Early fault prediction method for insulated overhead conductor
By training artificial intelligence and LSTM-GRU models based on historical signal data for signal data prediction, the problem of failures that cannot be predicted in the existing technology is solved, and the prediction and early warning of early failures of insulated overhead wires is realized, reducing the risk of failure and loss.
Patent Information
- Application Number
- CN202411344669.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-25
- Publication Date
- 2024-12-13
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art can only conduct real-time detection and cannot predict future data and faults, resulting in failures and losses that may have been caused when local discharge is detected.
By obtaining the signal data of insulated overhead wires in real time, training the artificial intelligence model and LSTM-GRU model based on historical signal data, signal data prediction and early fault diagnosis are carried out, determining whether there are early faults and sending early warnings.
The prediction of early failures of insulated overhead wires is achieved, reducing insulation performance degradation and equipment failure caused by local discharge, and reducing maintenance costs and the risk of unexpected power outages.
Smart Images
Figure CN119125775A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of insulated overhead conductor fault detection, in particular to an insulated overhead conductor early fault prediction method. Background Art
[0002] Insulated overhead conductors have obvious advantages over bare conductors in terms of weather resistance, safety and adaptability. They are particularly suitable for complex environments and places with high requirements for power supply reliability. They greatly reduce the number of power outages for maintenance and can meet the current high requirements for power quality. They have been widely used in recent years. However, many insulated overhead conductors often have a higher failure rate because they work in very harsh environments. Due to reasons such as terrain and weather, insulation damage that occurs in the early stage is also more difficult to detect, which has caused great trouble to the normal operation and repair of the line. Therefore, timely detection of early faults in insulated overhead conductors has become an issue that needs to be solved urgently.
[0003] Partial discharge is one of the main causes of insulation degradation in electrical equipment. By monitoring and analyzing the characteristic parameters of partial discharge, it is possible to understand the insulation status in real time, diagnose the cause of the fault, track the development trend of the fault, and perform power outages or take preventive measures when necessary. In insulated overhead conductors, initial partial discharge may not cause an immediate fault, but long-term accumulation will gradually weaken the insulation performance and may eventually lead to overall breakdown.
[0004] The prior art (the invention application with the publication number CN115113004A) discloses a device and method for detecting partial discharge of overhead lines and equipment along the distribution network, including: obtaining a second synchronous pulse signal and a first partial discharge signal through a master detection unit; obtaining a first synchronous pulse signal and a second partial discharge signal through a slave detection unit; analyzing the second synchronous pulse signal and the first synchronous pulse signal through a main control computer to obtain time data, calculating the propagation time based on the time data, synchronizing the time data based on the propagation time to obtain a reference zero point, and calculating the position of the partial discharge based on the reference zero point according to the first partial discharge signal and the second partial discharge signal; the application can effectively realize the rapid diagnosis, positioning and maintenance of partial discharge defects. However, the prior art can only perform real-time detection, and cannot predict future data and faults, resulting in the possibility that faults and losses may have been caused when partial discharge is detected.
[0005] Therefore, the present invention proposes a method for predicting early faults of insulated overhead conductors to solve the above technical problems. Summary of the invention
[0006] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an early fault prediction method for insulated overhead conductors, which is used to solve the technical problem that the prior art can only perform real-time detection but cannot predict future data and faults, resulting in failures and losses that may have already occurred when partial discharge is detected.
[0007] To achieve the above object, a first aspect of the present invention provides a method for predicting early faults of an insulated overhead conductor, comprising:
[0008] Step 1: Acquire signal data of insulated overhead conductors in real time;
[0009] Step 2: Train the artificial intelligence model based on historical signal data to obtain a signal data prediction model;
[0010] Step 3: Predict the signal data through the signal data prediction model to obtain the prediction result;
[0011] Step 4: Train the LSTM-GRU model based on historical signal data to obtain an early fault diagnosis model;
[0012] Step 5: Diagnose the prediction results based on the early fault diagnosis model to obtain the diagnosis results; determine whether there is an early fault in the insulated overhead wire based on the diagnosis results; if yes, send an early warning message to the client; if not, continue to detect and judge.
[0013] Preferably, the real-time acquisition of signal data of the insulated overhead conductor includes:
[0014] S1: Using Rogowski coil as current sensor to collect PD signal on insulated overhead conductor through high frequency current method;
[0015] S2: obtaining an initial PD signal by preprocessing the collected PD signal; wherein the preprocessing includes: amplification and filtering;
[0016] S3: Convert the initial PD signal into a digital signal through an ADC digital-to-analog converter to obtain signal data.
[0017] It should be noted that the Rogowski coil is widely used because of its strong portability, easy installation and good on-site anti-interference ability. The Rogowski coil can be installed on the grounding wire, neutral point wiring or cable body of the power equipment to capture the partial discharge pulse current signal flowing through. The Rogowski coil has linear characteristics and a wide dynamic detection range, so it has sensitive detection capabilities for tiny high-frequency PD currents.
[0018] Preferably, the training of the artificial intelligence model based on historical signal data includes:
[0019] Integrate the historical signal data at the initial moment into standard input data; integrate the historical signal data at the preset prediction moment into standard output data; wherein the historical signal data is the historical data of the signal data of the insulated overhead wire;
[0020] An artificial intelligence model is trained based on standard input data and standard output data to obtain a signal data prediction model; wherein the artificial intelligence model includes: a convolutional neural network or a deep belief network.
[0021] It should be noted that the preset prediction moment is obtained based on the actual preset time scale; wherein the preset time scale refers to the time period between the preset prediction moment and the initial moment; the preset time scale is set according to the actual detection requirements of the insulated overhead wires.
[0022] Preferably, predicting the signal data by using a signal data prediction model includes:
[0023] The signal data at the initial moment is input into the signal data prediction model to obtain a prediction result; wherein the prediction result refers to the signal data at the initial moment corresponding to the preset prediction moment.
[0024] Preferably, the training of the LSTM-GRU model based on historical signal data includes:
[0025] S1: Sample each historical signal data based on a preset sampling frequency to obtain a data point corresponding to each historical signal data; group the data points to obtain several groups of training data; wherein a data point refers to the intensity or amplitude of a signal at a certain time point;
[0026] S2: Mark the number of training data groups corresponding to each historical signal data as M; generate an M*N feature matrix by extracting the eigenvalues of each group of training data; where N is the number of eigenvalues; the eigenvalues include: permutation entropy, singular value entropy, approximate entropy, sample entropy, Petrosian fractal dimension, Katz fractal dimension, fractal dimension, mean, standard deviation, floating upper limit, floating lower limit, block amplitude and seven numerical extraction percentiles;
[0027] S3: taking the characteristic matrix corresponding to each historical signal data as input data, and taking the fault result corresponding to the characteristic matrix as output data; wherein the fault result includes: the existence of an early fault of the insulated overhead line or the absence of an early fault of the insulated overhead line;
[0028] S4: Train the LSTM-GRU model based on the input data and output data to obtain an early fault diagnosis model; wherein the LSTM-GRU model is a model of a five-layer network structure constructed in combination with the attention mechanism; the five-layer network structure includes: input layer, LSTM-GRU layer, ATTENTION layer, fully connected layer and output layer.
[0029] It should be noted that the permutation entropy is used to measure the timing characteristics of the PD signal and analyze its randomness; the singular value entropy is used to measure the structure, complexity and information content of the PD signal; the approximate entropy is used to evaluate the complexity and irregularity of the PD signal; the sample entropy is used to measure the instability of the PD signal timing; the Petrosian fractal dimension is used to calculate the fractal dimension of the PD signal and describe the roughness and irregularity of the signal; the Katz fractal dimension is used to measure the roughness of the PD signal and the shape complexity of the curve; the fractal dimension is used to estimate the fractal dimension of the PD signal timing; the mean is the average value of each group of data; the standard deviation is used to measure the degree of dispersion of the data; the floating upper limit is the sum of the mean and the standard deviation; the floating lower limit is the difference between the mean and the standard deviation; the block amplitude is used to measure the amplitude range of this group of data; the seven numerical extraction percentiles refer to extracting the values of 0%, 1%, 25%, 50%, 75%, 99%, and 100% between the maximum and minimum values in each group of data;
[0030] The fault results corresponding to the feature matrix are marked by experts or technicians based on professional knowledge and experience; the ATTENTION layer is an introduced attention mechanism; the LSTM-GRU layer is composed of a long short-term memory recurrent neural network (LSTM) and a gated recurrent unit (GRU).
[0031] Traditional neural networks (such as RNN) pay equal attention to all input data and do not pay more attention to important data. The attention mechanism gives different weights to different parts of the input, allowing the network to process input data more flexibly, greatly reducing the interference of useless data on training. At the same time, introducing the attention mechanism into the neural network can not only improve the performance of the model, but also help the network better understand the input data and improve the interpretability and explainability of the model.
[0032] Preferably, the diagnosing the prediction result based on the early fault diagnosis model includes:
[0033] A corresponding feature matrix is extracted based on signal data at a preset prediction time, and the feature matrix is input into an early fault diagnosis model to obtain a diagnosis result; wherein the diagnosis result includes: the existence of an early fault in the insulated overhead line or the absence of an early fault in the insulated overhead line.
[0034] Preferably, judging whether there is an early fault of the insulated overhead conductor based on the diagnosis result includes:
[0035] Determine whether the diagnosis result is that there is an early fault in the insulated overhead line; if yes, generate an early warning and send the early warning information to the client; if no, continue to detect and judge; the client includes: a mobile phone or a computer.
[0036] Compared with the prior art, the present invention has the following beneficial effects:
[0037] The present invention obtains signal data of an insulated overhead wire in real time; trains an artificial intelligence model based on historical signal data to obtain a signal data prediction model; predicts the signal data through the signal data prediction model to obtain a prediction result; and trains an LSTM-GRU model based on the historical signal data to obtain an early fault diagnosis model; diagnoses the prediction result based on the early fault diagnosis model to obtain a diagnosis result; and determines whether the insulated overhead wire has an early fault based on the diagnosis result, thereby solving the problem that the prior art can only perform real-time detection but cannot predict future data and faults, resulting in a fault and loss that may have been caused when partial discharge is detected. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 paying creative work.
[0039] Figure 1 A schematic diagram of the method steps of an embodiment of the present invention;
[0040] Figure 2 This is a waveform diagram of Longhuai 23 line in a normal state in an embodiment of the present invention;
[0041] Figure 3 This is a waveform diagram when partial discharge is detected on Longhuai 23 line in an embodiment of the present invention. DETAILED DESCRIPTION
[0042] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0043] See also Figure 1 The first embodiment of the present invention provides a method for predicting early faults of an insulated overhead conductor, comprising:
[0044] Step 1: Acquire signal data of insulated overhead conductors in real time;
[0045] Step 2: Train the artificial intelligence model based on historical signal data to obtain a signal data prediction model;
[0046] Step 3: Predict the signal data through the signal data prediction model to obtain the prediction result;
[0047] Step 4: Train the LSTM-GRU model based on historical signal data to obtain an early fault diagnosis model;
[0048] Step 5: Diagnose the prediction results based on the early fault diagnosis model to obtain the diagnosis results; determine whether there is an early fault in the insulated overhead wire based on the diagnosis results; if yes, send an early warning message to the client; if not, continue to detect and judge.
[0049] Real-time acquisition of signal data of insulated overhead conductors, including:
[0050] S1: Using Rogowski coil as current sensor to collect PD signal on insulated overhead conductor through high frequency current method;
[0051] S2: obtaining an initial PD signal by preprocessing the collected PD signal; wherein the preprocessing includes: amplification and filtering;
[0052] S3: Convert the initial PD signal into a digital signal through an ADC digital-to-analog converter to obtain signal data.
[0053] Training AI models based on historical signal data, including:
[0054] Integrate the historical signal data at the initial moment into standard input data; integrate the historical signal data at the preset prediction moment into standard output data; wherein the historical signal data is the historical data of the signal data of the insulated overhead wire;
[0055] An artificial intelligence model is trained based on standard input data and standard output data to obtain a signal data prediction model; wherein the artificial intelligence model includes: a convolutional neural network or a deep belief network.
[0056] The signal data is predicted by the signal data prediction model, including:
[0057] The signal data at the initial moment is input into the signal data prediction model to obtain a prediction result; wherein the prediction result refers to the signal data at the initial moment corresponding to the preset prediction moment.
[0058] For example: 1. Integrate historical signal data;
[0059] Collect signal data over a period of time, such as data over the past year.
[0060] Assuming that data is collected once an hour, there are 60 data points per hour and 1440 data points per day.
[0061] The data is organized into a table format with timestamps and corresponding signal strengths.
[0062] 2. Train AI models;
[0063] Assume that the preset prediction time is two hours later;
[0064] We choose to use a convolutional neural network (CNN) as the model.
[0065] The signal data at the initial moment is integrated into standard input data; the signal data two hours after the signal data at the initial moment is integrated into standard output data; the artificial intelligence model is trained based on the standard input data and the standard output data to obtain a signal data prediction model.
[0066] 3. Real-time signal data input model;
[0067] The signal data at the current moment is obtained in real time and input into the previously trained signal data prediction model.
[0068] Assuming the current time is 9:00 AM, the model will predict the signal data for 11:00 AM.
[0069] 4. Get the prediction results;
[0070] The signal data prediction model outputs the signal data at 11:00 am.
[0071] Training the LSTM-GRU model based on historical signal data includes:
[0072] S1: Sample each historical signal data based on a preset sampling frequency to obtain a data point corresponding to each historical signal data; group the data points to obtain several groups of training data; wherein a data point refers to the intensity or amplitude of a signal at a certain time point;
[0073] S2: Mark the number of training data groups corresponding to each historical signal data as M; generate an M*N feature matrix by extracting the eigenvalues of each group of training data; where N is the number of eigenvalues; the eigenvalues include: permutation entropy, singular value entropy, approximate entropy, sample entropy, Petrosian fractal dimension, Katz fractal dimension, fractal dimension, mean, standard deviation, floating upper limit, floating lower limit, block amplitude and seven numerical extraction percentiles;
[0074] For example, the data set used in this embodiment is provided by the ENET Center of the Technical University of Ostrava, Czech Republic, with a total of 9682 sets of insulated overhead wire signal data. Among them, the public data includes 2904 three-phase insulated overhead wire signal data, totaling 8712 single-phase signals, which are used as the data set for training the model. Among the 8712 single-phase signals, 512 contain partial discharges, accounting for about 6% of the total. Each single-phase signal of the original signal is 800,000 data points detected in one cycle (20ms) at a frequency of 50Hz, and the sampling frequency is 40MHz.
[0075] The present invention reconstructs a waveform of 200,000 data points by taking 1 out of every 4 data points, and the sampling frequency is 10 MHz. Since the data points of each phase of the insulated overhead conductor are as many as 100,000, it is impossible to use such a large amount of data input to construct a neural network in actual training. The present invention takes 2,000 data points as a group, which is divided into 100 groups of data, and extracts eigenvalues from each group of data. Due to the complex structure of the PD signal of the insulated overhead conductor, the extracted eigenvalues are relatively large, and each group of 2,000 data still belongs to a large order of magnitude. When extracting eigenvalues during the training process, different features are expressed in different ways, resulting in the fact that data with larger values have a much greater impact on classification than data with smaller values.
[0076] Therefore, in order to ensure that each performance parameter index has an equal probability when the classifier is processed, it is necessary to normalize the feature data. The present invention adopts a linear normalization method for data preprocessing. First, the positive and negative values of the data are determined, and then normalization is performed, and finally 2000 data are standardized to between -1 and 1. This processing method can improve the accuracy and stability of neural network model training, and finally obtain 19 eigenvalues, including: permutation entropy, singular value entropy, approximate entropy, sample entropy, Petrosian fractal dimension, Katz fractal dimension, fractal dimension, mean, standard deviation, floating upper limit, floating lower limit, block amplitude and seven numerical extraction percentiles.
[0077] After analysis, each set of data obtains the above 19 eigenvalues, and each phase of insulated overhead wire signal data has a 100*19 characteristic matrix.
[0078] S3: taking the characteristic matrix corresponding to each historical signal data as input data, and taking the fault result corresponding to the characteristic matrix as output data; wherein the fault result includes: the existence of an early fault of the insulated overhead line or the absence of an early fault of the insulated overhead line;
[0079] S4: Train the LSTM-GRU model based on the input data and output data to obtain an early fault diagnosis model; wherein the LSTM-GRU model is a model of a five-layer network structure constructed in combination with the attention mechanism; the five-layer network structure includes: input layer, LSTM-GRU layer, ATTENTION layer, fully connected layer and output layer.
[0080] Diagnose the predicted results based on the early fault diagnosis model, including:
[0081] A corresponding feature matrix is extracted based on signal data at a preset prediction time, and the feature matrix is input into an early fault diagnosis model to obtain a diagnosis result; wherein the diagnosis result includes: the existence of an early fault in the insulated overhead line or the absence of an early fault in the insulated overhead line.
[0082] Based on the diagnostic results, determine whether there is an early fault in the insulated overhead conductor, including:
[0083] Determine whether the diagnosis result is that there is an early fault in the insulated overhead line; if yes, generate an early warning and send the early warning information to the client; if no, continue to detect and judge; the client includes: a mobile phone or a computer.
[0084] For example, suppose that partial discharge signal detection is performed on some insulated overhead wires in a certain area; the insulated overhead wire line is Longhuai 23 line; the details are as follows:
[0085] This test is a long-term fixed-point continuous data collection. The PD signal of the insulated overhead wire is collected every two hours and analyzed and processed to obtain the corresponding signal data;
[0086] The signal data with the real-time detection time of 13:05 is input into the trained data prediction model, and the signal data two hours later, that is, the signal data of 15:05, is predicted;
[0087] The feature matrix corresponding to the signal data at 15:05 is input into the trained early fault diagnosis model to obtain the diagnosis result;
[0088] If the diagnosis results show the presence of partial discharge, indicating an early fault in the insulated overhead line, an early warning message will be sent to the manager’s mobile phone or computer client;
[0089] If the diagnosis results show that there is no partial discharge, it means that there is no early fault in the insulated overhead line, which means that the insulated overhead conductor is normal.
[0090] The specific diagnostic results are shown in the following table:
[0091]
[0092] The waveform analysis results are obtained by experts or technicians based on professional knowledge and experience. The details are as follows:
[0093] See also Figure 2-Figure 3 , compress the measured signal data into CSV format, read it through the read_csv function of the pandas library in Python, and then display the waveform through the matplotlib library;
[0094] By comparison Figure 2 and Figure 3 The waveform of the insulated overhead wire measured shows that the waveform affected by partial discharge and causing early failure has the following characteristics:
[0095] 1. Amplitude change: The typical manifestation of partial discharge is tip discharge. When the insulated overhead wire is affected by partial discharge, it will cause drastic fluctuations in the waveform amplitude, which is manifested as prominent peaks or troughs in the waveform, increasing the instability of the waveform.
[0096] 2. Frequency range diffusion: Due to the high-frequency components generated by partial discharge, the frequency distribution of the waveform may change, resulting in the diffusion of the frequency range.
[0097] 3. Waveform deformation: Partial discharge may cause the original shape of the waveform to change, which is manifested as distortion or deformation of the waveform, thus deviating from the normal waveform structure.
[0098] 4. Enhanced signal interference: The additional noise introduced by partial discharge may reduce the signal-to-noise ratio of the waveform, thereby affecting the signal resolution and measurement accuracy.
[0099] From the comparison of the waveforms before and after the Longhuai 23 line, it can be seen that when the PD signal appears, a large number of high-amplitude glitches will appear in the detection waveform, and the original sinusoidal waveform will no longer be maintained. Frequency signal components will appear, resulting in signal distortion.
[0100] This embodiment further verifies the accuracy of the early fault diagnosis model for early fault prediction of insulated overhead conductors through waveform analysis results; by predicting the occurrence of partial discharge of insulated overhead conductors in advance, it not only helps to prevent sudden accidents, reduce the problems of insulation performance degradation, shortened equipment life, increased safety hazards caused by partial discharge, but also reduces the risks of reduced system stability, increased power loss and increased environmental impact, thereby reducing maintenance costs and the risk of unexpected power outages.
[0101] The above embodiments are only used to illustrate the technical method 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 method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.
Claims
1. A method for predicting early faults of insulated overhead conductors, characterized in that: include: Step 1: Acquire signal data of insulated overhead conductors in real time; Step 2: Train the artificial intelligence model based on historical signal data to obtain a signal data prediction model; Step 3: Predict the signal data through the signal data prediction model to obtain the prediction result; Step 4: Train the LSTM-GRU model based on historical signal data to obtain an early fault diagnosis model; Step 5: Diagnose the prediction results based on the early fault diagnosis model to obtain the diagnosis results; determine whether the insulated overhead wire has an early fault based on the diagnosis results; if yes, send an early warning message to the client; If not, continue to test and judge.
2. The method for predicting early faults of insulated overhead conductors according to claim 1, characterized in that: The real-time acquisition of signal data of the insulated overhead conductor includes: S1: Using Rogowski coil as current sensor to collect PD signal on insulated overhead conductor through high frequency current method; S2: obtaining an initial PD signal by preprocessing the collected PD signal; wherein the preprocessing includes: amplification and filtering; S3: Convert the initial PD signal into a digital signal through an ADC digital-to-analog converter to obtain signal data.
3. The method for predicting early faults of insulated overhead conductors according to claim 1, characterized in that: The training of the artificial intelligence model based on historical signal data includes: Integrate the historical signal data at the initial moment into standard input data; integrate the historical signal data at the preset prediction moment into standard output data; wherein the historical signal data is the historical data of the signal data of the insulated overhead wire; An artificial intelligence model is trained based on standard input data and standard output data to obtain a signal data prediction model; wherein the artificial intelligence model includes: a convolutional neural network or a deep belief network.
4. The method for predicting early faults of insulated overhead conductors according to claim 1, characterized in that: The predicting of the signal data by using the signal data prediction model includes: The signal data at the initial moment is input into the signal data prediction model to obtain a prediction result; wherein the prediction result refers to the signal data at the initial moment corresponding to the preset prediction moment.
5. The method for predicting early faults of insulated overhead conductors according to claim 1, characterized in that: The LSTM-GRU model is trained based on historical signal data, including: S1: sampling each historical signal data based on a preset sampling frequency to obtain a data point corresponding to each historical signal data; grouping the data points to obtain several groups of training data; S2: Mark the number of training data groups corresponding to each historical signal data as M; generate an M*N feature matrix by extracting the eigenvalues of each group of training data; where N is the number of eigenvalues; the eigenvalues include: permutation entropy, singular value entropy, approximate entropy, sample entropy, Petrosian fractal dimension, Katz fractal dimension, fractal dimension, mean, standard deviation, floating upper limit, floating lower limit, block amplitude and seven numerical extraction percentiles; S3: taking the characteristic matrix corresponding to each historical signal data as input data, and taking the fault result corresponding to the characteristic matrix as output data; wherein the fault result includes: the existence of an early fault of the insulated overhead line or the absence of an early fault of the insulated overhead line; S4: Train the LSTM-GRU model based on the input data and output data to obtain an early fault diagnosis model; wherein the LSTM-GRU model is a model of a five-layer network structure constructed in combination with the attention mechanism; the five-layer network structure includes: input layer, LSTM-GRU layer, ATTENTION layer, fully connected layer and output layer.
6. The method for predicting early faults of insulated overhead conductors according to claim 1, characterized in that: The diagnosing the prediction result based on the early fault diagnosis model includes: A corresponding feature matrix is extracted based on signal data at a preset prediction time, and the feature matrix is input into an early fault diagnosis model to obtain a diagnosis result; wherein the diagnosis result includes: the existence of an early fault in the insulated overhead line or the absence of an early fault in the insulated overhead line.
7. The method for predicting early faults of insulated overhead conductors according to claim 1, characterized in that: The method of judging whether there is an early fault in the insulated overhead conductor based on the diagnosis result includes: Determine whether the diagnosis result is that there is an early fault in the insulated overhead line; if yes, generate an early warning and send the early warning information to the client; if no, continue to detect and judge; the client includes: a mobile phone or a computer.