Insulator anomaly detection method and system

Abstract

The invention discloses an insulator anomaly detection method and system, and relates to the technical field of insulator anomaly detection, and the method comprises the steps: setting an insulator anomaly detection time interval, arranging a sensor for monitoring original temperature data including detection of a target insulator string, and carrying out the processing; gray mapping is carried out on the preprocessed original temperature data, the original temperature data are converted into gray images, and meanwhile, a deep learning model is constructed and trained; and using the trained deep learning model to identify and output an insulator string area in the picture, and judging whether the temperature of the insulator is abnormal or not according to the temperature data of the insulator string area. The method saves the detection time, manpower and material resources each time, can avoid the interference of other irrelevant factors on the judgment, can judge the position of the insulator string more efficiently than manual labeling, and is convenient for temperature anomaly judgment.

Description

Technical Field

[0001] The present invention relates to the technical field of insulator abnormality detection, and in particular to an insulator abnormality detection method and system. Background Art

[0002] As a key component of the transmission line, the operating status of the composite insulator is directly related to the stability of the entire power grid system. Once the composite insulator has obvious mechanical and electrical performance failures, the impact range is wide and the loss will be very serious. Composite insulators are prone to aging due to long-term corona discharge, arc erosion and harsh environment. Through infrared detection, abnormal heating of insulators can be found, and then the operating status of the insulator can be inferred. Early detection of defects and hidden dangers will help avoid accidents.

[0003] At present, the traditional manual inspection method is to use a manual handheld infrared thermometer or a drone-mounted thermometer to measure temperature. Since the insulator string is placed at a high position, it is difficult for inspectors to approach, and it is difficult to use the manual inspection method to distinguish abnormalities. Using drone-mounted thermometers to measure temperature also has problems such as long time periods, large workloads, and difficulty in timely detection of heating hazards. Summary of the invention

[0004] In view of the problems existing in the existing insulator abnormality detection method, the present invention is proposed. Therefore, the problem to be solved by the present invention is how to provide an insulator abnormality detection method and system.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, the present invention provides an insulator abnormality detection method, which includes setting an insulator abnormality detection time interval, arranging sensors to monitor and process the original temperature data of the target insulator string; performing grayscale mapping on the pre-processed original temperature data, converting the original temperature data into a grayscale image, and constructing and training a deep learning model; using the trained deep learning model to identify and output the insulator string area in the image, and judging whether the insulator has a temperature abnormality based on the temperature data of the insulator string area.

[0007] As a preferred solution of the insulator abnormality detection method of the present invention, the processing includes the following steps: cleaning the collected original temperature data, removing outliers, checking whether there are obvious errors or outliers in the data and removing them; filling the missing parts in the data by interpolation, converting the data into data with zero mean and unit variance, and using Z-score standardization. The calculation formula is:

[0008]

[0009] Among them, T i is the standardized original temperature data, X is the original temperature data, μ is the average value of the original temperature data, and σ1 is the standard deviation of the original temperature data; the image is resized and randomly cropped to simulate different observation conditions and viewing angles, the data set is expanded, and the image is rotated, mirrored or cropped to generate more images, and the multiple grayscale images obtained are combined into an insulator string database.

[0010] As a preferred solution of the insulator abnormality detection method of the present invention, the grayscale mapping of the pre-processed original temperature data includes the following steps: determining the range of the temperature data, that is, the minimum temperature value T min and the maximum temperature T max , map the temperature data, and use the mapping formula to convert the temperature value into a grayscale value. The calculation formula is:

[0011]

[0012] Among them, G is the grayscale value, and T is the value of the original temperature data; if the grayscale value after mapping exceeds 255, it is cropped or adjusted to ensure that all grayscale values ​​are within the valid range; after each temperature value is converted into a grayscale value, the grayscale value is placed in a matrix, and each matrix element corresponds to a pixel in the image, and its grayscale value is the mapped temperature value; the matrix is ​​saved as an image file, and the grayscale image displays the pixel grid and coordinate axis, and the coordinates of the insulator area are manually calibrated, the insulator area is divided, and a label library of the insulator string image is established to form a data set.

[0013] As a preferred solution of the insulator abnormality detection method described in the present invention, the establishment of a label library of insulator string images and the formation of a data set include the following steps: recording the regional coordinates of the insulator string according to the grayscale image as the label of the corresponding image, and calculating the minimum circumscribed rectangle or contour coordinates of each detected insulator string contour; formatting the extracted insulator string regional coordinates into the label format required for model training, encoding each pixel into a corresponding category label; matching the prepared label with the corresponding grayscale image to ensure that each image has a corresponding label; dividing the labeled image data set into a training set, a validation set, and a test set, randomly checking the images and labels in the data set to ensure that the labels correspond to the image content, and saving the processed data set in a format suitable for model training.

[0014] As a preferred solution of the insulator anomaly detection method of the present invention, the deep learning model is constructed and trained, including using a pre-trained MobileNet model as a basic model, loading the weights pre-trained on the ImageNet large-scale data set to optimize the model; removing the original fully connected layer at the top of the MobileNet model, adding a global average pooling layer and a fully connected layer, freezing the first three convolution blocks of the MobileNet model, and retaining the subsequent convolution blocks; setting the loss function of the model to DIoU Loss, measuring the degree of overlap between the bounding box predicted by the model and the true bounding box and the center distance between the two boxes when the degree of overlap is the same, and the calculation formula is:

[0015]

[0016] Among them, IoU is the intersection over union ratio between the predicted box and the labeled box, the predicted box is P, and the labeled box is G; R(P,G) is the penalty term; ρ(P,G) represents the distance between the center end of the predicted box and the labeled box, p and g are the center points of the two boxes, and c represents the diagonal length of the minimum enclosing rectangle of the two boxes; the insulator string dataset is divided into training set, validation set and test set with a ratio of 7:2:1, the training set is input into the model, and the DIoU of the model is calculated Loss; use the back-propagation algorithm to calculate the gradient of the loss function to the model parameters, and use the Adam optimizer to update the parameters; the initial learning rate is 0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-8, amsgrad=False, eight_decay=0.001; regularly evaluate the model on the validation set to determine whether the model begins to overfit and decide whether to stop training early or adjust the model structure; use the test set to evaluate the trained model, use the real labels of the test set to compare with the predicted results of the model, and calculate the model loss function to verify the model; pre-process the sensor temperature data to be detected and convert it into a grayscale image to input into the deep learning model, and obtain the model output insulator string area bounding box coordinates and confidence score.

[0017] As a preferred solution of the insulator abnormality detection method of the present invention, the method of judging whether the insulator has abnormal temperature according to the temperature data of the insulator string area includes the following steps: determining the coordinate range of the insulator string area according to the bounding box coordinates output by the deep learning model, assuming that the bounding box coordinates of the insulator string area are described by the lower left corner coordinates (x1, y1) and the upper right corner coordinates (x2, y2), considering the maximum and minimum values ​​of the temperature, and calculating the normalization factor, the calculation formula is:

[0018]

[0019] Where Z is the normalization factor, Tmax is the maximum temperature, T min is the minimum value of temperature; within the coordinate range of the insulator string area, the distribution characteristics of the temperature data are analyzed and the weighted average temperature is calculated. The calculation formula is:

[0020]

[0021] In the formula, is the weighted average temperature, T ij is the temperature value at the coordinate (i, j), and all temperature values ​​in the area are traversed to obtain the highest temperature value T max ; Calculate the standard deviation of the temperature and the maximum value of the temperature curvature within the coordinate range of the insulator string area, identify the temperature mutation point, and the calculation formula is:

[0022]

[0023] Where σ2 is the standard deviation of the temperature within the coordinate range of the insulator string area, N is the number of pixels in the bounding box, and C max is the maximum value of the temperature curvature; the temperature difference limit is calculated comprehensively, and whether the temperature of the insulator string is abnormal is determined based on the calculated temperature difference limit.

[0024] As a preferred solution of the insulator abnormality detection method of the present invention, the comprehensive calculation of the temperature difference limit value includes comprehensively calculating the temperature difference limit value of the insulator string according to the standard deviation of the temperature within the coordinate range of the obtained insulator string area and the maximum value of the temperature curvature, and the calculation formula is:

[0025]

[0026] In the formula, ΔT odd is the temperature difference limit of the insulator string, k1 is the adjustment factor, C th is the starting point of the curvature influence; if the temperature difference in the insulator string area If the temperature deviates from the normal range, there is an overheating or safety hazard. Check immediately and increase the monitoring frequency of the area to real-time monitoring. If the temperature difference in the insulator string area It is determined that the insulator temperature is within the normal range and normal monitoring continues without intervention.

[0027] In the second aspect, the present invention provides an insulator abnormality detection system, which includes: a collection module, which is used to set the insulator abnormality detection time interval, arrange sensors to monitor the original temperature data of the target insulator string, and process it; a mapping module, which is used to perform grayscale mapping on the pre-processed original temperature data, convert the original temperature data into a grayscale image, and construct and train a deep learning model at the same time; a detection module, which is used to use the trained deep learning model to identify and output the insulator string area in the image, and judge whether the insulator has temperature abnormalities based on the temperature data of the insulator string area.

[0028] In a third aspect, the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, the steps of the insulator abnormality detection method are implemented.

[0029] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: the computer program implements the steps of an insulator abnormality detection method when executed by a processor.

[0030] The beneficial effect of the present invention is that it is convenient to measure temperature by using a fixed infrared temperature sensor, without the need to manually climb the pole tower or control the drone flight, thus saving time, manpower and material resources for each detection. The acquired temperature data is gray-mapped to suppress unnecessary features in the image, and the insulator string area is judged by shape rather than color, which can avoid interference of other irrelevant factors on the model judgment. The insulator string area is extracted by a deep learning model, which is more efficient than manual labeling to judge the position of the insulator string, facilitating subsequent temperature abnormality judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0032] Figure 1 The figure is a flow chart of the insulator abnormality detection method.

[0033] Figure 2 Grayscale image of sensor data mapping for insulator anomaly detection method. DETAILED DESCRIPTION

[0034] In order to make the above-mentioned purposes, features and advantages of the present invention more understandable, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0035] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0036] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0037] Example 1

[0038] Reference Figure 1 , which is the first embodiment of the present invention, provides an insulator abnormality detection method, comprising:

[0039] S1: setting the insulator abnormality detection time interval, arranging sensors to monitor the original temperature data of the target insulator string, and processing them;

[0040] Specifically, the collected raw temperature data is cleaned to remove outliers, and the data is checked for obvious errors or outliers to remove;

[0041] For the missing parts in the data, fill them by interpolation, convert the data into data with zero mean and unit variance, and use Z-score standardization. The calculation formula is:

[0042]

[0043] Among them, T i is the normalized original temperature data, X is the original temperature data, μ is the mean value of the original temperature data, and σ1 is the standard deviation of the original temperature data;

[0044] The images are resized and randomly cropped to simulate different observation conditions and perspectives. The data set is expanded using data augmentation technology. The images are rotated, mirrored, or cropped to generate more images. The multiple grayscale images obtained are combined into an insulator string database.

[0045] Determine the range of temperature data, that is, the minimum temperature value T min and the maximum temperature T max , map the temperature data, and use the mapping formula to convert the temperature value into a grayscale value. The calculation formula is:

[0046]

[0047] Among them, G is the gray value, T is the value of the original temperature data;

[0048] If the grayscale value after mapping exceeds 255, crop or adjust the mapping method to ensure that all grayscale values ​​are within the valid range;

[0049] After converting each temperature value into a grayscale value, the grayscale value is placed in a matrix. This matrix represents the grayscale image. Each matrix element corresponds to a pixel in the image, and its grayscale value is the mapped temperature value.

[0050] The matrix is ​​saved as an image file for use in deep learning models. The grayscale image displays the pixel grid and coordinate axis. The coordinates of the insulator area are manually calibrated, the insulator area is divided, and a label library of the insulator string image is established to form a data set.

[0051] The regional coordinates of the insulator string are recorded according to the grayscale image as the label of the corresponding image. For each detected insulator string contour, its minimum circumscribed rectangle or contour coordinates are calculated;

[0052] Format the extracted insulator string area coordinates into the label format required for model training, and encode each pixel into the corresponding category label;

[0053] Match the produced labels with the corresponding grayscale images to ensure that each image has a corresponding label;

[0054] Divide the labeled image dataset into training set, validation set and test set, randomly check the images and labels in the dataset to ensure that the labels correspond to the image content, and save the processed dataset in a format suitable for model training.

[0055] S2: Grayscale mapping is performed on the preprocessed raw temperature data to convert the raw temperature data into a grayscale image, and a deep learning model is constructed and trained at the same time;

[0056] Specifically, we use the pre-trained MobileNet model as the base model and load the weights pre-trained on the large-scale ImageNet dataset to optimize the model.

[0057] Remove the top fully connected layer of the MobileNet model and add a global average pooling layer and a new fully connected layer to output the results of object detection and localization. The function of the global average pooling layer is to reduce the dimension of the input feature map while retaining the main features of the feature map to help reduce overfitting.

[0058] The function of the fully connected layer is to map the feature vector output by the global average pooling layer to the predicted result of the positioning coordinates.

[0059] Then freeze the first three convolutional blocks of the MobileNet model and keep the following convolutional blocks for fine-tuning.

[0060] Set the model's optimizer to Adam optimizer and the loss function to DIoU Loss. DIoU Loss is used to measure the degree of overlap between the bounding box predicted by the model and the true bounding box, as well as the center distance between the two boxes when the degree of overlap is the same. The specific definition is as follows:

[0061]

[0062] Among them, IoU is the intersection over union ratio between the predicted box and the labeled box, the predicted box is P, and the labeled box is G; R(P,G) is the penalty term; ρ(P,G) represents the distance between the center end of the predicted box and the labeled box, p and g are the center points of the two boxes, and c represents the diagonal length of the minimum enclosing rectangular box of the two boxes. When the distance between the two boxes is infinite, the center point distance and the diagonal length of the circumscribed rectangular box are infinitely close, and R(P,G) approaches 1;

[0063] The insulator string dataset is divided into training set, validation set and test set with a ratio of 7:2:1. In each training step, the training set is input into the model and the DIoU Loss of the model is calculated;

[0064] Then use the back-propagation algorithm to calculate the gradient of the loss function with respect to the model parameters, and use the Adam optimizer to update the parameters;

[0065] The initial learning rate is 0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-8, amsgrad=False, eight_decay=0.001;

[0066] During the training process, set the appropriate number of training rounds and monitor the performance of the model on the training set and validation set. By regularly evaluating the model on the validation set, it is possible to determine whether the model has begun to overfit and decide whether to stop training early or adjust the model structure.

[0067] Use the test set to evaluate the trained model, use the actual labels of the test set to compare with the model's prediction results, and calculate the model loss function to verify the reliability and generalization ability of the model;

[0068] The sensor temperature data to be detected is preprocessed and converted into a grayscale image, which is input into the deep learning model to obtain the model output insulator string area bounding box coordinates and confidence score.

[0069] S3: Use the trained deep learning model to identify and output the insulator string area in the image, and determine whether the insulator temperature is abnormal based on the temperature data of the insulator string area.

[0070] Specifically, the coordinate range of the insulator string area is determined according to the bounding box coordinates output by the deep learning model. Assuming that the bounding box coordinates of the insulator string area are described by the coordinates of its lower left corner (x1, y1) and upper right corner (x2, y2), the normalization factor is calculated considering the maximum and minimum values ​​of the temperature. The calculation formula is:

[0071]

[0072] Where Z is the normalization factor, T max is the maximum temperature, T min is the minimum value of temperature;

[0073] Within the coordinate range of the insulator string area, the distribution characteristics of the temperature data are analyzed and the weighted average temperature is calculated. The calculation formula is:

[0074]

[0075] In the formula, is the weighted average temperature, T ij is the temperature value at the coordinate (i, j), and all temperature values ​​in the area are traversed to obtain the highest temperature value T max ;

[0076] Calculate the standard deviation of the temperature and the maximum value of the temperature curvature within the coordinate range of the insulator string area, and identify the temperature mutation point. The calculation formula is:

[0077]

[0078] Where σ2 is the standard deviation of the temperature within the coordinate range of the insulator string area, N is the number of pixels in the bounding box, and C max is the maximum value of temperature curvature;

[0079] The temperature difference limit is calculated comprehensively, and whether the temperature of the insulator string is abnormal is determined based on the calculated temperature difference limit.

[0080] The temperature difference limit of the insulator string is calculated comprehensively based on the standard deviation of the temperature within the coordinate range of the insulator string area and the maximum value of the temperature curvature. The calculation formula is:

[0081]

[0082] In the formula, ΔT odd is the temperature difference limit of the insulator string, k1 is the adjustment factor, C th is the starting point of the curvature influence;

[0083] If the temperature difference in the insulator string area If the temperature deviates from the normal range, an abnormal temperature alarm is triggered, and there is an overheating or safety hazard, an immediate inspection is carried out and the monitoring frequency of the area is increased to real-time monitoring;

[0084] If the temperature difference in the insulator string area It is determined that the insulator temperature is within the normal range and routine monitoring continues without special intervention.

[0085] Furthermore, the present embodiment also provides an insulator abnormality detection system, including: a collection module, used to set the insulator abnormality detection time interval, arrange sensors to monitor the original temperature data of the target insulator string, and process it; a mapping module, used to perform grayscale mapping on the pre-processed original temperature data, convert the original temperature data into a grayscale image, and simultaneously construct and train a deep learning model; a detection module, used to use the trained deep learning model to identify and output the insulator string area in the image, and determine whether the insulator has a temperature abnormality based on the temperature data of the insulator string area.

[0086] This embodiment also provides a computer device, which is applicable to the case of an insulator abnormality detection method, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement all or part of the steps of the method described in the embodiment of the present invention as proposed in the above embodiment.

[0087] This embodiment also provides a storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method in any optional implementation of the above embodiment is executed. Wherein, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0088] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0089] As can be seen from the above, this method ensures the timeliness and authenticity of the data and enhances the accuracy and reliability of the entire detection system. The abstract temperature data is visualized and converted into intuitive image forms to facilitate visual analysis and deep learning algorithm processing. The readability of the data is enhanced and the model's ability to recognize temperature distribution patterns is improved. The level of intelligent detection is improved and the reliance on manual intervention is reduced. The detection efficiency is greatly improved, and the system can accurately determine whether there is a temperature anomaly and detect potential overheating problems in a timely manner. The fault response time is shortened, and false alarms and missed alarms are reduced, ensuring the safe and stable operation of the power system and reducing maintenance costs.

[0090] Example 2

[0091] Reference Figure 2 , which is the second embodiment of the present invention, and this embodiment provides an insulator abnormality detection method. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0092] Table 1 Technical characteristics comparison table

[0093]

[0094] As can be seen from the above, this method ensures the timeliness and authenticity of the data and enhances the accuracy and reliability of the entire detection system. The abstract temperature data is visualized and converted into intuitive image forms to facilitate visual analysis and deep learning algorithm processing. The readability of the data is enhanced and the model's ability to recognize temperature distribution patterns is improved. The level of intelligent detection is improved and the reliance on manual intervention is reduced. The detection efficiency is greatly improved, and the system can accurately determine whether there is a temperature anomaly and detect potential overheating problems in a timely manner. The fault response time is shortened, and false alarms and missed alarms are reduced, ensuring the safe and stable operation of the power system and reducing maintenance costs.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for detecting abnormality of an insulator, characterized in that: include, Set the insulator abnormality detection time interval, arrange sensors to monitor the original temperature data of the target insulator string, and process it; Grayscale mapping is performed on the preprocessed raw temperature data to convert the raw temperature data into grayscale images, and a deep learning model is constructed and trained at the same time; Use the trained deep learning model to identify and output the insulator string area in the image, and determine whether the insulator temperature is abnormal based on the temperature data of the insulator string area.

2. The insulator abnormality detection method according to claim 1, characterized in that: The processing comprises the following steps: Clean the collected raw temperature data, remove outliers, and check if there are obvious errors or outliers in the data and remove them; For the missing parts in the data, fill them by interpolation, convert the data into data with zero mean and unit variance, and use Z-score standardization. The calculation formula is: Among them, T i is the normalized original temperature data, X is the original temperature data, μ is the mean value of the original temperature data, and σ1 is the standard deviation of the original temperature data; The images are resized and randomly cropped to simulate different observation conditions and viewing angles, expand the data set, rotate, mirror or crop the images to generate more images, and the multiple grayscale images obtained are combined into an insulator string database.

3. The insulator abnormality detection method according to claim 2, characterized in that: The grayscale mapping of the pre-processed original temperature data comprises the following steps: Determine the range of temperature data, that is, the minimum temperature value T min and the maximum temperature T max , map the temperature data, and use the mapping formula to convert the temperature value into a grayscale value. The calculation formula is: Among them, G is the gray value, T is the value of the original temperature data; If the grayscale value after mapping exceeds 255, crop or adjust the mapping to ensure that all grayscale values ​​are within the valid range; After converting each temperature value into a grayscale value, the grayscale value is placed in a matrix, where each matrix element corresponds to a pixel in the image, and its grayscale value is the mapped temperature value; The matrix is ​​saved as an image file, and the grayscale image displays the pixel grid and coordinate axis. The coordinates of the insulator area are manually calibrated, the insulator area is divided, and a label library of the insulator string image is established to form a data set.

4. The insulator abnormality detection method according to claim 3, characterized in that: The step of establishing a label library of insulator string images and forming a data set comprises the following steps: The regional coordinates of the insulator string are recorded according to the grayscale image as the label of the corresponding image. For each detected insulator string contour, its minimum circumscribed rectangle or contour coordinates are calculated; Format the extracted insulator string area coordinates into the label format required for model training, and encode each pixel into the corresponding category label; Match the produced labels with the corresponding grayscale images to ensure that each image has a corresponding label; Divide the labeled image dataset into training set, validation set and test set, randomly check the images and labels in the dataset to ensure that the labels correspond to the image content, and save the processed dataset in a format suitable for model training.

5. The insulator abnormality detection method according to claim 4, characterized in that: The deep learning model construction and training includes using a pre-trained MobileNet model as a base model and loading weights pre-trained on the ImageNet large-scale dataset to optimize the model; Remove the original fully connected layer at the top of the MobileNet model, add a global average pooling layer and a fully connected layer, freeze the first three convolutional blocks of the MobileNet model, and keep the subsequent convolutional blocks; The loss function of the model is set to DIoU Loss, which measures the degree of overlap between the bounding box predicted by the model and the true bounding box and the center distance of the two boxes when the degree of overlap is the same. The calculation formula is: Among them, IoU is the intersection over union ratio between the predicted box and the labeled box, the predicted box is P, and the labeled box is G; R(P,G) is the penalty term; ρ(P,G) represents the distance between the center of the predicted box and the labeled box, p and g are the center points of the two boxes, and c represents the diagonal length of the minimum enclosing rectangle of the two boxes; The insulator string dataset is divided into training set, validation set and test set with a ratio of 7:2:

1. The training set is input into the model and the DIoU Loss of the model is calculated. Use the back-propagation algorithm to calculate the gradient of the loss function with respect to the model parameters, and use the Adam optimizer to update the parameters; The initial learning rate is 0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-8, amsgrad=False, eight_decay=0.001; Regularly evaluate the model on the validation set to determine whether the model begins to overfit and whether to stop training early or adjust the model structure; Use the test set to evaluate the trained model, use the actual labels of the test set to compare with the model's prediction results, and calculate the model loss function to verify the model; The sensor temperature data to be detected is preprocessed and converted into a grayscale image and input into the deep learning model to obtain the model output insulator string area bounding box coordinates and confidence score.

6. The insulator abnormality detection method according to claim 5, characterized in that: The method of judging whether the insulator has abnormal temperature according to the temperature data of the insulator string area comprises the following steps: According to the bounding box coordinates output by the deep learning model, the coordinate range of the insulator string area is determined. Assuming that the bounding box coordinates of the insulator string area are described by the coordinates of its lower left corner (x1, y1) and upper right corner (x2, y2), the normalization factor is calculated considering the maximum and minimum values ​​of the temperature. The calculation formula is: Where Z is the normalization factor, T max is the maximum temperature, T min is the minimum value of temperature; Within the coordinate range of the insulator string area, the distribution characteristics of the temperature data are analyzed and the weighted average temperature is calculated. The calculation formula is: In the formula, is the weighted average temperature, T ij is the temperature value at the coordinate (i, j), and all temperature values ​​in the area are traversed to obtain the highest temperature value T max ; Calculate the standard deviation of the temperature and the maximum value of the temperature curvature within the coordinate range of the insulator string area, and identify the temperature mutation point. The calculation formula is: Where σ2 is the standard deviation of the temperature within the coordinate range of the insulator string area, N is the number of pixels in the bounding box, and C max is the maximum value of temperature curvature; The temperature difference limit is calculated comprehensively, and whether the temperature of the insulator string is abnormal is determined based on the calculated temperature difference limit.

7. The insulator abnormality detection method according to claim 6, characterized in that: The comprehensive calculation of the temperature difference limit value includes comprehensively calculating the temperature difference limit value of the insulator string according to the obtained standard deviation of the temperature within the coordinate range of the insulator string area and the maximum value of the temperature curvature, and the calculation formula is: Where, ΔT odd is the temperature difference limit of the insulator string, k1 is the adjustment factor, C th is the starting point of the curvature influence; If the temperature difference in the insulator string area If the temperature deviates from the normal range, an abnormal temperature alarm is triggered, and there is an overheating or safety hazard, an immediate inspection is carried out and the monitoring frequency of the area is increased to real-time monitoring; If the temperature difference in the insulator string area It is determined that the insulator temperature is within the normal range and normal monitoring continues without intervention.

8. An insulator abnormality detection system, based on the insulator abnormality detection method according to any one of claims 1 to 7, characterized in that: include, A collection module is used to set the insulator abnormality detection time interval, arrange sensors to monitor the original temperature data of the target insulator string, and process it; The mapping module is used to perform grayscale mapping on the preprocessed raw temperature data, convert the raw temperature data into grayscale images, and build and train a deep learning model at the same time; The detection module is used to use the trained deep learning model to identify and output the insulator string area in the image, and determine whether the insulator temperature is abnormal based on the temperature data of the insulator string area.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the insulator abnormality detection method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the insulator abnormality detection method according to any one of claims 1 to 7 are implemented.

Images