GIS state detection method and system, intelligent terminal and storage medium

Abstract

The invention provides a GIS state detection method and system, an intelligent terminal and a storage medium, and relates to the field of electrical equipment detection.The method comprises the steps that the leakage risk coefficient of sulfur hexafluoride gas is analyzed and determined according to a metal shell image set; according to the leakage risk coefficient, the maintenance interval duration, the current environmental parameters and the expected operation load, performing gas monitoring threshold optimization by taking maximization of the early warning accuracy of the sulfur hexafluoride gas as a target to obtain an optimal monitoring threshold; and setting an adaptive early warning strategy according to the optimal monitoring threshold value, and carrying out monitoring and early warning on the GIS equipment. The objective of the invention is to solve the technical problems that a conventional gas state detection method of GIS equipment lacks a dynamic monitoring mechanism based on actual working conditions of the equipment, so that misjudgment or missed judgment is easy to occur under complex working conditions, and the detection precision and reliability are insufficient. The objective of self-adaptive early warning can be achieved, the accuracy and reliability of gas detection of GIS equipment are effectively improved, and intelligent operation and maintenance requirements are met.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of electrical equipment detection, and in particular to a GIS state detection method and system, an intelligent terminal and a storage medium. BACKGROUND

[0002] Gas insulated switchgear (GIS) is widely used in high-voltage power transmission and transformation systems due to its compact structure, excellent insulation performance and strong adaptability to harsh environments. The GIS is usually filled with sulfur hexafluoride gas as the main insulation medium and arc extinguishing medium. The density state of the sulfur hexafluoride gas directly affects the insulation level and operational safety of the equipment. Therefore, real-time monitoring of the state of the sulfur hexafluoride gas has become a key means to ensure the safe and stable operation of GIS equipment.

[0003] In the prior art, the monitoring of sulfur hexafluoride gas usually relies on physical detection means such as pressure, temperature and density sensors, and alarm judgment is made according to pre-set static thresholds. However, such methods generally have the following shortcomings: on the one hand, the monitoring thresholds are usually fixed and lack a dynamic adjustment mechanism for multiple factors such as the operating environment of the equipment, the actual load state and the aging degree of the equipment, leading to false positives or false negatives under complex working conditions such as high-frequency breaking, power load fluctuations or drastic environmental changes; on the other hand, traditional detection methods ignore the actual leakage risks caused by corrosion, cracks or aging of the seal of the equipment body (such as a metal enclosed shell), thereby limiting the accuracy and reliability of the overall detection system. SUMMARY

[0004] The purpose of the present application is to provide a GIS state detection method, system, intelligent terminal and storage medium to solve the technical problem that the gas state detection method of traditional GIS equipment lacks a dynamic monitoring mechanism based on the actual working conditions of the equipment, leading to false positives or false negatives under complex working conditions, resulting in insufficient accuracy and reliability of the detection, including: In a first aspect, the present application provides a GIS state detection method, comprising: based on the last operation and maintenance record of the GIS equipment, obtaining a metal shell image set and a maintenance interval duration, and determining a leakage risk coefficient of sulfur hexafluoride gas according to the metal shell image set; monitoring and obtaining the current environmental parameters of the area where the GIS equipment is located within a preset time zone, and the expected operating load of the GIS equipment; according to the leakage risk coefficient, the maintenance interval duration, the current environmental parameters and the expected operating load, optimizing the gas monitoring threshold to maximize the early warning accuracy of the sulfur hexafluoride gas, and obtaining the optimal monitoring threshold; setting an adaptive early warning strategy according to the optimal monitoring threshold, and monitoring and warning the sulfur hexafluoride gas in the GIS equipment within the preset time zone.

[0005] Preferably, the GIS state detection method further comprises: acquiring a metal shell image set, wherein the metal shell image set comprises metal shell images of a plurality of key positions; performing defect identification on the metal shell images of the plurality of key positions by using a convolutional neural network to obtain a plurality of defect features; and combining an experience weight model to determine a leakage risk coefficient according to the plurality of defect features.

[0006] Preferably, the GIS state detection method further comprises: acquiring a gas monitoring index of sulfur hexafluoride gas and a monitoring index adjustment space, wherein the gas monitoring index comprises gas density, gas pressure, gas temperature, moisture content, and decomposition product concentration; training an early warning accuracy predictor of the sulfur hexafluoride gas according to the leakage risk coefficient, the maintenance interval length, the current environmental parameters, and the expected operating load; and performing gas monitoring threshold optimization by using the early warning accuracy predictor based on the monitoring index adjustment space to maximize the early warning accuracy of the sulfur hexafluoride gas, and outputting an optimal monitoring threshold.

[0007] Preferably, the GIS state detection method further comprises: taking the leakage risk coefficient, the maintenance interval length, the current environmental parameters, and the expected operating load as retrieval constraints, expanding the retrieval constraints according to a preset tolerance interval to obtain retrieval conditions; screening historical monitoring data of sulfur hexafluoride gas of similar GIS devices according to the retrieval conditions, collecting a sample gas monitoring threshold set, and statistically analyzing the probability of false positives and false negatives of sulfur hexafluoride gas index early warning under different sample gas monitoring thresholds to calculate a sample early warning accuracy and obtain a sample early warning accuracy set; and training a generative adversarial network to convergence by using the sample gas monitoring threshold set and the sample early warning accuracy set to obtain the early warning accuracy predictor.

[0008] Preferably, the GIS state detection method further comprises: using the sample gas monitoring threshold set and the sample early warning accuracy set as training data, and equally dividing the training data into P parts, randomly selecting P times in the P parts of data sets to obtain a first training set, and iteratively selecting P times to obtain P training sets, wherein P is an integer greater than or equal to 10; training a generative adversarial network to convergence by using the P training sets to obtain P early warning accuracy predictor branches, and combining the P early warning accuracy predictor branches to obtain the early warning accuracy predictor.

[0009] Preferably, the GIS state detection method further comprises: randomly selecting in the monitoring index adjustment space to obtain a plurality of initial monitoring thresholds; taking a standard monitoring threshold as a reference, calculating the deviation amplitudes of the plurality of initial monitoring thresholds respectively, determining a plurality of deviation values by weighting, and calculating a plurality of branch selection quantities Q, wherein Q is the integer value of the product of the deviation value and P, Q is greater than or equal to 1 and less than or equal to P; based on the plurality of branch selection quantities Q, predicting the plurality of initial monitoring thresholds by using the early warning accuracy predictor respectively, and outputting a plurality of predicted early warning accuracies; according to the plurality of predicted early warning accuracies, optimizing the gas monitoring threshold with the goal of maximizing the early warning accuracy of sulfur hexafluoride gas, and outputting an optimal monitoring threshold.

[0010] Preferably, the GIS state detection method further comprises: setting an initial monitoring threshold as an initial solution, sorting the plurality of initial monitoring thresholds according to the plurality of predicted early warning accuracies from large to small to obtain an initial solution sequence; setting the first K solutions of the initial solution sequence as good solutions and the last M solutions as poor solutions, and randomly performing equal value clustering on the M poor solutions with the K good solutions as the center to obtain K solution sets, wherein M is N times of K, and N is greater than or equal to 10; in the K solution sets, adjusting the poor solutions in the same solution set according to a preset optimization step size with the good solutions as the adjustment direction to obtain K updated solution sets, wherein if the adjusted poor solution does not meet the monitoring index adjustment space, it is removed and a monitoring threshold is randomly selected for supplementation; identifying the K updated solution sets, if the predicted early warning accuracy of the updated poor solution is greater than that of the good solution in the same solution set, replacing the good solution with the poor solution; performing iterative optimization until a preset optimization number is reached, outputting K current solution sets, selecting the solution set with the maximum sum of predicted early warning accuracies as the optimal solution set, and setting the good solution of the optimal solution set as the optimal monitoring threshold.

[0011] In a second aspect, the present application further provides a GIS state detection system for executing the GIS state detection method of the first aspect, comprising: a leakage risk coefficient determination module for obtaining a metal shell image set and a maintenance interval length based on the last operation and maintenance record of the GIS device, and determining the leakage risk coefficient of sulfur hexafluoride gas according to the metal shell image set; a data monitoring and collection module for monitoring and obtaining the current environmental parameters of the area where the GIS device is located in a preset time zone, and the expected operating load of the GIS device; a gas monitoring threshold optimization module for optimizing the gas monitoring threshold with the goal of maximizing the early warning accuracy of sulfur hexafluoride gas according to the leakage risk coefficient, the maintenance interval length, the current environmental parameters and the expected operating load, and obtaining an optimal monitoring threshold; and a device monitoring and early warning module for setting an adaptive early warning strategy according to the optimal monitoring threshold, and monitoring and early warning the sulfur hexafluoride gas in the GIS device in the preset time zone.

[0012] In a third aspect, the present application further provides an intelligent terminal, comprising: at least one processor; a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of any one of the methods in the first aspect.

[0013] In a fourth aspect, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program, and the computer program implements the steps of any one of the methods in the first aspect when executed.

[0014] Embodiments of the present application have the following advantages: By means of the GIS device last time operation and maintenance record, the metal shell image set and maintenance interval duration are obtained, and the leakage risk coefficient of sulfur hexafluoride gas is determined according to the metal shell image set analysis; on the other hand, the current environmental parameters of the area where the GIS device is located in the preset time zone are monitored and obtained, and the expected running load of the GIS device is monitored and obtained; then, according to the leakage risk coefficient, the maintenance interval duration, the current environmental parameters and the expected running load, the gas monitoring threshold optimization is carried out to maximize the early warning accuracy of sulfur hexafluoride gas, and the optimal monitoring threshold is obtained; finally, the adaptive early warning strategy is set according to the optimal monitoring threshold, and the sulfur hexafluoride gas in the GIS device is monitored and warned in the preset time zone. That is, by fusing metal shell image recognition, environmental parameter sensing and running load modeling, a multi-dimensional state sensing and evaluation mechanism is constructed, and the monitoring threshold of sulfur hexafluoride gas is dynamically optimized, so that the adaptive early warning target can be realized, the accuracy and reliability of GIS device gas detection can be effectively improved, and the needs of intelligent operation and maintenance and fine management can be met. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 A step flow chart of the GIS state detection method of the present application; Figure 2 A structural schematic diagram of the GIS state detection system of the present application; Figure 3 A structural schematic diagram of the intelligent terminal provided by the present application; Figure 4 A structural schematic diagram of the computer readable storage medium provided by the present application.

[0016] In the drawings, the components represented by the numbers are described as follows: The leakage risk coefficient determination module 11, the data monitoring and collection module 12, the gas monitoring threshold optimization module 13, the equipment monitoring and early warning module 14, the intelligent terminal 500, the memory 510, the processor 520, the first computer program 511, the computer readable storage medium 600, and the second computer program 611. DETAILED DESCRIPTION

[0017] The GIS state detection method and system provided by the application solve the technical problem that the gas state detection method of the conventional GIS equipment lacks a dynamic monitoring mechanism based on the actual working condition of the equipment, which leads to misjudgment or missed judgment under complex working conditions, and causes insufficient accuracy and reliability of detection. By fusing metal shell image recognition, environmental parameter sensing and running load modeling, a multi-dimensional state sensing and evaluation mechanism is constructed, and the monitoring threshold of sulfur hexafluoride gas is dynamically optimized, so that the goal of adaptive early warning can be achieved, and the accuracy and reliability of GIS equipment gas detection can be effectively improved, thereby meeting the needs of intelligent operation and maintenance and fine management.

[0018] The technical solutions in the application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the application, rather than all the embodiments of the application. It should be understood that the application is not limited by the example embodiments described herein. Based on the embodiments of the application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the application. In addition, it should be noted that, for the convenience of description, only the parts related to the application are shown in the drawings, not all.

[0019] Embodiment one, please refer to the attached Figure 1 The application provides a GIS state detection method applied to a GIS state detection system, and specifically includes the following steps: S10: Based on the last operation and maintenance record of the GIS equipment, a metal shell image set and a maintenance interval length are obtained, and a leakage risk coefficient of sulfur hexafluoride gas is determined according to the metal shell image set.

[0020] Further, the step S10 of the application further includes: S11: Obtain a metal shell image set, wherein the metal shell image set includes metal shell images of a plurality of key positions; S12: Use a convolutional neural network to perform defect recognition on the metal shell images of the plurality of key positions to obtain a plurality of defect features; and S13: Combine an experience weight model to determine a leakage risk coefficient according to the plurality of defect features.

[0021] Specifically, first, the last maintenance record of the GIS device is read, the last maintenance record refers to the relevant data and documents of the last maintenance, repair or inspection of the GIS device, which usually includes maintenance date, maintenance content, problems found and treatment results, etc., and is an important basis for evaluating the current state of the device. Then, according to the last maintenance record of the GIS device, the metal shell image set and the maintenance interval length are obtained, wherein the metal closed shell of the GIS device is fully shot by using a high-definition camera or an industrial camera, and the images of multiple key positions are collected, such as the joint, the sealing ring area, the position where corrosion or cracks may occur, etc. These image data constitute an image set, which is used for subsequent image processing and analysis to identify physical defects and abnormalities on the surface of the device; the maintenance interval length is the time interval from the last maintenance date to the current time, which reflects the duration of the device without maintenance, and is an important parameter for evaluating the aging degree and potential risk of the device. The longer the maintenance interval, the greater the possibility of device aging, sealing performance degradation or increased leakage risk.

[0022] Then, the metal shell image set is obtained, wherein the metal shell image set includes metal shell images of several key positions, and the key positions refer to areas in the metal shell that are prone to defects or have a high risk of leakage, such as joints, welding points, sealing ring areas, surface corrosion areas or positions where cracks may occur. Then, a convolutional neural network is used to identify defects in the metal shell images of the several key positions. The convolutional neural network is a deep learning model that automatically extracts spatial features in images through multiple convolutional layers and pooling layers, and classifies or detects through fully connected layers. The convolutional neural network model uses its convolution kernel to slide on the image and automatically identify edges, textures, color changes and other details in the image. These features are crucial for detecting defects such as cracks, corrosion spots and wear. The network distinguishes between possible defect areas and normal areas in the image through trained weights and biases, locates and classifies defects, and obtains several defect features, including defect type (such as crack, corrosion, pit, paint removal, etc.), defect location (specific coordinates or area on the actual device), defect size (area, length or volume of the defect), defect severity and defect morphology features (such as crack direction, corrosion spot shape, etc. auxiliary identification information).

[0023] Then, the experience weight model is set, which is a mathematical model based on expert experience and historical data, used to quantify the influence of different defect characteristics on the risk of sulfur hexafluoride gas leakage. Different defects have different contributions to the risk of leakage. For example, a crack may have a higher risk weight than a slight surface corrosion. The weight model can use weighted summation, a scoring system, or machine learning methods to convert multi-dimensional defect characteristics into a single risk indicator. Different base weights are assigned according to the type of defect (e.g., a crack has a higher weight than corrosion). The weight is adjusted according to the size of the defect (the larger the defect, the higher the weight); and the weight is corrected in combination with the location of the defect and the importance of the key parts of the equipment (the weight of the defect in the key parts is higher). Finally, all defect characteristics are calculated according to the weight model to obtain a numerical leakage risk coefficient, which reflects the overall leakage risk of the current metal shell. The larger the coefficient value, the worse the sealing state of the equipment, and the higher the risk of leakage.

[0024] S20: Monitor and acquire the current environmental parameters of the area where the GIS device is located within the preset time zone, and the expected operating load of the GIS device.

[0025] Specifically, a preset time zone is configured, which refers to a specific time period set in advance, such as a few hours of a day, a work cycle, or a maintenance cycle. Monitoring and data collection are limited within this time range to ensure the timeliness and relevance of the data. Then, the current environmental parameters of the area where the GIS device is located within the preset time zone are monitored and acquired, as well as the expected operating load of the GIS device. The current environmental parameters refer to real-time monitoring of the geographical location and surrounding environment of the GIS device, collecting environmental variables that can affect the operation of the device and the state of sulfur hexafluoride gas, including but not limited to temperature, humidity, pressure, etc. The environmental temperature directly affects the pressure and density of sulfur hexafluoride gas, thereby affecting the insulation performance and the accuracy of leakage detection. Changes in humidity can cause condensation inside the device, accelerating the corrosion of metal parts and affecting the sealing performance of the device. Changes in environmental air pressure affect the baseline of gas state parameters and require adjustment of the gas monitoring threshold. These environmental parameters are usually collected in real time by sensor devices installed on site (such as temperature and humidity sensors, air pressure sensors, etc.) and transmitted to the monitoring system through wireless or wired networks. The expected operating load refers to the electrical load and operating state that the GIS device may bear within the preset time zone, calculated based on the power grid scheduling plan, device operation history, and prediction analysis, including current load size, switch action frequency, etc. The larger the current and load during device operation, the higher the device working pressure and fault risk. The number of operations of switch devices such as circuit breakers may exacerbate mechanical wear and gas leakage risk.

[0026] By collecting environmental parameters and expected load information in real time, potential risks can be identified in advance in combination with environmental and load changes, thereby providing reliable real-time data basis for intelligent operation and maintenance systems.

[0027] S30: According to the leakage risk coefficient, the maintenance interval length, the current environmental parameter and the expected operation load, a gas monitoring threshold value is optimized to maximize the early warning accuracy of sulfur hexafluoride gas, and an optimal monitoring threshold value is obtained.

[0028] Further, the step S30 of the present application further comprises: S31: Obtain a gas monitoring index of sulfur hexafluoride gas, and a monitoring index adjustment space, wherein the gas monitoring index comprises gas density, gas pressure, gas temperature, moisture content and decomposition product concentration.

[0029] Specifically, first, the gas monitoring index of sulfur hexafluoride gas is obtained, and the gas monitoring index comprises gas density, gas pressure, gas temperature, moisture content and decomposition product concentration, wherein the gas density reflects the mass of sulfur hexafluoride gas in a unit volume, and abnormal change of the density may mean gas leakage or abnormal compression state, which is an important parameter for determining whether the gas meets the insulation and sealing requirements; the gas pressure refers to the pressure state of the gas in the GIS device, and a decrease in pressure usually indicates a leakage risk, and excessively high pressure may affect the safe operation of the device, and the pressure value determines the gas state together with the density and temperature; the temperature of the gas affects its physical properties and pressure, and abnormal fluctuation of the temperature may be related to abnormal operation of the device or environmental change, and the temperature is a reference for adjusting other monitoring indexes; moisture (humidity) is an important impurity in sulfur hexafluoride gas, and excessive moisture content may lead to a decrease in insulation performance and corrosion of the device, accelerating the aging of the device, and monitoring the moisture content helps to discover internal sealing problems in time. Sulfur hexafluoride will decompose to produce harmful gases (such as SO2, HF, etc.) under electrical fault or overheating conditions, and the decomposition product concentration is a direct index of partial discharge or fault of the device, and an increase in the decomposition product concentration usually indicates that the device has a safety hazard. Then, the monitoring index adjustment space of sulfur hexafluoride gas is obtained, and the monitoring index adjustment space refers to the ability to dynamically adjust the warning threshold or alarm range of the above monitoring indexes based on the current state of the device, environmental conditions and operation load.

[0030] S32: According to the leakage risk coefficient, the maintenance interval length, the current environmental parameter and the expected operation load, a gas monitoring threshold value is optimized to maximize the early warning accuracy of sulfur hexafluoride gas, and an optimal monitoring threshold value is obtained.

[0031] Further, the step S32 of the present application further comprises: S321: taking the leakage risk coefficient, the maintenance interval length, the current environment parameter and the expected operation load as retrieval constraints, expanding the retrieval constraints according to a preset tolerance interval, obtaining a retrieval condition; S322: screening historical monitoring data of sulfur hexafluoride gas of the same type GIS device according to the retrieval condition, collecting a sample gas monitoring threshold set, and statistically calculating the probability of false positives and false negatives of the sulfur hexafluoride gas index under different sample gas monitoring thresholds, to obtain a sample early warning accuracy rate set.

[0032] Specifically, first, the leakage risk coefficient, the maintenance interval length, the current environment parameter and the expected operation load are set as constraint conditions for data retrieval; then, a preset tolerance interval is configured, which is a reasonable variation range set by the system for each type of input parameter, used to expand the sample data matching space, avoid insufficient sample size due to too harsh conditions, and unable to perform effective threshold optimization, for example, if the leakage risk coefficient is 0.6 and the tolerance interval is ±0.1, then the historical data matching is allowed within the range of 0.5-0.7; if the current environment temperature is 35°C and the tolerance interval is set to ±5°C, then the matching data accepts records with an environment temperature range of 30-40°C. Then, the retrieval constraints are expanded according to the preset tolerance interval, that is, a reasonable upper and lower floating range is set for each variable to form an interval type retrieval condition, and all expanded parameters are combined into a multi-dimensional condition vector as a standard for querying historical data or simulation data, used to retrieve similar records or working condition samples in the historical database. By setting the tolerance interval to expand the matching range, suitable historical data or model samples can be efficiently obtained to provide accurate and sufficient data support for the next step of threshold optimization.

[0033] Then, according to the retrieval condition, the historical monitoring data of sulfur hexafluoride gas of the same type GIS device is screened to collect a sample gas monitoring threshold set; then the probability of false positives and false negatives of the sulfur hexafluoride gas index under different sample gas monitoring thresholds is statistically calculated, and the sample early warning accuracy rate is obtained by subtracting the false positive rate and the false negative rate from 1, to obtain a sample early warning accuracy rate set.

[0034] S323: using the sample gas monitoring threshold set and the sample early warning accuracy rate set to train the generative adversarial network to convergence, to obtain an early warning accuracy rate predictor.

[0035] Further, the step S323 of the present application further comprises: S3231: using the sample gas monitoring threshold set and the sample early warning accuracy set as training data, and equally dividing into P parts, in the P part data set, selecting P times with replacement, obtaining a first training set, and iteratively selecting P times to obtain P training sets, wherein P is an integer greater than or equal to 10; S3232: using the P training sets to train the generative adversarial network to convergence, obtaining P early warning accuracy prediction branches, and combining to obtain an early warning accuracy predictor.

[0036] Specifically, first, the sample gas monitoring threshold set and the sample early warning accuracy set are used as training data, and the training data is equally divided into P parts, wherein P is an integer greater than or equal to 10; then, in the P part data set, P times are selected with replacement to obtain a first training set, and P times are iteratively selected to obtain P training sets. By cross-validation method, multiple training data subsets for model training and verification are generated, providing diverse and robust training samples for subsequent model training.

[0037] Then, using the P training sets, the generative adversarial network is trained to convergence respectively, obtaining P early warning accuracy prediction branches, wherein the generative adversarial network is composed of a generator and a discriminator, the generator predicts an "accuracy" output by inputting a set of gas thresholds, simulating the real early warning accuracy; the discriminator is used to distinguish between the accuracy predicted by the generator and the real sample accuracy data. This kind of adversarial structure can effectively learn complex mapping relationship and perform well in nonlinear and multivariate correlation modeling; in the training process, the generator is responsible for predicting the early warning accuracy according to the input threshold, and the discriminator judges the difference between the prediction result and the real accuracy, and through the game optimization between the generator and the discriminator, the prediction ability of the generator for the accuracy is gradually improved, until the model converges, and finally multiple accuracy prediction branches are obtained. The training process can effectively capture the complex nonlinear relationship between the monitoring threshold and the early warning performance, and provide reliable data support for subsequent threshold optimization and adaptive adjustment; P early warning accuracy prediction branches are obtained, and an early warning accuracy predictor is combined according to the P early warning accuracy prediction branches.

[0038] S33: based on the monitoring index adjustment space, taking maximizing the early warning accuracy of sulfur hexafluoride gas as the target, using the early warning accuracy predictor to optimize the gas monitoring threshold, and outputting the optimal monitoring threshold.

[0039] Further, the step S33 of the present application further comprises: S331: randomly selecting in the monitoring index adjustment space to obtain a plurality of initial monitoring thresholds; S332: taking a standard monitoring threshold as a benchmark, respectively calculating deviation amplitudes of the plurality of initial monitoring thresholds, determining a plurality of deviation values by weighting, and calculating a plurality of branch selection quantities Q, wherein Q is an integral value of a product of the deviation value and P, Q is greater than or equal to 1 and less than or equal to P; S333: based on the plurality of branch selection quantities Q, respectively predicting the plurality of initial monitoring thresholds by using the early warning accuracy predictor, and outputting a plurality of predicted early warning accuracies.

[0040] Specifically, first, a batch of threshold sets are randomly selected in the monitoring index adjustment space by a certain strategy (such as uniform sampling or Gaussian sampling), which are used for subsequent performance evaluation and optimization screening. The goal of this step is to form an initial threshold set with representativeness and diversity, to prepare for the selection of the optimal threshold. Then, taking a standard monitoring threshold (such as the threshold set by the factory or the most commonly used threshold in the existing working condition) as a benchmark, the deviation amplitudes of the plurality of initial monitoring thresholds are respectively calculated, and the deviation amplitude is the ratio of the difference between the two to the index standard value. At the same time, in order to more reasonably evaluate the relative representativeness of different thresholds, a weighting factor (such as some parameters being more sensitive to accuracy) is introduced to weight each deviation amplitude to obtain the final deviation value, and a plurality of deviation values are obtained. Then, the deviation value of each initial threshold is multiplied by P, and the result is rounded to obtain the branch selection quantity Q. Finally, based on the plurality of branch selection quantities Q, Q early warning accuracy prediction branches are randomly selected in the P early warning accuracy prediction branches of the early warning accuracy predictor, and the plurality of initial monitoring thresholds are respectively predicted to obtain a plurality of predicted early warning accuracies, wherein the predicted early warning accuracy is the average of the output results of the Q early warning accuracy prediction branches.

[0041] By setting the corresponding branch selection quantity according to the actual deviation of the initial monitoring threshold relative to the standard threshold, that is, for thresholds with larger deviations, more prediction branches are allocated to enhance the robustness of the accuracy evaluation; and for thresholds with smaller deviations, only a small number of branches are used for rapid evaluation, thereby effectively saving the computing resources, improving the system operation efficiency and response speed, and enhancing the adaptability and intelligent level of the monitoring threshold optimization process.

[0042] S334: according to the plurality of predicted early warning accuracies, performing gas monitoring threshold optimization with the goal of maximizing the early warning accuracy of sulfur hexafluoride gas, and outputting an optimal monitoring threshold.

[0043] Further, the step S334 of the present application further comprises: S3341: set initial monitoring thresholds as initial solutions, sort the initial monitoring thresholds according to the prediction warning accuracies from large to small to obtain an initial solution sequence; S3342: set the first K solutions in the initial solution sequence as good solutions, set the last M solutions as poor solutions, and perform random equal-value clustering on the M poor solutions with the K good solutions as the center to obtain K solution sets, wherein M is N times of K, and N is greater than or equal to 10; S3343: in the K solution sets, adjust the poor solutions in the same solution set according to a preset optimization step length with the good solutions as the adjustment direction to obtain K updated solution sets, wherein if the adjusted poor solution does not satisfy the monitoring index adjustment space, the poor solution is removed and a monitoring threshold is randomly selected for supplement; S3344: identify the K updated solution sets, and if the prediction warning accuracy of the updated poor solution is greater than that of the good solution in the same solution set, replace the good solution with the poor solution; S3345: perform iterative optimization until a preset optimization number is reached, output K current solution sets, select a solution set with the maximum sum of prediction warning accuracies as an optimal solution set, and set the good solution of the optimal solution set as an optimal monitoring threshold.

[0044] Specifically, first, initial monitoring thresholds are set as initial solutions, the initial monitoring thresholds are sorted according to the warning accuracies from large to small according to the prediction warning accuracies to obtain an initial solution sequence; then, based on the heuristic idea, high-performance solutions and low-performance solutions are classified respectively to enhance the search directionality, that is, the first K solutions in the initial solution sequence are set as good solutions, the last M solutions are set as poor solutions, M is N times of K, N is greater than or equal to 10, and the poor solutions are divided into K solution sets using the equal-value random clustering method with the good solutions as the center. Further, in the K solution sets, the poor solutions in the same solution set are adjusted according to a preset optimization step length with the good solutions as the adjustment direction, wherein if the adjusted poor solution does not satisfy the monitoring index adjustment space, the poor solution is removed and a monitoring threshold is randomly selected for supplement to obtain K updated solution sets, that is, the poor solutions are allowed to move towards the good solutions to improve the overall accuracy of the solution sets, and the new thresholds generated randomly are used to replace the poor solutions to ensure the completeness of the solution sets, increase the diversity of local search, and prevent falling into local optimum. Then, the K updated solution sets are identified, and if the prediction warning accuracy of the updated poor solution is greater than that of the good solution in the same solution set, the good solution is replaced with the poor solution, new excellent solutions are introduced by dynamically optimizing the good solution set, and the global search efficiency is improved; then, iterative optimization is continued, and the solution set clustering, guided adjustment and replacement operations are repeatedly performed until a preset iteration number is reached, K current solution sets are output, a solution set with the maximum sum of prediction warning accuracies is selected as an optimal solution set, and the good solution of the optimal solution set is set as an optimal monitoring threshold.

[0045] By constructing multiple initial monitoring threshold solutions and fusing prediction accuracy feedback information, with solution set clustering and guided poor solution adjustment strategy, a multi-solution collaborative evolution optimization framework with self-learning ability is realized; through the structure design of taking good solutions as the guide and poor solutions as the search main body, not only the optimal threshold combination can be dynamically discovered, but also the convergence speed and global optimal coverage rate of the threshold optimization process are significantly improved; in addition, through the branch allocation mechanism driven by the deviation degree and the dynamic elimination and replacement mechanism, the calculation resource overhead is effectively controlled, the optimal balance between accuracy improvement and computing efficiency is realized, and the actual needs of intelligent adaptive optimization of GIS device gas monitoring threshold are met.

[0046] S40: According to the optimal monitoring threshold, an adaptive alarm strategy is set, and the sulfur hexafluoride gas in the GIS device is monitored and alarmed in the preset time zone.

[0047] Specifically, first, an adaptive alarm strategy is set according to the optimal monitoring threshold, wherein the strategy setting needs to consider the floating range of threshold adjustment and the prediction error tolerance to prevent excessive sensitivity or delayed reaction; then, the sulfur hexafluoride gas in the GIS device is monitored and alarmed in the preset time zone. Compared with the traditional fixed threshold strategy, this method can significantly reduce the false alarm and missed alarm rate and enhance the perception ability of the device state change in complex environment, thereby improving the timeliness and accuracy of fault alarm, realizing intelligent operation and maintenance and risk control based on state.

[0048] In summary, the GIS state detection method provided by the application has the following technical effects: By obtaining the metal shell image set and the maintenance interval length based on the last operation record of the GIS device, and determining the leakage risk coefficient of the sulfur hexafluoride gas according to the metal shell image set, on the other hand, the current environmental parameters of the area where the GIS device is located and the expected running load of the GIS device are monitored and obtained; then, according to the leakage risk coefficient, the maintenance interval length, the current environmental parameters and the expected running load, the gas monitoring threshold optimization is carried out to maximize the alarm accuracy of the sulfur hexafluoride gas, and the optimal monitoring threshold is obtained; finally, an adaptive alarm strategy is set according to the optimal monitoring threshold, and the sulfur hexafluoride gas in the GIS device is monitored and alarmed in the preset time zone. That is, by fusing metal shell image recognition, environmental parameter perception and running load modeling, a multi-dimensional state perception and evaluation mechanism is constructed, and the monitoring threshold of the sulfur hexafluoride gas is dynamically optimized, so that the adaptive alarm goal can be realized, the accuracy and reliability of GIS device gas detection can be effectively improved, and the needs of intelligent operation and fine management can be met.

[0049] Embodiment two, based on the same inventive concept as the GIS state detection method in the preceding embodiment, the present application also provides a GIS state detection system, please refer to the attached Figure 2 , comprising: a leakage risk coefficient determination module 11, configured to obtain a metal shell image set and a maintenance interval length based on the last operation and maintenance record of the GIS device, and determine the leakage risk coefficient of sulfur hexafluoride gas according to the metal shell image set analysis; a data monitoring and collection module 12, configured to monitor and obtain the current environmental parameters of the area where the GIS device is located within a preset time zone, and the expected running load of the GIS device; a gas monitoring threshold optimization module 13, configured to optimize the gas monitoring threshold according to the leakage risk coefficient, the maintenance interval length, the current environmental parameters and the expected running load, with the goal of maximizing the early warning accuracy of sulfur hexafluoride gas, and obtain the optimal monitoring threshold; a device monitoring and early warning module 14, configured to set an appropriate early warning strategy according to the optimal monitoring threshold, and monitor and early warn the sulfur hexafluoride gas in the GIS device within the preset time zone.

[0050] Further, the GIS state detection system is also used to: obtain a metal shell image set, wherein the metal shell image set includes metal shell images of several key parts; use a convolutional neural network to identify defects in the metal shell images of the several key parts, and obtain several defect features; combine an experience weight model to evaluate and determine the leakage risk coefficient according to the several defect features.

[0051] Further, the GIS state detection system is also used to: obtain a gas monitoring index of sulfur hexafluoride gas, and a monitoring index adjustment space, wherein the gas monitoring index includes gas density, gas pressure, gas temperature, moisture content and decomposition product concentration; train an early warning accuracy predictor of sulfur hexafluoride gas according to the leakage risk coefficient, the maintenance interval length, the current environmental parameters and the expected running load; based on the monitoring index adjustment space, use the early warning accuracy predictor to optimize the gas monitoring threshold with the goal of maximizing the early warning accuracy of sulfur hexafluoride gas, and output the optimal monitoring threshold.

[0052] Further, the GIS state detection system is further configured to: take the leakage risk coefficient, the maintenance interval length, the current environmental parameter, and the expected operation load as retrieval constraints, expand the retrieval constraints according to a preset tolerance interval, obtain retrieval conditions; according to the retrieval conditions, filter historical monitoring data of sulfur hexafluoride gas of similar GIS devices, collect a sample gas monitoring threshold set, and statistically obtain the probability of false positives and false negatives of sulfur hexafluoride gas index early warning under different sample gas monitoring thresholds, calculate a sample early warning accuracy rate, and obtain a sample early warning accuracy rate set; and train the generative adversarial network to convergence using the sample gas monitoring threshold set and the sample early warning accuracy rate set, and obtain an early warning accuracy rate predictor.

[0053] Further, the GIS state detection system is further configured to: take the leakage risk coefficient, the maintenance interval length, the current environmental parameter, and the expected operation load as retrieval constraints, expand the retrieval constraints according to a preset tolerance interval, obtain retrieval conditions; according to the retrieval conditions, filter historical monitoring data of sulfur hexafluoride gas of similar GIS devices, collect a sample gas monitoring threshold set, and statistically obtain the probability of false positives and false negatives of sulfur hexafluoride gas index early warning under different sample gas monitoring thresholds, calculate a sample early warning accuracy rate, and obtain a sample early warning accuracy rate set; and train the generative adversarial network to convergence using the sample gas monitoring threshold set and the sample early warning accuracy rate set, and obtain an early warning accuracy rate predictor.

[0054] Further, the GIS state detection system is further configured to: randomly select a plurality of initial monitoring thresholds in the monitoring index adjustment space; take a standard monitoring threshold as a reference, respectively calculate the deviation amplitudes of the plurality of initial monitoring thresholds, determine a plurality of deviation values by weighting, and calculate a plurality of branch selection quantities Q, wherein Q is the integer value of the product of the deviation value and P, Q is greater than or equal to 1 and less than or equal to P; based on the plurality of branch selection quantities Q, respectively predict the plurality of initial monitoring thresholds using the early warning accuracy rate predictor, output a plurality of predicted early warning accuracy rates; and according to the plurality of predicted early warning accuracy rates, perform gas monitoring threshold optimization with the maximum sulfur hexafluoride gas early warning accuracy rate as the target, and output an optimal monitoring threshold.

[0055] Furthermore, the GIS status detection system is also used for: setting initial monitoring thresholds as initial solutions; sorting multiple initial monitoring thresholds according to their early warning accuracy rates from largest to smallest, based on the multiple prediction and early warning accuracy rates, to obtain an initial solution sequence; designating the first K solutions of the initial solution sequence as good solutions and the last M solutions as poor solutions; and performing random equal-value clustering on the M poor solutions centered on the K good solutions to obtain K solution sets, where M is N times K, and N is greater than or equal to 10; within the K solution sets, adjusting in the direction of good solutions according to a preset optimization step size... The suboptimal solutions within a solution set are adjusted to obtain K updated solution sets. If the adjusted suboptimal solutions do not meet the adjustment space of the monitoring index, they are discarded, and a monitoring threshold is randomly selected to supplement them. The K updated solution sets are identified. If the prediction and warning accuracy of the updated suboptimal solutions is greater than the prediction and warning accuracy of the good solutions in the same solution set, the suboptimal solutions are used to replace the good solutions. Iterative optimization is performed until a preset number of optimizations is reached. K current solution sets are output. The solution set with the largest sum of prediction and warning accuracies is selected as the optimal solution set, and the good solutions of the optimal solution set are set as the optimal monitoring threshold.

[0056] Example 3, please refer to Figure 3 , Figure 3 This is a schematic diagram of an embodiment of a smart terminal provided by the present invention. For example... Figure 3 As shown, this embodiment of the invention provides a smart terminal 500, including a memory 510, a processor 520, and a first computer program 511 stored in the memory 510 and executable on the processor 520. When the processor 520 executes the first computer program 511, it performs the following steps: based on the most recent maintenance record of the GIS equipment, it acquires a set of metal casing images and maintenance interval duration, and analyzes and determines the leakage risk coefficient of sulfur hexafluoride gas based on the metal casing image set; it monitors and acquires the current environmental parameters of the area where the GIS equipment is located within a preset time zone, as well as the expected operating load of the GIS equipment; based on the leakage risk coefficient, maintenance interval duration, current environmental parameters, and expected operating load, it optimizes the gas monitoring threshold to maximize the early warning accuracy of sulfur hexafluoride gas, and obtains the optimal monitoring threshold; based on the optimal monitoring threshold, it sets an adaptive early warning strategy to monitor and warn of sulfur hexafluoride gas in the GIS equipment within the preset time zone.

[0057] Example 4, please refer to Figure 4 , Figure 4 This is a schematic diagram illustrating an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. For example... Figure 4As shown, the embodiment provides a computer readable storage medium 600, which stores a second computer program 611, which, when executed by a processor, implements the following steps: based on the last operation and maintenance record of the GIS device, obtaining a metal shell image set and a maintenance interval length, and determining a leakage risk coefficient of sulfur hexafluoride gas according to the metal shell image set; monitoring and obtaining the current environmental parameters of the region where the GIS device is located within a preset time zone, and the expected running load of the GIS device; according to the leakage risk coefficient, the maintenance interval length, the current environmental parameters and the expected running load, taking maximizing the early warning accuracy of sulfur hexafluoride gas as the target, performing gas monitoring threshold optimization to obtain an optimal monitoring threshold; and setting an adaptive early warning strategy according to the optimal monitoring threshold, and monitoring and early warning the sulfur hexafluoride gas in the GIS device within the preset time zone.

[0058] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0059] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0060] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The function specified in one flow or multiple flows and / or blocks Figure 1 The means for implementing the function specified in one flow or multiple flows and / or blocks.

[0061] These computer program instructions can also be stored in a computer readable storage medium, which can guide the computer or other programmable data processing devices to work in a specific way, so that the instructions stored in the computer readable storage medium produce a product including instruction means, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The function specified in one flow or multiple flows and / or blocksFigure 1 the function specified in the one or more blocks.

[0062] These computer program instructions can also be loaded into computer or other programmable data processing devices, so that a series of operation steps are performed on the computer or other programmable data devices to generate computer-implemented processes, thus the instructions executed on the computer or other programmable devices provide processes for implementing the flow Figure 1 the flow or flows and / or blocks Figure 1 the steps of the function specified in the one or more blocks.

[0063] Although the preferred embodiments of the application have been described, those skilled in the art will be able to make additional modifications and variations to these embodiments without departing from the spirit and scope of the application.

[0064] Obviously, various modifications and changes can be made to the present application by those skilled in the art without departing from the spirit and scope of the present application. Accordingly, it is intended that the present application embrace all such modifications and changes as fall within the scope of the present application and its equivalents.

Claims

1. A GIS condition detection method, characterized by, The method comprises: based on the last operation and maintenance record of the GIS device, obtaining the metal shell image set and the maintenance interval time length, and determining the leakage risk coefficient of sulfur hexafluoride gas according to the metal shell image set; monitoring and obtaining the current environmental parameters of the area where the GIS device is located within a preset time zone, and the expected running load of the GIS device; According to the leakage risk coefficient, the maintenance interval time length, the current environmental parameters and the expected running load, the optimal monitoring threshold is obtained by optimizing the gas monitoring threshold with the maximum sulfur hexafluoride gas warning accuracy as the target; According to the optimal monitoring threshold, the adaptive warning strategy is set, and the sulfur hexafluoride gas in the GIS device is monitored and warned within the preset time zone.

2. The GIS condition detection method of claim 1, wherein, According to the metal shell image set, the leakage risk coefficient of sulfur hexafluoride gas is determined, which comprises: obtain the metal shell image set, wherein the metal shell image set includes the metal shell images of several key positions; Using convolutional neural network to identify defects of the metal shell images of the several key positions, obtaining several defect features; Combined with the experience weight model, the leakage risk coefficient is evaluated and determined according to the several defect features.

3. The GIS condition detection method of claim 1, wherein, According to the leakage risk coefficient, the maintenance interval time length, the current environmental parameters and the expected running load, the optimal monitoring threshold is obtained by optimizing the gas monitoring threshold with the maximum sulfur hexafluoride gas warning accuracy as the target, which comprises: obtain the gas monitoring index of sulfur hexafluoride gas, and the monitoring index adjustment space, wherein the gas monitoring index includes gas density, gas pressure, gas temperature, moisture content and decomposition concentration; According to the leakage risk coefficient, the maintenance interval time length, the current environmental parameters and the expected running load, the warning accuracy predictor of sulfur hexafluoride gas is trained; Based on the monitoring index adjustment space, the optimal monitoring threshold is output by optimizing the gas monitoring threshold with the maximum sulfur hexafluoride gas warning accuracy as the target, using the warning accuracy predictor.

4. The GIS condition detection method of claim 3, wherein, According to the leakage risk coefficient, the maintenance interval time length, the current environmental parameters and the expected running load, the warning accuracy predictor of sulfur hexafluoride gas is trained, which comprises: Taking the leakage risk coefficient, the maintenance interval time length, the current environmental parameters and the expected running load as the retrieval constraints, the retrieval constraints are expanded according to the preset tolerance interval to obtain the retrieval conditions; According to the retrieval conditions, the historical monitoring data of sulfur hexafluoride gas of similar GIS devices is screened, the sample gas monitoring threshold set is collected, and the probability of false alarm and missed alarm of sulfur hexafluoride gas index under different sample gas monitoring threshold is counted, and the sample warning accuracy is calculated to obtain the sample warning accuracy set; Using the sample gas monitoring threshold set and the sample warning accuracy set, the generative adversarial network is trained to convergence to obtain the warning accuracy predictor.

5. The GIS condition detection method of claim 4, wherein, Using the sample gas monitoring threshold set and the sample warning accuracy set, the generative adversarial network is trained to convergence to obtain the warning accuracy predictor, which comprises: Adopting the sample gas monitoring threshold set and the sample early warning accuracy set as training data, and equally dividing into P parts, in P part data set has put back to select P times, obtain the first training set, iterative selection P times, obtain P training set, wherein, P is greater than or equal to 10 integer; Adopting the P training set, respectively training generation confrontation network to convergence, obtain P early warning accuracy prediction branch, combination obtains early warning accuracy predictor.

6. The GIS condition detection method of claim 5, wherein, Based on the monitoring index adjustment space, to maximize the early warning accuracy of sulfur hexafluoride gas as the target, the early warning accuracy predictor is used for gas monitoring threshold optimization, and the optimal monitoring threshold is output, including: Randomly selecting in the monitoring index adjustment space, obtain a plurality of initial monitoring threshold values; Taking the standard monitoring threshold value as the benchmark, the deviation amplitudes of the plurality of initial monitoring threshold values are calculated respectively, a plurality of deviation values are determined by weighting, and a plurality of branch selection quantities Q are calculated and obtained, wherein Q is the integer value of the product of the deviation value and P, Q is greater than or equal to 1 and less than or equal to P; Based on the plurality of branch selection quantities Q, the early warning accuracy predictor is used to predict the plurality of initial monitoring threshold values respectively, and a plurality of predicted early warning accuracies are output. According to the plurality of predicted early warning accuracies, the gas monitoring threshold value is optimized to maximize the early warning accuracy of sulfur hexafluoride gas as the target, and the optimal monitoring threshold value is output.

7. The GIS condition detection method of claim 6, wherein, According to the plurality of predicted early warning accuracies, the gas monitoring threshold value is optimized to maximize the early warning accuracy of sulfur hexafluoride gas as the target, and the optimal monitoring threshold value is output, including: Set the initial monitoring threshold value as the initial solution, sort the plurality of initial monitoring threshold values according to the plurality of predicted early warning accuracies from large to small to obtain the initial solution sequence; Set the first K solutions of the initial solution sequence as good solutions, set the last M solutions as bad solutions, and randomly perform equal value clustering on the M bad solutions with the K good solutions as the center to obtain K solution sets, wherein M is N times of K, and N is greater than or equal to 10; In the K solution sets, adjust the bad solutions in the same solution set according to the preset optimization step size with the good solution as the adjustment direction to obtain K updated solution sets, wherein if the adjusted bad solution does not satisfy the monitoring index adjustment space, it is removed and a monitoring threshold value is randomly selected for supplementing; Identify the K updated solution sets, if the predicted early warning accuracy of the updated bad solution is greater than the predicted early warning accuracy of the good solution in the same solution set, replace the good solution with the bad solution; Iterative optimization is performed until a preset optimization number is reached, K current solution sets are output, the solution set with the maximum sum of predicted early warning accuracies is selected as the optimal solution set, and the good solution of the optimal solution set is set as the optimal monitoring threshold value.

8. A GIS condition detection system characterized by, The steps of a GIS state detection method according to any one of claims 1 to 7, comprising: A leakage risk coefficient determination module is configured to determine a leakage risk coefficient of sulfur hexafluoride gas based on the metal shell image set and the maintenance interval length obtained from the last operation and maintenance record of the GIS device. The data monitoring and collecting module is configured to monitor and acquire current environmental parameters of a region where a GIS device is located and expected operation load of the GIS device in a preset time zone; The gas monitoring threshold optimization module is configured to perform gas monitoring threshold optimization according to the leakage risk coefficient, the maintenance interval length, the current environmental parameters and the expected operation load, to maximize the early warning accuracy of sulfur hexafluoride gas, and to acquire an optimal monitoring threshold. The device monitoring and early warning module is configured to set an adaptive early warning strategy according to the optimal monitoring threshold, and to monitor and early warn the sulfur hexafluoride gas in the GIS device in the preset time zone.

9. A smart terminal, characterized by The memory is configured to store a computer software program. The processor is configured to read and execute the computer software program, and to realize the steps of the GIS state detection method according to any one of claims 1 to 7. The storage medium stores a computer software program, and the computer software program is executed by the processor to realize the steps of the GIS state detection method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, comprising: ​

Images