight value, attracting ants to search;

[0031] Global pheromone and local pheromone are introduced. The initial value of the global pheromone concentration is set to 2, the initial value of the local pheromone concentration is set to 5, the pheromone evaporation coefficient is set to 1, and the local pheromone attenuation coefficient is set to 7. An evaluation index for the diversity of the ant population is introduced, and the Simpson index is used to quantify the population diversity. When the Simpson index is lower than 6, 20% of the ant individuals are regenerated through mutation operations. After 500 iterations of the ant colony algorithm, it converges to the global optimal solution, and the best inspection path for each inspection area is obtained. The best inspection path is converted into JSON format data, including the inspection area ID, the longitude and latitude of the path nodes, the distance between nodes, and the estimated time-consuming attributes. The JSON data is pushed to the / api / inspection / tasks interface of the mobile terminal APP through an HTTP POST request, and different inspection areas and paths are marked with different colors on the Amap API. The maintenance personnel can query the task details through the voice interaction function, such as "Query today's inspection tasks" and "Navigate to the next inspection point". The mobile terminal APP integrates a position perception module, which uses GPS / Beidou dual-mode positioning and combines the Chinese precision CORS system for RTK differential positioning. The horizontal positioning accuracy is better than 5 cm, and the vertical positioning accuracy is better than 10 cm. At the same time, the extended Kalman filter algorithm is used to fuse the GPS positioning data and the mobile phone IMU inertial navigation data, and the position update frequency reaches 20 Hz. The position perception module collects the on-site images through the mobile phone camera, and uses the YOLOv5s object detection algorithm based on MobileNet-SSD to perform real-time detection and positioning of the insulators. The average detection accuracy mAP reaches 95%, and the detection speed reaches 25 FPS. The front image of the insulator is extracted through affine transformation and perspective transformation and sent to the Tesseract OCR model based on CNN for nameplate information recognition. The average recognition accuracy reaches 95%. Finally, the recognition result is compared with the data in the asset management system to verify the correctness of the inspection object. The system matches the trajectory data with the best inspection path, and uses the dynamic time warping (DTW) algorithm to calculate the trajectory deviation. Once the deviation exceeds 10 meters, a warning prompt is sent to the maintenance personnel in a timely manner to prompt the maintenance personnel to adjust the inspection path until all inspection tasks are completed. The verification result will also be fed back to the maintenance personnel in real time to avoid inspection omissions and errors.

[0032] In the S104 step during the inspection process, a photogrammetry module is used to obtain the image data of the transmission line, extract the geometric model of the insulator, and determine whether there are defects in the insulator, including breakage or fouling. If there are defects, the defect type and position coordinates are uploaded to the fault repair database, and the local cleaning or replacement process is triggered.

[0033] Specifically, according to the transmission line image data obtained by the photogrammetry module, an image segmentation algorithm is used to extract the insulator region. Through 3D reconstruction technology, a 3D geometric model of the insulator is constructed based on the insulator region image. A deep learning algorithm is used to judge the defects of the insulator geometric model, identify whether there are damages or contaminations on the insulator. If it is judged that the insulator has damages, an edge detection algorithm is used to determine the position coordinates of the damaged area. If it is judged that the insulator has contaminations, a color segmentation algorithm is used to determine the position coordinates of the contaminated area. According to the defect judgment result, the defect type is obtained, and the defect type and position coordinate information are uploaded to the fault maintenance database. According to the defect information in the fault maintenance database, it is judged whether it is necessary to trigger the local cleaning or replacement process. If local cleaning is required, according to the position coordinates of the contaminated area, robotic automatic cleaning technology is used for positioning and cleaning. If insulator replacement is required, according to the position coordinates of the damaged area, robotic automatic replacement technology is used for positioning and replacement.

[0034] In one embodiment, first, a high-definition camera is carried by a drone to photograph the transmission line to obtain a high-definition image with a resolution of 4000×3000. Then, a region-based watershed segmentation algorithm is used to segment the image to extract the insulator region, and the segmentation accuracy can reach the pixel level. Multi-view images are used to perform 3D reconstruction on the insulator region, and a 3D geometric model of the insulator is obtained by constructing a triangular mesh model, and the reconstruction accuracy can reach the millimeter level. Then, a deep learning algorithm based on a convolutional neural network is used to judge the defects of the insulator geometric model. Through the training sample data set, it is identified whether there are damages or contaminations on the insulator, and the judgment accuracy rate can reach more than 95%. If it is judged that the insulator has damages, the Canny edge detection algorithm is used to determine the position coordinates of the damaged area, and the positioning accuracy can reach the centimeter level. If it is judged that the insulator has contaminations, a color segmentation algorithm based on the HSV color space is used to determine the position coordinates of the contaminated area, and the segmentation accuracy can reach the pixel level. According to the defect judgment result, the defect type and position coordinate information are uploaded to the fault maintenance database. Big data analysis technology is used to mine and analyze the historical maintenance data to judge whether it is necessary to trigger the local cleaning or replacement process. If local cleaning is required, according to the position coordinates of the contaminated area, robotic automatic cleaning technology is used for positioning and cleaning, and the cleaning efficiency can be increased by more than 50%. If insulator replacement is required, according to the position coordinates of the damaged area, robotic automatic replacement technology is used for positioning and replacement, and the replacement accuracy reaches the millimeter level.

[0035] In step S105, for the defective insulators, measure the width and length of the cracks on the surface of the insulators, determine the crack propagation path through the ant colony search module, predict the crack propagation trend and the remaining life of the insulators. If the remaining life is lower than the threshold, record the insulator number and location information in the fault maintenance database, and determine the maintenance time window according to the crack propagation rate, and schedule the maintenance resources through the heuristic rule module to minimize the risk brought by crack propagation.

[0036] Specifically, a three-dimensional laser scanner is used to scan the surface of the defective insulator. The scanned point cloud data is denoised, meshed, and surface reconstructed to obtain a three-dimensional model of the insulator surface with an accuracy of 0.1 mm. Then, through the geometric morphology analysis module, crack feature extraction is performed on the three-dimensional model to obtain the geometric parameters of the crack, such as length, width, and depth. The geometric parameters of the crack are input into the ant colony search module. This module uses the ant colony optimization algorithm (ACO) based on graph theory to divide the insulator surface into an N×N grid. Each grid corresponds to a node, and there are connections between adjacent grids. The connection weight is proportional to the crack propagation probability. Randomly release m ants, and each ant moves between nodes according to the pheromone concentration and heuristic factor according to the probability transition rule. After completing one cycle, update the pheromone, and repeat the iteration K times until the global pheromone concentration converges to obtain the path with the maximum crack propagation probability. According to the crack propagation probability distribution map, combined with the material properties, load spectrum, environmental temperature and humidity factors of the insulator, use the Paris formula based on fracture mechanics to predict the crack propagation trend in the future. The Paris formula describes the relationship between the crack propagation rate da / dN and the stress intensity factor range ΔK: da / dN = C(ΔK)^m, where C and m are constants related to the material, obtained by fitting through crack propagation experiments. The stress intensity factor range ΔK is related to the load, crack size, and specimen geometric size, and is obtained by finite element analysis. Discretize the Paris formula and recursively calculate the crack size: a_(i + 1)=a_i + C(ΔK_i)^m×ΔN_i, where ΔN_i is the number of cycles of the i-th cyclic loading, determined according to the set time interval and cyclic frequency. Input the crack propagation trend data into the remaining life prediction model. This model uses the Weibull distribution to describe the fatigue life distribution of the insulator, and its probability density function is: f(t)=(m / η)(t / η)^(m - 1)exp, where t is the fatigue life, m is the shape parameter, and η is the scale parameter, obtained by fitting from experimental data. Based on the critical crack size a_c predicted by the crack propagation model, take the α quantile t_α of the Weibull distribution as the remaining life prediction value at the α confidence level. Set the threshold of the insulator remaining life, and compare the predicted remaining life with the threshold. If the remaining life is lower than the threshold, automatically enter the number and location information of this insulator into the fault maintenance database, and determine the latest time window for maintenance according to the crack propagation model. Input the insulator information and maintenance time window in the fault maintenance database into the heuristic rule module. This module uses the multi-objective particle swarm optimization algorithm (MOPSO), with the maintenance task as the particle and the maintenance efficiency, maintenance cost, and risk level as the optimization objectives.Search for the Pareto optimal solution set in the solution space. Initialize a group of random particles, where each particle contains the priority of the maintenance task, the maintenance time, and the decision variables for the allocation of maintenance resources. The particles update their velocities and positions according to their own historical optimal positions and the global optimal position, and iterate T times until convergence. For the finally obtained Pareto front solution set, select the final maintenance plan according to the decision preference, generate the priority ranking of the maintenance tasks and the resource allocation plan, and achieve the balance of the maintenance efficiency, maintenance cost, and risk control objectives under the premise of meeting the maintenance time window constraint. Automatically send the maintenance task plan to the intelligent terminals of the maintenance personnel through the mobile collaborative office platform, guide the maintenance personnel to carry out the maintenance work according to the optimal plan, collect the on-site information, associate the on-site information with the digital insulator ledger and the electronic work ticket, generate a structured maintenance report, and through big data analysis technology, conduct text mining and semantic analysis on the historical maintenance reports, extract the failure modes and maintenance experience strategies, which are used to optimize the maintenance decision-making model and the crack propagation prediction model. At the same time, use blockchain technology to trace and audit the maintenance process to ensure the authenticity and traceability of the maintenance data, and realize the closed-loop management of the entire process of fault maintenance.

[0037] In one embodiment, the latest time window for maintenance is determined. For example, if the crack growth rate is 0.1 mm / month and the critical crack size is 10 mm, then the maintenance time window is (10 - current crack length) / 0.1 months. Use the Creaform Go, SCAN50 3D laser scanner to scan the surface of the insulator. The scanning distance is 30 cm, the scanning angle is 45°, the point cloud density is set to 0.05 mm, and the scanning speed is 4.8 million points per second. Use Geomagic Studio software to process the point cloud data, adopt the moving least squares method for noise reduction, set the curvature sampling rate to 30%, and the mesh density to 0.1 mm to generate a high-precision 3D model. Use the Image Processing Toolbox of MATLAB to extract cracks from the 3D model, adopt the Roberts operator for edge detection, set the threshold to 0.02, and extract the point cloud subset of the crack area. Calculate the geometric parameters of the crack, such as length, width, and depth. Input the crack parameters into the COMSOL Multiphysics finite element analysis software, adopt the crack growth module, select the tetrahedron mesh type, the maximum mesh size is 0.5 mm, set the load to cyclic tensile load, the stress ratio R = 0.1, and the frequency is 10 Hz. Substitute the calculated stress intensity factor range ΔK into the Paris formula to obtain the crack growth rate. Substitute the crack growth rate into the Weibull distribution fitting program written in Python, adopt the maximum likelihood estimation method, and iterate 50 times to obtain the shape parameter m = 2.3 and the scale parameter η = 2.5e6. Calculate the remaining life of the insulator at a 95% confidence level to be 1.2 years. Upload the insulator number, crack parameters, and remaining life data to the HBase distributed database of Alibaba Cloud, and use the MapReduce parallel computing framework to clean and aggregate the data to generate a maintenance task priority list. Use the Global Optimization Toolbox of MATLAB to implement the MOPSO algorithm, set the number of particles to 50, the maximum number of iterations to 200, the inertia weight to 0.8, the acceleration constants c1 = c2 = 2.0, the search dimension to 5, including task priority, maintenance time window, required personnel, spare parts, and tool resources, and set the convergence determination threshold to 1e-6 to obtain the Pareto optimal solution set. Use the Huawei Cloud ModelArts platform to build a task scheduling optimization model, select the deep reinforcement learning algorithm DDPG, the state variables include the task urgency and resource occupancy, the action variables include the task start time and resource allocation plan, the reward function comprehensively considers factors such as maintenance timeliness, cost savings, and risk avoidance, and adds a soft constraint penalty term, and train 500 episodes, and the learning rate is 0.001, The size of the ReplayMemory is 10,000. Output the optimal scheduling strategy, convert the strategy into a task card in JSON format and send it to the smart terminal of the maintenance personnel. Use the mobile collaborative office App developed with the Flutter framework for display. Use the mobile phone camera to scan the QR code on the insulator, associate the defect information with the task card, use voice input to fill in the maintenance process record and completion confirmation, and upload it to the knowledge graph platform of Huawei Cloud. Adopt natural language processing technology for semantic understanding and extraction, and update the fault diagnosis rule base and maintenance standard library;.

[0038] According to the width and length of the crack, use the multi-scale image segmentation module to extract the crack area on the surface of the insulator, construct a three-dimensional crack model, and simulate the crack propagation process;

[0039] A digital microscope is used to perform high-magnification imaging on the surface of the insulator to obtain crack image data with micron-level resolution. The image is denoised by wavelet transform, and the multi-scale texture features of the crack are extracted. Then, Gabor filtering is used to enhance the crack edges, and image binarization is achieved by setting a dynamic threshold. Finally, morphological closing operation is used to eliminate noise points to obtain a complete crack region. Using semantic segmentation algorithms based on deep learning, such as U-Net and DeepLab, the preprocessed crack image is segmented. By training a convolutional neural network on a large number of labeled samples, the multi-scale features of the crack are automatically learned to achieve precise segmentation of cracks with arbitrary widths and lengths, and the two-dimensional contour of the crack is obtained. The extracted crack contour data is input into a three-dimensional reconstruction module. Using structured light three-dimensional scanning technology, the three-dimensional surface topography information is obtained by projecting coded stripes, and the fringe phase is decoded using the phase-shift unwrapping algorithm. Then, the surface three-dimensional point cloud is reconstructed through phase unwrapping and the principle of triangulation. The point cloud is meshed and texture mapped to construct a three-dimensional crack model of the insulator with millimeter-level accuracy, and the crack region is accurately mapped onto the three-dimensional crack model to obtain the real crack morphology. According to the three-dimensional crack model, the key geometric parameters of the crack are extracted. The crack depth is calculated by measuring the maximum height difference of the crack region on the three-dimensional model;The crack tip curvature radius is obtained by fitting the spline curve at the crack tip and taking the curvature radius of the curve at the tip. According to the material type and load conditions of the insulator, the fracture criterion and the stress intensity factor calculation formula are selected. For ceramic insulators, linear elastic fracture mechanics and the KI stress intensity factor are used. For composite insulators, elastoplastic fracture mechanics and the J integral are used. Analyze the stress and strain distribution at the crack tip to determine the crack initiation and propagation conditions. Input the geometric parameters and material properties of the crack into the ABAQUS finite element analysis software, and use the extended finite element method to numerically simulate the crack propagation process. When modeling, first perform hexahedral structured mesh division on the entire insulator. The mesh density at the crack tip and its extension direction should be higher than other regions. Introduce crack propagation elements at the crack tip, set the crack propagation criterion, define the crack propagation direction and the increment step size. Consider the coupled action of electrical stress, mechanical stress, and thermal stress on the insulator during operation, set the boundary conditions and loading methods of multiple physical fields, and use the Newton-Raphson iterative algorithm for nonlinear solution. When the stress intensity factor or J integral satisfies the fracture criterion, start the crack propagation simulation until the crack propagates to the specified length or the insulator fails. Use the long short-term memory neural network time series prediction model to perform machine learning on the crack propagation data obtained from the finite element simulation. Take the crack propagation length, propagation direction, and propagation rate as sample features, and the corresponding load condition parameters as sample labels to construct a training dataset. By setting the number of neurons and activation functions in the input layer, hidden layer, and output layer, optimize the depth and width of the network, introduce Dropout regularization and early stopping mechanism to prevent overfitting, and use the stochastic gradient descent method to train the network parameters, continuously iterate and optimize until the requirements of prediction accuracy or convergence speed are met. Apply the trained prediction model, input the load parameters under actual operating conditions, and realize the intelligent prediction of the crack propagation trend and remaining life of the insulator. Combine the crack propagation prediction results with the electrical and mechanical performance degradation models of the insulator to establish the coupling relationship between crack propagation and breakdown probability and strength reduction. Use the Weibull distribution to describe the relationship between the breakdown probability of ceramic insulators and electric field strength and temperature factors, and use the Paris formula to describe the relationship between the strength reduction of composite insulators and crack propagation rate. Establish a physical mechanism model for insulator condition assessment and life prediction, identify the key factors affecting insulator reliability, formulate a risk-based condition-based maintenance strategy, accumulate big data on insulator crack propagation and performance degradation, use machine learning algorithms for knowledge mining, refine the laws of crack occurrence and development, form a knowledge base for crack intelligent diagnosis and early warning, guide the hierarchical control of insulator defects, and provide a decision-making basis for the condition-based maintenance and replacement of insulators.;

[0040] In one embodiment, a Keyence VHX-7000 series digital microscope is used. The optical zoom is set to 200 times and the resolution is 0.1 μm. A 10 cm × 10 cm area on the insulator surface is scanned and imaged to generate a high-definition image containing 20 million pixels. Based on the Daubechies 8th-order basis function of wavelet transform, the image is decomposed into 5 levels. Soft threshold denoising is applied to the high-frequency sub-bands, and the threshold is determined by 3 times the estimated noise variance. Gabor filters with a scale of 5×5 and directions of 0°, 45°, 90°, and 135° are used to extract crack textures. The differential Gaussian algorithm is used to detect crack edges, and the threshold is taken as 1.5 times the mean edge gradient. The U-Net convolutional neural network adopts a structure of 4 layers of downsampling and 4 layers of upsampling, with a convolutional kernel size of 3×3, and the number of feature map channels is 64, 128, 256, and 512 in sequence. The ReLU activation function and BatchNorm normalization are used, and the Adam optimizer is used to train for 100 epochs with an initial learning rate of 0.001, which decays by 10% every 10 epochs, achieving pixel-level segmentation of the crack contour with an mIoU of 0.98. The structured light three-dimensional scanning system uses a DLP projector to generate a sinusoidal grating stripe with a period of 1920 μm and a phase shift number of 4. The industrial camera has a resolution of 1280×1024 and a field of view of 50 mm×40 mm, and the calibration reprojection error is less than 0.05 mm. Phase profilometry is applied to solve the point cloud with a density of 100 points per square millimeter. Poisson reconstruction is used to obtain a 0.5-mm-thick insulator surface mesh model, and the model accuracy is evaluated by coverage, consistency, and integrity, with an error less than 0.2 mm. The three-dimensional crack contour is matched with the insulator model to extract geometric parameters such as crack depth, surface area, and volume. A cubic B-spline curve is fitted to the grid nodes at the crack tip, and the derivative is taken to obtain the curvature radius. The ABAQUS finite element model is based on the eight-node hexahedral element C3D8R, with a crack tip grid size of 0.1 mm, an energy integral convergence control parameter of 0.05, and isotropic elastic material parameters, with an elastic modulus of 110 GPa and a Poisson's ratio of 0.28. For the Paris formula, C = 1.8×10^-10 and m = 3.2. The crack propagation adopts the maximum energy release rate criterion, and the propagation direction is perpendicular to the maximum tensile stress plane, with a propagation step size of 0.05 mm. Electro-thermo-mechanical multi-physics coupling is applied with a voltage of 150 kV, a tensile force of 500 N, and a steady-state temperature of 80 °C, and the iterative error is less than 10^-3. The calculation is carried out for 12 hours until the insulator fails. The input layer of the LSTM network has 56 neurons, corresponding to 7 features of crack propagation data × 8 time steps, the hidden layer has 128 neurons, with 2 stacked layers, and the output layer has 1 neuron to predict the crack propagation length in the next time step. The learning rate is 0.01, and it is verified once every 10 epochs, with an earlystop patience of 5 times, and the prediction error converges to within 3%. The shape parameter m of the Weibull distribution is 7.85, the scale parameter η is 15.2, 63.The 2% breakdown voltage is 140 kV. When predicting that the remaining life is more than 1000 thermal cycles, the crack size should be controlled within 4 mm. Combining the Paris formula with the crack propagation prediction model, it is inferred that the critical crack depth when the insulator strength degrades to 70% is 8 mm. Association rule mining is carried out on the insulator crack parameters and defect modes, with a support of 0.4, a confidence of 0.85, and a lift of 2.5. It is found that glaze cracks and pores are the main causes of flashover, and it is recommended to optimize the glaze formula and firing process accordingly;.

[0041] Combined with the simulation data of the crack during the simulated propagation process, a multi-objective optimization model for crack repair is established to determine the optimal repair strategy and resource allocation plan;

[0042] According to the simulation data of the crack during the simulated propagation process, a multi-objective optimization method is adopted to establish a mathematical model for crack repair. By defining the repair strategy and resource allocation plan as optimization variables, and obtaining the crack propagation rate and repair cost as optimization objectives, the objective function and constraint conditions of the multi-objective optimization model are obtained. The non-dominated sorting genetic algorithm NSGA-II is used to solve the above multi-objective optimization model. By randomly generating the initial population of the repair strategy and resource allocation plan, the corresponding crack propagation rate and repair cost target values of each individual are obtained. According to the non-dominated sorting and crowding degree calculation, the fitness and sorting results of the population are obtained. According to the fitness and sorting results of the population, a binary tournament selection operator is used to select excellent individuals from the current population. Through simulated binary crossover and polynomial mutation operators, the selected individuals are subjected to crossover and mutation operations to obtain new repair strategies and resource allocation plans. The newly generated individuals are merged with the current population, and non-dominated sorting and crowding degree calculation are used to obtain a new generation of population. By iterating the above selection, crossover mutation, and population update processes until the maximum iteration number or convergence condition is reached, the Pareto front of the optimal repair strategy and resource allocation plan is obtained. According to the preference of the decision maker, a satisfactory solution is selected from the Pareto optimal solution set as the best crack repair strategy and resource allocation plan. By analyzing the crack propagation rate and repair cost of this plan, its effectiveness in extending the service life of the component and saving maintenance costs is judged, and the optimal decision for crack repair is determined.

[0043] In one embodiment, according to the simulation data of crack propagation, the NSGA-II algorithm is used to perform multi-objective optimization on the crack repair strategy and resource allocation scheme. First, 100 initial populations are randomly generated. Each individual contains 10 repair strategy variables and 5 resource allocation variables. The value range of the strategy variables is from 0 to 1, and the value range of the resource variables is from 0 to 1 million yuan. Then, each individual is substituted into the crack propagation model to obtain the crack propagation rate and repair cost. Next, non-dominated sorting is performed on the population, and crowding degree calculation is used among individuals to obtain fitness and ranking. After that, the binary tournament selection operator is used, with a crossover probability of 8 and a mutation probability of 1, to perform crossover and mutation on the individuals to obtain new solutions. The new solutions are merged with the current population to update the population. After 500 iterations, the Pareto front is obtained. Finally, according to the decision maker's preference, an optimal solution that balances the expansion rate and cost is selected: the crack propagation rate is reduced by 20%, the repair cost is 800,000 yuan, the service life of the component can be effectively extended by more than 10 years, and the maintenance cost can be saved by about 30%, realizing the scientific optimization of crack repair decision-making.

[0044] After the replacement or repair of the insulator in step S106 is completed, the temperature distribution on the surface of the insulator is detected through the thermal image data on the surface of the insulator, the frequency domain features of the thermal image data are extracted, and the convolutional autoencoder network is used to judge the current quality of the insulator. If there is an abnormal temperature distribution, the thermal image data is uploaded to the fault maintenance database, and the repair process is triggered to dynamically adjust the maintenance strategy.

[0045] Specifically, an infrared thermal imager is used to collect thermal images of the surface of insulators after replacement or repair. The shooting distance and angle are selected, and large insulators are photographed in regions to obtain thermal image data at different perspectives and distances. Then, through an image stitching algorithm based on SIFT feature point matching, an all-round and high-resolution temperature distribution map of the insulator surface is generated. Using two-dimensional Fourier transform 2D-FFT and wavelet transform mathematical tools, time-frequency domain analysis is performed on the thermal image data. The spectral energy distribution characteristics of the thermal image are extracted in the frequency domain, such as the proportion of low-frequency, medium-frequency, and high-frequency energies, as well as spectral entropy and spectral uniformity texture features. The multi-scale and multi-directional features of the thermal image are extracted in the wavelet domain, such as wavelet energy and the mean value of wavelet coefficients. The features extracted in the frequency domain and wavelet domain are constructed into a feature vector of the insulator surface temperature distribution. The feature vector of the thermal image is input into a convolutional autoencoder network based on the U-Net structure for quality judgment. The encoder part consists of 4 convolutional layers and 4 max-pooling layers, with a convolutional kernel size of 3×3 and a ReLU activation function. The decoder part consists of 4 transposed convolutional layers and 4 upsampling layers, gradually restoring the detailed information of the original thermal image. The network is trained using the Adam optimizer with an initial learning rate of 0.001, and the loss function is mean squared error (MSE). The evaluation index is peak signal-to-noise ratio (PSNR). The quality of the insulator is judged by the size of the reconstruction error. The higher the PSNR value, the better the reconstruction quality and the higher the insulator quality. A standard template library of insulator temperature distribution is constructed. Thermal image samples of insulators under different models, materials, and defect modes are collected. The K-means clustering algorithm is used to group the samples. First, the HOG and LBP texture features of the thermal image samples are extracted, and then the features are compressed to 20 dimensions using the PCA algorithm. Then, K-means clustering is performed, and the Silhouette coefficient determines that the optimal number of clusters is 5. The central thermal image of each cluster is calculated as the standard template, and its frequency domain and wavelet domain features are extracted. When judging whether a newly collected thermal image is abnormal, the Euclidean distance and cosine similarity between it and each standard template are calculated respectively. If the minimum distance is greater than a threshold (such as 10) or the maximum similarity is less than the threshold of 0.6, it is determined that the temperature distribution is abnormal. When it is monitored that the insulator has an abnormal temperature distribution, the edge upload process of the thermal image data is triggered. The collected thermal image data is preprocessed by cropping and normalization, and then H.It is compressed by 265 encoding, and then encrypted by the AES-256 algorithm. The compressed and encrypted data is packaged into the JSON format and sent to the fault repair database in the cloud in real time through the MQTT protocol, and an abnormal alarm message is generated. According to the abnormal temperature distribution of the insulator, the maintenance strategy and repair process are dynamically adjusted. A method combining Q-learning and simulated annealing algorithm is adopted. The abnormal degree and frequency of the insulator temperature are used as the state space, the detection frequency, equipment replacement, and process parameters are used as the action space, and the reward function is to improve the qualified rate and reduce the cost to learn the optimal maintenance strategy. At the same time, a multi-agent reinforcement learning framework is adopted to realize the collaborative maintenance optimization of multiple insulators, improve the operation and maintenance level of the entire transmission line through the game between agents, establish a quality management platform covering the whole life cycle of the insulator, including raw material quality inspection, production process monitoring, operation status monitoring, defect intelligent diagnosis, status evaluation and prediction, maintenance strategy optimization, and scrap disposal traceability function modules, and use blockchain technology to record and store the quality data of each link in an immutable manner to achieve the closed-loop of insulator quality management.

[0046] In one embodiment, the temperature measurement range of the thermal imager is -40°C to 2000°C, the thermal sensitivity is 0.02°C, the spatial resolution is 1024×768, and the frame rate is 30Hz. The FLIRT1020 infrared thermal imager is used to collect thermal images at a position 2 meters away from the insulator at an angle of 30°. The resolution is set to 1024×768, the frame rate is 30Hz, the temperature measurement range is -20°C to 150°C, and the thermal sensitivity is 0.02°C @ 30°C. For large insulators with a diameter exceeding 50 cm, they are divided into 4 quadrants, and thermal images are collected separately. Each quadrant overlaps by 20%. Then, the SIFT algorithm is used to extract feature points, and the homography matrix is estimated by the RANSAC algorithm. Finally, a complete thermal image is stitched together, and the stitching accuracy is better than 0.5 pixels. The Signal Processing Toolbox of MATLAB is used to perform frequency domain analysis on the thermal image. The 2D-FFT transform is used to obtain the amplitude spectrum and phase spectrum. The energy ratio features of low frequency 0 - 0.1π, medium frequency 0.1π, and high frequency 0.5π are extracted. The gray level co-occurrence matrix is calculated to obtain 14 texture features such as contrast, correlation, energy, and entropy. The db4 wavelet basis function is used to perform 5-level wavelet decomposition on the thermal image. The mean, variance, and energy of the HL, LH, and HH sub-band coefficients are extracted to obtain 9 wavelet domain features. The U-Net convolutional autoencoder network is trained. The encoder performs 4 times of downsampling, the decoder performs 4 times of upsampling, the convolution kernel size is 3×3, the stride is 1, the padding method is same, the activation function uses ReLU, the maximum pooling size is 2×2, the Adam optimizer is used, the initial learning rate is 0.001, it decays by 50% every 10 epochs, the minimum learning rate is 1e-6, the batch_size is 16, and it is trained for 100 epochs. The early stopping method is used to prevent overfitting. The PSNR of the best model on the validation set reaches 35.2dB. The K-means clustering algorithm is used to cluster the HOG and LBP feature vectors of 500 normal insulators and 200 abnormal insulators. The elbow method is used to determine the optimal number of clusters k = 5. The average PSNR between each cluster center and the samples is calculated, and the threshold is taken as 34dB. When the PSNR of the reconstructed image of the test insulator is lower than this threshold, it is determined that the temperature is abnormal and uploaded to the Huawei Cloud server. First, it is encoded and compressed by 50% using H.265, then encrypted using AES-256, and transmitted through the MQTT protocol, reducing the network bandwidth by 75%. The Q-learning algorithm is used for policy learning. The state space is the insulator health (0 - 100), the action space is the detection frequency 1 time / day, 7 times / week, 30 times / month. The reward function is to reward 100 points for a 10% reduction in the failure rate, and deduct 500 points for exceeding 3%. The learning rate α = 0.1, the discount factor γ = 0.9, the exploration probability ε = 0.7, the maximum number of rounds is 1000, and the average reward after convergence is 85 points. On this basis, the simulated annealing strategy is added, the initial temperature T0 = 100, and the attenuation rate λ = 0.With the inner loop L = 50 and the outer loop D = 20, the global optimal solution is found, and the average reward is increased to 90 points. For three 500 kV transmission lines each 100 km long, 30 fault monitoring agents are deployed. Using the multi - Agent - Q - learning framework, after 20,000 rounds of training, the failure rate is reduced by 15% and the operation and maintenance cost is reduced by 12%. During this process, the Hyperledger Fabric blockchain platform is used to record 30 million insulator quality data, the block size is 2 MB, the consensus mechanism is PBFT, the throughput reaches 1500 TPS, the shared ledger storage capacity is 6 TB, and the data traceability and privacy protection level reach the HIPAA standard.

[0047] In step S107, according to the multi - dimensional data of the insulator, including coating aging degree, glaze layer distortion, surface cracks, and temperature anomalies, the comprehensive health index of the insulator is obtained, and the Markov chain prediction model is used to estimate the state transition probability under different maintenance strategies. The multi - objective particle swarm optimization module is used to generate the optimal maintenance decision sequence for the whole life cycle of the insulator.

[0048] Specifically, from multi-source heterogeneous systems such as fault diagnosis systems, condition monitoring systems, and environmental monitoring systems, multi-dimensional data such as the aging degree of insulator coatings, glaze distortion conditions, surface crack parameters, and abnormal temperature monitoring are extracted. Using the method of ontology mapping, an ontology in the field of insulator health assessment is constructed to define the semantic associations of each data source. Through ontology reasoning, automatic association and conversion at the data semantic level are achieved. For data quality problems, data cleaning techniques such as noise recognition, missing value filling, and outlier detection are used, and data correction rules are constructed in combination with electrical domain knowledge to form a structured and standardized health record database for the entire life cycle of insulators. According to the insulator health status assessment standard, key indicators reflecting the health level of insulators are extracted from multi-dimensional monitoring data. The subjective and objective combined weights of each indicator are determined by combining the Delphi method and the entropy weight method. The comprehensive health index of insulators is calculated by weighted average. Based on the health index, a multi-classification model for the health status of insulators is constructed using the support vector machine algorithm. Multi-classification is achieved by finding the maximum margin hyperplane, and a kernel function is introduced to solve the non-linear problem. Precision, recall, and F1 value evaluation indicators are used, and the optimal model is selected through cross-validation. A Markov chain for the evolution of the insulator health status is constructed using the hidden Markov model. The comprehensive health index of the insulator is discretized into normal, sub-healthy, mildly deteriorated, moderately deteriorated, and severely deteriorated states. The state transition probability matrix is learned from the monitoring data through the Baum-Welch algorithm, the observation probability matrix is estimated based on the correspondence between the insulator health status and multi-dimensional data, and a reduction algorithm is introduced to reduce the dimension of the state space. Thus, the dynamic evolution trend and steady-state distribution of the insulator health status under different maintenance strategies are predicted. A multi-objective optimization model for insulator maintenance decision-making is constructed. The objective function includes three aspects: improving the insulator health level, minimizing the maintenance cost, and maximizing the remaining life of the insulator. The constraint conditions include insulator health state transition, upper limit of maintenance resources, and attenuation of the remaining life of the insulator. The decision variables are insulator maintenance and replacement maintenance measures. The objective function and constraint conditions are quantitatively expressed through mathematical formulas. The improved multi-objective particle swarm optimization algorithm (MOPSO) is used to solve the maintenance decision optimization model. The Pareto optimal solution set archive and crowding degree measurement are introduced to ensure the convergence and distribution diversity of the solutions. The position and velocity of the particles are updated according to the standard formula, and are updated through the inertia weight, acceleration constant, historical optimal position of the particle, and global optimal position. The key parameters of the algorithm are optimized through sensitivity analysis and experimental design. Thus, an optimal maintenance decision sequence for the entire life cycle of the insulator is generated. The optimal maintenance decision sequence is decoded into specific maintenance times, maintenance methods, and maintenance parameter maintenance measures at each stage of the entire life cycle of the insulator to form a dynamic maintenance schedule covering the entire life cycle. When dynamically adjusting the maintenance plan, a rolling optimization strategy is adopted. According to the insulator state monitoring data and environmental factor changes, the state transition probability, maintenance cost, and remaining life parameters in the multi-objective optimization model are periodically updated.Generate the optimal maintenance plan during the rolling period, dynamically adjust the weight coefficients of the optimization objectives according to the execution feedback, realize the adaptive optimization of maintenance decision-making, realize the real-time monitoring and early warning of the insulator state, develop an insulator asset management system based on big data analysis and artificial intelligence, integrate the full life cycle data of insulator design, manufacturing, operation, maintenance, and scrapping, adopt machine learning and deep learning algorithms, such as random forest, XGBoost, LSTM, establish a prediction model and an abnormal diagnosis model for the insulator degradation mechanism, the knowledge base adopts a hybrid reasoning method based on ontology and rules, integrates the structured and unstructured knowledge of equipment nameplates, defect cases, and diagnostic tests, forms a knowledge graph for the full life cycle management of insulators, uses the knowledge graph to realize intelligent diagnosis, prediction and decision-making, continuously optimize the insulator health state evaluation model and the maintenance decision-making optimization model, provide intelligent analysis and auxiliary decision-making functions for key decisions such as insulator selection, transformation, maintenance, and replacement, and improve the management level of power equipment and the asset utilization efficiency.

[0049] In one embodiment, data on the aging degree of the insulator coating is extracted from the fault diagnosis system. For example, the average temperature in the coating area of the infrared image is 68°C, which is 13°C higher than the normal value, and the aging degree is severe. The distortion of the glaze layer is judged by X-ray imaging. Canny edge detection is performed on the image, and the proportion of the distorted area is calculated to be 2%, exceeding the warning value of 5%. The surface crack parameters are obtained by laser three-dimensional scanning. The region growing algorithm is used to segment the cracks, and the length, width, and depth of the cracks are extracted as 5mm, 8mm, and 2mm respectively. The environmental monitoring system collects data on temperature, humidity, wind speed, and pollution degree around the insulator. The abnormal detection model is used to judge that the temperature data is abnormal, deviating from the normal value by 5 standard deviations. The above multi-source heterogeneous data is mapped through the ontology to construct an ontology in the field of insulator health assessment, including categories such as coating aging degree, glaze layer distortion degree, crack risk degree, and environmental anomaly degree, and the semantic associations between concepts are defined. For example, the functional relationship between the coating aging degree and the infrared temperature. Noise identification is performed on the monitoring data. If the high-frequency component of the temperature data exceeds the threshold, it is judged as noise. Missing values are filled using the nearest neighbor interpolation method, and outliers are detected by the isolation forest algorithm. Combining electrical domain knowledge, data correction rules are constructed. For example, when the wind speed exceeds 30m / s, the pollution degree data is multiplied by a correction coefficient of 5. Finally, a structured health record database for the entire life cycle of the insulator is formed. Key indicators are extracted from the health record database, and 10 experts are consulted using the Delphi method to obtain the subjective weight vector [3, 25, 2, 25]. Then, the objective weight vector [28, 32, 15, 25] is calculated by the entropy weight method. The combined weight vector is [29, 285, 175, 25], and the weighted average is used to obtain the comprehensive health index of the insulator as 75 points. Based on the health index, a multi-classification model for the health state is constructed using the support vector machine. The state space is "normal, sub-healthy, slightly deteriorated, moderately deteriorated, severely deteriorated", the sample size is 1000, the feature dimension is 10, the radial basis kernel function is used, the penalty coefficient C = 5, and the kernel function parameter g = 08. Through 5-fold cross-validation, the precision of the model is 92%, the recall rate is 95%, and the F1 value is 95%. The comprehensive health index of the insulator is discretized into 5 states, and a hidden Markov model is constructed. The initial state probability vector is [7, 2, 06, 03, 01]. The Baum-Welch algorithm is used to iteratively optimize the state transition probability matrix and the observation probability matrix, and the dimension of the observation data is reduced from 10 to 6 through the reduction algorithm to predict the evolution trend of the insulator health state in the next 10 years. The health state stabilizes at the sub-healthy level after 5 years. A multi-objective optimization model for insulator maintenance decision-making is constructed and solved using the MOPSO algorithm. 200 particles are initialized, the inertia weight w = 7, the acceleration constants c1 = c2 = 5, and the maximum number of iterations is 500.The archive capacity is 50. Through the Pareto optimal solution set and crowding degree measurement, the optimal maintenance decision sequence with a maintenance cost of 200,000 yuan, a 10% improvement in health level, and a 5-year extension of remaining life is obtained. This decision sequence includes preventive maintenance in the 3rd year to replace the aging coating, condition-based maintenance in the 5th year to repair the glaze layer defects, periodic maintenance in the 8th year to tighten the fittings, develop an insulator asset management system, establish a degradation mechanism prediction model using the random forest algorithm, with the feature importance ranking as coating aging degree 35, glaze layer distortion degree 28, crack risk degree 22, and environmental anomaly degree 15. Establish an anomaly diagnosis model using the XGBoost algorithm for multi-classification of fault types, with an accuracy rate reaching 95%. The knowledge graph contains 500 concepts and 1500 associations. Based on ontology and rule reasoning, intelligent diagnosis and prediction are realized. The knowledge system contains 300 diagnosis rules and 150 successful cases. Through case-based reasoning and rule-based reasoning, it provides intelligent analysis and auxiliary decision-making for key decisions such as insulator selection, transformation, maintenance, and replacement. The average time consumption for each decision is shortened from 2 hours to 10 minutes.,

[0050] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or replacements to the relevant technical features, and the technical solutions after these changes or replacements will all fall within the protection scope of the present invention.

[0051] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and scope of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for generating a distribution network maintenance strategy based on a heuristic algorithm, characterized in that: S101. Analyze the aging degree of the coating on the surface of the insulator in the distribution network, including using Fourier transform infrared spectroscopy analysis technology to obtain chemical bond vibration characteristic data of the coating material, construct an aging degree assessment model, determine whether the aging degree of the insulator is greater than a threshold, and if so, identify that the coating needs to be replaced, enter the number and location information of the insulator into the fault maintenance database, and determine the maintenance priority according to the aging degree of the coating; S102, using an X-ray diffractometer to analyze the physical composition of the glaze layer of the insulator, obtaining diffraction spectrum data about the crystal structure of the glaze layer, extracting features and classifying the diffraction spectrum data through a convolutional neural network model, and determining whether the glaze layer is distorted. If it is determined that the glaze layer is distorted, the insulator number and location information are entered into a fault maintenance database, and a glaze repair or replacement plan is determined according to the type of distortion; S103, for each insulator to be repaired in the fault maintenance database, by analyzing the geographical location of the tower and the electromagnetic interference level, the insulator to be repaired is divided into multiple maintenance areas through the fuzzy C-means clustering module, and the optimal maintenance path for each maintenance area is determined, the optimal maintenance path is sent to the terminal of the maintenance personnel, and the maintenance personnel are located and tracked in real time through the position sensing module; S104. During the maintenance process, the photogrammetry module is used to obtain image data of the transmission line, extract the geometric model of the insulator, and determine whether the insulator has defects, including damage or contamination. If so, the defect type and location coordinates are uploaded to the fault maintenance database, and a local cleaning or replacement process is triggered; S105. For defective insulators, measure the width and length of cracks on the surface of the insulators, determine the crack expansion path through the ant colony search module, predict the crack expansion trend and the remaining life of the insulators, and enter the insulator number and location information into the fault maintenance database if the remaining life is lower than the threshold. Determine the maintenance time window according to the crack expansion rate, and schedule maintenance resources through the heuristic rule module; S106. After the insulator is replaced or repaired, the temperature distribution on the surface of the insulator is detected through the thermal image data on the surface of the insulator, the frequency domain features of the thermal image data are extracted, and the current quality of the insulator is judged using a convolutional autoencoder network. If there is an abnormal temperature distribution, the thermal image data is uploaded to the fault maintenance database, and the repair process is triggered to dynamically adjust the maintenance strategy. S107. Based on the multi-dimensional data of the insulator, including coating aging degree, glaze deformation, surface cracks and temperature anomaly, the health index of the insulator is obtained, and the state transition probability under different maintenance strategies is estimated using the Markov chain prediction model. The multi-objective particle swarm optimization module is used to generate the optimal maintenance decision sequence for the entire life cycle of the insulator.

2. The method for generating a distribution network maintenance strategy based on a heuristic algorithm according to claim 1, characterized in that: The S101 step analyzes the aging degree of the coating on the surface of the insulator in the distribution network by using Fourier transform infrared spectroscopy analysis technology, performs chemical composition analysis on the coating material on the surface of the insulator, obtains chemical bond vibration characteristic data of the coating material, removes a sample from the coating on the surface of the insulator, grinds it into powder with a mortar, presses the powder sample into a tablet, and puts it into an FTIR instrument for testing. The test data is baseline corrected and smoothed preprocessed by using FTIR spectrum analysis software, and then compared with the data in the standard spectrum library to calculate the similarity. The similarity is measured by Euclidean distance and correlation coefficient indicators. According to the size of the similarity, the aging degree of the coating material is judged, and the chemical bond vibration characteristics obtained by FTIR analysis are spliced ​​with the surface image characteristics of the insulator coating to form a sample feature input. The aging degree category is used as the output, and the cross entropy loss function and Adam optimizer are used for training.

3. The method for generating a distribution network maintenance strategy based on a heuristic algorithm according to claim 1, characterized in that: In the step S102, an X-ray diffractometer is used to analyze the physical composition of the glaze layer of the insulator to obtain diffraction spectrum data of the crystal structure of the glaze layer. The setting parameters of the X-ray diffractometer include X-ray wavelength, incident angle, scanning step length and scanning speed. Through data preprocessing and feature engineering, key characteristic parameters of the diffraction spectrum are extracted. The key characteristic parameters include that the diffraction peak position corresponds to the crystal plane spacing, the peak intensity corresponds to the crystal plane orientation, the peak width corresponds to the grain size, and the background intensity corresponds to the amorphous phase content.

4. The method for generating a distribution network maintenance strategy based on a heuristic algorithm according to claim 1, characterized in that: The step S103 extracts information of each insulator to be repaired from the fault repair database, performs spatial visualization analysis on the tower position using the ArcGIS system, and calculates the distance matrix between the towers using the Geodesic distance calculation function.

5. The method for generating a distribution network maintenance strategy based on a heuristic algorithm according to claim 1, characterized in that: The step S104 uses a 3D laser scanner to scan the defective surface of the insulator, denoises, meshes, and reconstructs the point cloud data obtained by the scan to obtain a 3D model of the insulator surface with an accuracy of 0.1 mm, and then uses a geometric morphology analysis module to extract crack features from the 3D model to obtain geometric parameters of the length, width, and depth of the crack. The geometric parameters of the crack are input into an ant colony search module. The ant colony search module uses an ant colony optimization algorithm based on graph theory to divide the insulator surface into N×N grids, each grid corresponds to a node, and there are connections between adjacent grids. The connection weight is proportional to the probability of crack extension, and m ants are released randomly.

6. The method for generating a distribution network maintenance strategy based on a heuristic algorithm according to claim 1, characterized in that: The step S105 uses the Paris formula based on fracture mechanics to predict the crack growth trend. The Paris formula describes the relationship between the crack growth rate da / dN and the stress intensity factor range ΔK: da / dN=C(ΔK)^m, where C and m are material constants obtained by fitting the crack growth test. The stress intensity factor range ΔK is associated with the load, crack size, and sample geometry, obtained by finite element analysis. The Paris formula is discretized and the crack size is recursively calculated: , where a i and a i+1 , represents: the crack size after the i-th and i+1-th cycles, ΔK i Indicates: the stress intensity factor range of the i-th cycle loading, ΔN i is the number of cycles of the i-th cyclic loading, which is determined according to the set time interval and cycle frequency. The crack growth trend data is input into the remaining life prediction model. The remaining life prediction model uses Weibull distribution to describe the fatigue life distribution of the insulator, and its probability density function is: f(t)=(m / η)(t / η)^(m-1)exp, where t is the fatigue life, m is the shape parameter, and η is the scale parameter.

7. The method for generating a distribution network maintenance strategy based on a heuristic algorithm according to claim 1, characterized in that: For large insulators with a diameter of more than 50 cm, the S106 step divides them into four quadrants, collects thermal images respectively, and overlaps each quadrant by 20%. Then, the SIFT algorithm is used to extract feature points, the homography matrix is ​​estimated by the RANSAC algorithm, and the thermal images are spliced ​​together. The thermal images are analyzed in the frequency domain using MATLAB's Signal Processing Toolbox, and the amplitude spectrum and phase spectrum are obtained by 2D-FFT transformation.

Images