Composite insulator detection method based on image fusion analysis method
Through image fusion analysis method and deep learning technology, infrared and ultraviolet detection correction models are established, which solves the problem of factor interference in composite insulator detection, and accurately captures the heating and discharge characteristics of composite insulators and judges the defect level, which improves the detection Accuracy and reliability.
Patent Information
- Application Number
- CN202411058349.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-19
AI Technical Summary
Existing infrared and ultraviolet detection technologies are interfered by a variety of factors in the judgment of composite insulator temperature rise and discharge state, resulting in measurement errors and misjudgment, making it difficult to accurately judge insulator defects, affecting the stability and safety of transmission lines.
Using the detection method based on image fusion analysis method, the infrared and ultraviolet detection correction model is established, combined with deep learning and artificial intelligence, and analyzing environmental factors and instrument parameters, the accurate capture of the heating and discharge characteristics of composite insulators and the judgment of defect level are achieved.
It significantly improves the accuracy of composite insulator detection, reduces misjudgment and misjudgment, enhances the reliability of insulator state evaluation, and ensures the safety and reliability of transmission lines.
Smart Images

Figure CN118980718A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of power systems, and in particular relates to a composite insulator detection method based on an image fusion analysis method. Background Art
[0002] As an important component of overhead transmission lines, composite insulators are subjected to the combined effects of electrical and mechanical forces. Affected by factors such as humid environments, acid rain and mist, and corona discharge, the core rods inside the insulators are very likely to produce local discharge heating inside the sheath due to acid corrosion and electrical corrosion. This will accelerate the defects of the core rods and cause them to decay, which in turn leads to a decrease in the mechanical strength of the insulators. Under harsh working conditions, string breakage and line drop accidents may also occur, resulting in inestimable consequences. Accurately judging the temperature rise and discharge state of the composite insulators in operation and judging whether they have defects is of great significance for maintaining the stable operation of transmission lines and improving the reliability and safety of transmission lines. With the widespread application of drones in intelligent inspections, it has become normal to use drones equipped with infrared and ultraviolet detection equipment to conduct line inspections. However, this type of non-destructive testing technology is greatly affected by factors such as shooting distance, viewing angle, and the surrounding environment on site, and measurement errors are difficult to control.
[0003] In infrared temperature measurement, the shooting distance and viewing angle of the drone, as well as environmental factors such as light intensity, temperature and humidity have a great impact on the accuracy of the temperature rise measurement results of composite insulators. When measuring temperature, drones outside the line of sight usually cannot directly obtain the precise distance and viewing angle between them and the composite insulators. In addition, due to the influence of the zoom function of the infrared temperature measurement lens, it is even more difficult to estimate the temperature measurement distance and viewing angle, and it is difficult to judge whether the infrared temperature measurement picture is qualified. At the same time, composite insulators are usually long in string, and the actual distance between the drone and each section of the composite insulator is not equal during actual temperature measurement. However, in the infrared temperature measurement software, there is only one distance value representing the distance between the drone and the entire string of insulators. When analyzing the data, the distance between the drone and the far end of the composite insulator will be reduced, making the temperature measurement result at the low voltage end larger, and may lead to deviations in the temperature rise calculation and misjudgment of the defect level.
[0004] For partial discharge of external insulation of composite insulators, UV imaging is usually used. This method does not require power outage and is safe, intuitive, and fast. It is also simple to operate and does not require direct contact with the equipment. Due to different manufacturers, although the basic imaging principle of UV imagers is the same, their instrument design, internal structure, optical devices and other hardware are not completely consistent, and the software such as image processing technology and data statistical methods are also different. There are definitely differences in the detection results of UV imagers produced by different manufacturers. Therefore, in order to better realize the UV detection of composite insulator discharge, it is necessary to calibrate the UV imaging detection results. Existing research and DL / T 345-2010 standard only stipulate the wind speed suitable for detection, but do not analyze the specific conditions that affect the UV detection results such as detection distance and angle, instrument gain, temperature, humidity and wind speed. The calibration of different types of UV imagers is also a gap that needs to be filled and a research focus.
[0005] Infrared thermal imagers detect thermal radiation and current-induced heating type defects caused by electrical faults, while UV imagers detect UV radiation and corona discharge type defects caused by abnormal voltage intensity. The two optical detectors are suitable for different types of insulator defects. Infrared detection technology and UV detection technology each have advantages and disadvantages.
[0006] The main problems with current infrared and ultraviolet detection are as follows: (1) Infrared temperature measurement is affected by multiple factors, including distance and angle, temperature, humidity, and light intensity. Distance and angle interfere with the dispersion and attenuation of infrared radiation; temperature and humidity affect the propagation and reception of radiation in the infrared thermal imager; and light intensity causes uneven distribution of insulator surface temperature. (2) Ultraviolet imaging detection is also affected by multiple factors, including distance and angle, temperature, humidity, wind speed, and ultraviolet imager gain. Changes in distance directly affect the intensity of ultraviolet radiation, and angle adjustments can cause ultraviolet radiation to refract, resulting in a smaller spot area for detection. Temperature and humidity can affect the state of the insulator, causing deviations in the spot area during measurement. Wind speed can cause the flow of suspended particles in the air around the insulator, which may interfere with the propagation of ultraviolet radiation. The gain setting of the ultraviolet imager can affect the noise level, thereby affecting the accuracy of the results. (3) Since the equipment is in operation during detection, the impact of factors such as line load on the detection results must be considered; (4) Analysis and diagnosis of the detection results is also an important link. How to intelligently analyze data in batches, eliminate the influence of various factors, and obtain accurate detection reports is an inevitable trend in the development of intelligent inspection. Summary of the invention
[0007] In order to address the shortcomings of the prior art, the present invention aims to provide a composite insulator detection method based on image fusion analysis, which reduces the interference of factors such as the surrounding environment, detection distance and viewing angle on the temperature rise and discharge detection results by analyzing the heating and discharge characteristics of defective composite insulators under different factors, and significantly improves the accuracy of composite insulator detection.
[0008] In order to achieve the above object, the technical solution adopted by the present invention is:
[0009] A composite insulator detection method based on image fusion analysis method, the method specifically comprises the following steps:
[0010] 1) Establishment of infrared temperature measurement correction model for composite insulators: Use Python and TensorFlow framework to train a neural network model, and organize the test experiment data into a data set; after the training is completed, establish a correction model with the temperature rise measurement value of the infrared thermal imager, the detection distance and viewing angle, and the light intensity as input and the corrected temperature rise as output;
[0011] 2) Use the monocular ranging method to perform error analysis and verification during dynamic ranging;
[0012] 3) Establishment of UV detection correction model for composite insulators: Calculate relevant correction coefficients based on the collected sample data and establish a UV detection correction model;
[0013] 4) Insulator fault detection based on infrared and ultraviolet "dual light" images and artificial intelligence algorithms;
[0014] 5) Conduct defect diagnosis on composite insulators.
[0015] Wherein, step 1) specifically includes the following contents:
[0016] 1.1) Data preparation: Conduct infrared temperature rise and ultraviolet discharge measurement tests on composite insulators and collect experimental data, including the shooting angle, shooting distance, light intensity, infrared thermal imager temperature measurement value, and thermocouple sensor temperature measurement value recorded during the experiment; and preprocess the data, including cleaning the data, processing outliers and missing values; then organize the experimental data into a data set containing features and target variables;
[0017] 1.2) Data partitioning: Divide the data set into training set and test set;
[0018] 1.3) Construct a neural network model: select the corresponding neural network structure and define the hyperparameters including the number of neural network layers, number of neurons, and activation function;
[0019] 1.4) Training model: Use the training set to train the neural network model, adjust the weights and biases of the model through the back propagation algorithm and optimizer, so that the model fits the training data; select an appropriate loss function to measure the error of the model, and use the optimizer to minimize the loss function;
[0020] 1.5) Model evaluation: Use the test set to evaluate the performance of the trained model and calculate the prediction error of the model on the test set, including mean square error, root mean square error, and mean absolute error.
[0021] In step 1.2), the training set is set to 80% of the total data, and the remaining data is used as the test set.
[0022] In step 2), the error compensation for attitude solution is to adopt a combined measurement method, and convert the estimation of attitude angle into the estimation of attitude error angle through multi-sensor information fusion technology, and eliminate it;
[0023] For attitude measurement, Kalman filtering technology is used to fuse the information of the auxiliary attitude measurement system composed of the UAV accelerometer and magnetometer with the gyro attitude measurement system to obtain the optimal attitude estimation value.
[0024] Step 3) specifically includes the following steps:
[0025] 3.1) Collect sample data, including spot area and temperature, humidity, air pressure and wind speed related to environmental factors;
[0026] 3.2) Calculate the wind speed correction coefficient K based on the collected sample data W , Angle correction coefficient K T , gain correction factor K U , distance correction factor K d , Temperature correction coefficient K M and humidity correction factor K N ; Use various correction coefficients to establish a comprehensive correction model, take environmental factors as input, perform correction calculation on the original spot area, and obtain the spot area value under uniform environmental conditions;
[0027] The normalized correction model formula is:
[0028] S=(K W K T K U K d K N K M )S 0
[0029] Among them, S is the spot area under a certain condition after correction, S 0It is the spot area under a certain condition in actual detection. By correcting the model, the influence of environmental factors on the spot area is eliminated, so as to achieve data comparability under different conditions, making the detection of the spot area under different environmental conditions more accurate and reliable.
[0030] The fault detection in step 4) specifically includes:
[0031] 4.1) Select the characteristic quantity for composite insulator defect diagnosis:
[0032] First, the temperature and spot area detected by infrared and ultraviolet light are taken as candidate attributes. The temperature rise data is read and converted into the actual temperature rise value according to the detection distance, angle and light intensity. The actual spot area is obtained according to the ultraviolet detection correction model, and the spot area is uniformly normalized. Then, the data is sorted according to the values corresponding to these attributes and classified by different thresholds.
[0033] 4.2) Classification of composite insulator defect levels:
[0034] According to relevant technical specifications, and based on temperature rise and UV spot area, the defect grade standards of insulators are divided; defect grades are divided into "serious defect", "general defect", "minor defect" and "good"; considering the temperature rise and UV spot area conditions, the defect grade is determined according to the preset defect grade; when multiple conditions deviate from the normal range at the same time, the degree of defect grade is further comprehensively analyzed and determined to ensure the accuracy and reliability of the determination result;
[0035] 4.3) Generate composite insulator defect decision tree model:
[0036] By extracting the characteristic parameters of the original data for preliminary analysis, the input data dimension of the composite insulator is reduced; the classifier construction stage can be divided into two processes: training and evaluation. The classifier is trained using training data and evaluated using test data. The classification algorithm rules are modified through training data to make the classification results consistent with the known data categories; in the use stage, the characteristic parameters of unknown data are extracted and classified as classifier input data; in the process of decision tree construction, the preset influencing variables are temperature rise, ultraviolet spot area and its influencing factors;
[0037] Due to the complex data structure and incompleteness of some data entries, the CART algorithm is used to process mixed data using the Gini coefficient (gini ratio) as a measure of information gain. The data is then substituted into the Python algorithm environment and a preliminary decision tree is obtained through machine learning. In order to prevent overfitting, the decision tree will be pruned.
[0038] The basis for defect grade classification in step 4.2) is:
[0039] According to the existing benchmark values and standard ranges, the temperature rise and UV spot area parameters are compared with the normal state; if the temperature rise is much higher than the normal range and the UV spot area shows an expanding trend, it indicates a serious defect; if the temperature rise is moderate and the UV spot area increases slightly, it is a general defect; when the temperature rise is low, but the UV spot area does not change significantly, it is a minor defect; if the temperature rise and UV spot area are both within the normal range, it is judged to be in good condition.
[0040] Step 5) specifically includes the following contents:
[0041] 5.1) After inputting the corresponding insulator model, infrared imaging, ultraviolet imaging, and measured environmental information, decision tree diagnosis can be performed; by transferring the input results to the trained decision tree file, after diagnosis, the decision tree returns the diagnosis result, which is displayed in text form;
[0042] 5.2) Composite insulator defects found during inspection shall be included in the scope of equipment defect management system, handled in accordance with the equipment defect management process and corresponding treatment plans shall be given.
[0043] The treatment scheme for the three defect levels in step 5.2) is:
[0044] Minor defects: There are abnormal corona discharges and a certain gradient in the temperature field, which will affect the equipment but will not cause accidents. It is required to be recorded and the development of the defects should be observed;
[0045] General defects: If the corona discharge is abnormally prominent, the equipment is overheated, and the temperature difference is large, the inspection cycle should be shortened and the power outage maintenance opportunity should be used to arrange maintenance in a planned manner to eliminate the defects;
[0046] Serious defects: The corona discharge of the composite insulator is serious, and the maximum temperature exceeds the specified maximum allowable temperature. A power outage should be arranged immediately. If one of the following situations occurs, the insulator is considered to have failed and should be replaced: ① The umbrella cover is brittle, powdered or cracked; ② The umbrella cover has leakage marks and corrosion, and the cumulative length is greater than 10% of the creepage distance of the insulator or the corrosion depth is greater than 30% of the material thickness at the location; ③ The sealing of the connecting parts of the insulator fails, and debonding, cracks and slippage occur; ④ The insulator is burned by the arc on the surface of the umbrella skirt after flashover.
[0047] The beneficial effects of the present invention are:
[0048] 1. The present invention adopts a comprehensive data correction model, combines multiple influencing factors such as environmental conditions, instrument parameters, distance and angle, and establishes a correction model through experimental data to effectively correct infrared temperature measurement and ultraviolet discharge detection results, thereby improving the accuracy of insulator status judgment.
[0049] 2. The present invention applies deep learning to the front-end flight system of the UAV to achieve accurate measurement of the detection distance and viewing angle, solving the problem of difficulty in obtaining accurate distance and viewing angle during UAV inspection of composite insulators.
[0050] 3. The present invention achieves multi-angle and multi-parameter data fusion by fusing and analyzing the data of infrared thermal imager and ultraviolet imager, thereby accurately capturing the heating and discharge characteristics of composite insulators, and using this as a basis for judging the defect level of composite insulators. The influence of multiple factors such as distance, angle, temperature, humidity, etc. is taken into account, making the non-destructive testing of insulators more comprehensive and accurate. Through the design of an intelligent diagnostic system, the infrared and ultraviolet detection results are comprehensively used for defect diagnosis, which improves the accuracy of insulator status assessment and reduces misjudgment and missed judgment. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] Figure 1 Constructing a temperature correction model diagram for the present invention;
[0052] Figure 2 It is a normalized correction model diagram of ultraviolet imaging quantitative parameters under the influence of multiple factors of the present invention;
[0053] Figure 3 Establishing a flow chart for the composite insulator defect decision tree model of the present invention;
[0054] Figure 4 This is the core flow chart of the decision tree analysis system of the present invention. DETAILED DESCRIPTION
[0055] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not intended to limit the scope of use of the present invention.
[0056] like Figure 1-4 As shown, the present invention proposes a composite insulator detection method based on image fusion analysis method, which specifically includes the following steps:
[0057] Step 1) Establishment of infrared temperature measurement correction model for composite insulators:
[0058] In order to consider the correlation between the temperature measurement value of the infrared thermal imager and the shooting distance, shooting angle of view, and light intensity, a neural network model was trained using Python and TensorFlow framework, and the data of the test experiment was organized into a data set, which includes the detection distance and angle of view, light intensity, infrared temperature rise measurement value and thermocouple temperature rise measurement value (i.e. actual temperature rise); after the training is completed, a correction model is established with the infrared thermal imager temperature rise measurement value, detection distance and angle of view, and light intensity as input and the corrected temperature rise as output. Specifically, it includes:
[0059] 1.1) Data preparation: Conduct infrared temperature rise and ultraviolet discharge measurement tests on composite insulators and collect experimental data, including the shooting angle, shooting distance, light intensity, infrared thermal imager temperature measurement value, and thermocouple sensor temperature measurement value recorded during the experiment; and preprocess the data, including cleaning the data, processing outliers and missing values, etc.; then organize the experimental data into a data set containing features and target variables.
[0060] This application uses infrared thermal imagers, ultraviolet imagers, thermocouple temperature sensors, etc. to conduct infrared and discharge tests on normal and typical defective composite insulators in AC withstand voltage tests, and obtains parameters such as temperature rise at the high-voltage end of the insulator relative to the low-voltage end, ultraviolet spot area, etc. under different detection instrument parameters, environmental conditions, distance between the detection instrument and the detection target, and observation angle. By adjusting the built-in parameters of the infrared thermal imager and ultraviolet imager, the fusion display and analysis of ultraviolet images and visible light images can be achieved.
[0061] 1.2) Data division: The data set is divided into a training set and a test set; the training set is set to 80% of the total data volume, and the remaining data is used as the test set.
[0062] 1.3) Build a neural network model: Select a suitable neural network structure, such as multi-layer perceptron (MLP), convolutional neural network (CNN) or recurrent neural network (RNN), etc. And define the hyperparameters of the neural network, such as the number of layers, number of neurons, activation function, etc.
[0063] 1.4) Training model: Use the training set to train the neural network model, adjust the weights and biases of the model through the back-propagation algorithm and optimizer, so that the model can fit the training data; select an appropriate loss function to measure the error of the model, and use the optimizer to minimize the loss function.
[0064] 1.5) Model evaluation: Use the test set to evaluate the performance of the trained model and calculate the prediction error of the model on the test set, including mean square error, root mean square error, mean absolute error, etc.
[0065] The present invention calibrates the composite insulator data set of infrared experimental temperature measurement, prepares a data set, and uses a YOLOv5 deep learning algorithm for deep learning training, so that the trained algorithm can accurately identify the composite insulator under the infrared temperature measurement background. On this basis, a monocular ranging model is established according to the structural characteristics of the composite insulator, and the distance and angle between the drone and each section of the composite insulator are calculated in combination with the characteristic dimensions of the composite insulator identified by the deep learning algorithm.
[0066] The correlation between the temperature measured by the infrared thermal imager and the detection distance and viewing angle, light intensity, temperature and humidity was quantitatively analyzed by principal component analysis and grey correlation analysis. A correction model was established with the temperature measured by the infrared thermal imager, detection distance and viewing angle, light intensity, temperature and humidity as input and the temperature measured by the thermistor temperature sensor as output.
[0067] Step 2) Error analysis and verification of the monocular ranging method during dynamic ranging.
[0068] In visual measurement, in addition to the need to determine the camera's internal parameters, the camera's external parameters also have a great impact on the measurement accuracy, mainly manifested in the camera's attitude angle. In distance measurement based on monocular vision, a small deviation in the camera's attitude angle will cause a large measurement error at a long distance. The camera's attitude angle includes pitch angle, yaw angle, and roll angle. During the shooting process, it is impossible for the camera to be completely on a completely horizontal plane with the object being photographed, so it is necessary to consider the error analysis and verification of the camera's various shooting attitudes.
[0069] The main function of the UAV heading attitude measurement system is to measure the pitch angle, yaw angle and roll angle of the UAV carrier. The attitude measurement system is the premise for the UAV to achieve attitude control. On the one hand, it provides three-dimensional attitude information for the UAV flight control system, and on the other hand, it also provides a three-dimensional attitude reference for airborne stabilization equipment such as camera gimbals. The present invention mainly analyzes the principle scheme of UAV attitude measurement and the related errors of the sensors used, and provides a theoretical feasibility demonstration for the realization of high performance, miniaturization, low cost and low power consumption of the UAV attitude measurement system.
[0070] The present invention takes into account the camera attitude error in the flight scene. If a high-precision gyroscope is used to compensate for the camera attitude, a simple micromechanical gyroscope will have a large time accumulation error after long-term operation in calculating the attitude.
[0071] The error compensation of attitude solution in the present invention adopts the idea of combined measurement, converts the estimation of attitude angle into the estimation of attitude error angle through multi-sensor information fusion technology, and eliminates it.
[0072] Commonly used information fusion and filtering methods include comprehensive averaging, neural network, Wiener filtering, complementary filtering and Kalman filtering. The above methods are not suitable for attitude measurement of small UAVs. Therefore, the present invention adopts Kalman filtering technology for attitude measurement, and fuses the information of the auxiliary attitude measurement system composed of the UAV accelerometer and magnetometer with the gyro attitude measurement system to obtain the optimal attitude estimation value.
[0073] Step 3) Establishment of UV detection correction model for composite insulators: Calculate relevant correction coefficients based on the collected sample data and establish a UV detection correction model.
[0074] In order to ensure accurate and comparable UV spot area data under different environmental conditions, the present invention aims to establish a spot area correction model under multiple factors. Ultraviolet imaging technology has important applications in corona discharge detection. However, the influence of factors such as distance, angle, wind speed and UV imager gain on the spot area will affect the spot area detection results. Therefore, the correction coefficient method is adopted to establish a composite insulator UV detection spot area correction model through systematic experimental research and data analysis. Specifically:
[0075] 3.1) Collect sample data, including spot area and data related to environmental factors such as temperature, humidity, air pressure and wind speed. Specifically, the experiment can be carried out in the laboratory using a corona generator and a UV imager to record the spot area. Change the distance, angle, wind speed, UV imager gain, temperature and humidity and other factors to obtain the curve of the spot area changing with these factors and obtain sample data.
[0076] 3.2) According to the collected sample data, calculate the wind speed correction coefficient (K W ), angle correction factor (K T ), gain correction factor (K U ), distance correction factor (K d ), temperature correction factor (K M ) and humidity correction factor (K N ). Use each correction coefficient to establish a comprehensive correction model, take environmental factors as input, correct the original spot area and calculate the spot area value under uniform environmental conditions. The following formula is the normalized correction model formula, S is the spot area under a certain condition after correction, S 0 It is the spot area under a certain condition in actual detection.
[0077] S=(K W K T K U K d K N K M )S 0
[0078] By correcting the model, the influence of environmental factors on the spot area can be eliminated, thereby achieving data comparability under different conditions, making the detection of the spot area under different environmental conditions more accurate and reliable. This will provide a more reliable basis for the analysis of corona discharge and the determination of the defect level of insulators, and will also help decision-making and problem solving in engineering practice.
[0079] The spot area of the present invention is used as a discharge quantitative parameter of the ultraviolet imager detection equipment discharge, and is an important indicator for detecting discharge. The ultraviolet imaging method for detecting discharge is affected by the inherent characteristics of the ultraviolet imager itself (such as instrument gain), environmental conditions (wind speed, temperature, humidity), the distance between the detection instrument and the detection target, and the observation angle. By collecting sample data of the spot area under the influence of factors such as temperature, humidity, wind speed, angle, ultraviolet imager gain, and detection distance, the correction coefficient is calculated using a mathematical fitting method. A multi-factor correction model is established to correct the spot area under different environmental conditions, and the corrected data is subjected to error analysis to verify its effectiveness.
[0080] Step 4) Insulator fault detection based on infrared and ultraviolet "dual light" images and artificial intelligence algorithms. Including:
[0081] 4.1) Select the characteristic quantity for composite insulator defect diagnosis:
[0082] Composite insulator defects are mainly reflected by three characteristics: temperature rise and ultraviolet spot area. These three characteristics are affected by temperature, air pressure, voltage level, ambient humidity, and wind speed. In order to apply decision trees to insulator defect diagnosis, the temperature and spot area detected by infrared and ultraviolet light are first used as candidate attributes. Directly read the temperature rise data in the infrared image and convert it into the actual temperature rise value according to the detection distance, angle, and light intensity. According to the actual spot area obtained by the ultraviolet detection correction model, the spot area is uniformly normalized. Then sort the data according to the values corresponding to these attributes and classify them by different thresholds. In the process of building the decision tree, the preset influencing variables are temperature rise, ultraviolet spot area, and their influencing factors.
[0083] 4.2) Classification of composite insulator defect levels:
[0084] According to the methods and standards in DL / T664-2016 "Application Specifications for Infrared Diagnosis of Live Equipment" and "Technical Specifications for Live Detection Instruments for Power Equipment - Part 3: Technical Specifications for Ultraviolet Imagers", the defect grade standards for insulators are divided based on conditions such as temperature rise and ultraviolet spot area. The defect grades are divided into "serious defects", "general defects", "minor defects" and "good".
[0085] According to the existing benchmark values and standard ranges, compare parameters such as temperature rise and UV spot area with the normal state. If the temperature rise is significantly higher than the normal range and the UV spot area shows an expanding trend, it may indicate a serious defect. If the temperature rise is moderate and the UV spot area increases slightly, it may be a general defect. In the case of a low temperature rise, but the UV spot area does not change significantly, it may be a minor defect. If the temperature rise and UV spot area are both within the normal range, it can be judged to be in good condition.
[0086] The three conditions of temperature rise and ultraviolet spot area are comprehensively considered and judged according to the preset defect level. When multiple conditions deviate from the normal range at the same time, the degree of defect level can be further comprehensively analyzed to ensure the accuracy and reliability of the judgment result. Through the above method, combined with key parameters such as temperature rise and ultraviolet spot area, the defect level of composite insulators is reasonably divided, providing a scientific basis for equipment maintenance and management.
[0087] 4.3) Generate composite insulator defect decision tree model:
[0088] By extracting the characteristic parameters of the original data for preliminary analysis, the input data dimension of the composite insulator is reduced. The classifier construction stage can be divided into two processes: training and evaluation. The classifier is trained using training data and evaluated using test data. The classification algorithm rules are modified through training data to make the classification results consistent with the known data categories. In the use stage, the characteristic parameters of unknown data are extracted and classified as classifier input data.
[0089] Combined with the defect map database established by test data, the composite insulator defect diagnosis algorithm and judgment criteria are studied. Threshold division or intelligent algorithm with low computational time cost is used to determine the equipment defect level. According to temperature rise conditions, ultraviolet quantitative parameters and environmental factors, the insulator defect level is determined.
[0090] Due to the complex data structure and incomplete data entries, the CART algorithm is used to process mixed data using the Gini coefficient ratio as a measure of information gain. The data is substituted into the Python algorithm environment and a preliminary decision tree is obtained through machine learning. In order to prevent overfitting, the decision tree will be pruned.
[0091] The present invention combines the latest relevant specifications and standards for infrared temperature detection and ultraviolet discharge detection of live equipment to formulate a composite insulator defect grade judgment criterion that comprehensively considers temperature and discharge characteristics. The infrared temperature measurement and ultraviolet discharge correction model are used to obtain the precise values of temperature rise and discharge spot area in the infrared and ultraviolet "double light" image. The decision tree diagnosis method based on the classification and regression tree algorithm is used, and the test data is used as the training sample to establish a composite insulator defect diagnosis model, which realizes the evaluation and diagnosis of defect types.
[0092] Step 5) diagnose defects on the composite insulator.
[0093] 5.1) After inputting the corresponding insulator model, infrared imaging, ultraviolet imaging, and measured environmental information, the decision tree diagnosis can be performed. The input result is transferred to the trained decision tree .pickle file through the .get() statement. After the diagnosis, the decision tree will return the diagnosis result, which will be displayed in text form.
[0094] 5.2) Composite insulator defects found during testing should be included in the scope of the equipment defect management system and handled in accordance with the equipment defect management process. The treatment solutions for the three defect levels are:
[0095] Minor defects: There are abnormal corona discharges and a certain temperature gradient, which will affect the equipment but will not cause accidents. They should be recorded and the development of the defects should be observed.
[0096] General defects: If the corona discharge is abnormally prominent, the equipment is overheated, and the temperature difference is large, the inspection cycle should be shortened and the power outage maintenance opportunity should be used to arrange maintenance in a planned manner to eliminate the defects.
[0097] Serious defects: The corona discharge of the composite insulator is serious, and the maximum temperature exceeds the specified maximum allowable temperature. The power outage should be arranged immediately. If one of the following situations occurs, the insulator can be considered to be failed and should be replaced: ① The umbrella cover is embrittled, powdered or cracked; ② The umbrella cover has leakage marks and corrosion, and the cumulative length is greater than 10% of the creepage distance of the insulator or the corrosion depth is greater than 30% of the material thickness at the location; ③ The sealing of the connecting parts of the insulator fails, debonding, cracks and slippage occur; ④ The insulator is burned by the arc on the surface of the umbrella skirt after flashover.
[0098] The present invention inputs the corresponding insulator model, infrared imaging map, ultraviolet imaging map, and on-site measured environmental information, obtains the temperature rise of the composite insulator and the discharge spot area according to the infrared temperature measurement model, compares the relevant parameters of typical defects in the defect library, and realizes automatic identification of defect types and automatic production of detailed treatment plans.
[0099] The existing live nondestructive testing methods for overhead transmission line insulators mainly control the flight of drones through visible light cameras, determine the shooting position, and carry infrared or ultraviolet equipment for testing. Infrared thermal imagers detect thermal radiation and current-induced thermal defects caused by electrical faults, while ultraviolet imagers detect ultraviolet radiation and corona discharge defects caused by abnormal voltage intensity. The two optical detectors are suitable for different types of insulator defects. According to EPRI statistics, infrared testing can only determine 28% of insulator defects, while ultraviolet testing can determine 50% of insulator defects. If infrared testing and ultraviolet testing are performed in series, problems such as low detection accuracy and long detection cycle will occur.
[0100] In order to improve the accuracy and efficiency of detection, the present invention applies deep learning technology, image processing technology, and monocular ranging technology to combine the visible light image of the insulator with the infrared temperature image and the ultraviolet discharge image. This comprehensive analysis method effectively reduces the interference of various factors such as the environment, acquisition angle, and detection distance on the infrared and ultraviolet detection results, and also fully considers the heating and discharge characteristics of the insulator under different factors. Finally, a decision tree diagnosis model is established, and a typical defect diagnosis system for composite insulators is designed. Comprehensive, accurate, and efficient perception of composite insulator defect level determination is achieved, which improves the intelligent operation and maintenance level of transmission lines and ensures the reliability of transmission systems. It not only improves the accuracy of detection, but also improves the detection efficiency, providing an advanced solution for insulator inspection in the power industry.
[0101] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.
Claims
1. A composite insulator detection method based on image fusion analysis, characterized in that: The method specifically comprises the following steps: 1) Establishment of infrared temperature measurement correction model for composite insulators: Use Python and TensorFlow framework to train a neural network model, and organize the test experiment data into a data set; after the training is completed, establish a correction model with the temperature rise measurement value of the infrared thermal imager, the detection distance and viewing angle, and the light intensity as input and the corrected temperature rise as output; 2) Use the monocular ranging method to perform error analysis and verification during dynamic ranging; 3) Establishment of UV detection correction model for composite insulators: Calculate relevant correction coefficients based on the collected sample data and establish a UV detection correction model; 4) Insulator fault detection based on "dual light" images of infrared and ultraviolet and artificial intelligence algorithms; 5) Conduct defect diagnosis on composite insulators.
2. The composite insulator detection method based on image fusion analysis according to claim 1 is characterized in that: Step 1) specifically includes the following contents: 1.1) Data preparation: Conduct infrared temperature rise and ultraviolet discharge measurement tests on composite insulators and collect experimental data, including the shooting angle, shooting distance, light intensity, infrared thermal imager temperature measurement value, and thermocouple sensor temperature measurement value recorded during the experiment; and preprocess the data, including cleaning the data, processing outliers and missing values; then organize the experimental data into a data set containing features and target variables; 1.2) Data partitioning: Divide the data set into training set and test set; 1.3) Construct a neural network model: select the corresponding neural network structure and define the hyperparameters including the number of neural network layers, number of neurons, and activation function; 1.4) Training model: Use the training set to train the neural network model, adjust the weights and biases of the model through the back propagation algorithm and optimizer, so that the model fits the training data; select an appropriate loss function to measure the error of the model, and use the optimizer to minimize the loss function; 1.5) Model evaluation: Use the test set to evaluate the performance of the trained model and calculate the prediction error of the model on the test set, including mean square error, root mean square error, and mean absolute error.
3. The composite insulator detection method based on image fusion analysis according to claim 2 is characterized in that: In step 1.2), the training set is set to 80% of the total data, and the remaining data is used as the test set.
4. The composite insulator detection method based on image fusion analysis according to claim 1 is characterized in that: In step 2), the error compensation for attitude solution is to adopt a combined measurement method, and convert the estimation of attitude angle into the estimation of attitude error angle through multi-sensor information fusion technology, and eliminate it; For attitude measurement, Kalman filtering technology is used to fuse the information of the auxiliary attitude measurement system composed of the UAV accelerometer and magnetometer with the gyro attitude measurement system to obtain the optimal attitude estimation value.
5. The composite insulator detection method based on image fusion analysis method according to claim 1 is characterized in that: Step 3) specifically includes the following steps: 3.1) Collect sample data, including spot area and temperature, humidity, air pressure and wind speed related to environmental factors; 3.2) Calculate the wind speed correction coefficient K based on the collected sample data W , Angle correction coefficient K T , gain correction factor K U , distance correction factor K d , Temperature correction coefficient K M and humidity correction factor K N ; Use various correction coefficients to establish a comprehensive correction model, take environmental factors as input, perform correction calculation on the original spot area, and obtain the spot area value under uniform environmental conditions; The normalized correction model formula is: S=(K W K T K U K d K N K M )S0 Among them, S is the spot area under a certain condition after correction, and S0 is the spot area under a certain condition in actual detection. By correcting the model, the influence of environmental factors on the spot area is eliminated, thereby achieving data comparability under different conditions, making the detection of the spot area under different environmental conditions more accurate and reliable.
6. The composite insulator detection method based on image fusion analysis according to claim 1 is characterized in that: The fault detection in step 4) specifically includes: 4.1) Select the characteristic quantity for composite insulator defect diagnosis: First, the temperature and spot area detected by infrared and ultraviolet light are taken as candidate attributes. The temperature rise data is read and converted into the actual temperature rise value according to the detection distance, angle and light intensity. The actual spot area is obtained according to the ultraviolet detection correction model, and the spot area is uniformly normalized. Then, the data is sorted according to the values corresponding to these attributes and classified by different thresholds. 4.2) Classification of composite insulator defect levels: According to relevant technical specifications, and based on temperature rise and UV spot area, the defect grade standards of insulators are divided; defect grades are divided into "serious defect", "general defect", "minor defect" and "good"; temperature rise and UV spot area conditions are comprehensively considered, and the defect grade is determined according to the preset defect grade; when multiple conditions deviate from the normal range at the same time, the degree of defect grade is further comprehensively analyzed and determined to ensure the accuracy and reliability of the determination result; 4.3) Generate composite insulator defect decision tree model: By extracting the characteristic parameters of the original data for preliminary analysis, the input data dimension of the composite insulator is reduced; the classifier construction stage can be divided into two processes: training and evaluation. The classifier is trained using training data and evaluated using test data. The classification algorithm rules are modified through training data to make the classification results consistent with the known data categories; in the use stage, the characteristic parameters of unknown data are extracted and classified as classifier input data; in the process of decision tree construction, the preset influencing variables are temperature rise, ultraviolet spot area and its influencing factors; Due to the complex data structure and incompleteness of some data entries, the CART algorithm is used to process mixed data using the Gini coefficient (gini ratio) as a measure of information gain. The data is then substituted into the Python algorithm environment and a preliminary decision tree is obtained through machine learning. In order to prevent overfitting, the decision tree will be pruned.
7. The composite insulator detection method based on image fusion analysis method according to claim 6 is characterized in that: The basis for defect grade classification in step 4.2) is: According to the existing benchmark values and standard ranges, the temperature rise and UV spot area parameters are compared with the normal state; if the temperature rise is much higher than the normal range and the UV spot area shows an expanding trend, it indicates a serious defect; if the temperature rise is moderate and the UV spot area increases slightly, it is a general defect; when the temperature rise is low, but the UV spot area does not change significantly, it is a minor defect; if the temperature rise and UV spot area are both within the normal range, it is judged to be in good condition.
8. The composite insulator detection method based on image fusion analysis method according to claim 1 is characterized in that: Step 5) specifically includes the following contents: 5.1) After inputting the corresponding insulator model, infrared imaging, ultraviolet imaging, and measured environmental information, decision tree diagnosis can be performed; by transferring the input results to the trained decision tree file, after diagnosis, the decision tree returns the diagnosis result, which is displayed in text form; 5.2) Composite insulator defects found during inspection shall be included in the scope of equipment defect management system, handled in accordance with the equipment defect management process and corresponding treatment plans shall be given.
9. The composite insulator detection method based on image fusion analysis method according to claim 8 is characterized in that: The treatment scheme for the three defect levels in step 5.2) is: Minor defects: There are abnormal corona discharges and a certain gradient in the temperature field, which will affect the equipment but will not cause accidents. It is required to be recorded and the development of the defects should be observed; General defects: If the corona discharge is abnormally prominent, the equipment is overheated, and the temperature difference is large, the inspection cycle should be shortened and the power outage maintenance opportunity should be used to arrange maintenance in a planned manner to eliminate the defects; Serious defects: If the corona discharge of the composite insulator is serious and the maximum temperature exceeds the specified maximum allowable temperature, a power outage should be arranged immediately; If any of the following situations occurs, the insulator is deemed to have failed and should be replaced: ① the sheath becomes brittle, powdered or cracked; ② leakage marks and corrosion appear on the sheath, and the cumulative length is greater than 10% of the creepage distance of the insulator or the corrosion depth is greater than 30% of the material thickness at the location; ③ the sealing of the connecting parts of the insulator fails, and debonding, cracks and slippage occur; ④ the insulator is burned by arc on the surface of the sheath after flashover.
Citation Information
Cited By
Silencer production and manufacturing data processing method and system
CN120044042A
High-voltage overhead transmission line insulator string detection device and control method thereof
CN120908561A
Electrified equipment ultraviolet real-time test and diagnosis method and system
CN122506324A
