GIS equipment multi-mode diagnosis method and device

Abstract

The invention relates to a GIS equipment multi-mode diagnosis method and device. The method comprises the following steps: acquiring acquired data of GIS equipment, and preprocessing the acquired data; performing time-space dimension multi-modal feature extraction and fusion on the preprocessed acquisition data to obtain a high-dimensional feature vector; obtaining a comprehensive health index based on the high-dimensional feature vector, and obtaining a degradation track prediction result of the comprehensive health index; and performing hierarchical diagnosis according to the comprehensive health index and the degradation track prediction result to obtain a diagnosis result of the GIS equipment. According to the invention, the diagnosis reliability of the GIS equipment can be improved.

Description

Technical Field

[0001] This application relates to the field of power equipment condition monitoring technology, and in particular to a multimodal diagnostic method and device for GIS equipment. Background Technology

[0002] Gas-insulated switchgear (GIS) equipment is a critical node in the power grid. Internal faults such as insulation deterioration and poor contact can be insidious and have serious consequences.

[0003] However, the reliability of traditional diagnostic methods for GIS equipment is insufficient. Summary of the Invention

[0004] Therefore, it is necessary to provide a multimodal diagnostic method and apparatus for GIS equipment that can improve the diagnostic reliability of GIS equipment, addressing the aforementioned technical problems.

[0005] In a first aspect, this application provides a multimodal diagnostic method for GIS equipment, the method comprising:

[0006] Acquire the data collected by the GIS equipment and preprocess the collected data;

[0007] Multimodal feature extraction and fusion in the spatiotemporal dimensions of the preprocessed collected data are performed to obtain high-dimensional feature vectors;

[0008] The comprehensive health index is obtained based on high-dimensional feature vectors, and the degradation trajectory prediction results of the comprehensive health index are obtained.

[0009] Based on the comprehensive health index and degradation trajectory prediction results, a graded diagnosis is performed to obtain the diagnostic results from GIS equipment.

[0010] In one embodiment, the method further includes:

[0011] Obtain the confidence interval of the comprehensive health index, and conduct a comprehensive confidence assessment of the diagnostic results based on the confidence interval.

[0012] In one embodiment, multimodal feature extraction and fusion of the collected data in the spatiotemporal dimensions are performed to obtain a high-dimensional feature vector, including:

[0013] Spatial features are extracted from the collected data that have spatial distribution characteristics to obtain spatial features;

[0014] Time series features are extracted from the slow variables in the collected data to obtain time series features;

[0015] Based on the attention mechanism, the correlation weights of spatial features and time series features are obtained, and data features are fused according to the correlation weights to obtain a high-dimensional feature vector.

[0016] In one embodiment, a comprehensive health index is obtained based on a high-dimensional feature vector, and the predicted degradation trajectory of the comprehensive health index is obtained, including:

[0017] Map high-dimensional feature vectors to a comprehensive health index;

[0018] Based on a hybrid prediction model, the evolution trajectory of the comprehensive health index within a preset time period is predicted in a rolling manner, and the prediction results of the degradation trajectory are obtained.

[0019] In one embodiment, the hybrid prediction model includes a long short-term memory network and a particle filter; wherein the long short-term memory network is used to learn the deterioration trend of the comprehensive health index, and the particle filter is used to correct the prediction results of the long short-term memory network.

[0020] In one embodiment, the preprocessing includes:

[0021] The non-stationary signals in the collected data are separated to separate interference from the actual fault characteristics. The physical quantities in the collected data are dynamically removed from the outliers, and the data sampling frequency is adjusted.

[0022] Secondly, this application provides a multimodal diagnostic device for GIS equipment, the device comprising:

[0023] The data acquisition module is used to acquire data collected by GIS equipment and to preprocess the collected data.

[0024] The spatiotemporal feature fusion module is used to extract and fuse multimodal features in the spatiotemporal dimensions of the preprocessed collected data to obtain high-dimensional feature vectors.

[0025] The dynamic degradation trajectory prediction module is used to obtain the comprehensive health index and the degradation trajectory prediction results of the comprehensive health index based on high-dimensional feature vectors.

[0026] The diagnostic module is used to perform graded diagnosis based on the comprehensive health index and degradation trajectory prediction results, and to obtain the diagnostic results of GIS equipment.

[0027] Thirdly, this application provides a GIS equipment diagnostic system, which includes a terminal layer, an edge computing layer, and a cloud computing platform layer; the terminal layer is communicatively connected to the edge computing layer, and the edge computing layer is communicatively connected to the cloud computing platform layer.

[0028] The terminal layer is used to acquire data collected by GIS devices and to preprocess the collected data.

[0029] The edge computing layer is used to extract and fuse multimodal features in the spatiotemporal dimensions of the preprocessed collected data to obtain high-dimensional feature vectors;

[0030] The cloud computing platform layer is used to obtain a comprehensive health index based on high-dimensional feature vectors and to obtain the degradation trajectory prediction results of the comprehensive health index. The cloud computing platform layer is also used to perform hierarchical diagnosis based on the comprehensive health index and degradation trajectory prediction results and to obtain the diagnostic results of GIS equipment.

[0031] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0032] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method.

[0033] The aforementioned multimodal diagnostic method and apparatus for GIS equipment acquires and preprocesses the collected data from the GIS equipment; extracts and fuses multimodal features in the spatiotemporal dimensions of the preprocessed data to obtain a high-dimensional feature vector; obtains a comprehensive health index based on the high-dimensional feature vector and acquires the degradation trajectory prediction result of the comprehensive health index; and performs hierarchical diagnosis based on the comprehensive health index and the degradation trajectory prediction result to obtain the diagnostic result of the GIS equipment, thereby improving the diagnostic reliability of the GIS equipment. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a flowchart illustrating a multimodal diagnostic method for GIS equipment in one embodiment;

[0036] Figure 2 This is a schematic diagram of the steps in a GIS equipment health status classification and diagnosis method based on multimodal information spatiotemporal fusion and dynamic degradation trajectory prediction in one embodiment;

[0037] Figure 3 This is a schematic diagram of a dual-stream spatiotemporal fusion network structure based on an attention mechanism in one embodiment;

[0038] Figure 4 This is a schematic diagram illustrating the principle of dynamic degradation trajectory prediction based on particle filtering and LSTM in one embodiment.

[0039] Figure 5 This is a schematic diagram of the structure of a multimodal diagnostic device for GIS equipment in one embodiment;

[0040] Figure 6 This is a logical architecture diagram of a GIS equipment diagnostic system in one embodiment;

[0041] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0043] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0044] In traditional technologies, GIS equipment is a critical node in the power grid, and its internal faults, such as insulation degradation and poor contact, are often insidious and have serious consequences. Existing diagnostic technologies mostly face challenges: 1) Shallow fusion: These technologies often rely on simple data or feature aggregation, failing to uncover deep correlations across time and space. For example, the spatial location information of partial discharge signals is not effectively correlated with the temporal synchronicity of changes in gas composition within specific gas chambers. 2) Static diagnosis: Models based on fixed thresholds or historical samples cannot characterize the continuous degradation process of equipment performance over time, making true early warning difficult. 3) Lack of confidence: Diagnostic results lack reliable confidence assessments. When sensor data contains noise or is partially missing, the reliability of diagnostic conclusions cannot be guaranteed, impacting operational and maintenance decisions. Therefore, there is an urgent need for an intelligent diagnostic solution that can deeply mine the inherent correlations of multimodal information, dynamically track equipment degradation paths, and provide quantified confidence levels.

[0045] Based on the aforementioned traditional technologies, embodiments of this application provide a multimodal diagnostic method, apparatus, and system for GIS equipment. Furthermore, it proposes a GIS equipment health status classification diagnostic method based on spatiotemporal fusion of multimodal information and dynamic degradation trajectory prediction. This method achieves accurate, proactive, and reliable classification diagnostics of GIS equipment health status through spatiotemporal feature fusion, dynamic degradation trajectory prediction, and comprehensive confidence assessment, thereby solving the problems of insufficient depth of multi-source information fusion, static and rigid diagnostic models, and unclear result confidence in traditional technologies. Embodiments of this application relate to the field of power equipment condition monitoring and intelligent operation and maintenance, and particularly to a method and system for classifying the health status of gas-insulated switchgear (GIS) based on the fusion of spatiotemporal features of multiple physical quantities and dynamic degradation trajectory prediction and confidence assessment.

[0046] It should be noted that the beneficial effects or technical problems solved by the embodiments of this application are not limited to this one, but may also be other implicit or related problems. For details, please refer to the description of the embodiments below.

[0047] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0048] In one exemplary embodiment, such as Figure 1 As shown, a multimodal diagnostic method for GIS equipment is provided, including the following steps 202 to 208. Wherein:

[0049] Step 202: Obtain the data collected by the GIS equipment and preprocess the collected data.

[0050] The data collected by the GIS equipment can be used to comprehensively characterize its operating status, covering four categories: electrical quantities, physical quantities, chemical quantities, and mechanical quantities. Electrical quantities may include partial discharge original pulse signals, insulation leakage current waveforms, etc.; physical quantities may include SF6 gas density, trace moisture content, multi-point temperature, etc.; chemical quantities may include SF6 decomposition products and their concentrations; mechanical quantities may include vibration acceleration, current, displacement or stroke signals during the operation of the operating mechanism, etc. This application is not limited to these.

[0051] For example, data collected by GIS equipment is acquired, and different types of preprocessing are performed on some of the data collected.

[0052] Step 204: Extract and fuse multimodal features in the spatiotemporal dimensions of the preprocessed collected data to obtain high-dimensional feature vectors.

[0053] For example, multimodal features of time and space dimensions are extracted from preprocessed collected data based on memory networks, and multimodal features of time and space dimensions are fused based on attention mechanisms, thereby ensuring accurate diagnosis of the health status of GIS equipment.

[0054] Step 206: Obtain the comprehensive health index based on the high-dimensional feature vector, and obtain the prediction results of the degradation trajectory of the comprehensive health index.

[0055] The comprehensive health index (HI) can also be simply referred to as the health index.

[0056] Optionally, a health index can be constructed based on a high-dimensional feature vector to obtain a comprehensive health index. Furthermore, the degradation trajectory prediction result, i.e. the evolution trajectory of HI, can be obtained based on the degradation trend pattern of historical HI data. The confidence interval corresponding to the evolution trajectory of HI can also be obtained simultaneously, thereby achieving advanced diagnosis of the health status of GIS equipment.

[0057] Step 208: Perform a graded diagnosis based on the comprehensive health index and degradation trajectory prediction results, and obtain the diagnostic results from the GIS equipment.

[0058] Specifically, the health status of GIS equipment is classified and diagnosed based on the comprehensive health index and degradation trajectory prediction results, thereby obtaining the detection results of the classified health status of GIS equipment incorporating prediction information, and thus realizing reliable classified diagnosis of GIS equipment.

[0059] In the aforementioned multimodal diagnostic method for GIS equipment, the collected data from the GIS equipment is acquired and preprocessed. Multimodal features in the spatiotemporal dimensions of the preprocessed data are extracted and fused to obtain high-dimensional feature vectors. A comprehensive health index is obtained based on these high-dimensional feature vectors, and the degradation trajectory prediction results of the comprehensive health index are also obtained. A graded diagnosis is performed based on the comprehensive health index and the degradation trajectory prediction results to obtain the diagnostic results for the GIS equipment. Thus, through spatiotemporal feature fusion and dynamic degradation trajectory prediction, accurate, proactive, and reliable graded diagnosis of the health status of GIS equipment is achieved, thereby improving the diagnostic reliability of GIS equipment.

[0060] In one exemplary embodiment, the method further includes: obtaining a confidence interval for a comprehensive health index, and assessing the comprehensive confidence level of the diagnostic results based on the confidence interval.

[0061] Specifically, by obtaining the confidence interval of the comprehensive health index, the confidence of the diagnostic results is comprehensively assessed based on the confidence interval, thereby quantifying the confidence of the diagnostic results and further improving the reliability of the diagnostic results.

[0062] For example, the confidence intervals of the evolution trajectory of HI are obtained, and a comprehensive confidence assessment is performed: a comprehensive confidence score C is calculated for each diagnostic conclusion. C = w1 * C data +w2*C model +w3*C pred .in:

[0063] C data : Data confidence based on the current data quality and consistency of each sensor;

[0064] C model Model confidence based on the failure mode probabilities output by the fusion model;

[0065] C pred : Prediction confidence based on the width of the confidence interval for predicted trajectory using particle filtering.

[0066] The weights w1, w2, and w3 can be adjusted according to the actual situation. High-confidence diagnostic results will directly trigger automated operation and maintenance suggestions, while low-confidence diagnostic results will prompt manual intervention for review.

[0067] Optionally, refer to Figure 2 The complete execution steps of the GIS equipment health status hierarchical diagnosis method based on multimodal information spatiotemporal fusion and dynamic degradation trajectory prediction can include: multimodal heterogeneous data acquisition and adaptive preprocessing; spatiotemporal multimodal feature extraction and fusion; dynamic degradation trajectory prediction based on particle filtering and LSTM; hierarchical health status diagnosis and confidence assessment incorporating prediction information; and visualization output of diagnosis results and decision support.

[0068] In an exemplary embodiment, the steps of tiered health status diagnosis and confidence assessment incorporating predictive information may include tiered diagnosis and comprehensive confidence assessment, wherein the specific steps of tiered diagnosis are as follows:

[0069] Health status is divided into four levels:

[0070] Healthy (HS1): HI > 0.9, and the predicted trajectory is stable.

[0071] Note (HS2): 0.8 < HI ≤ 0.9, or the predicted trajectory shows a slight downward trend. Initiate refined tracking.

[0072] Warning (HS3): If 0.6 < HI ≤ 0.8, and the predicted trajectory will continue to decline to near the warning threshold within the next 72 hours, it is determined to be in the potential fault development phase, and the most likely fault mode and probability are output.

[0073] Alarm (HS4): HI ≤ 0.6, or the predicted trajectory shows that HI will drop sharply. This indicates a critical fault or that a fault has already occurred. The above-mentioned comprehensive confidence assessment procedure is then performed.

[0074] In an exemplary embodiment, step 204 involves extracting and fusing multimodal features in the spatiotemporal dimensions of the collected data to obtain a high-dimensional feature vector, including:

[0075] Spatial features are extracted from the collected data that have spatial distribution characteristics. Time series features are extracted from the collected data. Based on the attention mechanism, the correlation weights of spatial features and time series features are obtained. Data features are fused according to the correlation weights to obtain a high-dimensional feature vector.

[0076] For example, the steps of spatiotemporal multimodal feature extraction and fusion (spatiotemporal feature fusion module) include temporal and spatial feature extraction and spatiotemporal feature fusion, wherein:

[0077] Spatial feature extraction includes: for data with spatial distribution characteristics, such as multiple ultrasonic / ultra-high frequency partial discharge sensor arrays, an improved convolutional neural network (CNN) is used to extract the spatial propagation characteristics of the discharge signal and the coarse location information of the discharge source.

[0078] Time series feature extraction includes: for slow variables such as gas concentration and temperature, extracting time-domain statistical features such as mean, variance, and trend within a sliding window, and using Long Short-Term Memory (LSTM) networks to capture their long-term dependencies.

[0079] Spatiotemporal feature fusion includes: using a two-stream spatiotemporal fusion network based on an attention mechanism, the network structure of which is as follows: Figure 3 As shown: The input layer obtains spatial and temporal feature inputs. One network processes spatial features, such as partial discharge localization, while the other processes temporal features, such as gas concentration changes. The spatial and temporal feature extraction networks in the feature extraction layer perform the aforementioned feature extraction respectively. Through an attention fusion layer with a built-in attention gate, the correlation weights of spatial and temporal features at each time step are dynamically calculated, achieving deep fusion of data features and outputting a weighted, high-dimensional feature vector. For example, when the temporal feature stream detects a sharp increase in H2S concentration, the attention gate assigns higher weights to the partial discharge features related to the spatial location of the gas chamber, thereby achieving a precise association of "a specific physicochemical change occurring at a specific location and time."

[0080] In an exemplary embodiment, step 206, obtaining a comprehensive health index based on a high-dimensional feature vector and obtaining a prediction result of the degradation trajectory of the comprehensive health index, includes: mapping the high-dimensional feature vector to a comprehensive health index; and based on a hybrid prediction model, rollingly predicting the evolution trajectory of the comprehensive health index within a preset time period and obtaining a prediction result of the degradation trajectory.

[0081] In one exemplary embodiment, the hybrid prediction model includes a long short-term memory network and a particle filter; wherein the long short-term memory network is used to learn the deterioration trend of the comprehensive health index, and the particle filter is used to correct the prediction results of the long short-term memory network.

[0082] For example, the steps of dynamic degradation trajectory prediction based on particle filtering and LSTM include health index construction and degradation trajectory prediction, the principle of which is as follows: Figure 4 As shown:

[0083] The construction of the health index (HI) includes mapping the fused high-dimensional feature vector to a one-dimensional comprehensive health index (HI), which takes values ​​in the range of [0, 1], where 1 represents a completely healthy state and 0 represents a complete failure.

[0084] The degradation trajectory prediction includes, optionally, employing a hybrid prediction model combining particle filtering and LSTM (Long Short-Term Memory) to eliminate static threshold judgments. LSTM learns the degradation trend patterns of historical HI data, while particle filtering is used to correct the LSTM prediction results in real time and quantify the uncertainty of the prediction. This model can predict the evolution trajectory of HI and its confidence interval over a future period (e.g., 72 hours).

[0085] In one exemplary embodiment, the preprocessing includes:

[0086] The non-stationary signals in the collected data are separated to separate interference from the actual fault characteristics. The physical quantities in the collected data are dynamically removed from the outliers, and the data sampling frequency is adjusted.

[0087] For example, the steps of multimodal heterogeneous data acquisition and adaptive preprocessing include data acquisition and adaptive preprocessing. The adaptive preprocessing includes: using an adaptive denoising algorithm based on Variational Mode Decomposition (VMD) to process non-stationary signals such as partial discharge and vibration, effectively separating interference from actual fault characteristics. For data such as gas and temperature, a dynamic outlier removal method based on the Grubbs criterion is used, and the data sampling frequency is dynamically adjusted according to the equipment load.

[0088] In one exemplary embodiment, diagnostic results and decision support are visualized. The system visualizes the diagnostic results, including the device's current health level, key anomalies, failure type probability, risk level assessment, and maintenance recommendations.

[0089] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0090] Based on the same inventive concept, this application also provides a GIS equipment multimodal diagnostic device for implementing the aforementioned GIS equipment multimodal diagnostic method. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more GIS equipment multimodal diagnostic device embodiments provided below can be found in the limitations of the GIS equipment multimodal diagnostic method described above, and will not be repeated here.

[0091] In one exemplary embodiment, such as Figure 5 As shown, a multimodal diagnostic device 900 for GIS equipment is provided, comprising: a data acquisition module 901, a spatiotemporal feature fusion module 902, a dynamic degradation trajectory prediction module 903, and a diagnostic module 904, wherein:

[0092] The acquisition module 901 is used to acquire data collected by GIS equipment and to preprocess the acquired data;

[0093] The spatiotemporal feature fusion module 902 is used to extract and fuse multimodal features in the spatiotemporal dimensions of the preprocessed collected data to obtain high-dimensional feature vectors.

[0094] The dynamic degradation trajectory prediction module 903 is used to obtain the comprehensive health index and the degradation trajectory prediction results of the comprehensive health index based on the high-dimensional feature vector;

[0095] The diagnostic module 904 is used to perform graded diagnosis based on the comprehensive health index and degradation trajectory prediction results, and to obtain the diagnostic results of the GIS equipment.

[0096] In one embodiment, the GIS equipment multimodal diagnostic device 900 further includes a comprehensive confidence assessment module for obtaining the confidence interval of the comprehensive health index and performing a comprehensive confidence assessment of the diagnostic results based on the confidence interval.

[0097] In one embodiment, the spatiotemporal feature fusion module 902 is further configured to: extract spatial features from data with spatial distribution characteristics in the collected data to obtain spatial features; extract time series features from slow variables in the collected data to obtain time series features; obtain the correlation weights of spatial features and time series features based on an attention mechanism; and fuse data features according to the correlation weights to obtain a high-dimensional feature vector.

[0098] In one embodiment, the dynamic degradation trajectory prediction module 903 is further configured to: map the high-dimensional feature vector to a comprehensive health index; and based on a hybrid prediction model, predict the evolution trajectory of the comprehensive health index over a future preset time period to obtain the degradation trajectory prediction result.

[0099] In one embodiment, the hybrid prediction model includes a long short-term memory network and a particle filter; wherein the long short-term memory network is used to learn the deterioration trend of the comprehensive health index, and the particle filter is used to correct the prediction results of the long short-term memory network.

[0100] In one embodiment, the acquisition module 901 is further configured to: separate interference and real fault characteristics from non-stationary signals in the acquired data, dynamically remove outliers from physical quantities in the acquired data, and adjust the data sampling frequency.

[0101] Each module in the aforementioned GIS equipment multimodal diagnostic device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0102] Based on the same inventive concept, embodiments of this application also provide a GIS equipment diagnostic system for implementing the aforementioned multimodal diagnostic method for GIS equipment. In an exemplary embodiment, such as Figure 6 As shown, a GIS equipment diagnostic system is provided, including: a terminal layer, an edge computing layer, and a cloud computing platform layer; the terminal layer is communicatively connected to the edge computing layer, and the edge computing layer is communicatively connected to the cloud computing platform layer;

[0103] The terminal layer is used to acquire data collected by GIS devices and to preprocess the collected data.

[0104] The edge computing layer is used to extract and fuse multimodal features in the spatiotemporal dimensions of the preprocessed collected data to obtain high-dimensional feature vectors;

[0105] The cloud computing platform layer is used to obtain a comprehensive health index based on high-dimensional feature vectors and to obtain the degradation trajectory prediction results of the comprehensive health index. The cloud computing platform layer is also used to perform hierarchical diagnosis based on the comprehensive health index and degradation trajectory prediction results and to obtain the diagnostic results of GIS equipment.

[0106] For example, this invention proposes an intelligent diagnostic system that implements the above method. This system consists of three logical layers: a terminal layer (endpoint), an edge computing layer (edge), and a cloud computing platform layer (cloud). Data flows from bottom to top, undergoing collection, preprocessing, fusion, and deep analysis; while diagnostic models, strategies, and instructions are distributed from top to bottom, guiding the behavior of the edge and the terminal. Its architecture is as follows: Figure 6 As shown.

[0107] The terminal layer (end) is deployed on or near the GIS equipment and consists of various high-performance intelligent sensors and acquisition units, possessing multimodal data acquisition, edge preprocessing, and local feature extraction capabilities. The multimodal data acquisition module can simultaneously acquire multiple types of raw monitoring data, such as partial discharge pulse waveforms, leakage current, temperature, micro-water content, SF6 decomposition products, and circuit breaker mechanical action signals.

[0108] Edge preprocessing enables each sensor or acquisition unit to possess preliminary edge computing capabilities. For example, the partial discharge acquisition unit has a built-in variational mode decomposition algorithm that can adaptively denoise the raw signal locally, extracting effective discharge pulses and significantly reducing the amount of data uploaded. The gas monitoring unit has a built-in anomaly detection algorithm that can identify and remove instantaneous abnormal data points caused by sudden environmental changes. Local feature extraction can extract primary features of the data, such as the discharge quantity, discharge frequency, and phase of the partial discharge signal; and the mean and gradient of temperature data, forming a concise feature vector for uploading, rather than the original massive amount of data.

[0109] The edge computing layer (edge) is deployed in edge computing gateways or servers, and has a built-in spatiotemporal feature fusion module and lightweight diagnostic model, responsible for analysis and control with high real-time requirements. Its core functions mainly include data aggregation and alignment, real-time spatiotemporal feature fusion, local real-time diagnosis and early warning, and data compression and uploading.

[0110] The data collection and alignment process includes receiving feature data uploaded by all terminals from the same GIS interval or multiple air chambers, and performing spatiotemporal alignment based on high-precision clock synchronization technology to ensure that data of different physical quantities are comparable in time and space.

[0111] The real-time spatiotemporal feature fusion edge layer incorporates a lightweight "attention-based dual-stream spatiotemporal fusion network." Spatial stream: Processes data from sensor arrays (such as multiple UHF sensors), quickly performing coarse spatial localization of the discharge source using a lightweight CNN. Temporal stream: Processes time-series features of slowly varying data such as gas and temperature. Attention fusion gate: Dynamically calculates the weights of spatial and temporal features to achieve deep fusion, outputting a high-dimensional fused feature vector in real time that comprehensively reflects the current device state.

[0112] Local real-time diagnostics and early warning: A simplified health status (HS) rapid diagnostic model runs at the edge layer. This model compares the FFV (Front-End Value) with preset thresholds, enabling it to determine whether a device is in an "alarm (HS4)" or "early warning (HS3)" state within seconds or minutes. In the event of an emergency, local alarms or linkage control signals can be triggered directly without going through the cloud, ensuring rapid fault response.

[0113] Data compression and upload: Package and compress the complete FFV, diagnostic intermediate results and raw data samples, and upload them to the cloud platform periodically or according to event triggers.

[0114] The cloud computing platform layer (cloud) is deployed in the group or provincial-level data center, with built-in dynamic degradation trajectory prediction and comprehensive confidence assessment modules. It has powerful storage and computing capabilities, aggregating data from multiple substations across the network for in-depth big data mining, model training and updates, and long-term trend prediction. Its core functions mainly include big data storage and management, dynamic degradation trajectory prediction, comprehensive confidence assessment and accurate diagnosis, model training and optimization, and panoramic visualization and decision support.

[0115] Among them, big data storage and management can store massive amounts of historical FFV, raw data, diagnostic records, etc., forming a database covering the entire lifecycle of the equipment.

[0116] The dynamic degradation trajectory prediction function is based on a core "particle filtering and LSTM hybrid prediction model" running on the cloud platform. For each GIS device, the model calculates its historical health index (HI) trajectory based on its historical FFV sequence. LSTM is used to learn the device's personalized degradation patterns. Particle filtering is used to make rolling predictions of future HI values ​​and provide confidence intervals for the prediction results. This is key to achieving the leap from "condition-based maintenance" to "predictive maintenance".

[0117] Comprehensive confidence assessment and accurate diagnosis: After receiving the data uploaded by the edge layer, the cloud platform combines its own prediction results to conduct a more accurate and comprehensive diagnosis and calculate a comprehensive confidence score C. This score takes into account data quality, model output probability and prediction uncertainty, so that the diagnostic conclusion is no longer a "black box" but decision support information with reliability assessment.

[0118] Model Training and Optimization: The cloud platform utilizes massive amounts of data from across the internet to regularly retrain and optimize the edge layer fusion model, diagnostic model, and its own predictive model. When new fault modes are discovered and confirmed, the system can automatically update the knowledge base, enabling continuous self-evolution of the model and constant improvement in diagnostic capabilities over time.

[0119] Panoramic visualization and decision support: Provides maintenance personnel with a unified web interface to display a panoramic view of the health status of all GIS devices in the network, hierarchical early warning information, predictive maintenance suggestions, and confidence scores.

[0120] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores diagnostic data for GIS equipment. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a multimodal diagnostic method for GIS equipment.

[0121] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0122] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0123] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0124] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method described above.

[0125] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0126] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0127] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A GIS equipment multi-modal diagnostic method, characterized in that, The method comprises: acquiring acquisition data of the GIS device, and preprocessing the acquisition data; extracting and fusing multi-modal features of the preprocessed acquisition data in space-time dimensions to obtain a high-dimensional feature vector; obtaining a comprehensive health index based on the high-dimensional feature vector, and obtaining a degradation trajectory prediction result of the comprehensive health index; performing hierarchical diagnosis according to the comprehensive health index and the degradation trajectory prediction result to obtain a diagnosis result of the GIS device.

2. The method of claim 1, wherein, The method further comprises: obtaining a confidence interval of the comprehensive health index, and comprehensively evaluating the confidence of the diagnosis result according to the confidence interval.

3. The method of claim 1, wherein, The extraction and fusion of multi-modal features of the acquisition data in space-time dimensions to obtain a high-dimensional feature vector comprises: extracting spatial features from data with spatial distribution characteristics in the acquisition data to obtain spatial features; extracting time series features from slowly varying variables in the acquisition data to obtain time series features; obtaining correlation weights of the spatial features and the time series features based on an attention mechanism, fusing data features according to the correlation weights, and obtaining the high-dimensional feature vector.

4. The method of claim 1, wherein, The obtaining of a comprehensive health index based on the high-dimensional feature vector, and the obtaining of a degradation trajectory prediction result of the comprehensive health index, comprises: mapping the high-dimensional feature vector to the comprehensive health index; rolling forward to predict the evolution trajectory of the comprehensive health index in a future preset time period based on a hybrid prediction model to obtain the degradation trajectory prediction result.

5. The method of claim 4, wherein, The hybrid prediction model comprises a long short-term memory network and a particle filter; wherein the long short-term memory network is used to learn the degradation trend of the comprehensive health index, and the particle filter is used to correct the prediction result of the long short-term memory network.

6. The method according to any one of claims 1 to 5, characterized in that, The preprocessing comprises: separating interference and true fault features from non-stationary signals in the acquisition data, removing dynamic abnormal points from physical quantities in the acquisition data, and adjusting the data sampling frequency.

7. A GIS device multi-modal diagnostic apparatus, characterized by, The device comprises: an acquisition module configured to acquire acquisition data of the GIS device, and preprocess the acquisition data; a space-time feature fusion module configured to extract and fuse multi-modal features of the preprocessed acquisition data in space-time dimensions to obtain a high-dimensional feature vector; a dynamic degradation trajectory prediction module configured to obtain a comprehensive health index based on the high-dimensional feature vector, and obtain a degradation trajectory prediction result of the comprehensive health index; a diagnosis module configured to perform hierarchical diagnosis according to the comprehensive health index and the degradation trajectory prediction result to obtain a diagnosis result of the GIS device.

8. A GIS equipment diagnostic system characterized by, The system comprises a terminal layer, an edge computing layer, and a cloud computing platform layer; the terminal layer is in communication connection with the edge computing layer, and the edge computing layer is in communication connection with the cloud computing platform layer; The terminal layer is configured to acquire acquisition data of the GIS device, and preprocess the acquisition data; The edge computing layer is configured to extract and fuse multi-modal features of the preprocessed acquisition data in space-time dimensions to obtain a high-dimensional feature vector; The cloud computing platform layer is configured to obtain a comprehensive health index based on the high-dimensional feature vector, and obtain a degradation trajectory prediction result of the comprehensive health index; and the cloud computing platform layer is further configured to perform hierarchical diagnosis according to the comprehensive health index and the degradation trajectory prediction result, and obtain a diagnosis result of the GIS device.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by a processor, implements the steps of the method of any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that, The computer program, when executed by a processor, implements the steps of the method of any one of claims 1 to 6.

Images