A power distribution line fault detection method, system and device

Abstract

The application belongs to the technical field of power systems, and discloses a power distribution line fault detection method, system and equipment, which comprises the following steps: based on the process data of insulation deterioration of the power distribution line, screening out the core influence parameters of insulation deterioration evolution, and constructing a line insulation deterioration evolution model; based on the line insulation deterioration evolution model, identifying the risk section of the power distribution line existing insulation deterioration; acquiring historical image data of the power distribution line, combining a two-stage model training method, constructing and training a multi-scale synchronous insulation defect detection model; collecting real-time inspection images of the risk section, and inputting the real-time inspection images into the trained multi-scale synchronous insulation defect detection model to identify and locate the defect type and defect position of the risk section. The application can comprehensively reflect the dynamic evolution law of line insulation deterioration under multi-factor coupling, and solves the problems of existing model parameter ambiguity and single factor coverage.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and in particular to a method, system and equipment for detecting faults in power distribution lines. Background Technology

[0002] Power lines are the core carriers of power transmission and distribution systems, and their insulation performance determines the safe and stable operation of the power system. During long-term service, the insulation materials of power lines are subjected to the coupled effects of multiple factors such as environmental, electrical, and mechanical stresses, leading to deterioration, manifested as a decrease in insulation resistance, and in severe cases, causing line faults, large-scale power outages, and resulting in economic losses and social impact. Therefore, accurately describing the dynamic laws of line insulation deterioration and establishing deterioration models and analytical formulas are of great significance for assessing insulation condition, early warning of faults, and operational and maintenance decisions.

[0003] Currently, existing line insulation degradation models fall into two categories: one is empirical linear or simple exponential models, which are simple in structure and easy to calculate, but neglect the coupling of environmental factors and the cumulative effect of aging, and cannot accurately describe the nonlinear evolution process, resulting in large prediction errors; the other is complex models based on physicochemical mechanisms, which, although considering the influence of multiple factors, have many parameters, unclear physical meanings, and complex analytical derivations, requiring a large amount of experimental data and complex calculations, thus having poor practicality and being difficult to use for actual line monitoring. In addition, the analytical derivation of existing models mostly adopts a single fitting method, without combining the physical nature of insulation degradation, leading to a disconnect between the model and reality, and making it difficult to meet the engineering requirements for prediction accuracy and practicality.

[0004] In general, existing line insulation degradation models have the following shortcomings:

[0005] (1) Existing models do not adequately consider the coupling effect of multiple factors and cannot accurately describe the nonlinear evolution process of insulation degradation. Traditional linear or simple exponential models ignore the synergistic effect of environmental factors (temperature, humidity, pollutants, etc.) and electrical and mechanical stress, resulting in a large deviation between the predicted results and the actual degradation law.

[0006] (2) The physical mechanism model has too many parameters and the physical meaning is vague, resulting in poor practicality. Existing complex models based on physicochemical mechanisms have problems such as parameter redundancy and cumbersome derivation process. They rely on a large amount of experimental data and professional computing equipment, which makes it difficult to meet the needs of rapid evaluation in engineering field.

[0007] (3) The analytical derivation is disconnected from the physical nature of insulation degradation. Most existing models use a single data fitting method without taking into account the microscopic mechanisms of insulation material aging, such as molecular chain breakage and oxidation reaction. This results in the model being unable to accurately reflect the influence weight of each parameter on the degradation process, reducing the reliability and interpretability of the prediction results.

[0008] Therefore, how to provide a method, system and equipment for detecting power distribution line faults is an urgent problem to be solved. Summary of the Invention

[0009] This invention provides a method, system, and device for detecting power distribution line faults to solve the problems mentioned above in the prior art.

[0010] According to a first aspect of the present invention, a method for detecting faults in power distribution lines is provided.

[0011] In one embodiment, the power distribution line fault detection method includes the following steps:

[0012] Based on the process data of insulation degradation of distribution lines, the core influencing parameters of insulation degradation evolution were screened out, and based on the irreversibility and exponential decay characteristics of insulation aging, a line insulation degradation evolution model was constructed.

[0013] Based on the line insulation degradation evolution model, risk sections of distribution lines with insulation degradation are identified;

[0014] Historical image data of power distribution lines were acquired, and a multi-scale synchronous insulation defect detection model was constructed and trained by combining a two-stage model training method.

[0015] Real-time inspection images of risk sections are collected and input into the trained multi-scale synchronous insulation defect detection model to identify and locate the defect types and locations in the risk sections.

[0016] According to a second aspect of the present invention, a power distribution line fault detection system is provided.

[0017] In one embodiment, the power distribution line fault detection system includes:

[0018] The evolution model construction module is used to screen out the core influencing parameters of insulation degradation evolution based on the process data of insulation degradation of distribution lines, and to construct a line insulation degradation evolution model based on the irreversibility and exponential decay characteristics of insulation aging.

[0019] The risk identification module is used to identify risk sections of distribution lines with insulation degradation based on the line insulation degradation evolution model;

[0020] The detection model construction module is used to acquire historical image data of power distribution lines and, combined with a two-stage model training method, construct and train a multi-scale synchronous insulation defect detection model.

[0021] The defect identification module is used to acquire real-time inspection images of risk sections and input them into the trained multi-scale synchronous insulation defect detection model to identify and locate the defect type and location of the risk section.

[0022] According to a third aspect of the present invention, a computer device is provided.

[0023] In some embodiments, the computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method described above.

[0024] According to a fourth aspect of the present invention, a computer-readable storage medium is provided.

[0025] In one embodiment, a computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the above method.

[0026] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0027] (1) The physical meaning of the model parameters is clear and the factors covered are comprehensive: The parameter model constructed by this invention includes multi-dimensional parameters such as initial insulation performance, inherent degradation rate, environmental influence, aging accumulation, and correction factors. Each parameter corresponds to the specific physical meaning in the insulation degradation process, which can comprehensively reflect the dynamic evolution law of line insulation degradation under multi-factor coupling, and solve the problem of existing model parameters being vague and the factors covered being single.

[0028] (2) The analytical derivation is scientific, rigorous and practical: Based on the physical and chemical mechanism of insulation degradation, the present invention derives the general solution of the model through differential equation derivation, and then combines multi-factor correction and data fitting to obtain the final analytical expression. The derivation process is simple and easy to understand and logically rigorous. At the same time, by fitting parameters through experimental data, it is ensured that the analytical expression can be highly consistent with the actual insulation degradation process. The calculation is simple and does not require complicated calculation equipment, which is convenient for practical engineering applications.

[0029] (3) High prediction accuracy and significant application value: The analytical formula of this invention can accurately predict the changing trend of line insulation performance at different times, reduce prediction error, and provide reliable theoretical support for power line insulation status assessment and fault early warning. At the same time, by clarifying the degree of influence of each parameter on insulation degradation, it can provide scientific basis for line insulation operation and maintenance decisions, such as insulation material selection, operation and maintenance cycle formulation, and environmental protection measure optimization, thereby reducing the line fault rate and improving the safe and stable operation level of the power system.

[0030] (4) The detection method is highly innovative and has a wide range of applications: The fault detection technology proposed in this invention, based on a multi-scale detection unit network and a two-stage model training method, significantly improves the detection accuracy and comprehensiveness of line equipment faults under complex backgrounds through innovative means such as data augmentation, multi-scale feature fusion, and multi-frame regression. At the same time, this method is not only applicable to the detection of insulation degradation of power distribution lines, but can also be extended to the fault detection of other power equipment and industrial fields, and has broad application prospects.

[0031] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0032] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0033] Figure 1 This is a flowchart illustrating a power distribution line fault detection method according to an exemplary embodiment;

[0034] Figure 2 This is a schematic block diagram of a power distribution line fault detection system according to an exemplary embodiment;

[0035] Figure 3 This is a flowchart illustrating the analytical derivation of a model in a power distribution line fault detection method, according to an exemplary embodiment.

[0036] Figure 4 This is an SSMD network architecture diagram in a power distribution line fault detection method according to an exemplary embodiment;

[0037] Figure 5 This is a flowchart illustrating a two-stage training fault detection method for a power distribution line fault detection method according to an exemplary embodiment;

[0038] Figure 6 This is a schematic diagram of the structure of a computer device according to an exemplary embodiment. Detailed Implementation

[0039] The following description and accompanying drawings fully illustrate specific embodiments described herein to enable those skilled in the art to practice them. Some portions and features of certain embodiments may be included in or replace portions and features of other embodiments. The scope of the embodiments herein includes the entire scope of the claims and all available equivalents thereof. The various embodiments described herein are presented in a progressive manner, with each embodiment focusing on its differences from other embodiments; similar or identical parts between embodiments can be referred to interchangeably.

[0040] The modules in the apparatus or system of this application can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0041] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0042] Figure 1 An embodiment of a power distribution line fault detection method of the present invention is shown.

[0043] In this optional embodiment, the power distribution line fault detection method includes the following steps:

[0044] Step S101: Based on the process data of insulation degradation of power distribution lines, the core influencing parameters of insulation degradation evolution are screened out, and based on the irreversibility and exponential decay characteristics of insulation aging, a line insulation degradation evolution model is constructed.

[0045] Step S102: Based on the line insulation degradation evolution model, identify risk sections of the distribution line where insulation degradation exists;

[0046] Step S103: Obtain historical image data of power distribution lines, and combine the two-stage model training method to construct and train a multi-scale synchronous insulation defect detection model.

[0047] Step S104: Collect real-time inspection images of the risk section and input them into the trained multi-scale synchronous insulation defect detection model to identify and locate the defect type and defect location of the risk section.

[0048] In this optional embodiment, the process of selecting core influencing parameters of insulation degradation evolution based on the process data of distribution line insulation degradation, and constructing a line insulation degradation evolution model based on the irreversibility and exponential decay characteristics of insulation aging, includes the following steps: Based on the process data of distribution line insulation degradation, and combined with the physical mechanism and on-site operating environment characteristics, core influencing parameters of insulation degradation evolution are selected; based on the core influencing parameters, and combined with the irreversibility and exponential decay characteristics of insulation aging, a differential equation for insulation degradation evolution is established and the general solution is solved to obtain a preliminary insulation degradation evolution model; using a multi-factor correction method, the preliminary insulation degradation evolution model is corrected to obtain the final analytical expression of the insulation degradation evolution parameter model; based on experimental values ​​obtained at different times through prior experiments, the unknown parameters in the final analytical expression of the insulation degradation evolution parameter model are fitted and solved to determine the final form of the final analytical expression of the insulation degradation evolution parameter model; the final form is verified for relative error until a line insulation degradation evolution model that meets the preset error requirements is obtained.

[0049] In this optional embodiment, the process data of distribution line insulation degradation, combined with physical mechanisms and on-site operating environment characteristics, is used to screen out the core influencing parameters of insulation degradation evolution. This includes the following steps: Based on the process data of distribution line insulation degradation and combined with the microscopic mechanism of insulation degradation, basic parameters including initial insulation, inherent degradation rate, and aging accumulation are determined; based on the actual operating environment of the line, environmental factor parameters including temperature, humidity, pollution, and electric field are determined; according to the characteristics of engineering monitoring and operation and maintenance, correction factor parameters including test error, operation and maintenance intervention, and service life are determined; the basic parameters, environmental factor parameters, and correction factor parameters are integrated and summarized to form a complete set of core influencing parameters of insulation degradation evolution.

[0050] In this optional embodiment, the step of establishing a differential equation for insulation degradation evolution and solving the general solution based on the core influencing parameters and the irreversibility and exponential decay characteristics of insulation aging to obtain a preliminary insulation degradation evolution model includes the following steps: Based on the irreversibility and exponential decay characteristics of insulation aging, and considering the positive correlation between the decay rate of insulation performance characterization parameters over time and the current insulation performance value and the degree of aging accumulation, a differential equation is established; the differential equation is separated into variables, and the integral result is obtained by performing equation transformations and integration on the separated differential equation; the integral result is subjected to exponential operations, and the general solution of the differential equation is obtained through formula simplification, which serves as the preliminary insulation degradation evolution model.

[0051] In this optional embodiment, the step of using a multi-factor correction method to correct the preliminary insulation degradation evolution model and obtain the final analytical expression of the insulation degradation evolution parameter model includes the following steps: calculating the environmental impact term based on the inherent degradation rate of the insulation material itself and the contribution of environmental factors to the degradation rate; determining the correction term used to correct the preliminary insulation degradation evolution model based on the correction factor parameters; substituting the environmental impact term and the correction term into the preliminary insulation degradation evolution model, and deriving the final analytical expression of the line insulation degradation evolution parameter model.

[0052] In this optional embodiment, the step of fitting and solving the unknown parameters in the final analytical expression of the insulation degradation evolution parameter model based on experimental values ​​obtained in advance through experiments at different times, and determining the final form of the final analytical expression of the insulation degradation evolution parameter model, includes the following steps: obtaining experimental values ​​at different times recorded in advance through accelerated insulation degradation tests, wherein the experimental values ​​include quantified values ​​of environmental factors, correction coefficients, and insulation performance characterization parameters; constructing an objective function for parameter fitting with the goal of minimizing the mean square error between the experimental values ​​and the predicted values ​​of the final analytical expression of the insulation degradation evolution parameter model; solving the objective function using the least squares method, and obtaining the optimal fitting value of the unknown parameters of the final analytical expression of the insulation degradation evolution parameter model through iterative calculation; substituting the optimal fitting value of the unknown parameters into the final analytical expression of the insulation degradation evolution parameter model to obtain the analytical expression of the line insulation degradation evolution with determined parameter values, and using it as the final form of the final analytical expression of the insulation degradation evolution parameter model.

[0053] In this optional embodiment, the process of verifying the relative error of the final form until a line insulation degradation evolution model that meets the preset error requirements is obtained includes the following steps: selecting verification data that is independent of the experimental values ​​used for fitting the unknown parameters, and substituting the quantified values ​​of environmental factors and correction coefficients in the verification data into the analytical expression after determining the final form to calculate the predicted insulation performance value at each time point; calculating the relative error of each data point in the verification data based on the quantified values ​​of environmental factors in the verification data and the calculated predicted insulation performance value; determining whether the relative errors of all verification data points meet the preset error requirements, and if so, determining that the analytical expression is valid and completing the construction of the line insulation degradation evolution model; if the relative error of any data point exceeds the preset error requirements, readjusting the fitting range of the unknown parameters and refitting until the relative errors of all verification data points meet the preset error requirements.

[0054] In this optional embodiment, the step of acquiring historical image data of the power distribution line and constructing and training a multi-scale synchronous insulation defect detection model by combining it with a two-stage model training method includes the following steps: acquiring historical image data of the power distribution line, performing multi-strategy data augmentation processing to obtain augmented historical image data; dividing the augmented historical image data into a first dataset and a second dataset; constructing a multi-scale synchronous insulation defect detection network, and using the first dataset and the second dataset to perform two-stage model training on the multi-scale synchronous insulation defect detection network to obtain a multi-scale synchronous insulation defect detection model.

[0055] In this optional embodiment, the process of acquiring historical image data of power distribution lines and performing multi-strategy data enhancement processing to obtain enhanced historical image data includes the following steps: performing rotation, scaling, translation, and shearing transformations on the images in the historical image data based on affine transformation to obtain affine transformed historical image data; using Gaussian noise injection technology to filter the images in the affine transformed historical image data to generate image data of different resolutions to obtain filtered historical image data; and using physical field constraint enhancement technology based on electromagnetic field-visual coupling to enhance the filtered historical image data to obtain enhanced historical image data.

[0056] In this optional embodiment, the physical-driven data enhancement technology based on electromagnetic field-visual coupling enhances the filtered historical image data with physical field constraints to obtain enhanced historical image data, including the following steps: For the filtered historical image data, extract the line operating voltage and the line phase-to-phase or line-to-ground distance corresponding to each image, and calculate the distance field from each pixel to the defect area in the image; calculate the normalized electromagnetic field strength value of each pixel based on the distance field from each pixel to the defect area in the image; generate a synthetic image under electromagnetic field constraints based on the normalized electromagnetic field strength value of each pixel, combined with the physical laws of electromagnetic fields and the pixel values ​​of each pixel in the original image, and use the synthetic image as the enhanced historical image data.

[0057] In this optional embodiment, the construction of a multi-scale synchronous insulation defect detection network and the two-stage model training of the multi-scale synchronous insulation defect detection network using the first and second datasets to obtain the multi-scale synchronous insulation defect detection model include the following steps: constructing a multi-scale synchronous insulation defect detection network based on a preset basic backbone feature extraction structure; constructing a joint loss function based on a temporal consistency constraint method; and performing two-stage model training of the multi-scale synchronous insulation defect detection network based on the first and second datasets and in combination with the joint loss function to obtain the multi-scale synchronous insulation defect detection model.

[0058] Figure 2An embodiment of a power distribution line fault detection system according to the present invention is shown.

[0059] In this optional embodiment, the power distribution line fault detection system includes:

[0060] The evolution model construction module 201 is used to screen out the core influencing parameters of insulation degradation evolution based on the process data of insulation degradation of distribution lines, and to construct a line insulation degradation evolution model based on the irreversibility and exponential decay characteristics of insulation aging.

[0061] Risk identification module 202 is used to identify risk sections of power distribution lines with insulation degradation based on the line insulation degradation evolution model;

[0062] The detection model construction module 203 is used to acquire historical image data of power distribution lines and, in combination with a two-stage model training method, construct and train a multi-scale synchronous insulation defect detection model.

[0063] The defect identification module 204 is used to acquire real-time inspection images of risk sections and input them into the trained multi-scale synchronous insulation defect detection model to identify and locate the defect type and defect location of the risk section.

[0064] To facilitate understanding of the above technical solutions of the present invention, the following further explains the above technical solutions of the present invention from the perspective of architecture and principle, as follows:

[0065] A parametric model for the evolution of line insulation degradation is proposed. This model uses insulation performance parameters as dependent variables and degradation time as independent variables. It combines initial insulation state, degradation rate, environmental influences, and cumulative aging effects to construct a multi-factor coupled nonlinear parametric model. The specific expression is as follows:

[0066] ;

[0067] The meanings of each parameter are as follows:

[0068] ① Insulation performance characterization parameters It is used to quantify the insulation performance of the line insulation material at time t. It can be selected according to the actual monitoring needs. Common ones include insulation resistance (unit MΩ), dielectric loss tangent (dimensionless), breakdown field strength (unit kV / mm), volume resistivity (unit Ω·m), etc. The higher the value, the better the insulation performance; as the degradation process progresses, Gradually decrease.

[0069] ② Initial insulation performance parameters , The baseline values ​​of insulation performance characterization parameters at a given time (i.e., before the insulation material deteriorates) are obtained from standard tests to ensure consistency with... The units must be consistent. For example, insulation resistance can be measured using a megohmmeter on a new insulation sample to obtain... The value;

[0070] ③Inherent degradation rate parameter A dimensionless parameter that characterizes the rate of deterioration of insulating materials due to aging (such as material oxidation, molecular chain breakage, etc.) when there are no external environmental factors. The value is determined by the properties of the insulating material, and different types of insulating materials (such as XLPE, EPR, oil paper, etc.) are different. The values ​​are different; The higher the value, the faster the insulation material itself deteriorates.

[0071] ④ Environmental factor-related parameters, including the number of environmental factors. , No. Deterioration impact coefficient of various environmental factors , Time of the first Quantitative values ​​of environmental factors Among the environmental factors selected, those that significantly affect insulation degradation include temperature (°C), humidity (%RH), and atmospheric pollutant concentration (mg / m³, such as...). (e.g., dust), ultraviolet radiation intensity (unit: W / m²), electric field strength (unit: kV / m), etc. The dimensionless coefficient represents the first... The extent to which environmental factors affect the rate of insulation degradation When the value is >0, environmental factors will accelerate insulation degradation. The higher the value, the more pronounced the accelerating effect; When the value is less than 0, environmental factors (such as low humidity environments) will inhibit insulation degradation. The actual quantitative value of this environmental factor at time t can be obtained in real time through environmental monitoring equipment.

[0072] ⑤ Aging cumulative index A dimensionless parameter that characterizes the cumulative effect of insulation degradation. . At that time, the cumulative effect of insulation degradation is linear, that is, the degradation rate does not change with time; As time goes on, the rate of aging increases, which manifests as accelerated deterioration (e.g., in the later stages of aging of insulation materials, the deterioration rate increases significantly). At this time, the rate of aging accumulation slows down over time, which is manifested as "decelerated degradation" (such as the slow degradation rate in the early stage of insulation material aging).

[0073] ⑥ Correction factors related parameters, including the number of correction factors. , Time of the first Correction coefficients for various correction factors Correction factors are used to correct model prediction biases. Common correction factors include insulation material service life correction (considering the basic differences in material performance under different service durations), test error correction (correcting deviations caused by instrument errors and operational errors during testing), and maintenance intervention correction (considering the impact of cleaning, protection, and other measures on insulation degradation during line maintenance). This is a dimensionless correction factor, with a value range of [value range missing]. , There is no corrective effect at this time. Increase the model's predicted value at times. The model prediction value is reduced by fitting the deviation between the experimental data and the model prediction value.

[0074] To enhance the applicability and accuracy of the core model, a degradation stage segmentation model and an environment coupling weight model are set up as auxiliary models. These two auxiliary models work in conjunction with the core model. The degradation stage segmentation model determines the stage based on the output of the core model, while the environment coupling weight model is used to correct the environmental impact coefficient of the core model, as detailed below:

[0075] 1. Degradation Stage Classification Model: Insulation degradation is a continuous nonlinear process. The changes in insulation performance and maintenance requirements differ significantly at different degradation stages. Therefore, a degradation stage classification model is constructed based on the output of the core model. By combining the calibrated thresholds (which are determined by graded calibration based on standard failure indicators of insulation materials, on-site operation monitoring data, and operation and maintenance procedures, with independent calibration for different insulation materials to adapt to actual degradation patterns), accurate division of degradation stages is achieved. The model expression is as follows:

[0076] ;

[0077] Model parameter description:

[0078] (1) for Insulation degradation stage indicators at all times Indicates the initial stage of degradation. During this stage, the insulation material only undergoes slight aging, the molecular chains do not break significantly, the insulation performance declines slowly, no special operation and maintenance is required, only routine monitoring is needed;

[0079] (2) Indicates mid-stage degradation. During this stage, the aging of insulation materials intensifies, some molecular chains break, and insulation performance declines rapidly. It is necessary to increase the monitoring frequency and take timely maintenance measures such as cleaning and protection.

[0080] (3) Indicates late stage of degradation. During this stage, a large number of molecular chains of the insulating material break down, the insulating structure is severely damaged, the insulating performance is on the verge of failure, and it is easy to cause line faults. The insulating material needs to be replaced immediately or a major overhaul needs to be carried out.

[0081] (4) Deterioration threshold ( , The threshold is determined by standard failure indices of insulation materials and multiple sets of accelerated aging test data. The thresholds for different types of insulation materials can be flexibly adjusted to ensure the accuracy of stage division.

[0082] 2. Environmental Coupling Weight Model: In the actual process of line insulation degradation, multiple environmental factors do not act independently, but rather exhibit significant coupling effects (e.g., the insulation degradation rate is much higher under high temperature and high humidity conditions than under single high temperature or high humidity conditions). Existing models neglect this coupling effect, leading to deviations in the quantification of environmental impact coefficients. Therefore, an environmental coupling weight model is constructed to quantify the coupling strength of different environmental factors and correct the core model. To improve the prediction accuracy of the core model, the model expression is:

[0083] ;

[0084] Model parameter description:

[0085] (1) : No. The coupling weights of various environmental factors are dimensionless. The sum of the coupling weights of all environmental factors is , The larger the value, the more significant the impact of the environmental factor on insulation degradation in the coupling process;

[0086] (2) : No. The contribution rate of environmental factors to degradation, dimensionless. The results were obtained by fitting a single environmental factor accelerated aging test, specifically by controlling other environmental factors to remain constant and only changing the first environmental factor. The quantitative values ​​of various environmental factors are used to record the rate of insulation performance degradation, and the contribution rate of each environmental factor to insulation degradation is obtained by fitting.

[0087] (3) : No. The percentage of actual duration of each environmental factor's influence, dimensionless. By recording the actual duration of each environmental factor's influence over a period of time (e.g., one year) using environmental monitoring equipment, the proportion of each environmental factor's influence duration to the total monitoring duration is calculated. (For example, if the duration of high-temperature environment (≥35℃) in a certain region is 300 hours in a year, and the total monitoring duration is 8760 hours, then the temperature...) );

[0088] (4) The revised version of the first... The environmental factor degradation impact coefficient is dimensionless. Substitute the core model to replace the original At this point, the overall degradation rate parameter of the core model is updated to... This enables precise quantification of environmental coupling effects.

[0089] Furthermore, this invention also provides an analytical derivation method based on the above-mentioned parameter model. This method combines the physicochemical mechanism of insulation degradation and obtains the analytical expression of the model through steps such as differential equation derivation, parameter correction, and data fitting. Specifically, it includes the following steps:

[0090] Step 1: Determine the core parameters affecting insulation degradation evolution; based on the actual process of line insulation degradation, screen the core parameters affecting insulation degradation, including initial insulation performance parameters. Inherent degradation rate parameter Environmental factor parameters ( , , aging cumulative index Correction factor parameters ( , Clarify the physical meaning of each parameter and determine the methods for obtaining each parameter, such as experimental measurement, environmental monitoring, and data fitting, to lay the foundation for subsequent model building and analytical derivation.

[0091] Specifically, core parameters are selected based on the actual degradation process, prioritizing physical mechanisms and on-site operating environment characteristics. Specifically: 1. Based on the microscopic mechanisms of insulation degradation, three basic parameters are determined: initial insulation, inherent degradation rate, and aging accumulation; 2. Key environmental parameters such as temperature, humidity, pollution, and electric field are incorporated into the actual operating environment of the line; 3. Correction parameters such as test errors, maintenance interventions, and service life are added to align with engineering monitoring and maintenance. Finally, parameters with clear physical meaning, measurability, and calculability are selected for model construction.

[0092] Step 2: Establish the differential equation for the evolution of insulation degradation and solve the general solution; based on the irreversibility and exponential decay characteristics of insulation aging, determine the insulation performance characterization parameters. Over time The evolution of insulation follows a nonlinear differential equation. Based on the physical nature of insulation degradation, the rate of insulation performance decay is positively correlated with the current insulation performance value and the degree of aging accumulation. Therefore, the differential equation is established as follows: ;

[0093] in, To comprehensively assess the degradation rate parameter, the effects of environmental and corrective factors are temporarily ignored, and the initial conditions are as follows: hour, (That is, the insulation performance has not deteriorated at the initial moment).

[0094] Separating variables from the above differential equation, we can determine the variables containing... Move the term to the left side of the equation, including Moving the term to the right side of the equation, we get: ;

[0095] Integrating both sides of the equation simultaneously, the integration interval is... That is, from the initial time ( ) to any time ( ): ;

[0096] According to the integral formula, the result of the integral on the left side is: The integral result on the right side is Therefore, we get: ;

[0097] By performing exponential operations on both sides of the equation and rearranging, we obtain the general solution of the differential equation, which is the preliminary model of insulation degradation evolution: ;

[0098] This preliminary model reflects the exponential decay of insulation performance over time, including initial insulation performance, overall degradation rate, and cumulative aging effect. However, it does not consider the influence of environmental factors and correction factors and needs further revision.

[0099] Step 3: Introduce multi-factor correction and derive the analytical expression of the parameter model; to ensure that the model can accurately describe the actual insulation degradation process of the line, it is necessary to introduce environmental impact terms, aging accumulation terms, and correction terms to revise the preliminary model. The specific process is as follows:

[0100] (1) Introducing the environmental impact term: The actual line insulation degradation is affected by the coupling of multiple environmental factors. Therefore, the comprehensive degradation rate parameter k is corrected to the sum of the inherent degradation rate and the contribution of each environmental factor, that is:

[0101] ;

[0102] in, This refers to the inherent degradation rate of the insulating material itself. The total contribution of all environmental factors to the degradation rate reflects the coupling effect of multiple environmental factors.

[0103] (2) Retaining the aging accumulation term: in the preliminary model This term effectively characterizes the cumulative effect of insulation degradation, therefore it is retained. The aging cumulative index is determined by fitting subsequent experimental data to reflect the nonlinear cumulative characteristics of the deterioration of different insulation materials.

[0104] (3) Introduction of correction terms: Considering that insulation degradation is also affected by factors such as operation and maintenance intervention, test errors, and material service life during actual service, these factors cannot be fully covered by environmental factors and inherent degradation rates. Therefore, a correction term is introduced. Correction coefficients for various correction factors The product form is incorporated into the model to correct prediction biases and ensure that the model is consistent with the actual degradation process.

[0105] Substituting the above correction terms into the preliminary model, we obtain the final analytical expression for the line insulation degradation evolution parameter model, namely: ;

[0106] Step 4: Fit the model parameters and determine the final form of the analytical expression; the model's analytical expression contains multiple unknown parameters ( , , , The specific value needs to be determined by fitting experimental data. The specific process is as follows:

[0107] (1) Conduct accelerated insulation degradation test: Select a sample with the same insulation material as the actual line, simulate the environmental conditions in actual operation, such as different combinations of temperature, humidity and pollutant concentration, set up multiple test groups, and record the data at different times. Corresponding insulation performance characterization parameters Quantitative values ​​of various environmental factors Simultaneously, the sources of error and operational simulation measures during the experiment were recorded to provide correction coefficients. This provides a basis for the fitting.

[0108] (2) Constructing the objective function: With the goal of minimizing the mean square error between the experimental values ​​and the model predictions, construct the objective function for parameter fitting, i.e.: ;

[0109] (3) Solving the objective function: The objective function is solved using the least squares method. Through iterative calculation, the result is obtained. , , , The least squares method is characterized by its simple calculation and high fitting accuracy. It is suitable for solving multi-parameter fitting problems and ensures that the fitted parameters can match the model predictions and experimental values ​​to the greatest extent possible.

[0110] (4) Determine the final form of the analytical expression: Substitute the optimal parameters obtained by fitting into the analytical expression to obtain the analytical expression of line insulation degradation evolution with determined parameter values, and complete the derivation of the analytical expression.

[0111] Step 5: Verify the validity and accuracy of the analytical expression; to ensure that the derived analytical expression can accurately describe the evolution law of line insulation degradation, its validity and accuracy need to be verified. The specific process is as follows:

[0112] (1) Selecting verification data: Select another set of test data (or actual line operation monitoring data) that are different from the test conditions in step 4 as verification data to ensure the independence of verification data and avoid verification deviation caused by the overlap between fitted data and verification data.

[0113] (2) Calculate the predicted values: Quantify the environmental factors in the validation data. Correction factors and correction coefficients Substituting into the final analytical expression, we can calculate the results at different times. Corresponding insulation performance prediction value .

[0114] (3) Calculate the relative error: based on the experimental values ​​in the validation data. and the calculated predicted value The relative error is calculated using the following formula. : ;

[0115] (4) Error judgment: Set the allowable error range (in this invention, it is set to ). If the relative errors of all verification data points satisfy If the analytical expression is valid and accurate, it can accurately predict the process of insulation degradation in the line; if there is a relative error in the data points... If the error is not met, return to step 4, adjust the parameter fitting range, and repeat the parameter fitting until the error requirement is met.

[0116] Furthermore, this invention proposes an intelligent analysis and detection method for related line equipment faults based on a multi-scale detection unit network: The detection system first captures the physical characteristics of the line equipment surface by acquiring detailed information from optical images. Optical sensors can record changes in light reflection on the line equipment surface, thereby obtaining potential fault information. This image data is then processed by a multi-scale detection unit network. Based on a convolutional neural network (CNN), the multi-scale detection unit network extracts, analyzes, and classifies features from the images, enabling automatic identification of defects on the line equipment surface.

[0117] (1) Data Augmentation Method for Fault Detection of Related Line Equipment; In the training process of convolutional neural networks, this invention uses data augmentation to improve the generalization ability of the model. First, affine transformation is used to perform rotation, scaling, translation, and shearing transformations on the image. The affine transformation formula is as follows: ;

[0118] in, The coordinates of the transformed points. It is a two-dimensional affine transformation matrix. The coordinates of the point before the transformation. It is a translation vector.

[0119] Secondly, Gaussian noise injection technology is employed to simulate noise interference in complex environments. Gaussian noise is expressed by the following formula: ;

[0120] Among them, u x y Here is the noise value, and μ is the mean. Standard deviation; It represents a normal distribution.

[0121] In addition, brightness adjustment and Gaussian blur techniques were also applied in data augmentation. Brightness adjustment simulates scenes under different lighting conditions, while Gaussian blur uses a two-dimensional Gaussian function to filter the image, generating image data with varying degrees of sharpness.

[0122] ;

[0123] in, It is a Gaussian function. Standard deviation, These are the horizontal and vertical coordinates of the image, respectively.

[0124] Furthermore, this invention introduces a physically driven data enhancement method based on electromagnetic field-visual coupling, overcoming the limitations of traditional pure image transformation. An electromagnetic field-visual coupling data generation model is established, coupling the distribution of the electromagnetic physical field around the device with visual image features, and generating synthetic data that conforms to the characteristics of the real field based on the physical laws of electromagnetic fields.

[0125] ;

[0126] ;

[0127] In the formula, For pixels Normalized electromagnetic field strength; This refers to the operating voltage of the line. This refers to the phase-to-phase / ground distance of the line. This represents the distance field from the pixel to the defect region. , , For field strength mapping and fusion coefficients; These are the original image pixel values ​​(i.e., the image pixel values ​​generated by filtering the image using a two-dimensional Gaussian function to produce image data of different resolutions). This is a composite image under electromagnetic field constraints.

[0128] (2) Multi-scale detection unit network based on convolutional neural network; the multi-scale detection unit network proposed in this invention is as follows: Figure 4 As shown, features are extracted through a basic network, and predicted bounding boxes are generated through multiple auxiliary convolutional layers. This invention constructs a deep convolutional neural network based on the VGG-16 network. The first few layers of VGG-16 are mainly used to extract basic features from the input high-resolution line equipment image, including low-level features such as edges and corners. Subsequent convolutional layers gradually extract more complex geometric shapes and texture information.

[0129] The SSMD model uses the VGG-16 network as its basic backbone feature extraction structure. It undertakes the task of extracting shallow texture, edge details, and deep insulation defect semantic features of line images step by step, providing multi-level original feature support for subsequent defect detection. It is the feature foundation of the entire model.

[0130] In the backend of the VGG-16 base network, the SSMD model adds 6 independent auxiliary convolutional layers as a core extension structure, which is a key design that distinguishes it from traditional detection networks. The 6 auxiliary convolutional layers have clearly defined functions: 5 of them are dedicated to generating multi-scale prediction boxes, while the remaining layer is used for feature channel dimensionality reduction and feature information refinement, thereby optimizing feature dimensions and filtering redundant information to ensure subsequent detection accuracy.

[0131] The model, relying on newly added multi-layer auxiliary convolutional layers, constructs a multi-scale feature detection architecture that can independently generate detection prior boxes on feature maps of different resolutions. This allows it to adapt to target features of large, medium, and small scales related to line insulation defects, effectively adapting to complex background scenarios of transmission lines. Simultaneously, a multi-layer feature fusion mechanism is introduced, aggregating and fusing shallow and deep features from the backbone network with multi-scale features from the auxiliary convolutional layers across layers. This integrates feature information from different receptive fields, ultimately generating a total of 8732 prediction boxes globally. This achieves full coverage of candidate box selection for potential line defects, significantly improving the completeness, robustness, and accuracy of defect detection.

[0132] The loss function of SSMD consists of confidence loss. and positioning loss composition:

[0133] ;

[0134] In the formula: This represents the overall loss of the model. , These are the balancing weights for classification confidence loss and localization loss, respectively. Take 1, (Choose 1 or 2). Start model training according to the chosen value to ensure rapid convergence in the early stages of training. Monitor the rate of decrease in classification confidence loss and localization loss in real time and make dynamic adjustments to select the coefficient value that optimizes the overall performance. By setting reasonable weight coefficients, balance the constraint ratio between defect category classification accuracy and bounding box regression localization accuracy during model training, achieving collaborative optimization of classification and localization tasks. This completes the iterative update of parameters and model convergence training of the SSMD network.

[0135] Confidence loss The formula used to evaluate the accuracy of the model in target classification is as follows:

[0136] ;

[0137] in, This indicates the matching relationship between the predicted bounding box and the truth bounding box. The predicted probability for the target category.

[0138] Location loss Used to measure the error between the predicted bounding box and the true bounding box, employing a smoothing method. Loss function: ;

[0139] in, For the predicted bounding box coordinates, The coordinates of the truth box. To calculate the smoothing loss of coordinate differences, fine optimization with small deviations and stable training with large deviations are performed to improve the frame positioning accuracy of defects such as broken insulators and discharge marks.

[0140] (3) Accurate fault location based on multi-box regression mechanism; In the SSMD model (multi-scale synchronous insulation defect detection model), the multi-box regression mechanism is used for the regression calculation of bounding boxes. It achieves accurate target location through multi-box prediction. Specifically, a large number of predicted boxes are generated on feature maps of different scales, then confidence loss is used to filter the categories, localization loss is used to correct the position of the bounding boxes, and finally the optimal box is selected by intersection-union ratio to complete the accurate location.

[0141] The multi-frame regression mechanism is implemented through convolutional networks, improving detection accuracy while maintaining model detection speed. Its loss function consists of two parts: confidence loss and localization loss, used to measure the difference between the predicted result and the actual target.

[0142] SSMD's multi-box regression mechanism employs an Inception-style convolutional network, using 1×1 convolutional kernels to reduce dimensionality while maintaining the width and height of the feature maps. For targets of different sizes, SSMD uses feature maps of different scales for prediction to ensure the model's robustness in multi-scale scenarios.

[0143] The localization accuracy of multi-frame regression is mainly achieved through intersection-union ratio (IU). Conduct an assessment when If the prediction exceeds a preset threshold, the result is considered correct; otherwise, it is a false positive. The formula for calculating the crossover ratio (CRR) is:

[0144] ;

[0145] in, To predict the area of ​​the frame, This represents the actual area of ​​the bounding box.

[0146] (4) Two-stage model training method; To overcome the data shortage of the training model (SSMD multi-scale synchronous insulation defect detection model) and the traditional single-stage optimization method, this paper uses the necessary dataset as the main dataset (i.e., the first dataset) and the special dataset as the secondary dataset (i.e., the second dataset). The necessary dataset mainly consists of images of normal insulation and common typical defects under normal line operating conditions, serving as the main data for model training; the special dataset selects scarce samples such as extreme weather, complex backgrounds, and rare hidden insulation defects as supplementary data; the division is completed according to the principle of classifying common samples into the necessary dataset and special scarce and difficult samples into the special dataset, and the aerial images are trained in two stages. The processing process is as follows: Figure 5 As shown.

[0147] like Figure 5As shown, in the first stage of training, a basic dataset is trained using 18,000 aerial images and line equipment in 12 different backgrounds. In the second stage, two specialized datasets are constructed using aerial images of different types and scenes. By using these specialized datasets as input, the detection model can learn more parameters related to line equipment. Compared to older methods with one-stage training, two-stage fine-tuning feeds more data into both the training and detection models, producing optimal output.

[0148] To further improve the convergence accuracy, generalization ability, and adaptability to complex working conditions of the multi-scale synchronous insulation defect detection (SSMD) model, this scheme introduces a spatiotemporal consistency constraint training algorithm for joint optimization based on the conventional two-stage training paradigm. This differs from the traditional general training method that relies solely on gradient descent, exhibiting non-obviousness and significant technological advancements. A spatiotemporal consistency constraint loss is constructed, applying temporal feature correlation constraints to consecutive frames of aerial inspection images to suppress abrupt changes in defect features caused by environmental jitter and illumination variations. Its expression is: ;

[0149] In the formula: The first Frame and the Deep network features corresponding to the input frame image; This represents the number of training sample batches. The temporal consistency regularization constraint loss is used. This temporal constraint is weighted and fused with the original classification and localization total loss of the SSMD model to construct a joint training loss function:

[0150] ;

[0151] In the formula: The weighted sum of the model classification confidence loss and the bounding box localization loss; The spatiotemporal consistency constraint weight coefficient (values ​​from 0.1 to 0.5) is used to balance the training ratio of detection tasks and temporal constraint tasks. The spatiotemporal consistency constraint weight coefficient is first initialized to 0.2 based on empirical optimization, and then the detection accuracy and temporal stability are observed through model training. The coefficient is then dynamically adjusted in combination with the validation set results, and finally the value that optimizes the overall performance of the model is selected.

[0152] By strengthening the temporal stability of defect features in sequence frames through spatiotemporal consistency constraints, the detection accuracy and robustness can be significantly improved compared with conventional training methods. The training strategy is unique and creative, and the technological progress is remarkable.

[0153] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores static and dynamic information data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the above method embodiments.

[0154] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device to which the present invention is applied. A specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0155] In addition, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0156] In addition, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0157] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0158] This invention is not limited to the structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this invention is limited only by the appended claims.

Claims

1. A method for detecting faults in power distribution lines, characterized in that, The method includes the following steps: Based on the process data of insulation degradation of distribution lines, the core influencing parameters of insulation degradation evolution were screened out, and based on the irreversibility and exponential decay characteristics of insulation aging, a line insulation degradation evolution model was constructed. Based on the line insulation degradation evolution model, risk sections of distribution lines with insulation degradation are identified; Historical image data of power distribution lines were acquired, and a multi-scale synchronous insulation defect detection model was constructed and trained by combining a two-stage model training method. Real-time inspection images of risk sections are collected and input into the trained multi-scale synchronous insulation defect detection model to identify and locate the defect types and locations in the risk sections.

2. The method for detecting faults in power distribution lines according to claim 1, characterized in that, Based on the process data of power distribution line insulation degradation, the core influencing parameters of insulation degradation evolution are screened out, and based on the irreversibility and exponential decay characteristics of insulation aging, a line insulation degradation evolution model is constructed, including the following steps: Based on the process data of insulation degradation of power distribution lines, and combined with the physical mechanism and the characteristics of the field operating environment, the core influencing parameters of insulation degradation evolution were screened out. Based on the core influencing parameters and combined with the irreversibility and exponential decay characteristics of insulation aging, a differential equation for the evolution of insulation degradation is established and the general solution is solved to obtain a preliminary insulation degradation evolution model. By using a multi-factor correction method, the preliminary insulation degradation evolution model is modified to obtain the final analytical expression of the insulation degradation evolution parameter model. Based on the experimental values ​​obtained at different times through prior experiments, the unknown parameters in the final analytical expression of the insulation degradation evolution parameter model are fitted and solved to determine the final form of the final analytical expression of the insulation degradation evolution parameter model. The final form is subjected to relative error verification until a line insulation degradation evolution model that meets the preset error requirements is obtained.

3. The method for detecting power distribution line faults according to claim 2, characterized in that, Based on the process data of insulation degradation in power distribution lines, and combined with the physical mechanisms and characteristics of the field operating environment, the core influencing parameters of insulation degradation evolution were screened out, including the following steps: Based on the process data of insulation degradation of power distribution lines and combined with the micro-mechanism of insulation degradation, basic parameters including initial insulation, inherent degradation rate and aging accumulation were determined. Based on the actual operating environment of the line, environmental factor parameters including temperature, humidity, pollution and electric field are determined; Based on the characteristics of engineering monitoring and operation and maintenance, correction factor parameters including test error, operation and maintenance intervention and service life are determined; The basic parameters, environmental factors, and correction factors are integrated and summarized to form a complete set of core influencing parameters for insulation degradation evolution.

4. The method for detecting power distribution line faults according to claim 2, characterized in that, Based on the core influencing parameters and considering the irreversibility and exponential decay characteristics of insulation aging, the differential equation for insulation degradation evolution is established and the general solution is solved to obtain a preliminary insulation degradation evolution model. This process includes the following steps: Based on the irreversibility and exponential decay characteristics of insulation aging, and considering that the decay rate of insulation performance characterization parameters over time is positively correlated with the current insulation performance value and the degree of aging accumulation, a differential equation is established. The differential equation is separated into variables, and the integral result is obtained by performing equation transformations and integration on the separated differential equation. The integral results are processed by exponential operations, and the general solution of the differential equation is obtained by formula simplification, which serves as a preliminary insulation degradation evolution model.

5. The method for detecting faults in power distribution lines according to claim 2, characterized in that, The process of revising the preliminary insulation degradation evolution model using a multi-factor correction method to obtain the final analytical expression for the insulation degradation evolution parameter model includes the following steps: Based on the inherent degradation rate of the insulation material itself, and combined with the contribution of environmental factors to the degradation rate, the environmental impact item is calculated. Based on the correction factor parameters, correction terms are determined to correct the preliminary insulation degradation evolution model; The environmental impact term and the correction term are substituted into the preliminary insulation degradation evolution model, and the final analytical expression of the line insulation degradation evolution parameter model is obtained through derivation.

6. The method for detecting faults in power distribution lines according to claim 2, characterized in that, The process of fitting and solving the unknown parameters in the final analytical expression of the insulation degradation evolution parameter model based on experimental values ​​obtained at different times through prior experiments, and determining the final form of the final analytical expression of the insulation degradation evolution parameter model, includes the following steps: Obtain test values ​​at different times recorded in advance through accelerated insulation degradation tests, the test values ​​including quantified values ​​of environmental factors, correction factors and insulation performance characterization parameters; The objective function for parameter fitting is constructed with the goal of minimizing the mean square error between the experimental values ​​and the final analytical prediction values ​​of the insulation degradation evolution parameter model. The objective function is solved using the least squares method, and the optimal fitting values ​​of the unknown parameters of the final analytical expression of the insulation degradation evolution parameter model are obtained through iterative calculation. Substituting the optimal fitted value of the unknown parameters into the final analytical expression of the insulation degradation evolution parameter model, we obtain the analytical expression of the line insulation degradation evolution with determined parameter values, which serves as the final form of the final analytical expression of the insulation degradation evolution parameter model.

7. The method for detecting faults in power distribution lines according to claim 2, characterized in that, The process of verifying the relative error of the final form until a line insulation degradation evolution model that meets the preset error requirements is obtained includes the following steps: Select validation data that are independent of the experimental values ​​used to fit the unknown parameters, and substitute the quantified values ​​of environmental factors and correction coefficients in the validation data into the analytical expression after determining the final form, and calculate the predicted values ​​of insulation performance at each time point. Based on the quantified values ​​of environmental factors in the verification data and the calculated predicted values ​​of insulation performance, the relative error of each data point in the verification data is calculated. Determine whether the relative errors of all verification data points meet the preset error requirements. If they do, the analytical expression is deemed valid, and the construction of the line insulation degradation evolution model is completed. If the relative errors of some data points exceed the preset error requirements, readjust the fitting range of the unknown parameters and refit until the relative errors of all verification data points meet the preset error requirements.

8. The method for detecting faults in power distribution lines according to claim 1, characterized in that, The process of acquiring historical image data of power distribution lines and constructing and training a multi-scale synchronous insulation defect detection model using a two-stage model training method includes the following steps: Historical image data of power distribution lines are acquired, and multi-strategy data augmentation processing is performed to obtain enhanced historical image data. The enhanced historical image data is divided into a first dataset and a second dataset; A multi-scale synchronous insulation defect detection network was constructed, and the network was trained in two stages using the first and second datasets to obtain the multi-scale synchronous insulation defect detection model.

9. The method for detecting faults in power distribution lines according to claim 8, characterized in that, The process of acquiring historical image data of power distribution lines and performing multi-strategy data augmentation to obtain enhanced historical image data includes the following steps: Based on affine transformation, images in historical image data are rotated, scaled, translated, and sheared to obtain historical image data after affine transformation. By using Gaussian noise injection technology, the images in the affine transformed historical image data are filtered to generate image data of different resolutions, thus obtaining the filtered historical image data. Physically driven data augmentation technology based on electromagnetic field-visual coupling is used to enhance filtered historical image data with physical field constraints, resulting in enhanced historical image data.

10. The method for detecting faults in power distribution lines according to claim 9, characterized in that, The physical-driven data augmentation technique based on electromagnetic field-visual coupling applies physical field constraints to filtered historical image data to obtain enhanced historical image data, including the following steps: For the filtered historical image data, extract the line operating voltage and the line phase-to-phase or line-to-ground distance corresponding to each image, and calculate the distance field from each pixel to the defect area in the image. The normalized electromagnetic field strength value of each pixel is calculated based on the distance field from each pixel to the defect area in the image. Based on the normalized electromagnetic field intensity value of each pixel, and combined with the physical laws of electromagnetic fields and the pixel values ​​of each pixel in the original image, a synthetic image under electromagnetic field constraints is generated, and the synthetic image is used as enhanced historical image data.

11. The method for detecting faults in power distribution lines according to claim 8, characterized in that, The construction of a multi-scale synchronous insulation defect detection network, and the two-stage model training of the multi-scale synchronous insulation defect detection network using the first and second datasets to obtain the multi-scale synchronous insulation defect detection model, includes the following steps: Based on the pre-defined basic backbone feature extraction structure, a multi-scale synchronous insulation defect detection network is constructed. Based on the temporal consistency constraint method, a joint loss function is constructed; Based on the first and second datasets and combined with the joint loss function, a two-stage model training was performed on the multi-scale synchronous insulation defect detection network to obtain the multi-scale synchronous insulation defect detection model.

12. The method for detecting faults in power distribution lines according to claim 5, characterized in that, The final analytical expression for the insulation degradation evolution parameter model is as follows: ； In the formula, For the final insulation degradation evolution parameter model, The initial value of the insulation performance at the initial moment. This refers to the inherent degradation rate of the insulating material itself. Let be the influence coefficient of the degradation rate of the i-th environmental factor. This represents the actual quantified value of the environmental factor at time t. The cumulative aging index, for Time of the first The correction coefficients for each correction factor, where m is the total number of correction factor types and n is the number of environmental factor types.

13. The method for detecting faults in power distribution lines according to claim 6, characterized in that, The expression for the objective function fitted to the parameters is: ； In the formula, To minimize the objective function value by adjusting unknown parameters, The unknown parameters to be solved are: For the number of test data points, The result obtained by substituting the unknown parameters into the analytical expression The predicted insulation performance value at any given time, where r is the sequence number of the test data point. for The corresponding insulation performance characterization parameters at any given time.

14. The method for detecting faults in power distribution lines according to claim 11, characterized in that, The expression for the joint loss function is: ； ； ； ； ； In the formula, For the joint loss function, The loss is due to spatiotemporal consistency regularization constraints. Here, M represents the weighting coefficients for the spatiotemporal consistency constraint, and M represents the number of training sample batches. The first Frame and the The deep network features corresponding to the input image frame The loss function consists of classification confidence loss and localization loss. , Classification confidence loss With positioning loss The balance weight coefficient, This indicates the matching relationship between the predicted bounding box and the truth bounding box. The predicted probability for the target category. For the predicted bounding box coordinates, The coordinates of the truth box. To calculate the smoothing loss of coordinate differences.

15. A power distribution line fault detection system, characterized in that, The system includes: The evolution model construction module is used to screen out the core influencing parameters of insulation degradation evolution based on the process data of insulation degradation of distribution lines, and to construct a line insulation degradation evolution model based on the irreversibility and exponential decay characteristics of insulation aging. The risk identification module is used to identify risk sections of distribution lines with insulation degradation based on the line insulation degradation evolution model; The detection model construction module is used to acquire historical image data of power distribution lines and, combined with a two-stage model training method, construct and train a multi-scale synchronous insulation defect detection model. The defect identification module is used to acquire real-time inspection images of risk sections and input them into the trained multi-scale synchronous insulation defect detection model to identify and locate the defect type and location of the risk section.

16. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 14.

Images