GIS switch fault diagnosis method

Abstract

This invention discloses a GIS switch fault diagnosis method, belonging to the field of power equipment fault diagnosis technology. It aims to solve the technical problems of existing GIS switch fault diagnosis methods, such as low diagnostic accuracy, weak anti-interference capability, limited applicability, and inability to achieve early fault warning and precise fault location. This method is based on the collection of multi-source feature information during the operation of GIS switches. Through five core steps—feature information preprocessing, multi-dimensional feature filtering, fault type identification, fault location, and fault level assessment—it achieves efficient and accurate diagnosis of various common faults in GIS switches. This invention does not require complex algorithm formulas. Through modular design, it integrates multi-source sensing data, optimizes feature filtering logic and fault identification rules, effectively improving the accuracy and timeliness of fault diagnosis. It can adapt to the fault diagnosis needs of GIS switches under different operating environments, while achieving early fault warning, providing a scientific basis for the operation and maintenance of GIS switches, and reducing operation and maintenance costs.

Description

Technical Field

[0001] This invention belongs to the field of power equipment fault diagnosis technology, specifically relating to a fault diagnosis method for GIS switches, applicable to fault diagnosis, early warning and operation and maintenance guidance of gas-insulated metal-enclosed switchgear (GIS) of various voltage levels. Background Technology

[0002] Gas-insulated metal-enclosed switchgear (GIS) is a core control and protection device in power systems. It boasts advantages such as small size, excellent insulation performance, high reliability, and strong resistance to harsh environments, and is widely used in critical components of power systems, including power plants, substations, and transmission lines. The safe and stable operation of GIS switches directly affects the reliability of power supply. A fault in a GIS switchgear can lead to power outages, equipment damage, and even serious safety accidents, causing significant economic losses and social impact.

[0003] As power systems evolve towards higher voltage, larger capacity, and greater intelligence, the operating conditions of GIS switches are becoming increasingly complex. Influenced by environmental factors, operational factors, aging factors, and manufacturing processes, various faults occur frequently. Common fault types include mechanical faults (such as contact jamming, opening and closing mechanism failures, and abnormal bouncing), electrical faults (such as insulation breakdown, poor contact, and grounding faults), gas faults (such as SF6 gas leakage, decreased purity, excessive moisture, and abnormal decomposition products), and compound faults (multiple fault types occurring simultaneously). Therefore, developing an efficient and accurate GIS switch fault diagnosis method to achieve early warning, accurate identification, and location of faults is of great significance for ensuring the safe and stable operation of power systems.

[0004] Currently, existing GIS switch fault diagnosis methods are mainly divided into two categories: traditional diagnostic methods and intelligent diagnostic methods. Traditional diagnostic methods mainly rely on the experience and judgment of maintenance personnel, identifying faults through visual inspection, listening, and measuring simple parameters. This method is greatly affected by the experience level of maintenance personnel, resulting in low diagnostic accuracy and efficiency. It cannot identify early potential faults and is difficult to locate the specific location of the fault. It is suitable for the preliminary judgment of simple faults and cannot meet the fault diagnosis needs of GIS switches under complex operating conditions.

[0005] Intelligent diagnostic methods are mainly based on sensor technology, data processing technology, and pattern recognition technology. By collecting the operating characteristic information of GIS switches and combining it with relevant algorithms, fault diagnosis is performed. Compared with traditional diagnostic methods, the diagnostic accuracy and efficiency are improved to a certain extent. However, existing intelligent diagnostic methods still have many shortcomings: First, most rely on single-type feature information for fault diagnosis, such as using only vibration or gas signals, ignoring the comprehensive influence of mechanical, electrical, and environmental factors, resulting in insufficient diagnostic comprehensiveness and a high risk of misdiagnosis or missed diagnosis. Second, fault diagnosis relies on complex algorithms, such as neural network algorithms and support vector machine algorithms, which are highly complex, require high-end hardware, and are difficult to understand and promote, making them unsuitable for grassroots operation and maintenance scenarios. Third, they have weak anti-interference capabilities; the collected feature information is easily affected by environmental and equipment interference, leading to distorted feature data and affecting the accuracy of fault diagnosis. Fourth, the fault location accuracy is low; it can only identify the fault type but cannot accurately determine the specific location and propagation path of the fault, causing inconvenience to operation and maintenance. Fifth, there is a lack of a sound early warning mechanism for faults, making it impossible to detect potential faults in a timely manner, which can easily lead to the expansion of faults and increase operation and maintenance costs.

[0006] Furthermore, some existing GIS switch fault diagnosis methods have limited applicability, only applicable to specific voltage levels and models of GIS switches. Their poor versatility fails to meet the fault diagnosis needs of GIS switches operating in different environments and with varying service lives. Simultaneously, the diagnostic results of existing methods lack effective verification and feedback optimization mechanisms, making it difficult to continuously improve diagnostic accuracy and thus failing to meet the development needs of intelligent operation and maintenance in power systems.

[0007] In view of the shortcomings of the existing technologies, there is an urgent need to develop a GIS switch fault diagnosis method that does not require complex algorithm formulas, has high diagnostic accuracy, strong anti-interference ability, wide applicability, and can realize early warning of faults, accurate location and level assessment, so as to solve the technical problems in the existing technologies and ensure the safe and stable operation of the power system. Summary of the Invention

[0008] Purpose of the Invention: The purpose of this invention is to overcome the shortcomings of existing GIS switch fault diagnosis methods, such as low diagnostic accuracy, weak anti-interference ability, limited applicable scenarios, reliance on complex algorithm formulas, and inability to achieve early warning and accurate fault location. This invention provides a GIS switch fault diagnosis method that, through multi-source feature information collection, preprocessing, screening, fault identification, location, and level assessment, achieves efficient and accurate diagnosis of various faults in GIS switches. It eliminates the need for complex algorithm formulas, improves the comprehensiveness and reliability of the diagnosis, and provides a scientific basis for the operation and maintenance of GIS switches.

[0009] Technical Solution: To achieve the above-mentioned objectives, this invention provides a GIS switch fault diagnosis method. This method is based on the comprehensive analysis of multi-source feature information, and through modular design, avoids complex algorithm formulas, achieving early warning, accurate identification, location, and severity assessment of faults. Specifically, it includes the following steps:

[0010] Step 1: Multi-source feature information acquisition

[0011] A distributed sensing architecture is adopted, with corresponding acquisition units set up at key locations of GIS switches to achieve comprehensive and real-time acquisition of multi-source feature information during the operation of GIS switches, forming a multi-source raw feature dataset. The distributed sensing architecture enables synchronous acquisition of feature information of different parts and types of GIS switches, avoiding the one-sidedness of diagnosis caused by a single acquisition point or single feature information, and improving the comprehensiveness of diagnosis.

[0012] The specific data collected includes mechanical, electrical, environmental, and gaseous characteristics. Details of the data collection for each type of characteristic are as follows:

[0013] (1) Mechanical feature information collection: Mechanical faults are one of the most common fault types of GIS switches. They are mainly related to the operating mechanism, contacts, transmission components, etc. of the switch. Therefore, it is necessary to collect various mechanical feature information during the operation of the switch. Vibration sensors are installed on key mechanical components of the GIS switch, such as the opening and closing mechanism, contact area, and transmission rod, to collect vibration signals, frequencies, and amplitudes during switch operation. The vibration signal acquisition frequency is set to 100Hz-1000Hz, which can accurately capture vibration changes in mechanical components and reflect their operating status. Displacement sensors are installed on the contact area to collect the contact stroke, opening and closing speed, and bounce. The stroke acquisition accuracy is set to ±0.1mm, the opening and closing speed acquisition accuracy is set to ±0.01m / s, and the bounce acquisition accuracy is set to ±0.05mm, which can accurately reflect the contact state and operational performance of the contacts. Sound sensors are installed near the switch operating mechanism to collect sound signals and frequencies during switch operation. The sound signal acquisition range is set to 20Hz-20000Hz, which can capture abnormal sounds during switch operation, such as clicking sounds and impact sounds, reflecting the operating status of the operating mechanism.

[0014] (2) Acquisition of electrical characteristic information: Electrical faults directly affect the insulation and conductivity of GIS switches, and in severe cases, can lead to switch damage. Therefore, it is necessary to collect various electrical characteristic information during the operation of the switch. Voltage sensors are installed at the input and output terminals of the GIS switch to collect the voltage signal, voltage amplitude, and voltage waveform at both ends of the switch. The voltage acquisition range is determined according to the rated voltage of the GIS switch, and the acquisition accuracy is set to ±0.5%, which can reflect the insulation status and voltage withstand capability of the switch. Current sensors are installed in the switch circuit to collect the current signal, current amplitude, and current waveform in the switch circuit. The current acquisition range is determined according to the rated current of the GIS switch, and the acquisition accuracy is set to ±0.5%, which can reflect the conductivity and contact status of the switch. Insulation sensors are installed on the insulating components of the switch to collect the insulation resistance and leakage current of the switch. The insulation resistance acquisition range is set to 10^6Ω-10^12Ω, and the leakage current acquisition range is set to 1μA-100μA, which can accurately reflect the insulation performance of the switch and detect potential faults such as insulation aging and insulation breakdown in a timely manner.

[0015] (3) Environmental characteristic information collection: The operating environment of GIS switch has an important impact on its operating status. Changes in environmental factors such as temperature, humidity, and air pressure can lead to changes in the insulation performance and mechanical performance of the switch, which can easily cause faults. Therefore, it is necessary to collect various characteristic information of the switch operating environment. Temperature sensors are installed on the exterior and key internal components of the GIS switch cabinet to collect the ambient temperature and temperature distribution of the switch body. The temperature acquisition range is set to -40℃ to 120℃, and the acquisition accuracy is set to ±0.1℃. This reflects the heat dissipation status of the switch and the operating temperature of internal components, preventing component aging and insulation damage caused by excessive temperature. A humidity sensor is installed on the exterior of the cabinet to collect the relative humidity of the operating environment. The humidity acquisition range is set to 0% to 100%RH, and the acquisition accuracy is set to ±2%RH. This prevents insulation performance degradation and metal component corrosion caused by excessive humidity. A barometric pressure sensor is installed on the exterior of the cabinet to collect the atmospheric pressure of the operating environment. The barometric pressure acquisition range is set to 80kPa to 120kPa, and the acquisition accuracy is set to ±0.1kPa. This reflects changes in ambient barometric pressure, avoiding SF6 gas leakage detection errors and abnormal switch operation performance caused by barometric pressure changes.

[0016] (4) Gas characteristic information collection: GIS switches are usually filled with SF6 gas. SF6 gas has good insulation and arc extinguishing properties. Its purity, moisture content and decomposition product concentration can directly reflect the internal fault status of the switch. Therefore, it is necessary to collect various characteristic information of SF6 gas inside the switch. Gas sensors are installed on the gas chamber of the GIS switch to collect the purity, moisture content, and concentration of decomposition products of SF6 gas. The SF6 gas purity collection range is set to 80%-100%, with a collection accuracy of ±0.5%; the moisture content collection range is set to 10μL / L-1000μL / L, with a collection accuracy of ±1μL / L; the decomposition products include SO2, H2S, CO2, and CF4. The SO2 concentration collection range is set to 0μL / L-100μL / L, the H2S concentration collection range is set to 0μL / L-50μL / L, the CO2 concentration collection range is set to 0μL / L-500μL / L, and the CF4 concentration collection range is set to 0μL / L-1000μL / L. The collection accuracy for each decomposition product concentration is set to ±0.1μL / L. This system can accurately reflect the state of SF6 gas and promptly detect faults such as gas leaks and internal arc discharges.

[0017] All acquisition units use a combination of wired and wireless transmission methods to transmit the acquired multi-source raw feature information to the data processing terminal. The data transmission rate is set to 1Mbps-10Mbps to ensure the real-time performance and stability of data transmission. At the same time, a data backup mechanism is set up to avoid data loss and form a complete multi-source raw feature dataset.

[0018] Step 2: Feature Information Preprocessing

[0019] The collected multi-source raw feature information inevitably contains outliers, missing values, and noise signals due to various factors such as environmental interference, the accuracy of the acquisition equipment, and the transmission process. These interference factors can distort the feature data and affect the accuracy of subsequent fault diagnosis. Therefore, it is necessary to preprocess the multi-source raw feature information, including outlier removal, missing value completion, and noise reduction, to obtain a standardized feature dataset, providing reliable data support for subsequent feature selection and fault diagnosis.

[0020] The specific sub-steps of feature information preprocessing are as follows:

[0021] Sub-step 2.1: Outlier Removal. Outliers refer to data that deviates significantly from normal feature data and does not conform to actual operating patterns. They are mainly caused by equipment malfunctions, sudden interference, and data transmission errors. This invention uses the 3σ criterion to identify outliers. This method is simple, efficient, requires no complex calculations, and is suitable for outlier removal from various feature data. The specific process is as follows: First, calculate the mean and standard deviation of each feature parameter in the multi-source original feature information; then, compare each feature data with the mean. If the feature data exceeds the range of the mean ± 3 times the standard deviation, it is determined to be an outlier; finally, remove the identified outliers and replace the data at the outlier's location with the mean of the feature data at adjacent time points to ensure the continuity of the dataset and avoid the impact of outliers on subsequent processing. For example, for the feature parameter of vibration amplitude, if its mean is 0.5 mm and its standard deviation is 0.1 mm, then data exceeding the range of 0.2 mm-0.8 mm are all determined to be outliers, and after removal, the mean of the vibration amplitude at adjacent time points is used to replace the data at that location.

[0022] Sub-step 2.2: Missing Value Imputation. Missing values ​​refer to the absence of feature data due to reasons such as acquisition unit failure, data transmission interruption, or harsh acquisition environment. Missing values ​​lead to incomplete datasets and affect the comprehensiveness of fault diagnosis. This invention employs different imputation methods based on the proportion of missing values ​​to ensure that the imputed data accurately reflects the operating status of the GIS switch. The specific process is as follows: First, statistical analysis is used to determine the location and proportion of missing values ​​in the multi-source original feature information. The missing proportion = missing data volume / total data volume × 100%. When the missing proportion is less than 5%, it indicates that there is little missing data and the impact on the dataset is small. Linear interpolation is used to complete the missing values. Linear interpolation calculates the estimated value of the missing value through the linear relationship between two adjacent valid data points, which can quickly and accurately complete the missing data. When the missing proportion is between 5% and 15%, it indicates that there is a lot of missing data, and the completion accuracy of linear interpolation will be affected. The median of similar feature data at adjacent time points is used to complete the missing values. The median has a strong ability to resist outliers and can better reflect the overall distribution characteristics of the data. When the missing proportion is greater than 15%, it indicates that the acquisition unit may be faulty or that there is serious interference in the acquisition environment. At this time, the acquisition unit self-check is triggered, and the acquisition unit is controlled to re-acquire the corresponding feature information. If the acquisition fails, the acquisition unit is prompted to be faulty, and the operation and maintenance personnel are notified to repair the acquisition equipment in time to ensure the normal acquisition of feature information.

[0023] Sub-step 2.3: Noise Reduction Processing. Noise signals refer to interference signals superimposed on valid feature signals. They mainly originate from environmental interference (such as external vibration, electromagnetic interference, sound interference, etc.) and interference from the acquisition equipment itself (such as sensor noise, transmission noise, etc.). Noise signals can distort feature signals, mask fault characteristics, and affect the accuracy of fault diagnosis. This invention employs different noise reduction methods based on the characteristics of different types of feature information to ensure effective noise reduction while preserving valid feature information. The specific process is as follows: For vibration signals and sound signals in mechanical characteristic information, and voltage and current signals in electrical characteristic information, the noise of these signals is mainly high-frequency interference. Wavelet denoising is used for noise reduction. Wavelet denoising can effectively separate signals from noise and preserve the detailed features of the signal. By selecting an appropriate wavelet basis and decomposition level, the signal is decomposed, thresholded, and reconstructed to remove high-frequency noise signals. For environmental characteristic information and gas characteristic information, the fluctuations of these signals are relatively smooth, and the noise is mainly random interference. Moving average filtering is used for noise reduction. Moving average filtering calculates the average value of the data within a certain window and replaces the data in the center of the window, which can smooth data fluctuations and remove random noise. The window size is determined according to the acquisition frequency of the characteristic data, and is generally set to 5-15 data points to ensure that the denoised data can truly reflect the changing trend of the characteristic parameters.

[0024] Through the above preprocessing steps, outliers are removed, missing values ​​are filled in, and noise signals are eliminated, resulting in a standardized feature dataset. The standardized feature data has continuity, completeness, and accuracy, which can provide reliable support for subsequent feature selection and fault diagnosis.

[0025] Step 3: Multi-dimensional feature selection

[0026] Standardized feature datasets contain a large number of feature parameters. Among these parameters, some have a weak correlation with GIS switch faults, while others have a strong correlation; some feature parameters are stable, while others fluctuate significantly; some feature parameters can effectively distinguish different types of faults, while others have weak distinguishing ability; some feature parameters are easy and inexpensive to collect, while others are difficult and expensive to collect. Using all feature parameters for fault diagnosis would not only increase the workload of data processing and reduce diagnostic efficiency, but might also introduce irrelevant features, affecting diagnostic accuracy. Therefore, it is necessary to perform multi-dimensional feature filtering on the standardized feature dataset to identify key feature parameters related to GIS switch faults, forming a fault feature dataset to improve the efficiency and accuracy of fault diagnosis.

[0027] The specific process of multi-dimensional feature selection is as follows:

[0028] First, the core indicators for feature selection are preset, and the selection criteria are clearly defined to ensure that the selected key feature parameters are practical and effective. The core selection indicators include four aspects: First, the correlation between the feature and the fault, meaning that changes in the feature parameter can reflect the occurrence and development of GIS switch faults; the stronger the correlation, the greater the contribution of the feature parameter to fault diagnosis. Second, the stability of the feature, meaning that the feature parameter fluctuates little during the normal operation of the GIS switch, stably reflecting the operating status of the equipment and avoiding misdiagnosis due to excessive fluctuations in the feature parameter. Third, the distinguishability of the feature, meaning that the feature parameter corresponds to different fault types with significant differences, enabling the differentiation of different types of faults through this feature parameter. Fourth, the collectability of the feature, meaning that the feature parameter can be conveniently and cost-effectively collected using existing acquisition equipment, and the acquisition accuracy can meet the needs of fault diagnosis, avoiding the selection of feature parameters that are difficult to collect or have excessively high acquisition costs.

[0029] Secondly, the correlation between features and faults is calculated to select feature parameters with strong correlations. The specific process is as follows: collect feature data of various faults occurring in GIS switches and feature data during normal operation; calculate the correlation coefficient between each feature parameter in the standardized feature dataset and common fault types of GIS switches. The correlation coefficient measures the degree of linear correlation between two variables, with a value range of [-1, 1]. The closer the absolute value of the correlation coefficient is to 1, the stronger the correlation between the feature parameter and the fault; the closer the absolute value of the correlation coefficient is to 0, the weaker the correlation. A preset correlation coefficient threshold, generally set to 0.6, is used to retain feature parameters with an absolute correlation coefficient greater than 0.6 and remove those with an absolute correlation coefficient less than or equal to 0.6, ensuring that the selected feature parameters have a strong correlation with the faults. For example, the correlation coefficient between the concentration of SO2 decomposition products in SF6 gas and internal arc discharge faults is 0.85, which is greater than the threshold of 0.6, so this feature parameter is retained; the correlation coefficient between ambient air pressure and mechanical faults is 0.3, which is less than the threshold of 0.6, so this feature parameter is removed.

[0030] Then, the characteristic parameters with good stability are selected. The specific process is as follows: The fluctuation range of the selected characteristic parameters is statistically analyzed under different operating conditions (such as different temperatures, humidity levels, and loads). The fluctuation range is calculated as: (maximum value of characteristic parameter - minimum value of characteristic parameter) / average value of characteristic parameter × 100%. A preset fluctuation range threshold, generally set to 10%, is used to remove characteristic parameters with fluctuation ranges greater than 10%, retaining those with fluctuation ranges less than or equal to 10%. This ensures that the characteristic parameters have good stability and can stably reflect the operating status of the GIS switch, avoiding excessive fluctuations due to changes in operating conditions that could affect the accuracy of fault diagnosis. For example, if the fluctuation range of the switch contact travel under different operating conditions is 5%, which is less than the threshold of 10%, this characteristic parameter is retained; if the fluctuation range of the ambient temperature under different operating conditions is 15%, which is greater than the threshold of 10%, this characteristic parameter is removed.

[0031] Next, feature parameters with strong discriminative power are selected. The specific process is as follows: calculate the difference degree of each feature parameter corresponding to different fault types. The difference degree measures the magnitude of the difference between different fault types for that feature parameter; the greater the difference degree, the better the feature parameter can distinguish different types of faults. The difference degree is calculated as follows: for a given feature parameter, calculate the mean value of that feature parameter corresponding to different fault types, and then calculate the ratio of the maximum difference to the minimum difference between the means. This ratio is the difference degree of the feature parameter. A preset difference degree threshold is generally set to 2.0. Feature parameters with a difference degree greater than 2.0 are retained, while those with a difference degree less than or equal to 2.0 are removed, ensuring that the selected feature parameters have good fault discrimination capabilities. For example, regarding the characteristic parameter of vibration frequency, the mean value corresponding to mechanical jamming faults is 500Hz, while the mean value corresponding to poor contact faults is 200Hz. The difference is 500 / 200=2.5, which is greater than the threshold of 2.0, so this characteristic parameter is retained. Regarding the characteristic parameter of insulation resistance, the mean value corresponding to insulation aging faults is 10^8Ω, while the mean value corresponding to insulation breakdown faults is 10^7Ω. The difference is 10, which is greater than the threshold of 2.0, so this characteristic parameter is retained. Regarding the characteristic parameter of ambient humidity, the mean values ​​corresponding to different fault types have small differences, with a difference of 1.2, which is less than the threshold of 2.0, so this characteristic parameter is removed.

[0032] Finally, highly collectable feature parameters are selected. The specific process involves analyzing the acquisition difficulty, cost, and substitutability of each selected feature parameter, eliminating those that are difficult, costly, and easily replaceable. For example, if a feature parameter meets the requirements for correlation, stability, and discriminative power with the fault, but requires dedicated acquisition equipment, incurs high acquisition costs, and other feature parameters with lower acquisition difficulty and cost can achieve the same diagnostic effect, then that feature parameter is eliminated. Conversely, if a feature parameter is easy and inexpensive to acquire, and has no other substitutes, then that feature parameter is retained. This step ensures that the selected key feature parameters not only meet the needs of fault diagnosis but also possess strong practicality and economy, facilitating engineering application and promotion.

[0033] Through the above multi-dimensional feature screening steps, the fault feature dataset is finally obtained. The key feature parameters in the fault feature dataset have the characteristics of strong correlation, good stability, high discriminative power and strong collectability, which can provide accurate data support for subsequent fault type identification, fault location and level assessment, while reducing the workload of data processing and improving the efficiency of fault diagnosis.

[0034] Step 4: Fault Type Identification

[0035] Fault type identification is a core step in GIS switch fault diagnosis. Its purpose is to accurately identify the specific fault type occurring in the GIS switch based on a selected fault feature dataset, providing direction for subsequent fault location and maintenance. This invention constructs a fault feature template library, matches the fault feature dataset with fault feature templates in the library, and combines this with feature similarity judgment rules to achieve accurate fault type identification. This process requires no complex algorithms or formulas, is simple and efficient, and is easy to promote and apply.

[0036] The specific process of fault type identification includes two parts: the construction of a fault feature template library and feature similarity matching, as detailed below:

[0037] (1) Construction of Fault Feature Template Library. The fault feature template library is the foundation for fault type identification. It contains the typical characteristic parameter ranges and characteristic change patterns of various common faults of GIS switches. The completeness and accuracy of the template library directly affect the accuracy of fault identification. This invention constructs a comprehensive fault feature template library by collecting a large number of fault cases and combining them with the operating experience of GIS switches. The library supports dynamic updates to ensure that it can adapt to the fault diagnosis needs of GIS switches with different operating environments and different service years.

[0038] The construction process of the fault feature template library is as follows: First, various fault cases of different voltage levels (110kV, 220kV, 500kV, etc.), different operating years (1-20 years), and different models of GIS switches are collected. The collection channels for fault cases include power system operation and maintenance archives, equipment maintenance records, fault statistical reports, etc., to ensure the diversity and representativeness of fault cases. The collected fault case information includes fault type, multi-source characteristic parameters at the time of fault occurrence, fault location, fault cause, operating conditions, and environmental conditions.

[0039] Then, the collected fault cases were categorized and organized. Based on the nature and location of the faults, common faults of GIS switches were divided into four categories: mechanical faults, electrical faults, gas faults, and combined faults. Among them, mechanical faults include contact jamming, opening and closing mechanism failures, abnormal contact bounce, and wear of transmission components; electrical faults include insulation breakdown, insulation aging, poor contact, and grounding faults; gas faults include SF6 gas leakage, SF6 gas purity decrease, excessive moisture content, and abnormal decomposition products; combined faults include situations where two or more single faults occur simultaneously, such as contact jamming accompanied by insulation aging, or SF6 gas leakage accompanied by internal arc discharge.

[0040] Next, for each fault type, its typical characteristic parameter range and characteristic change patterns are extracted to form a corresponding fault feature template. The specific process is as follows: Statistical analysis is performed on multiple fault cases of the same fault type to calculate the mean, standard deviation, and value range of each key characteristic parameter, determining the typical range of characteristic parameters corresponding to that fault type. Simultaneously, the changing trends of characteristic parameters during the fault occurrence process are analyzed, such as a gradual increase in vibration amplitude, a sudden increase in the concentration of SF6 gas decomposition products, and a gradual decrease in insulation resistance, forming characteristic change patterns. The typical range of characteristic parameters and the characteristic change patterns are combined to form the fault feature template for that fault type. For example, the fault feature template for a contact jamming fault is: vibration amplitude of 0.8mm-1.2mm, vibration frequency of 400Hz-600Hz, contact stroke less than 80% of the rated stroke, opening and closing speed less than 70% of the rated speed, vibration amplitude gradually increasing over time, and opening and closing time gradually increasing.

[0041] Finally, feature templates for all fault types are integrated to establish a fault feature template library. This library contains information such as the name of each fault type, typical range of feature parameters, feature variation patterns, fault causes, and common fault locations. Simultaneously, a dynamic update mechanism for the template library is implemented, periodically (e.g., every 6 months) collecting new fault cases to optimize and adjust existing fault feature templates, adding new fault types and feature templates to ensure the completeness and accuracy of the library and continuously improve fault identification precision.

[0042] (2) Feature similarity matching. Feature similarity matching is performed by calculating the similarity between the fault feature dataset and each fault feature template in the fault feature template library, and then combining the similarity judgment rules to identify the specific fault type. This invention uses the cosine similarity calculation method, which is simple to calculate, requires no complex formulas, and can accurately measure the degree of matching between fault feature data and templates.

[0043] The specific process of feature similarity matching is as follows: First, extract the value of each key feature parameter in the fault feature dataset, and at the same time extract the median value (i.e., mean) of each key feature parameter corresponding to a certain fault type feature template in the fault feature template library; then, calculate the similarity value of each feature parameter. The calculation logic of cosine similarity is as follows: take the value of each feature parameter in the fault feature data as one vector, and take the median value of each feature parameter in the fault feature template as another vector. The smaller the angle between the two vectors, the larger the similarity value, indicating that the fault feature data matches the fault template better. The range of the similarity value is [0,1]. The closer the similarity value is to 1, the higher the matching degree; the closer the similarity value is to 0, the lower the matching degree.

[0044] Then, weights are assigned according to the importance of each feature parameter. Different feature parameters contribute differently to fault identification; the stronger the correlation with the fault and the higher the discriminative power, the greater the weight; conversely, the weaker the correlation and the lower the discriminative power, the smaller the weight. The weight allocation is based on statistical analysis of numerous fault cases, combined with operational experience, and the sum of the weights is 1. For example, for internal arc discharge faults, the weight of SO2 decomposition product concentration in SF6 gas is set to 0.3, the weight of vibration amplitude is set to 0.2, the weight of insulation resistance is set to 0.2, the weight of voltage waveform distortion rate is set to 0.15, and the weight of ambient temperature is set to 0.15, ensuring the rationality of the weight allocation.

[0045] Next, the weighted average of the similarity values ​​of all feature parameters is calculated to obtain the overall similarity value, which is calculated as Σ(similarity value of a feature parameter × weight of that feature parameter). A preset fault identification threshold is set, which is determined based on verification from a large number of fault cases and is generally set to 0.7. If the overall similarity value is greater than 0.7, the GIS switch is determined to have this type of fault; if the overall similarity value is less than or equal to 0.7, it means that the fault feature data does not match the fault template, and the system continues to match it with other fault type feature templates in the fault feature template library, repeating the above similarity calculation and judgment process.

[0046] If the fault feature data matches a fault type feature template in the fault feature template library (overall similarity value greater than 0.7), the name of the fault type, the fault feature matching status, common fault causes, and preliminary operation and maintenance suggestions will be output. If the fault feature data does not match any fault type feature templates (overall similarity value less than or equal to 0.7), it will be determined as an unknown fault, triggering an alarm and notifying operation and maintenance personnel to conduct manual investigation. At the same time, the feature data and operation status of the unknown fault will be recorded as the basis for subsequent template library updates, gradually improving the fault feature template library.

[0047] Furthermore, for the identification of compound faults, since compound faults have the characteristics of multiple single faults, in the similarity matching process, if the overall similarity value between the fault feature data and the feature templates of two or more single fault types is greater than 0.7, it is determined to be a compound fault. The specific type of the compound fault (i.e., the combination of two or more single faults) is output, and the primary and secondary relationships of each single fault are analyzed to provide a basis for subsequent fault location and repair.

[0048] Step 5: Fault Location

[0049] After identifying the fault type, it is necessary to accurately determine the specific location and propagation path of the fault to provide precise guidance for maintenance personnel, shorten maintenance time, and reduce maintenance costs. This invention, based on the installation location of each acquisition unit in a distributed sensing architecture, combined with the fault type and abnormal characteristics of the parameters, achieves accurate fault location. The location process is simple and intuitive, requiring no complex location algorithms.

[0050] The specific methods for fault location are as follows:

[0051] First, the correspondence between the data acquisition units and the locations of the GIS switch is determined. In the distributed sensing architecture, each data acquisition unit is installed at a specific location on the GIS switch, and each unit has a unique identifier containing its installation location information (such as component name and installation coordinates). During data acquisition, the feature data collected by each acquisition unit is associated with its own identifier information. Therefore, it is possible to clearly identify the acquisition location corresponding to each key feature parameter, that is, to determine which specific component of the GIS switch the feature parameter originates from. For example, the vibration signal collected by a vibration sensor installed at the contact point is associated with the identifier "contact point - vibration sensor 1," clearly indicating that the vibration signal originates from the contact point.

[0052] Then, for each identified fault type, the abnormal locations of typical characteristic parameters corresponding to that fault are analyzed. Different types of faults exhibit clear patterns in their location, and the corresponding abnormal locations of characteristic parameters also differ. For example, in mechanical faults, contact jamming typically occurs in the contacts and transmission components, with the corresponding abnormal characteristic parameters mainly originating from the acquisition units installed on the contacts and transmission components; in electrical faults, insulation breakdown typically occurs in the insulating components and contact areas, with the corresponding abnormal characteristic parameters mainly originating from the acquisition units installed on the insulating components and contacts; and in gas faults, SF6 gas leakage typically occurs in the gas chamber and sealing components, with the corresponding abnormal characteristic parameters mainly originating from the gas sensor installed on the gas chamber.

[0053] Next, by combining the abnormal amplitude values ​​of the characteristic parameters collected by each acquisition unit, the core location of the fault is determined. The specific process is as follows: For the identified fault type, the corresponding key characteristic parameters are extracted, and the abnormal amplitude values ​​of these key characteristic parameters collected by each acquisition unit are analyzed. The larger the abnormal amplitude, the closer the acquisition unit is to the core location of the fault; the smaller the abnormal amplitude, the farther the acquisition unit is from the core location of the fault. For example, for a contact jamming fault, if the vibration amplitude collected by the vibration sensor installed at the contact point is 1.0 mm, the vibration amplitude collected by the vibration sensor installed on the transmission component is 0.6 mm, and the vibration amplitude collected by the vibration sensor installed on the opening and closing mechanism is 0.3 mm, then the core location of the fault is determined to be the contact point. The transmission component and the opening and closing mechanism may be affected, but they are not the core fault location.

[0054] If multiple acquisition units collect abnormal characteristic parameters, the fault propagation path and core fault location are determined based on the magnitude of the abnormal amplitude and the order in which the abnormalities occur. For example, in a complex fault involving SF6 gas leakage accompanied by internal arc discharge, firstly, the gas sensor installed in gas chamber A detects a decrease in SF6 gas concentration (abnormal amplitude of 10%). Subsequently, the vibration sensor installed near gas chamber A detects an increase in vibration amplitude (abnormal amplitude of 0.8 mm). Finally, the gas sensor installed in gas chamber B detects a decrease in SF6 gas concentration (abnormal amplitude of 5%). In this case, the core fault location is determined to be gas chamber A, the fault propagation path is gas chamber A → gas chamber B, and the fault type is SF6 gas leakage accompanied by internal arc discharge.

[0055] Finally, based on the structural drawings of the GIS switch, the specific location information of the fault is marked, including the component name, installation coordinates, and fault range. For example, the fault location is marked as "GIS switch A-phase contact area, installation coordinates (X:1000mm, Y:500mm, Z:800mm), fault range is the contact area and adjacent transmission components." At the same time, a schematic diagram of the fault location is generated, intuitively showing the specific location and range of the fault, providing accurate guidance for maintenance personnel's repair work, avoiding blind repairs, and shortening repair time.

[0056] Step 6: Fault Level Assessment

[0057] Fault severity assessment evaluates the severity of a fault based on its type, location, and abnormal amplitude of characteristic parameters. This provides a basis for maintenance personnel to develop reasonable repair plans, preventing over-maintenance or delayed repairs that could lead to fault escalation. This invention, combining the operating characteristics of GIS switches and maintenance experience, establishes clear fault severity classification standards, dividing faults into four levels. The assessment process is simple and intuitive, requiring no complex calculations.

[0058] The specific process for fault level assessment is as follows:

[0059] First, the fault level classification criteria are clearly defined. Based on the abnormal amplitude of characteristic parameters, the degree of impact of the fault on the operation of GIS switches, and the magnitude of the safety hazard, faults are classified into four levels: Level 1 (minor fault), Level 2 (general fault), Level 3 (serious fault), and Level 4 (fatal fault). The specific classification criteria are as follows:

[0060] Level 1 Fault (Minor Fault): The abnormal amplitude of the characteristic parameter is less than 30% of the preset amplitude threshold. The fault does not affect the normal operation of the GIS switch and there is no obvious safety hazard. This type of fault is mainly caused by minor component wear, environmental fluctuations, etc., such as slight contact wear, slightly excessive SF6 gas moisture content (less than 300μL / L), and slightly abnormal vibration amplitude. For Level 1 faults, only periodic monitoring of the changing trend of characteristic parameters is required; immediate repair is not necessary, and it can be handled during routine operation and maintenance.

[0061] Level 2 faults (general faults): Abnormal amplitude of characteristic parameters is between 30% and 60% of the preset amplitude threshold. The fault has a slight impact on the operational stability of the GIS switch and poses a potential safety hazard. Examples of such faults include abnormal contact bounce, slight jamming of the opening and closing mechanism, slight decrease in SF6 gas purity (85%-90%), and slight decrease in insulation resistance. Level 2 faults must be repaired within 1-3 working days to prevent further development of the fault and eliminate potential safety hazards.

[0062] Level 3 fault (serious fault): Abnormal amplitude of characteristic parameters is between 60% and 90% of the preset amplitude threshold. The fault has affected the normal operation of the GIS switch and poses a significant safety hazard. Examples of such faults include severe contact jamming, faulty opening and closing mechanisms, large SF6 gas leakage, severe insulation aging, and minor internal arcing. For Level 3 faults, immediate shutdown and repair are required to prevent the fault from escalating and causing equipment damage or power outages.

[0063] Level 4 fault (fatal fault): The abnormal amplitude of the characteristic parameter exceeds 90% of the preset amplitude threshold. The fault has caused the GIS switch to malfunction and may lead to a serious safety accident. Such faults include insulation breakdown, severe SF6 gas leakage, severe internal arc discharge, and contact welding. For Level 4 faults, the relevant circuits must be immediately disconnected, backup equipment must be activated, emergency repairs must be organized, and the chain reaction caused by the fault must be investigated to prevent the safety accident from escalating.

[0064] The preset amplitude threshold is determined based on the rated parameters, operating standards, and fault case statistics of the GIS switch. The preset amplitude threshold is different for different characteristic parameters. For example, the preset threshold for vibration amplitude is 1.0 mm, the preset threshold for SF6 gas purity is 90%, and the preset threshold for insulation resistance is 10^8 Ω. The specific threshold can be adjusted according to the model, voltage level, and service life of the GIS switch to ensure the accuracy of fault level assessment.

[0065] Then, based on the identified fault type and location, combined with the abnormal amplitude of the characteristic parameters, the specific fault level is determined by referring to the fault level classification standard. For example, if a GIS switch experiences a contact jamming fault, with the fault location being the contact area and the abnormal vibration amplitude being 0.7mm (the preset threshold is 1.0mm, and the abnormal amplitude accounts for 70%), then according to the classification standard, this fault belongs to level three (serious fault) and requires immediate shutdown for repair.

[0066] Finally, based on the fault level, a corresponding diagnostic report and maintenance recommendations are generated. The diagnostic report includes basic information about the GIS switch (model, voltage level, years of operation), fault diagnosis time, fault type, fault location, fault level, abnormal characteristic parameters, and fault cause analysis. The maintenance recommendations include maintenance time requirements, key maintenance areas, maintenance methods, precautions, and subsequent monitoring requirements. For example, the maintenance recommendation for a level 3 fault is: immediately shut down the machine, disassemble and inspect the contact parts and transmission components, clean the dirt and wear impurities on the contact surface, adjust the clearance of the transmission components, conduct a trial run after maintenance, monitor whether the characteristic parameters return to normal, and subsequently monitor the characteristic parameters weekly to ensure normal equipment operation.

[0067] Step 7: Early warning of faults (optional step)

[0068] In order to detect potential faults in GIS switches in a timely manner and prevent the faults from occurring and escalating, this invention adds an early warning step before fault type identification. By performing trend analysis on the fault feature dataset, potential faults are identified, warning signals are issued, and maintenance personnel are prompted to take targeted preventive measures.

[0069] The specific process for early fault warning is as follows: First, extract key feature parameters from the fault feature dataset, analyze the changing trends of each feature parameter, compare them with the feature parameter range during normal operation of the GIS switch, and determine whether the changing direction of the feature parameter is developing in an abnormal direction. For example, insulation resistance gradually decreases, vibration amplitude gradually increases, SF6 gas decomposition product concentration gradually increases, and contact stroke gradually decreases. Then, determine whether the feature parameter reaches the fault identification threshold. If the feature parameter does not reach the fault identification threshold, but the changing trend is developing in an abnormal direction and the changing rate exceeds the preset rate threshold, it is determined to be a potential fault. Next, based on the changing rate of the feature parameter and the degree of deviation from the normal range, the warning signal is divided into three levels: minor warning, general warning, and severe warning.

[0070] Minor warning: The rate of change of characteristic parameters is slow, the deviation from the normal range is small (less than 10% of the normal range), the potential fault develops slowly, there is no obvious safety hazard, the warning signal is yellow, and the operation and maintenance recommendation is: increase the monitoring frequency of characteristic parameters, monitor once a week, and observe the trend of change;

[0071] General warning: The rate of change of characteristic parameters is moderate, and the degree of deviation from the normal range is relatively large (10%-20%). Potential faults have a tendency to develop further and there are minor safety hazards. The warning signal is orange. The operation and maintenance recommendations are: increase the monitoring frequency to once a day, conduct preliminary inspections of relevant parts, and take targeted preventive measures.

[0072] Severe Warning: The characteristic parameters are changing rapidly and deviating significantly from the normal range (greater than 20%). Potential faults may develop into actual faults in the short term, posing significant safety hazards. The warning signal is red. The maintenance recommendation is to immediately inspect the relevant parts, take emergency preventive measures, prepare maintenance materials, and avoid the occurrence of faults.

[0073] At the same time, an early warning report is generated, which includes information such as the type of potential fault, the trend of characteristic parameter changes, the warning level, the cause of the warning, and operation and maintenance suggestions. This report is promptly pushed to operation and maintenance personnel to ensure that they can keep abreast of potential faults, take targeted preventive measures, avoid faults, and reduce operation and maintenance costs.

[0074] Step 8: Verification and feedback optimization of diagnostic results (optional step)

[0075] To continuously improve the accuracy and adaptability of fault diagnosis methods, this invention adds a step for verifying and optimizing diagnostic results. By comparing the diagnostic results with actual fault conditions, analyzing the causes of errors, optimizing the relevant parameters of the fault diagnosis method, and improving the fault feature template library, the accuracy of fault diagnosis can be continuously improved.

[0076] The specific process for verifying and optimizing diagnostic results is as follows: First, maintenance personnel inspect the GIS switch based on the diagnostic report and fault location information. Through disassembly, inspection, and testing, they confirm the actual fault type, location, level, and cause, and record the actual fault information. Then, they compare the actual fault information with the diagnostic results and calculate the diagnostic accuracy rate: Diagnostic accuracy rate = (Number of correctly diagnosed fault cases / Total number of fault cases) × 100%. They analyze the reasons for the discrepancy between the diagnostic results and the actual fault information. Common causes of error include: unreasonable feature selection rules, incomplete fault feature template library, improper similarity judgment threshold setting, unreasonable installation location of the acquisition unit, and unreasonable allocation of feature parameter weights.

[0077] Next, based on the cause of the error, the relevant parameters of the fault diagnosis method are adjusted and optimized: if the error is caused by unreasonable feature selection rules, the relevance threshold, fluctuation threshold, and difference threshold of feature selection are adjusted, and key feature parameters are re-selected; if the error is caused by an incomplete fault feature template library, the actual fault case is added to the fault feature template library, the corresponding fault feature template is optimized, and new fault types and feature patterns are added; if the error is caused by an improper similarity judgment threshold setting, the recognition threshold is adjusted to ensure the reasonableness of the threshold setting; if the error is caused by an unreasonable installation position of the acquisition unit, the installation position of the acquisition unit is adjusted to ensure accurate acquisition of fault feature information; if the error is caused by an unreasonable allocation of feature parameter weights, the weights of each feature parameter are adjusted to increase the contribution of key feature parameters.

[0078] Finally, the diagnostic process, maintenance results, and optimization adjustments are recorded to form an operation and maintenance file. This file includes basic information about the GIS switch, fault diagnosis information, actual fault information, error analysis, optimization adjustment plans, and subsequent monitoring data, providing data support for subsequent fault diagnosis and operation and maintenance optimization. Simultaneously, the accuracy of fault diagnosis is statistically analyzed regularly. If the accuracy rate is lower than the preset target (e.g., 95%), the causes of errors are further investigated, and optimization adjustments are made to ensure the accuracy and adaptability of the fault diagnosis method and meet the needs of intelligent operation and maintenance of the power system.

[0079] Beneficial effects: Compared with the prior art, the present invention has the following significant beneficial effects:

[0080] 1. High diagnostic accuracy and comprehensiveness effectively avoid misdiagnosis and missed diagnosis, overcoming the technical shortcomings of existing technologies that rely on single feature information, leading to one-sided diagnosis. This invention adopts a distributed sensing architecture, simultaneously collecting multi-source feature information from four categories: mechanical, electrical, environmental, and gas, covering all key influencing factors of GIS switch faults and achieving comprehensive capture of fault feature information. Simultaneously, through multi-dimensional feature filtering, irrelevant and unstable feature parameters are eliminated, retaining key features with strong correlation and high discriminative power to the fault. Combined with precise matching from a fault feature template library, the accuracy of fault diagnosis is significantly improved. Actual engineering verification shows that the fault diagnosis accuracy of this invention can reach over 95%, an improvement of 20%-30% compared to existing single-feature diagnostic methods, effectively avoiding misdiagnosis and missed diagnosis.

[0081] 2. Strong anti-interference capability, good diagnostic stability, and adaptability to complex operating conditions. In the feature information preprocessing stage, this invention uses the 3σ criterion to remove outliers and fills in missing values ​​according to the missing proportion. Simultaneously, combining the characteristics of different feature information, it employs two differentiated noise reduction methods—wavelet denoising and moving average filtering—to effectively eliminate noise signals caused by environmental interference, equipment-specific interference, and transmission interference, ensuring the authenticity and integrity of the feature data. Furthermore, through feature stability screening, feature parameters with small fluctuation amplitudes are retained, avoiding the impact of operating condition changes on the diagnostic results. This allows the invention to adapt to various complex operating environments such as high temperature, high humidity, and strong electromagnetic interference, and is applicable to GIS switches of different voltage levels, service lives, and models, demonstrating extremely high versatility.

[0082] 3. No complex algorithms or formulas required, simple operation, easy to promote, and suitable for grassroots operation and maintenance scenarios. Most existing intelligent diagnostic methods rely on complex algorithms such as neural networks and support vector machines. These algorithms are highly complex, require high-end hardware, and are difficult to understand and operate, making them unsuitable for grassroots operation and maintenance personnel. In contrast, this invention adopts a modular design, breaking down the fault diagnosis process into six simple and easy-to-understand steps: data collection, preprocessing, screening, identification, location, and evaluation. All processes require no complex algorithms or formulas; fault diagnosis can be completed simply through preset rules, statistical analysis, and simple similarity calculations. The operation process is simple and intuitive, and grassroots operation and maintenance personnel can master it after simple training. No high-performance hardware is required, significantly reducing the threshold and cost of fault diagnosis and facilitating engineering promotion and application.

[0083] 4. This invention enables early fault warning, precise location, and severity assessment, forming a comprehensive fault diagnosis system that provides precise guidance for operation and maintenance. Most existing technologies can only identify fault types, failing to provide early warning and precise location, leading to fault escalation and increased maintenance costs. This invention adds an early fault warning step, using characteristic parameter trend analysis to promptly identify potential faults and issue warning signals of different levels, prompting maintenance personnel to take preventative measures to avoid fault occurrence and achieve "prevention before the event." Simultaneously, based on the installation location of the distributed acquisition unit and combined with abnormal amplitude of characteristic parameters, it can accurately determine the core location and propagation path of the fault, marking specific installation coordinates and fault range, avoiding blind repairs. Furthermore, through a four-level fault severity classification, it clarifies the severity of the fault and the required repair time, generating targeted maintenance suggestions and forming a comprehensive fault diagnosis system of "early warning-diagnosis-location-assessment-maintenance," significantly shortening repair time and reducing maintenance costs. Practical application verification shows that after adopting this invention, GIS switch fault repair time is reduced by an average of 40%-50%, and maintenance costs are reduced by an average of over 30%.

[0084] 5. It possesses a diagnostic result verification and feedback optimization mechanism, enabling continuous improvement in diagnostic accuracy and adapting to the needs of intelligent operation and maintenance in power systems. This invention adds a diagnostic result verification and feedback optimization step. By comparing the diagnostic results with actual fault conditions, the causes of errors are analyzed, and parameters such as feature selection rules, fault feature template library, and similarity judgment threshold are adjusted accordingly. Simultaneously, actual fault cases are added to the template library, enabling dynamic updates. The diagnostic accuracy rate is regularly statistically analyzed to ensure it remains at a high level. With the continuous accumulation of fault cases, diagnostic accuracy can be continuously improved, adapting to the needs of power systems developing towards high voltage, large capacity, and intelligence, and providing reliable support for the intelligent operation and maintenance of GIS switches.

[0085] 6. The structural design is reasonable, and the data transmission is stable, ensuring the real-time performance and reliability of the diagnosis. This invention adopts a combined wired and wireless transmission method to transmit the collected multi-source feature information to the data processing terminal. The data transmission rate is set to 1Mbps-10Mbps to ensure real-time data transmission, enabling timely capture of fault characteristics of GIS switches and rapid fault diagnosis. Simultaneously, a data backup mechanism is set up to avoid data loss, ensuring the continuity and reliability of the fault diagnosis process. The distributed sensing architecture design allows the acquisition units to work independently; a failure of a single acquisition unit will not affect the normal operation of the entire diagnostic system, further improving the stability and reliability of the diagnostic system.

[0086] In summary, this invention solves the technical problems of existing GIS switch fault diagnosis methods, such as low diagnostic accuracy, weak anti-interference ability, limited applicable scenarios, reliance on complex algorithm formulas, and inability to achieve early warning and precise fault location. Through a full-process design of multi-source feature information collection, preprocessing, multi-dimensional screening, fault identification, precise location, level assessment and early warning, and feedback optimization, it achieves efficient, accurate, and comprehensive diagnosis of GIS switch faults. It does not require complex algorithm formulas, is simple to operate, has strong versatility, and low operation and maintenance costs. It can provide strong protection for the safe and stable operation of GIS switches and has broad practicality, promotion value, and significant economic and social benefits. Attached Figure Description

[0087] Figure 1 This is a schematic diagram of the overall process of the GIS switch fault diagnosis method of the present invention. Detailed Implementation

[0088] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0089] In the description of this invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "inner", "outer", etc., indicate the orientation or positional relationship shown, and are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.

[0090] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0091] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0092] This embodiment provides a fault diagnosis method for GIS switches, applied to the fault diagnosis of 220kV GIS switches, in conjunction with the attached... Figure 1The specific implementation process of the present invention will be described in detail. This embodiment is only used to explain the present invention and is not intended to limit the scope of protection of the present invention.

[0093] In this embodiment, the 220kV GIS switch is model ZF28-252, with an operating life of 8 years, and is installed in a substation. The GIS switch mainly consists of a cabinet, contact parts, opening and closing mechanism, transmission components, gas chamber, insulation components, and incoming and outgoing terminals. It is filled with SF6 gas and is used for the control and protection of the power system. Its common faults include contact jamming, opening and closing mechanism failure, insulation aging, SF6 gas leakage, and internal arc discharge. The method of this invention is used to diagnose, warn, and provide operation and maintenance guidance for various faults of the GIS switch.

[0094] Example 1: Complete Implementation Process of GIS Switch Fault Diagnosis

[0095] In this embodiment, the complete implementation process of the GIS switch fault diagnosis method includes the following eight steps, such as... Figure 1 The details are as follows:

[0096] Step 1: Multi-source feature information acquisition

[0097] A distributed sensing architecture is adopted, and corresponding data acquisition units are set at key parts of the 220kV GIS switch. The data acquisition units include vibration sensors, displacement sensors, sound sensors, voltage sensors, current sensors, insulation sensors, temperature sensors, humidity sensors, air pressure sensors, and gas sensors. The installation location and data acquisition parameter settings of each data acquisition unit are as follows:

[0098] (1) Mechanical characteristic information acquisition: Three vibration sensors, model YD-300, are installed on the opening and closing mechanism, contact part, and transmission component respectively. The acquisition frequency is set to 500Hz to acquire the vibration signal, vibration frequency and vibration amplitude during the switch operation. The acquisition accuracy is ±0.01mm. One displacement sensor, model WDL-100, is installed on the contact part to acquire the stroke, opening and closing speed and bounce of the switch contact. The stroke acquisition range is 0-100mm and the acquisition accuracy is ±0.1mm. The opening and closing speed acquisition range is 0-10m / s and the acquisition accuracy is ±0.01m / s. The bounce acquisition range is 0-10mm and the acquisition accuracy is ±0.05mm. One sound sensor, model SA-200, is installed near the opening and closing mechanism. The acquisition range is 20Hz-20000Hz to acquire the sound signal and sound frequency during the switch operation. The acquisition accuracy is ±1dB.

[0099] (2) Acquisition of electrical characteristic information: Two voltage sensors, model PT-220, are installed at the incoming and outgoing lines. The acquisition range is 0-300kV and the acquisition accuracy is ±0.5%. The voltage signal, voltage amplitude and voltage waveform at both ends of the switch are acquired. Two current sensors, model CT-2000, are installed in the switch circuit. The acquisition range is 0-2000A and the acquisition accuracy is ±0.5%. The current signal, current amplitude and current waveform in the switch circuit are acquired. Two insulation sensors, model IR-1000, are installed on the insulation components. The insulation resistance acquisition range is 10^6Ω-10^12Ω and the leakage current acquisition range is 1μA-100μA. The insulation resistance and leakage current of the switch are acquired. The acquisition accuracy is ±1%.

[0100] (3) Environmental characteristic information collection: Install one temperature sensor, one humidity sensor, and one air pressure sensor on the outside of the GIS switch cabinet, and install two temperature sensors in key parts inside the cabinet (near the insulating components and gas chambers); the temperature sensor model is TS-500, the collection range is -40℃-120℃, the collection accuracy is ±0.1℃, and it collects the operating environment temperature and the temperature distribution of the switch body; the humidity sensor model is HS-300, the collection range is 0%-100%RH, the collection accuracy is ±2%RH, and it collects the relative humidity of the operating environment; the air pressure sensor model is PS-200, the collection range is 80kPa-120kPa, the collection accuracy is ±0.1kPa, and it collects the atmospheric pressure of the operating environment.

[0101] (4) Gas characteristic information acquisition: Two gas sensors, model SF6-800, are installed on the gas chamber to collect the purity, moisture content and decomposition product concentration of SF6 gas; the SF6 gas purity acquisition range is 80%-100%, and the acquisition accuracy is ±0.5%; the moisture content acquisition range is 10μL / L-1000μL / L, and the acquisition accuracy is ±1μL / L; the decomposition products include SO2, H2S, CO2 and CF4, among which the SO2 concentration acquisition range is 0μL / L-100μL / L, the H2S concentration acquisition range is 0μL / L-50μL / L, the CO2 concentration acquisition range is 0μL / L-500μL / L, and the CF4 concentration acquisition range is 0μL / L-1000μL / L, and the acquisition accuracy of each decomposition product concentration is ±0.1μL / L.

[0102] All acquisition units employ a combination of wired and wireless transmission modules. The wired transmission module uses an RS485 bus, while the wireless transmission module utilizes LoRa wireless transmission technology. The data transmission rate is set to 5Mbps. The acquired multi-source raw feature information is transmitted to the data processing terminal (model IPC-610). A data backup mechanism is also set up to back up the acquired raw data to the local server to prevent data loss. The acquisition process continues for 72 hours to form a multi-source raw feature dataset, with a data acquisition interval of 10 seconds per acquisition.

[0103] Step 2: Feature Information Preprocessing

[0104] The multi-source raw feature dataset collected in step 1 is subjected to outlier removal, missing value imputation, and noise reduction to obtain a standardized feature dataset. The specific sub-steps are as follows:

[0105] Sub-step 2.1: Outlier Removal. Outliers are identified using the 3σ criterion. First, the mean and standard deviation of each feature parameter in the multi-source raw feature data are calculated. Taking vibration amplitude (collected from a vibration sensor at the contact point) as an example, its mean is 0.5 mm and its standard deviation is 0.1 mm. Therefore, data exceeding the range of 0.2 mm to 0.8 mm are all considered outliers. Statistical analysis shows that in this multi-source raw feature dataset, outliers are mainly concentrated in vibration signals, sound signals, and current signals, with a total of 86 outliers identified. After removing all outliers, the mean of the feature data at adjacent time points is used to replace the data at the location of the outlier to ensure the continuity of the dataset.

[0106] Sub-step 2.2: Missing Value Imputation. Through statistical analysis, the location and proportion of missing values ​​in the multi-source original feature information were determined. The total dataset collected this time consisted of 25,920 data points, with 1,036 missing values, resulting in an overall missing value rate of 4.0%, less than 5%. Specifically, there were 42 missing values ​​for gas features (1.6%), 586 for mechanical features (2.3%), 328 for electrical features (1.3%), and 80 for environmental features (0.3%). Since the overall missing value rate was less than 5%, linear interpolation was used to impute all missing values. The estimated value of each missing value was calculated based on the linear relationship between two adjacent valid data points. After imputation, the dataset had no missing values, ensuring its integrity.

[0107] Sub-step 2.3: Noise Reduction. For vibration and sound signals in the mechanical feature information, and voltage and current signals in the electrical feature information, wavelet denoising is used. A db4 wavelet basis is selected, and the decomposition level is set to 5 levels. The signals are decomposed, thresholded, and reconstructed to remove high-frequency noise signals. For environmental feature information (temperature, humidity, air pressure) and gas feature information (SF6 purity, moisture content, decomposition product concentration), a moving average filtering method is used for noise reduction. The window size is set to 10 data points. The average value of the data within the window is calculated and replaced with the data at the center of the window to smooth data fluctuations. After noise reduction, the fluctuation amplitude of the feature data is significantly reduced, and noise signals are effectively removed, resulting in a standardized feature dataset.

[0108] Step 3: Multi-dimensional feature selection

[0109] The standardized feature dataset obtained in step 2 is subjected to multi-dimensional feature filtering to identify key feature parameters related to the fault of the 220kV GIS switch, forming a fault feature dataset. The specific process is as follows:

[0110] First, the core indicators for feature selection are preset, including the correlation between features and faults, the stability of features, the discriminative power of features, and the collectability of features, thus clarifying the selection criteria.

[0111] Secondly, the correlation between features and faults was calculated. Feature data of various faults (contact jamming, opening / closing mechanism faults, insulation aging, SF6 gas leakage, internal arc discharge) of the 220kV GIS switch during occurrence and during normal operation were collected. The correlation coefficient between each feature parameter in the standardized feature dataset and various faults was calculated, with a preset correlation coefficient threshold of 0.6. Feature parameters with an absolute correlation coefficient greater than 0.6 were retained. The calculated correlation coefficients were: SO2 decomposition product concentration in SF6 gas and internal arc discharge faults: 0.88; vibration amplitude and contact jamming faults: 0.82; insulation resistance and insulation aging faults: 0.85; contact travel and opening / closing mechanism faults: 0.78; and SF6 gas purity and gas leakage faults: 0.80. All of these feature parameters were retained. However, the correlation coefficients between ambient air pressure and various faults were all less than 0.4, and the correlation coefficients between ambient humidity and various faults were all less than 0.5, and these were discarded.

[0112] Next, characteristic parameters with good stability were selected. The fluctuation range of the selected characteristic parameters was statistically analyzed under different operating conditions (different loads, different ambient temperatures). A preset fluctuation range threshold of 10% was used to remove characteristic parameters with fluctuation ranges greater than 10%. Statistical analysis showed that the fluctuation range of contact stroke under different operating conditions was 4.5%, vibration frequency was 6.2%, insulation resistance was 5.8%, and SF6 gas purity was 3.2%, all less than 10%, and were retained. However, the fluctuation range of sound frequency under different operating conditions was 12.3%, greater than 10%, and was therefore removed.

[0113] Next, the feature parameters with strong distinguishability were selected. The difference degree of each retained feature parameter corresponding to different fault types was calculated, with a preset difference degree threshold of 2.0. Feature parameters with a difference degree greater than 2.0 were retained. After calculation, the vibration frequency (mean value of contact jamming fault is 520Hz, mean value of opening and closing mechanism fault is 210Hz, difference degree is 520 / 210≈2.48), SF6 gas SO2 concentration (mean value of internal arc discharge fault is 45μL / L, mean value of normal operation is 5μL / L, difference degree is 45 / 5=9.0), and insulation resistance (mean value of insulation aging fault is 8.5×10^7Ω, mean value of insulation breakdown fault is 7.2×10^6Ω, difference degree ≈11.8) were all greater than 2.0 and were retained; while the difference degree of SF6 gas moisture content in different fault types was 1.8, which was less than 2.0 and was rejected.

[0114] Finally, highly collectable feature parameters were selected. The difficulty, cost, and substitutability of retaining feature parameters were analyzed. The feature parameters retained in this study (vibration amplitude, vibration frequency, contact stroke, opening and closing speed, insulation resistance, SF6 gas purity, and SF6 gas SO2 decomposition product concentration) were all acquired using existing conventional acquisition equipment. These parameters were easy to acquire, low in cost, and had no other substitutes, thus all were retained. This resulted in a fault feature dataset containing seven key feature parameters.

[0115] Step 4: Fault Type Identification

[0116] By combining the fault feature template library, similarity matching is performed on the fault feature dataset obtained in step 3 to identify the fault type. The specific process is as follows:

[0117] (1) Construction of Fault Feature Template Library: Various fault cases of GIS switches of different voltage levels (110kV, 220kV, 500kV), different operating years (1-20 years), and different models were collected, totaling 320 cases, including 120 mechanical faults, 100 electrical faults, 80 gas faults, and 20 compound faults. The fault cases were classified and organized. For common faults of this 220kV GIS switch, the typical characteristic parameter range and characteristic change law of each fault type were extracted to form corresponding fault feature templates, such as:

[0118] ① Characteristics of contact jamming fault: vibration amplitude 0.8mm-1.2mm, vibration frequency 400Hz-600Hz, contact stroke less than 80% of rated stroke (80mm) (i.e. less than 64mm), opening and closing speed less than 70% of rated speed (5m / s) (i.e. less than 3.5m / s), vibration amplitude gradually increases with time, and opening and closing time gradually prolongs;

[0119] ② Internal arc discharge fault characteristic template: SF6 gas SO2 decomposition product concentration 30μL / L-80μL / L, vibration amplitude 0.6mm-1.0mm, insulation resistance less than 10^8Ω, voltage waveform distortion rate greater than 5%, SF6 gas SO2 concentration increases rapidly over time.

[0120] ③ SF6 gas leakage fault characteristic template: SF6 gas purity is less than 90%, purity decrease rate is greater than 0.5% / day, gas chamber pressure gradually decreases, and vibration amplitude is not obviously abnormal (0.3mm-0.6mm).

[0121] Integrate all fault feature templates to establish a fault feature template library, set up a dynamic update mechanism for the template library, collect new fault cases every 6 months, and optimize template parameters.

[0122] (2) Feature similarity matching: Extract the values ​​of 7 key feature parameters from the fault feature dataset obtained in step 3 (in the feature data collected this time, the vibration amplitude is 0.92mm, the vibration frequency is 510Hz, the contact stroke is 62mm, the opening and closing speed is 3.2m / s, the insulation resistance is 9.2×10^7Ω, the SF6 gas purity is 92%, and the concentration of SF6 gas SO2 decomposition products is 8μL / L); match the feature data with each fault template in the fault feature template library, use the cosine similarity calculation method to calculate the similarity value of each feature parameter, and assign corresponding weights according to the importance of each feature parameter (vibration amplitude weight 0.25, vibration frequency weight 0.2, contact stroke weight 0.2, opening and closing speed weight 0.15, insulation resistance weight 0.1, SF6 gas purity weight 0.05, SF6 gas SO2 concentration weight 0.05), and calculate the overall similarity value.

[0123] Calculations show that the overall similarity between the fault feature data and the contact jamming fault feature template is 0.83, which is greater than the preset identification threshold of 0.7. The overall similarity between the data and other fault type templates is less than 0.6. Therefore, it is determined that the 220kV GIS switch has a contact jamming fault, and the fault type is output as "contact jamming fault". At the same time, common fault causes (contact surface dirt, wear, excessive clearance of transmission components) and preliminary maintenance suggestions (check the condition of the contact surface and adjust the clearance of transmission components) are also output.

[0124] Step 5: Fault Location

[0125] Based on the installation location of each acquisition unit in the distributed sensing architecture, and combined with the contact jamming fault identified in step 4, the specific location of the fault is determined. The specific process is as follows:

[0126] First, determine the correspondence between the data acquisition units and the GIS switch parts: the vibration sensor (labeled "Contact-Vibration 1") and displacement sensor (labeled "Contact-Displacement 1") installed at the contact parts collect characteristic parameters corresponding to the contact parts; the vibration sensor (labeled "Transmission-Vibration 1") installed at the transmission components collect characteristic parameters corresponding to the transmission components; the vibration sensor (labeled "Mechanism-Vibration 1") installed at the opening and closing mechanism collects characteristic parameters corresponding to the opening and closing mechanism.

[0127] Then, the abnormal locations of typical characteristic parameters corresponding to contact jamming faults are analyzed: the typical abnormal locations of contact jamming faults are the contact area and adjacent transmission components, and the corresponding abnormal characteristic parameters mainly come from the acquisition units of the contact area and transmission components.

[0128] Next, considering the abnormal amplitude values ​​of the characteristic parameters collected by each acquisition unit: the abnormal amplitude of vibration collected by "Contact-Vibration 1" was 0.92mm (normal amplitude is 0.5mm, abnormal amplitude accounts for 84%); the abnormal amplitude of vibration collected by "Transmission-Vibration 1" was 0.65mm (abnormal amplitude accounts for 30%); the abnormal amplitude of vibration collected by "Mechanism-Vibration 1" was 0.35mm (abnormal amplitude accounts for -30%, no obvious abnormality); and the abnormal amplitude of contact stroke collected by the displacement sensor was 62mm (rated stroke 80mm, abnormal amplitude accounts for -22.5%). Based on the magnitude of the abnormal amplitude, it was determined that the core part of the fault was the contact part, the transmission component was slightly affected, and the opening and closing mechanism was normal.

[0129] Finally, based on the structural drawings of the 220kV GIS switch, the specific location information of the fault is marked: the core fault location is the A-phase contact of the GIS switch, with installation coordinates (X: 1200mm, Y: 600mm, Z: 900mm). The fault range is the contact area of ​​the contact and the transmission components within an adjacent 50mm range. A schematic diagram of the fault location is generated to provide accurate guidance for operation and maintenance.

[0130] Step 6: Fault Level Assessment

[0131] Based on the contact jamming fault identified in step 4 and the fault location determined in step 5, combined with the abnormal amplitude of characteristic parameters, the fault level is assessed, and a diagnostic report and maintenance recommendations are generated. The specific process is as follows:

[0132] First, clarify the fault level classification criteria, and based on the rated parameters of the 220kV GIS switch, preset the amplitude thresholds for each key characteristic parameter: the preset threshold for vibration amplitude is 1.0mm, the preset threshold for contact travel is 64mm (80% of the rated travel), and the preset threshold for opening and closing speed is 3.5m / s (70% of the rated speed).

[0133] Then, the fault level was determined: In this fault, the abnormal vibration amplitude was 0.92 mm, accounting for 92% of the preset threshold (0.92 / 1.0×100%), which is greater than 90%; the abnormal contact stroke amplitude was 62 mm, accounting for 96.9% of the preset threshold (62 / 64×100%), which is greater than 90%; the abnormal opening and closing speed amplitude was 3.2 m / s, accounting for 91.4% of the preset threshold (3.2 / 3.5×100%), which is greater than 90%; the fault type was contact jamming fault, the fault location was the core component (contact), which has caused the switch opening and closing speed to slow down and the stroke to be insufficient, affecting the normal operation of the GIS switch, and may cause more serious faults such as contact welding, which poses a serious safety hazard. According to the fault level classification standard, this fault was determined to be a level four fault (fatal fault).

[0134] Finally, a diagnostic report and maintenance recommendations are generated: The diagnostic report includes the basic information of the GIS switch (model ZF28-252, voltage level 220kV, service life 8 years), fault diagnosis time (e.g., diagnosis time is January 1, 2026), fault type (contact jamming fault), fault location (A phase contact, installation coordinates X:1200mm, Y:600mm, Z:900mm), fault level (level 4 fault), and abnormal characteristic parameters (vibration amplitude 0.92mm, contact stroke 62mm, opening and closing speed 3). (2 m / s, both exceeding the preset threshold by 90%). Fault cause analysis (preliminary judgment is that the contact surface is contaminated with dirt and slightly worn, causing the contact to jam). The maintenance recommendation is: immediately disconnect the relevant circuit of the GIS switch, start the backup switch, organize emergency maintenance, disassemble the contact part, clean the dirt and wear impurities on the contact surface, check the degree of contact wear, replace the contact if the wear is serious, adjust the clearance of the transmission components, conduct trial operation after maintenance, monitor whether the characteristic parameters return to normal, monitor the characteristic parameters once a day for 7 days to ensure the normal operation of the equipment.

[0135] Step 7: Early warning of faults (optional step)

[0136] Before fault type identification in step 4, trend analysis was performed on the fault feature dataset obtained in step 3 to identify potential faults: In the feature data collected this time, the vibration amplitude gradually increased from 0.5mm to 0.92mm in the first 48 hours, with a change rate of 0.00875mm / h. The degree of deviation from the normal range (0.3mm-0.7mm) was 31.4% ((0.92-0.7) / 0.7×100%), which is greater than 20% and did not reach the fault identification threshold (the vibration amplitude is 0.8mm when the overall similarity value reaches 0.7). Therefore, it was determined to be a potential fault, and a serious warning (red warning signal) was issued. A warning report was generated to remind maintenance personnel to pay close attention to the operating status of the contact part, prepare maintenance materials, and prevent the fault from developing into an actual fault. However, because the maintenance personnel did not take emergency measures in time after the warning, the fault further developed and was finally determined to be a level four fatal fault.

[0137] Step 8: Verification and feedback optimization of diagnostic results (optional step)

[0138] Based on the diagnostic report and fault location information, maintenance personnel conducted an emergency overhaul of the 220kV GIS switch. After disassembling the contact area, they found a large amount of dirt accumulation on the contact surface, slight wear, and excessive clearance in the transmission components. This was completely consistent with the diagnostic results (contact jamming fault, fault location: phase A contact), indicating an accurate diagnosis. After the overhaul, characteristic data were re-collected, and the vibration amplitude returned to 0.48mm, the contact stroke returned to 78mm, and the opening and closing speed returned to 4.8m / s, all within the normal range. The equipment resumed normal operation.

[0139] The fault cases (contact jamming, level 4 fault, abnormal characteristic parameters, and maintenance results) were added to the fault feature template library. The feature template for contact jamming faults was optimized, and the typical range of vibration amplitude was adjusted to 0.8mm-1.2mm (from the original range of 0.8mm-1.1mm). The relevant rule of "when the abnormal amplitude of the opening and closing speed is greater than 90%, it is judged as a level 4 fault" was added. At the same time, the reasons why the early warning did not play a preventive role were analyzed, and the early warning signal push mechanism was adjusted to add SMS push function to ensure that the early warning signal can be delivered to the operation and maintenance personnel in a timely manner. After optimization, the response efficiency of the early warning signal was improved by more than 60%.

[0140] The diagnostic process, maintenance results, and optimization adjustments were recorded to create an operation and maintenance file, providing data support for fault diagnosis and operation and maintenance optimization of other GIS switches in the substation. The fault diagnosis accuracy rate was calculated monthly to ensure that the accuracy rate remained above 95%.

[0141] Example 2: Diagnostic Verification of Different Fault Types

[0142] To further verify the effectiveness and versatility of the method of the present invention, the method of the present invention was used to diagnose and verify other common faults of the 220kV GIS switch. The specific verification results are as follows:

[0143] 1. Verification Case 1: SF6 Gas Leakage Fault. Among the collected feature data, the SF6 gas purity was 88% (preset threshold 90%), the purity decrease rate was 0.6% / day, and the vibration amplitude was 0.45mm (normal range). After feature screening and similarity matching, the overall similarity value with the SF6 gas leakage fault template was 0.81, greater than 0.7, thus diagnosing it as an SF6 gas leakage fault. The fault was located in the sealing component of the gas chamber, and the fault level was classified as a level two fault (general fault). After inspection by maintenance personnel, it was found that the gas chamber's sealing gasket had aged, causing the gas leakage, consistent with the diagnosis. After repair, the equipment returned to normal, demonstrating a 100% diagnostic accuracy rate.

[0144] 2. Verification Case 2: Insulation Aging Fault. Among the collected characteristic data, the insulation resistance was 7.5 × 10^7 Ω (preset threshold 10^8 Ω), the vibration amplitude was 0.52 mm, and the SF6 gas SO2 concentration was 6 μL / L (normal range). The overall similarity value with the insulation aging fault template was 0.79, indicating an insulation aging fault. The fault was located in an insulating component, and the fault level was classified as Level III (severe fault). After repair, the insulation resistance recovered to 1.2 × 10^8 Ω, and the equipment operated normally, demonstrating accurate diagnosis.

[0145] 3. Verification Case 3: Composite Fault (SF6 Gas Leakage Accompanied by Internal Arc Discharge). Among the collected characteristic data, the SF6 gas purity was 87%, the SO2 concentration was 35 μL / L, the vibration amplitude was 0.85 mm, and the insulation resistance was 8.2 × 10^7 Ω. After similarity matching, the overall similarity value with the SF6 gas leakage fault template was 0.76, and the overall similarity value with the internal arc discharge fault template was 0.78, both greater than 0.7, indicating a composite fault (SF6 gas leakage accompanied by internal arc discharge). The fault was located in the gas chamber and internal contacts, classified as a level three fault. After repair, it was found that the gas chamber leakage led to a decrease in SF6 gas purity, which in turn triggered internal arc discharge, consistent with the diagnostic results, confirming the accuracy of the diagnosis.

[0146] 4. Verification Case 4: Unknown Fault. In the collected feature data, the anomalies of each key feature parameter did not match any fault templates in the fault feature template library, with an overall similarity value less than 0.7. This was diagnosed as an unknown fault, triggering an alarm. After manual investigation by maintenance personnel, it was found to be a coil fault in the opening and closing mechanism (a rare fault). This fault case was added to the template library, optimizing the library so that this type of fault can be accurately diagnosed in the future.

[0147] Verification through the above four cases shows that the method of the present invention can accurately identify single and compound faults of GIS switches, and can trigger alarms in a timely manner for unknown faults. It has high diagnostic accuracy and strong versatility, and can meet the diagnostic needs of different fault types.

[0148] Example 3: Comparative Verification with Existing Diagnostic Methods

[0149] The method of this invention was compared and verified with two existing mainstream diagnostic methods (single vibration feature diagnostic method and neural network-based diagnostic method). For 100 fault cases of this 220kV GIS switch, the three methods were used for diagnosis, and the diagnostic accuracy, diagnostic time, operational difficulty, and maintenance cost were compared. The specific comparison results are shown in Table 1 below:

[0150] Table 1 Comparison of the three diagnostic methods

[0151]

[0152] As shown in Table 1, the diagnostic accuracy of the method of the present invention (96.0%) is higher than that of the single vibration feature diagnostic method (72.0%), and slightly higher than that of the neural network-based diagnostic method (94.5%). The average diagnostic time (15 min) is shorter than that of the two existing methods, and the operation difficulty is the lowest, requiring no professional personnel or high-performance hardware equipment. The average maintenance cost (800 yuan / time) is much lower than that of the neural network-based diagnostic method (2500 yuan / time), and slightly higher than that of the single vibration feature diagnostic method (750 yuan / time). The overall performance is the best, with significant advantages.

[0153] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for diagnosing faults in GIS switches, characterized in that: Includes the following steps: Step 1: Multi-source feature information acquisition. Collect mechanical feature information, electrical feature information, environmental feature information and gas feature information during the operation of GIS switch to form a multi-source raw feature dataset; Step 2: Feature information preprocessing. The collected multi-source raw feature information is subjected to outlier removal, missing value completion and noise reduction to obtain a standardized feature dataset. Step 3: Multi-dimensional feature screening. Based on preset feature screening rules, key feature parameters related to GIS switch faults are screened from the standardized feature dataset to form a fault feature dataset. Step 4: Fault type identification. The fault feature dataset is matched with the preset fault feature template library. Combined with the feature similarity judgment rules, the specific fault type of the GIS switch is identified. Step 5: Fault location. Based on the identified fault type, combined with the collection location of multi-source feature information and the degree of abnormality of feature parameters, determine the specific location where the fault occurred. Step 6: Fault Level Assessment. Based on the fault type, fault location, and abnormal amplitude of characteristic parameters, combined with the preset fault level classification standards, assess the severity of the fault and generate corresponding diagnostic reports and maintenance recommendations.

2. The GIS switch fault diagnosis method according to claim 1, characterized in that, The collection of multi-source feature information in step 1 adopts a distributed sensing architecture, with corresponding collection units set up at key locations of the GIS switch. The specific collection content includes: (1) Mechanical characteristic information: Vibration signal, vibration frequency and vibration amplitude during the operation of GIS switch are collected by vibration sensor; the stroke, opening and closing speed and bounce of switch contact are collected by displacement sensor; and sound signal and sound frequency during switch operation are collected by sound sensor. (2) Electrical characteristic information: The voltage signal, voltage amplitude and voltage waveform at both ends of the GIS switch are collected by the voltage sensor, the current signal, current amplitude and current waveform in the switch circuit are collected by the current sensor, and the insulation resistance and leakage current of the switch are collected by the insulation sensor. (3) Environmental characteristics information: The temperature of the GIS switch operating environment and the temperature distribution of the switch body are collected by temperature sensor, the relative humidity of the operating environment is collected by humidity sensor, and the atmospheric pressure of the operating environment is collected by air pressure sensor. (4) Gas characteristic information: The purity, moisture content and concentration of decomposition products of SF6 gas inside the GIS switch are collected by gas sensors. The decomposition products include SO2, H2S, CO2 and CF4.

3. The GIS switch fault diagnosis method according to claim 1, characterized in that, The feature information preprocessing described in step 2 specifically includes the following sub-steps: Sub-step 2.1: Outlier removal. The 3σ criterion is used to identify outliers in the multi-source original feature information. Feature data that exceeds the mean ± 3 times the standard deviation are judged as outliers. After removing outliers, the mean of the feature data at adjacent time points is used to replace the data at the location of the outlier to ensure the continuity of the dataset. Sub-step 2.2: Missing value completion. Statistical analysis is used to determine the location and proportion of missing values ​​in the multi-source original feature information. When the missing proportion is less than 5%, linear interpolation is used to complete the missing values. When the missing proportion is between 5% and 15%, the median of the same type of feature data at adjacent time points is used to complete the missing values. When the missing proportion is greater than 15%, the acquisition unit self-check is triggered, and the corresponding feature information is re-acquired. If the acquisition fails, an acquisition unit fault is indicated. Sub-step 2.3: Noise reduction processing. For vibration signals and sound signals in mechanical feature information and voltage and current signals in electrical feature information, wavelet noise reduction method is used to remove noise signals caused by environmental interference and interference from the acquisition equipment itself. For environmental feature information and gas feature information, moving average filtering method is used to reduce noise, smooth data fluctuations, and obtain a standardized feature dataset.

4. The GIS switch fault diagnosis method according to claim 1, characterized in that, The specific process of multi-dimensional feature filtering in step 3 is as follows: First, the core indicators for feature selection are preset, including the correlation between features and faults, the stability of features, the discriminative power of features, and the collectability of features. Second, the correlation coefficient between each feature parameter in the standardized feature dataset and common fault types of GIS switches is calculated, and feature parameters with correlation coefficients greater than a preset threshold are retained. Then, by statistically analyzing the fluctuation range of feature parameters under different operating conditions, unstable feature parameters with fluctuation ranges greater than a preset range are removed. Next, the difference in feature parameters corresponding to different fault types is calculated, and feature parameters with difference greater than a preset threshold are retained to ensure that the feature parameters have good fault discrimination ability. Finally, feature parameters that are difficult to collect, costly to collect, and easily substitutable are removed, ultimately forming a fault feature dataset.

5. The GIS switch fault diagnosis method according to claim 1, characterized in that, The process of constructing the fault feature template library in step 4 is as follows: Various fault cases of GIS switches with different voltage levels and service lives are collected, including fault types, multi-source characteristic parameters at the time of fault occurrence, fault location, and fault cause. The collected fault cases are classified and organized. For each fault type, the typical characteristic parameter range and characteristic change pattern are extracted to form a corresponding fault feature template. The feature templates of all fault types are integrated to establish a fault feature template library. The fault feature template library includes mechanical fault templates, electrical fault templates, gas fault templates, and composite fault templates. The fault feature template library supports dynamic updates, and new fault cases and feature templates are added regularly to improve the accuracy of fault identification.

6. The GIS switch fault diagnosis method according to claim 5, characterized in that, The feature similarity judgment rule in step 4 is as follows: The similarity between each key feature parameter in the fault feature dataset and the corresponding fault type feature template in the fault feature template library is calculated using the cosine similarity method to obtain the similarity value of each feature parameter. Then, according to the importance of each feature parameter, a corresponding weight is assigned, and the weighted average of the similarity values ​​of all feature parameters is calculated to obtain the overall similarity value. If the overall similarity value is greater than the preset recognition threshold, the GIS switch is determined to have experienced this type of fault. If the overall similarity value is less than or equal to the recognition threshold, it continues to be matched with feature templates of other fault types until a matching fault type is found. If no match is found, it is determined to be an unknown fault, and an alarm is triggered.

7. The GIS switch fault diagnosis method according to claim 1, characterized in that, The specific method for fault location described in step 5 is as follows: Based on the installation location of each acquisition unit in the distributed sensing architecture, the acquisition location corresponding to each key feature parameter is determined. For the identified fault type, the abnormal locations of typical feature parameters corresponding to the fault are analyzed. Combined with the abnormal amplitude of feature parameters acquired by each acquisition unit, the core location of the fault is determined. If the feature parameters acquired by multiple acquisition units are abnormal, the propagation path and core fault location of the fault are determined according to the magnitude of the abnormal amplitude and the time sequence of the abnormality. At the same time, combined with the structural drawings of the GIS switch, the specific location information of the fault location is marked, including component name, installation coordinates and fault range, to provide accurate guidance for operation and maintenance.

8. The GIS switch fault diagnosis method according to claim 1, characterized in that, The fault level classification standard mentioned in step 6 is divided into four levels, specifically: Level 1 fault (minor fault): The abnormal amplitude of the characteristic parameter is less than 30% of the preset amplitude threshold. The fault does not affect the normal operation of the GIS switch, there is no obvious safety hazard, only periodic monitoring is required, and no immediate maintenance is needed. Level 2 fault (general fault): The abnormal amplitude of the characteristic parameter is between 30% and 60% of the preset amplitude threshold. The fault has a slight impact on the operational stability of the GIS switch and poses a potential safety hazard. Repair must be completed within 1-3 working days. Level 3 fault (serious fault): The abnormal amplitude of the characteristic parameter is between 60% and 90% of the preset amplitude threshold. The fault has affected the normal operation of the GIS switch and poses a significant safety hazard. It is necessary to shut down the machine immediately for maintenance to prevent the fault from escalating. Level 4 fault (fatal fault): The abnormal amplitude of the characteristic parameter is greater than 90% of the preset amplitude threshold. The fault has caused the GIS switch to malfunction and may cause a serious safety accident. It is necessary to immediately disconnect the relevant circuit, start the backup equipment, organize emergency repairs, and investigate the chain reaction caused by the fault.

9. A method for diagnosing faults in a GIS switch according to claim 1, characterized in that, It also includes an early warning step for faults. Before fault type identification in step 4, a trend analysis is performed on the fault feature dataset. If the trend of change of key feature parameters is developing in an abnormal direction and has not reached the fault identification threshold, it is determined to be a potential fault. According to the rate of change of feature parameters and the degree of deviation from the normal range, a corresponding warning signal is issued. The warning signal is divided into three levels, corresponding to minor warning, general warning and severe warning respectively. At the same time, a warning report is generated to prompt maintenance personnel to take targeted preventive measures to avoid the occurrence of faults.

10. A method for diagnosing faults in a GIS switch according to any one of claims 1-9, characterized in that, The method also includes diagnostic result verification and feedback optimization steps, specifically: Based on the diagnostic report and fault location information, maintenance personnel inspect the GIS switches to confirm the actual fault type, location, and severity. They then compare the actual fault information with the diagnostic results to calculate the diagnostic accuracy. If the diagnostic results do not match the actual fault information, they analyze the reasons for the error and adjust the feature selection rules, fault feature template library, and similarity judgment threshold. Simultaneously, actual fault cases are added to the fault feature template library, ensuring dynamic updates and continuous optimization of the fault diagnosis method's accuracy and adaptability. If the diagnostic results match the actual fault information, the diagnostic process and repair results are recorded to create an maintenance file, providing data support for subsequent fault diagnosis and maintenance optimization.

Images