Operation state and characterization state parameter method of GIS isolation switch

Abstract

The invention discloses an operation state and characterization state parameter method of a GIS isolation switch, and relates to the field of GIS isolation switch operation. Multi-source heterogeneous sensors including vibration, temperature, strain gauges, fiber bragg gratings, acoustic cameras and the like are deployed at key parts, and data are collected in a self-adaptive mode according to working conditions; processing data by using a deep learning noise reduction auto-encoder, adaptive weighted fusion and DBSCAN; deeply mining features by using empirical wavelet transform, a heat conduction model and the like; constructing a composite state parameter by means of a deep belief network and principal component analysis in combination with particle swarm optimization; the state is evaluated through a convolutional neural network of transfer learning, LSTM prediction and early warning are carried out, and a fault tree and a Bayesian network are combined to locate a fault. According to the invention, data is comprehensively collected, intelligent processing and deep mining are carried out, scientific state parameters are constructed, and accurate state evaluation and early warning are realized in combination with transfer learning; the adaptability is improved through self-learning and multi-modal fusion, the operation and maintenance efficiency is improved through a VR / AR visual platform, the risk of the power system is reduced, and benefits are remarkable.

Description

Technical Field

[0001] The present invention relates to the technical field of GIS disconnector operation, and in particular to an operating state of a GIS disconnector and a method for characterizing state parameters. Background Art

[0002] GIS (Gas-Insulated Metal-Enclosed Switchgear) disconnectors are crucial equipment in power systems. They perform critical functions such as power isolation and switching operations, and their operating status is directly related to the safety and stability of the power system. However, there are currently many problems in monitoring and evaluating the operating status of GIS disconnectors.

[0003] Most existing monitoring methods rely solely on a single type of sensor, such as vibration or temperature sensors. This single-source monitoring approach has significant limitations and cannot fully and accurately reflect the true operating status of the disconnector. For example, while vibration sensors can detect equipment vibration, they are unable to detect issues such as thermal failure of contacts. Temperature sensors can only monitor temperature changes and are unable to effectively capture issues such as wear and loosening of mechanical structures. Furthermore, data from a single sensor is susceptible to external interference, reducing the reliability of monitoring results.

[0004] Traditional methods for data processing and analysis are relatively simple, often relying on threshold judgments and other methods. This approach fails to fully tap into the rich information contained in the data and is insufficient for early warning of faults. Often, monitoring systems only sound alarms after equipment has already experienced a serious fault, by which time significant losses have already occurred and may even impact the normal operation of the entire power system. Furthermore, traditional methods struggle to adapt to changes in equipment operating conditions and environments, lacking adaptability and intelligence.

[0005] There is currently a lack of a scientific and rational system for evaluating the condition of GIS disconnectors. Existing evaluation indicators are often too simplistic and fail to comprehensively consider various equipment factors, resulting in inaccurate evaluation results. This makes it difficult for operators to formulate reasonable maintenance plans based on the evaluation results. This can lead to either excessive maintenance resulting in wasted resources or insufficient maintenance leading to frequent equipment failures. Therefore, developing a GIS disconnector, accurately monitoring its operating status, and constructing a scientific and rational method for characterizing its condition parameters are urgent and crucial for improving the reliability and economic efficiency of power systems. Summary of the Invention

[0006] The present invention proposes an operating state of a GIS disconnector and a method for characterizing state parameters to solve the problems mentioned in the above-mentioned prior art.

[0007] In order to achieve the above object, the present invention adopts the following technical solution: a method for measuring the operating state and characterizing the state parameters of a GIS disconnector, comprising:

[0008] Multi-source heterogeneous data acquisition: Vibration, temperature, displacement, and current sensors are deployed at key locations on GIS disconnectors. Strain gauges and fiber Bragg grating sensors are also added. Vibration sensors adaptively adjust sampling frequencies, strain gauges measure stress, and fiber Bragg grating sensors measure deformation. Acoustic cameras are used to capture sound distribution images, and spectrum analysis is combined to collect equipment operating status data.

[0009] Intelligent data preprocessing: The noise reduction autoencoder (DAE) is used to denoise the vibration signal of the original data. An adaptive weighted fusion algorithm is used to dynamically assign weights to the data of each sensor. The spatial clustering algorithm DBSCAN is used to identify and process outliers.

[0010] Deep feature mining: Mining pre-processed data features, using empirical wavelet transform (EWT) combined with Hilbert-Huang transform (HHT) to analyze vibration frequency and energy distribution, introducing a heat conduction model, calculating the temperature gradient change rate through a formula, extracting stress, deformation, sound spectrum, and sound intensity data features to characterize the operating status of the disconnector;

[0011] Composite state parameter construction: Deep belief network (DBN) combined with principal component analysis (PCA) is used to construct comprehensive state parameters. DBN is used to mine feature relationships, PCA is used for dimensionality reduction, and the particle swarm optimization algorithm (PSO) is used to optimize the matrix. Based on the principal component contribution rate and feature importance, the analytic hierarchy process (AHP) is used to determine the weight of each principal component to construct the composite state parameters.

[0012] Dynamic status assessment and early warning: Build an adaptive status assessment model based on transfer learning. Using a convolutional neural network (CNN) pre-trained with historical data, transfer learning techniques are used to transfer parameters and fine-tune the model. Thresholds are set, and when they are exceeded, long short-term memory (LSTM) networks are used for prediction and early warning. Fault tree analysis and Bayesian networks are combined to infer fault causes and locate fault locations.

[0013] Furthermore, it also includes:

[0014] Real-time online self-learning steps: The monitoring process uses an incremental learning algorithm to update the DBN and CNN model parameters in real time based on newly collected data. When the equipment operating conditions or environment changes, the self-organizing map network (SOM) clustering analysis data is used to automatically adjust the model structure and parameters.

[0015] Furthermore, it also includes:

[0016] Multimodal data fusion enhancement steps: Use the tensor fusion network TFN to fuse vibration, temperature, stress, deformation, and sound data. By constructing a tensor model, the multimodal fusion features are uniformly represented. The fusion process uses the mutual information maximization principle to optimize the fusion weight.

[0017] Furthermore, the deep feature mining step introduces a heat conduction model, and the temperature gradient change rate is calculated through finite element analysis to reflect the temperature change trend. The formula is:

[0018]

[0019] The generative adversarial network (GAN) variant Cycle-GAN is introduced to enhance features. Cycle-GAN performs unsupervised conversion between different types of features, generating diverse feature data to expand the feature library.

[0020] Furthermore, in the process of PCA dimensionality reduction, the particle swarm optimization algorithm PSO is used to optimize the principal component transformation matrix. The calculation formula is: Y = XW, where X is the feature matrix after DBN processing, W is the principal component transformation matrix after PSO optimization, and Y is the principal component matrix after dimensionality reduction. The weight of each principal component is determined by the hierarchical analysis method AHP to construct the composite state parameter. w i is the weight of the i-th principal component, y i is the value of the i-th principal component, m is the number of principal components;

[0021] The quantum-behaved particle swarm optimization (QPSO) algorithm is combined with fuzzy C-means clustering (FCM) to optimize the principal component weights. The QPSO algorithm determines the initial value of the principal component weight, and the FCM clustering is used to analyze the principal components. The weights are adjusted according to the clustering results.

[0022] Furthermore, the dynamic status assessment and early warning step uses blockchain technology to record status assessment results and early warning information, and encrypts and stores each status assessment time, composite status parameter value, assessment status and early warning details.

[0023] Furthermore, the multi-source heterogeneous data collection step uses wireless energy harvesting technology to power some sensors, and deploys electromagnetic induction energy harvesters near the operating mechanism of the disconnector to collect electromagnetic energy; deploys temperature difference energy harvesters at the contact position to generate electrical energy using the temperature difference between the contact and the surrounding environment, and stores the collected energy through the energy management circuit and powers the sensors.

[0024] Furthermore, the intelligent data preprocessing step adopts an adaptive weighted fusion algorithm, and the calculation formula is: w i is the weight of the i-th sensor data, σ i is the standard deviation of the i-th sensor data, n is the number of sensors;

[0025] Homomorphic encryption technology is used to encrypt and process sensitive data, and data noise reduction, fusion and outlier processing operations are completed in an encrypted state.

[0026] Furthermore, the deep feature mining step uses the graph neural network (GNN) to model the relationship between the various components of the disconnector. The disconnector structure is represented in the form of a graph, where nodes represent components and edges represent the connection relationship between components. GNN learns node features and edge features to mine the overall operating status features.

[0027] Furthermore, the entire method develops a visual monitoring platform based on virtual reality (VR) and augmented reality (AR), uses VR technology to build a virtual model of the GIS disconnector, maps the equipment operating status in real time, and intuitively views the internal structure of the equipment and the status of each component through an immersive experience; through AR technology, the real-time status of the equipment and early warning information are displayed in the operator's field of view in the form of augmented reality.

[0028] Compared with the existing technology, the beneficial effects of the present invention are:

[0029] In terms of data collection, the use of multi-source heterogeneous sensors comprehensively collects data covering multiple dimensions, including vibration, temperature, stress, deformation, and sound, greatly improving the comprehensiveness and accuracy of the data. The addition of strain gauges and fiber Bragg grating sensors can capture even more subtle changes in equipment status. The use of acoustic cameras enriches the data sources from an acoustic perspective, providing a richer information foundation for subsequent status assessments.

[0030] During the intelligent data preprocessing stage, the deep learning denoising autoencoder effectively removes noise, the adaptive weighted fusion algorithm dynamically allocates weights based on sensor reliability, and the DBSCAN algorithm is combined to process outliers, thereby improving data quality, reducing the impact of external interference on monitoring results, and making subsequent analysis more reliable.

[0031] In the deep feature mining step, a combination of advanced algorithms, such as the empirical wavelet transform, the Hilbert-Huang transform, and the heat conduction model, can deeply mine data features from different angles and accurately reflect the operating status of the equipment. The introduction of Cycle-GAN and graph neural networks further expands the feature library and improves the comprehensiveness and accuracy of feature mining.

[0032] The composite state parameters are constructed using deep belief networks, principal component analysis and improved particle swarm optimization algorithms, combined with the hierarchical analysis method to determine the weights. The constructed composite state parameters can more accurately and comprehensively characterize the operating status of the disconnector and provide a scientific basis for state assessment.

[0033] Dynamic state assessment and early warning, combined with transfer learning, LSTM prediction, and blockchain technology, not only accurately assesses the current state but also provides early warning of faults, ensuring the traceability of assessment results and warning information. Furthermore, real-time online self-learning and multimodal data fusion enhancements make the system adaptive and intelligent, adapting to changes in equipment operating conditions and the environment. The VR / AR-based visual monitoring platform provides operators with an intuitive and convenient operating experience, improving operational efficiency and reducing operational risks in the power system, resulting in significant economic and social benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A schematic block diagram of the operating state of a GIS disconnector and a method for characterizing state parameters proposed by the present invention;

[0035] Figure 2 This is a schematic diagram of fault warning time comparison;

[0036] Figure 3 This is a schematic diagram of the comparison of sensor data reliability;

[0037] Figure 4 A schematic diagram showing the comparison of state assessment accuracy. DETAILED DESCRIPTION

[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0039] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0040] In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the said features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined. In addition, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The present invention will be further described in detail below with reference to the accompanying drawings.

[0041] Reference Figures 1 to 4 : A method for measuring the operating state and characterizing state parameters of a GIS disconnector, comprising:

[0042] Multi-source heterogeneous data collection steps: Traditional vibration, temperature, displacement, and current sensors are deployed in key locations of the GIS disconnector, such as the contacts, transmission connecting rods, and operating motors. Strain gauges are also added to monitor stress changes in key components, and fiber Bragg grating (FBG) sensors are used to precisely measure minute deformations. The vibration sensors utilize high-precision microelectromechanical systems (MEMS)-based sensors. The sampling frequency, which can reach up to 50kHz and adaptively adjusts to the disconnector's operating frequency, ensures the capture of subtle vibration characteristics. Strain gauge sensors measure stress using the Wheatstone bridge principle, collecting data every 10ms. Fiber Bragg grating sensors measure deformation by detecting wavelength changes using a demodulator, recording data once per second. Furthermore, an acoustic camera is used to capture sound distribution images during disconnector operation. Combined with sound spectrum analysis, comprehensive data on equipment operating status is collected.

[0043] Intelligent data preprocessing step: For the collected raw data, the vibration signal is denoised using a deep learning-based denoising autoencoder (DAE). DAE effectively removes noise and retains key signal features by adding noise to the input data and then letting the model learn to restore the original data. For data such as temperature, displacement, current, stress, and deformation, an adaptive weighted fusion algorithm is used to dynamically assign weights for fusion based on the reliability of each sensor data. The weight calculation formula is: where w i is the weight of the i-th sensor data, σ i is the standard deviation of the i-th sensor data, and n is the number of sensors. The density-based spatial clustering algorithm (DBSCAN) is used to identify and process outliers in the data. Data that deviates from the normal data density area is considered abnormal and corrected.

[0044] Deep feature mining step: Deeply mine features from preprocessed data. For vibration signals, the empirical wavelet transform (EWT) is used to decompose them into multiple intrinsic mode functions (IMFs), and the Hilbert-Huang transform (HHT) is combined to calculate the marginal spectrum characteristics of each IMF to accurately analyze the vibration frequency and energy distribution. For temperature data, a heat conduction model is introduced and the temperature gradient change rate is calculated through finite element analysis. The formula is: This technology can reflect temperature trends. Features such as maximum stress and strain concentration areas are extracted from stress and deformation data. Features such as the center of gravity of sound energy distribution and sound intensity at specific frequencies are extracted from sound spectrum images to comprehensively characterize the operating status of the disconnector.

[0045] Steps for constructing composite state parameters: Deep belief network (DBN) combined with principal component analysis (PCA) is used to construct comprehensive state parameters. First, DBN is used to perform unsupervised learning on multi-source features to explore the potential relationship between features, and then the learned features are input into PCA for dimensionality reduction. During the PCA dimensionality reduction process, the principal component transformation matrix is optimized by the improved particle swarm optimization algorithm (PSO) so that the data after dimensionality reduction retains the original information to the maximum extent. The calculation formula is: Y = XW, where X is the feature matrix after DBN processing, W is the principal component transformation matrix after PSO optimization, and Y is the principal component matrix after dimensionality reduction. According to the principal component contribution rate and feature importance, the hierarchical analysis method (AHP) is used to determine the weight of each principal component to construct the composite state parameters. where w i is the weight of the i-th principal component, y i is the value of the i-th principal component, and m is the number of principal components.

[0046] Dynamic state assessment and early warning steps: Construct an adaptive state assessment model based on transfer learning. Using a convolutional neural network (CNN) as the base model, pre-training is performed using a large amount of historical data on GIS disconnectors with known states. In actual application, transfer learning technology is used to migrate the parameters of the pre-trained model to the currently monitored GIS disconnector data, and the model is fine-tuned using a small amount of current equipment data. Different state thresholds are set. When the composite state parameter exceeds the abnormal state threshold, the long short-term memory network (LSTM) is used to predict the state parameters for a period of time in the future and issue an early warning. If an impending fault is predicted, fault tree analysis and Bayesian network are combined to infer the cause of the fault and locate the fault location.

[0047] The present invention also includes the following modules:

[0048] Real-time online self-learning steps: During the monitoring process, incremental learning algorithms, such as online stochastic gradient descent, are used to update the parameters of models such as deep belief networks (DBNs) and convolutional neural networks (CNNs) in real time based on newly acquired sensor data, including multi-dimensional information such as vibration, temperature, and stress. Online stochastic gradient descent calculates the gradient of each sample and continuously adjusts model parameters in extremely small steps, enabling the model to quickly adapt to dynamic data changes. Whenever new monitoring data is input, the algorithm rapidly analyzes the data characteristics and fine-tunes the model accordingly, ensuring that the model always matches the current equipment operating status. When equipment operating conditions or environments undergo significant changes, self-organizing maps (SOMs) play a crucial role. SOMs are unsupervised neural networks capable of clustering large amounts of complex data. By processing data collected under different operating conditions and environments, SOMs can identify underlying patterns and structures within the data. SOMs map high-dimensional data into a low-dimensional space, clustering similar data points into similar locations. The system automatically detects changes in data distribution and adjusts the structure and parameters of models like DBN and CNN accordingly. When ambient temperature fluctuates significantly or equipment loads change, the SOM promptly captures the impact of these changes on the monitored data and guides the model to make corresponding adjustments, ensuring the accuracy and timeliness of GIS disconnector status assessments and parameter calculations, ensuring reliability under complex and changing operating conditions.

[0049] The present invention also includes the following modules:

[0050] Multimodal data fusion enhancement steps: A tensor fusion network (TFN) is used to construct a tensor model to mine complex, high-order correlations between multimodal data, including vibration, temperature, stress, deformation, and sound. When analyzing the operating status of GIS disconnectors, anomalies in vibration data are often intrinsically linked to stress and deformation data, and temperature changes may also be correlated with sound characteristics. The tensor model's unique mathematical structure comprehensively captures these complex relationships. During the fusion process, the TFN integrates the features of each modality into a unified feature representation. Vibration data can sensitively reflect the mechanical vibration frequency of the equipment, while sound data can reveal abnormal noise during operation. The fusion provides a more comprehensive picture of the equipment's operating status. To further enhance fusion performance, the TFN optimizes fusion weights using the principle of maximizing mutual information. Mutual information measures the degree of information sharing between different modal data. By continuously adjusting weights, modal data that contributes more significantly to the disconnector's operating status is given a higher priority in the fused features. When the equipment is operating under high load, stress and deformation data may be more critical for condition assessment. In this case, the TFN automatically increases the fusion weights of these two modal data. Through the above operations, TFN significantly improves the ability of the fused features to characterize the operating status of GIS disconnectors, thereby significantly improving the accuracy of status assessment and providing a solid guarantee for reliable operation monitoring of the equipment.

[0051] In the present invention, the deep feature mining step introduces a variant of the generative adversarial network (GAN), Cycle-GAN, to achieve feature enhancement. Cycle-GAN consists of a generator and a discriminator, and has a unique dual-domain cyclic conversion mechanism. When processing the monitoring data of the GIS disconnector, it can perform unsupervised conversion between different types of features, such as vibration features and sound features. This is because Cycle-GAN does not need to rely on labeled sample data and can directly learn the mapping relationship between different feature domains from the original sensor data. Specifically, when vibration feature data is input, the generator will try to convert it into the form of sound feature data, while the discriminator is responsible for judging whether the generated data is realistic and whether it belongs to the sound feature domain. Through this adversarial training method, the generator continuously optimizes the conversion capability and can generate diverse feature data. These newly generated data are not simply repeated, but cover a wider feature space. For example, it can simulate the combination of vibration and sound features that may appear under different operating conditions. The generated diverse feature data effectively expands the feature library. When constructing state parameters, the rich feature information provides more comprehensive support. This enables the status parameters to more accurately capture the complex operating status of the GIS disconnector. Whether it is subtle changes during normal operation or abnormal signs before a potential fault occurs, they can be more accurately characterized, thereby improving the comprehensiveness and accuracy of the GIS disconnector operating status assessment, and providing more powerful technical support for reliable operation monitoring and fault warning of the equipment.

[0052] In the present invention, the composite state parameter construction step uses the quantum behavior particle swarm optimization (QPSO) algorithm combined with fuzzy C-means clustering (FCM) to deeply optimize the principal component weights. The QPSO algorithm is based on the idea of quantum behavior and gives particles a more flexible search capability in the solution space. When determining the initial value of the principal component weight, it simulates the uncertainty and probabilistic characteristics of quantum, enabling particles to search efficiently in a wider solution space. Each particle represents a set of possible principal component weight combinations, and by continuously updating its own position, it gradually approaches the global optimal solution during the search process. This unique search mechanism, compared to the traditional particle swarm optimization algorithm, can avoid falling into the local optimal solution, thereby providing a more potential initial value for subsequent weight adjustment. After determining the initial value, the fuzzy C-means clustering (FCM) algorithm comes on the scene. FCM is a partition-based clustering algorithm that regards principal components as data points and divides them into different clusters by calculating the similarity between each principal component. In this process, FCM allows a data point to belong to multiple clusters at the same time with a certain degree of membership, thereby more flexibly reflecting the complex relationship between principal components. Based on the clustering results, the algorithm makes targeted adjustments to the principal component weights. For example, if certain principal components exhibit similar characteristics within a cluster, their weights may be adjusted appropriately to ensure that the composite state parameter more accurately reflects the actual operating status of the GIS disconnector. The organic combination of QPSO and FCM allows for a more scientific and precise construction of the composite state parameter, providing a solid foundation for accurate assessment of the operating status of GIS disconnectors.

[0053] In this invention, blockchain technology is incorporated into the dynamic status assessment and early warning process to record status assessment results and early warning information. Blockchain technology, with its distributed ledger and encryption algorithm, provides strong guarantees for secure data storage and traceability. During the GIS disconnector status monitoring process, key information generated by each status assessment includes: the assessment time, which accurately records the time point of the device status monitoring and is crucial for analyzing the changing trends of the device status over time; the composite status parameter value, as core data reflecting the device's operating status, which is an important basis for assessing the device's operating status; the assessment status (normal / abnormal / faulty), which clearly defines the device's current status category, providing intuitive judgment for operations and maintenance personnel; and early warning details, including information such as the fault type, possible cause, and recommended measures, providing guidance for timely addressing potential issues. This information is processed using an encryption algorithm and stored in various nodes of the blockchain. The blockchain's distributed structure allows data to be stored in multiple locations, rather than centralized on a single server, greatly enhancing data security and reliability. Furthermore, the blockchain's chain structure and timestamp mechanism ensure data immutability. Each addition of status assessment information is like adding a link to the chain; once recorded, it cannot be tampered with. This enables operation and maintenance personnel to trace and query status information at any time throughout the equipment life cycle, accurately understand the historical status changes of the equipment, provide reliable data support for formulating scientific and reasonable maintenance plans and fault diagnosis strategies, and effectively improve the operation and maintenance management level and reliability of GIS disconnectors.

[0054] In this invention, the multi-source heterogeneous data collection step utilizes wireless energy harvesting technology to power some sensors. An electromagnetic induction energy harvester is deployed near the operating mechanism of the disconnector. When the disconnector operates, it generates a changing magnetic field. Based on Faraday's law of electromagnetic induction, the electromagnetic induction energy harvester converts the electromagnetic energy in this changing magnetic field into electrical energy. The induction coil within the harvester generates an induced electromotive force (EMF) when the magnetic field changes. Through rectification, filtering, and other circuit processing, the AC power is converted into stable DC power, enabling energy collection and storage. A thermoelectric energy harvester is employed at the contact point. During operation, the contact generates heat due to factors such as the flow of current, creating a temperature difference with the surrounding environment. The thermoelectric energy harvester utilizes the Seebeck effect to convert this temperature difference into electrical energy. The thermoelectric energy harvester is composed of two different semiconductor materials, forming a thermocouple. When a temperature difference exists between the two ends, electrons migrate in a directed manner within the materials, generating an electromotive force. The harvested energy is processed by an energy management circuit. This circuit also features energy storage, storing excess electrical energy in capacitors or small batteries for use by the sensor when energy harvesting is insufficient. At the same time, the energy management circuit can also regulate the output voltage to ensure a stable and appropriate power supply voltage for the sensor. This wireless energy harvesting technology not only effectively reduces wiring costs and avoids complex cable laying and maintenance work, but also increases the flexibility of sensor deployment, allowing sensors to be installed in locations where wiring is difficult, further improving the reliability and stability of the system.

[0055] In the present invention, in the intelligent data preprocessing link, homomorphic encryption technology is used to encrypt sensitive data (such as data related to key performance parameters of the equipment). Homomorphic encryption is an advanced cryptographic technology that allows calculations to be performed directly on encrypted data without prior decryption. In the GIS isolation switch data processing scenario, sensitive data such as key performance parameters of the equipment enter the homomorphic encryption process immediately after collection. The encryption process is based on complex mathematical algorithms to convert the original data into ciphertext form, and the ciphertext can remain encrypted in subsequent operations. After encryption is completed, the data enters the operation stages such as noise reduction, fusion and outlier processing. Taking noise reduction as an example, the denoising autoencoder (DAE) based on deep learning will perform operations on the ciphertext data. Homomorphic encryption technology ensures that the DAE can perform noise removal operations on the encrypted data without touching the original plaintext, just like processing plaintext data, accurately identifying and filtering out noise while retaining key signal features. In the data fusion link, the adaptive weighted fusion algorithm also performs weight calculation and data integration on the encrypted data. Even for encrypted data of different sources and types, the algorithm dynamically assigns weights based on the reliability of each sensor data, achieving data fusion while maintaining data encryption throughout the entire process. For outlier processing, the density-based spatial clustering algorithm (DBSCAN) identifies data points in the ciphertext data space that deviate from normal data density areas and treats them as outliers for correction. The application of homomorphic encryption technology ensures the security of sensitive data throughout the entire intelligent data preprocessing process, effectively preventing data leakage and providing strong support for the reliable operation of GIS disconnect switches and data security management.

[0056] In this invention, the deep feature mining step introduces a graph neural network (GNN) to model the relationships between the various components of the disconnector. The structure of the GIS disconnector is visually presented in the form of a graph, with nodes representing its components, such as contacts, transmission rods, and operating motors. Each node carries data features collected by sensors on the corresponding component, covering various aspects such as vibration, temperature, and stress. Edges represent the connections between components, reflecting the physical connections and interactions between them. The GNN learns these node and edge features through a unique information propagation mechanism. In each iteration, a node receives information from its neighboring nodes and edges and updates it based on its own existing features. For example, a contact node not only considers its own characteristic data, such as temperature and wear, but also receives information transmitted through edges from connected components, such as the transmission rod, such as vibration transmission and force feedback. In this way, the GNN captures the complex interactions between components and uncovers deep features hidden in the data. After multiple rounds of information propagation and feature updating, the GNN is able to learn features that better reflect the overall operating status of the GIS disconnector. These features are no longer limited to the isolated information of a single component, but comprehensively consider the collaborative work and mutual influence between components, thereby greatly improving the comprehensiveness and accuracy of feature mining, providing a more valuable information basis for subsequent state parameter construction and state evaluation, and making the understanding and analysis of the operating status of GIS disconnectors more in-depth and accurate.

[0057] In the present invention, a visualization monitoring platform based on virtual reality (VR) / augmented reality (AR) is developed. VR technology is used to construct a virtual model of the GIS disconnector through high-precision modeling software. The model accurately restores the internal structure of the equipment, including the position, shape and connection relationship of key components such as contacts, transmission connecting rods, and operating motors. With the help of equipment operation data collected in real time by sensors, such as vibration frequency, temperature changes, stress distribution, etc., it is deeply bound to the virtual model to achieve real-time mapping of the equipment operation status. Operators can enter an immersive virtual environment by wearing a VR helmet, view the internal details of the equipment from any angle, and intuitively feel the operating status of each component, as if they were inside the equipment for inspection, effectively improving their comprehensive understanding of the overall status of the equipment. AR technology plays an important role in on-site monitoring. When the operation and maintenance personnel are at the GIS disconnector site, they turn on the AR mode through smart devices equipped with AR functions, such as tablets or smart glasses. Real-time device status information, including composite status parameter values, assessment status (normal / abnormal / fault), and early warning information such as fault type and warning level, is accurately displayed as a virtual image or text overlaid on the actual location of the device in the operator's field of view. For example, when a contact temperature warning is issued due to excessive temperature, the AR interface clearly displays the temperature value and warning icon near the contact location. This allows operators to quickly obtain key information and make timely decisions without complex operations, greatly improving operation and maintenance efficiency and reducing troubleshooting and resolution time.

[0058] The above are only preferred specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solutions and inventive concepts of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for measuring the operating state and characterizing state parameters of a GIS disconnector, characterized in that: The following steps are involved: Multi-source heterogeneous data acquisition: Vibration, temperature, displacement, and current sensors are deployed at key locations on GIS disconnectors. Strain gauges and fiber Bragg grating sensors are also added. Vibration sensors adaptively adjust sampling frequencies, strain gauges measure stress, and fiber Bragg grating sensors measure deformation. Acoustic cameras are used to capture sound distribution images, and spectrum analysis is combined to collect equipment operating status data. Intelligent data preprocessing: The noise reduction autoencoder (DAE) is used to denoise the vibration signal of the original data. An adaptive weighted fusion algorithm is used to dynamically assign weights to the data of each sensor. The spatial clustering algorithm DBSCAN is used to identify and process outliers. Deep feature mining: Mining pre-processed data features, using empirical wavelet transform (EWT) combined with Hilbert-Huang transform (HHT) to analyze vibration frequency and energy distribution, introducing a heat conduction model, calculating the temperature gradient change rate through a formula, extracting stress, deformation, sound spectrum, and sound intensity data features to characterize the operating status of the disconnector; Composite state parameter construction: Deep belief network (DBN) combined with principal component analysis (PCA) is used to construct comprehensive state parameters. DBN is used to mine feature relationships, PCA is used for dimensionality reduction, and the particle swarm optimization algorithm (PSO) is used to optimize the matrix. Based on the principal component contribution rate and feature importance, the analytic hierarchy process (AHP) is used to determine the weight of each principal component to construct the composite state parameters. Dynamic status assessment and early warning: Build an adaptive status assessment model based on transfer learning. Use convolutional neural networks (CNN) to pre-train with historical data, adopt transfer learning technology to transfer parameters and fine-tune the model, set thresholds, and use long short-term memory networks (LSTMs) to predict and warn when the thresholds are exceeded. Combine fault tree analysis and Bayesian networks to locate fault causes and fault locations.

2. The method for measuring the operating state and characterizing the state parameters of a GIS disconnector according to claim 1, characterized in that: Also includes: Real-time online self-learning steps: The monitoring process uses an incremental learning algorithm to update the DBN and CNN model parameters in real time based on newly collected data. When the equipment operating conditions or environment changes, the self-organizing map network (SOM) clustering analysis data is used to automatically adjust the model structure and parameters.

3. The method for measuring the operating state and characterizing the state parameters of a GIS disconnector according to claim 1, characterized in that: Also includes: Multimodal data fusion enhancement steps: Use the tensor fusion network TFN to fuse vibration, temperature, stress, deformation, and sound data. By constructing a tensor model, the multimodal fusion features are uniformly represented. The fusion process uses the mutual information maximization principle to optimize the fusion weight.

4. The method for measuring the operating state and characterizing the state parameters of a GIS disconnector according to claim 1, characterized in that: The deep feature mining step introduces a heat conduction model and uses finite element analysis to calculate the temperature gradient change rate to reflect the temperature change trend. The formula is: The generative adversarial network (GAN) variant Cycle-GAN is introduced to enhance features. Cycle-GAN performs unsupervised conversion between different types of features, generating diverse feature data to expand the feature library.

5. The method for measuring the operating state and characterizing the state parameters of a GIS disconnector according to claim 1, characterized in that: The composite state parameter construction step is to optimize the principal component transformation matrix by particle swarm optimization algorithm PSO during PCA dimensionality reduction. The calculation formula is: Y = XW, X is the feature matrix after DBN processing, W is the principal component transformation matrix after PSO optimization, and Y is the principal component matrix after dimensionality reduction. The weight of each principal component is determined by the hierarchical analysis method AHP to construct the composite state parameter. w i is the weight of the i-th principal component, y i is the value of the i-th principal component, m is the number of principal components; The quantum-behaved particle swarm optimization (QPSO) algorithm is combined with fuzzy C-means clustering (FCM) to optimize the principal component weights. The QPSO algorithm determines the initial value of the principal component weight, and the FCM clustering is used to analyze the principal components. The weights are adjusted according to the clustering results.

6. The method for measuring the operating state and characterizing the state parameters of a GIS disconnector according to claim 1, characterized in that: The dynamic status assessment and early warning steps use blockchain technology to record status assessment results and early warning information, and encrypt and store each status assessment time, composite status parameter value, assessment status and early warning details.

7. The method for measuring the operating state and characterizing the state parameters of a GIS disconnector according to claim 1, characterized in that: The multi-source heterogeneous data collection step uses wireless energy harvesting technology to power some sensors, and deploys electromagnetic induction energy harvesters near the operating mechanism of the disconnector to collect electromagnetic energy; deploys temperature difference energy harvesters at the contact position to generate electrical energy using the temperature difference between the contact and the surrounding environment, and stores the collected energy through the energy management circuit and powers the sensors.

8. The method for measuring the operating state and characterizing the state parameters of a GIS disconnector according to claim 1, characterized in that: The intelligent data preprocessing step uses an adaptive weighted fusion algorithm, and the calculation formula is: w i is the weight of the i-th sensor data, σ i is the standard deviation of the i-th sensor data, n is the number of sensors; Homomorphic encryption technology is used to encrypt and process sensitive data, and data noise reduction, fusion and outlier processing operations are completed in an encrypted state.

9. The method for measuring the operating state and characterizing the state parameters of a GIS disconnector according to claim 1, characterized in that: The deep feature mining step uses a graph neural network (GNN) to model the relationships between the components of the disconnector. The disconnector structure is represented in the form of a graph, where nodes represent components and edges represent the connection relationships between components. GNN is used to learn node and edge features and mine the overall operating status features.

10. The method for measuring the operating state and characterizing the state parameters of a GIS disconnector according to claim 1, characterized in that: The entire method develops a visual monitoring platform based on virtual reality (VR) and augmented reality (AR). VR technology is used to build a virtual model of the GIS disconnector, mapping the equipment operating status in real time. The internal structure of the equipment and the status of each component can be intuitively viewed through an immersive experience. AR technology is used to display the real-time status and warning information of the equipment in the form of augmented reality in the operator's field of view.

Images