Operation and maintenance debugging monitoring system suitable for electric field

Abstract

The invention discloses an operation and maintenance debugging monitoring system suitable for an electric field, and relates to the technical field of operation and maintenance debugging monitoring, and the system comprises a collection unit which collects the operation data of an electric field device at a multi-dimensional monitoring point, the operation data comprises an electrical parameter, a mechanical vibration parameter, an environment parameter and chemical gas component data, and a multi-modal monitoring data set is constructed; the feature analysis module is used for carrying out spatial-temporal feature extraction on the multi-modal monitoring data set, generating a multi-dimensional feature tensor containing a time domain feature, a frequency domain feature and a spatial distribution feature, obtaining a spatial-temporal coupling relationship among the features through a tensor decomposition technology, and sending the spatial-temporal coupling relationship to the multi-modal monitoring data set; according to the method, a whole-process intelligent solution from monitoring, early warning to decision making is provided for electric field operation and maintenance, the equipment fault risk is greatly reduced, the power failure time is shortened, and the safety and economical efficiency of power grid operation are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of operation, maintenance, debugging and monitoring, and in particular relates to an operation, maintenance, debugging and monitoring system suitable for an electric field. Background Art

[0002] Against the backdrop of intelligent power system development, the operating conditions of electric field equipment are becoming increasingly complex, and the uncertainty of the operating environment has increased significantly. Existing electric field operation, maintenance, commissioning, and monitoring technologies are limited in their ability to integrate and analyze multi-source heterogeneous data, making it impossible to deeply explore potential correlations between data. Consequently, it is difficult to accurately deconstruct the spatiotemporal coupling relationships between equipment operating states, resulting in significant errors in equipment health assessments and a lack of timeliness and accuracy in fault risk predictions. Furthermore, traditional monitoring technologies cannot effectively integrate equipment topology for systematic analysis, making it difficult to formulate scientifically sound operation and maintenance decisions, and failing to meet the actual needs for safe and stable operation of electric field equipment. Summary of the Invention

[0003] The purpose of the present invention is to provide an operation, maintenance, debugging and monitoring system suitable for electric fields, which solves the technical problems that the existing technology cannot effectively combine the equipment topology structure for systematic analysis, is difficult to make scientific and reasonable operation and maintenance decisions, and cannot meet the actual needs of safe and stable operation of electric field equipment.

[0004] Applicable to the operation, maintenance, commissioning and monitoring system of the electric field, including:

[0005] An acquisition unit collects operating data of the electric field equipment at multi-dimensional monitoring points, wherein the operating data includes electrical parameters, mechanical vibration parameters, environmental parameters, and chemical gas composition data, and constructs a multimodal monitoring data set;

[0006] A feature analysis module extracts spatiotemporal features from the multimodal monitoring dataset, generates a multidimensional feature tensor containing time domain features, frequency domain features, and spatial distribution features, and obtains the spatiotemporal coupling relationship between features through tensor decomposition technology;

[0007] a calculation module, which constructs a device health state space based on the spatiotemporal coupling relationship, calculates the density distribution of historical failure cases in the state space, and determines a fault warning threshold surface of key parameters;

[0008] The risk assessment module maps the real-time monitoring data to the equipment health status space, measures the deviation between the real-time status point and the warning threshold surface using Mahalanobis distance, and generates a fault risk heat map;

[0009] The decision module combines the fault risk heat map with the equipment topology, uses a graph neural network to predict the fault propagation path, and generates an operation and maintenance decision matrix that includes inspection priorities and maintenance windows;

[0010] The verification unit verifies the validity of the operation and maintenance decision matrix through digital twin technology, and dynamically adjusts the parameter weights of the spatiotemporal feature extraction algorithm according to the verification results.

[0011] As a further solution of the present invention, the spatiotemporal feature extraction of the multimodal monitoring dataset is specifically performed in the following manner:

[0012] Build a hybrid neural network model consisting of a two-dimensional convolutional neural network layer, a long short-term memory network layer, and a fully connected layer;

[0013] The two-dimensional convolutional neural network layer is provided with multiple convolution layers and pooling layers with different convolution kernel sizes to extract data spatial features;

[0014] Use the gating mechanism of the long short-term memory network layer to process the time series characteristics of the data;

[0015] Finally, the spatiotemporal features are fused through the fully connected layer to output a multidimensional feature tensor containing time domain features, frequency domain features and spatial distribution features.

[0016] As a further solution of the present invention: the device health state space is constructed based on the spatiotemporal coupling relationship, specifically:

[0017] Use Riemannian manifolds to construct high-dimensional state space;

[0018] First, whiten the multidimensional feature tensor to eliminate the correlation between features;

[0019] Then the processed data is mapped to the Riemannian manifold space through exponential mapping;

[0020] In the Riemannian manifold space, the kernel density estimation method is used to calculate the probability density distribution of historical fault cases, and then the fault warning threshold surface of key parameters is determined.

[0021] As a further solution of the present invention: combining the fault risk heat map with the equipment topology structure and using a graph neural network to predict the fault propagation path, specifically:

[0022] Build a graph neural network consisting of graph convolutional layer, graph attention layer and output layer;

[0023] Use graph convolutional layers to aggregate neighborhood information of device nodes to update node features;

[0024] The attention coefficient between nodes is calculated through the graph attention layer to enhance the ability to identify key nodes in the device topology;

[0025] Finally, the output layer predicts the fault propagation path in the device topology based on the updated node features and outputs the path probability distribution.

[0026] As a further solution of the present invention: the deviation between the real-time state point and the warning threshold surface is measured by Mahalanobis distance to generate a fault risk heat map, specifically:

[0027] First, the real-time monitoring data is standardized to have zero mean and unit variance;

[0028] The local weighted regression method is used to perform local fitting on the warning threshold surface;

[0029] Calculate the Mahalanobis distance from the real-time state point to the fitted surface, and smooth the Mahalanobis distance calculation result by combining Gaussian kernel density estimation;

[0030] Using the geographical location of the equipment as the coordinate reference, a fault risk heat map is ultimately generated.

[0031] As a further solution of the present invention: the validity of the operation and maintenance decision matrix is ​​verified by digital twin technology, specifically:

[0032] Build a digital twin that includes equipment geometry, material properties, operational data, and environmental impacts;

[0033] Use 3D modeling software to build a geometric model of the equipment and set parameters according to material characteristics to form a material property model;

[0034] Synchronize equipment operation data to establish an operation data model, and simulate environmental factors to build an environmental impact model;

[0035] The operation and maintenance decision matrix is ​​simulated and executed in the digital twin, and the results including at least the operating status and failure probability are collected.

[0036] As a further solution of the present invention: the electrical parameters include at least three of partial discharge, insulation resistance, dielectric loss factor, leakage current, dynamic capacitance and capacitance change rate, ultra-high frequency discharge signal and pulse current signal; the mechanical vibration parameters include acceleration amplitude, vibration frequency distribution, and vibration phase characteristics; the environmental parameters include temperature, humidity, air pressure, and wind speed; and the chemical gas component data includes the concentration of SF6 decomposition products and the dissolved gas content in transformer oil.

[0037] As a further solution of the present invention: when the processed data is mapped to the Riemannian manifold space by exponential mapping, the method further includes:

[0038] Calculate the geodesic distance between the current data point and the origin of the Riemannian manifold space;

[0039] The scale parameter of the exponential mapping is dynamically adjusted according to the geodetic distance. When the distance is greater than a preset threshold, a contraction-type exponential mapping is used to suppress the influence of outliers.

[0040] When the distance is less than the preset threshold, the dilated exponential mapping is used to enhance the discrimination of normal state data;

[0041] The geometric relationship between data points is kept unchanged through parallel movement operations on the manifold.

[0042] As a further solution of the present invention: when determining the fault warning threshold surface of the key parameter, a topological data analysis method is used to extract the fault feature topological structure in the Riemann manifold space, specifically:

[0043] Construct the Vietoris-Rips complex of historical fault cases to capture the characteristics of holes and connected branches in the manifold space;

[0044] Compute persistent homology groups at different scales and identify fault signatures with topological persistence;

[0045] Correlate and map the persistent coherence characteristics with the physical failure modes of the equipment to determine the topological critical points of key parameters;

[0046] Generate geometrically invariant threshold surfaces in Riemannian manifold space via topological critical points.

[0047] As a further solution of the present invention: when the digital twin technology verifies the validity of the operation and maintenance decision matrix, it also includes:

[0048] Constructing a digital twin model consistent with the physical electric field device parameters, wherein the digital twin model includes a device electrical parameter model, a mechanical structure model, and an environmental impact model;

[0049] Input the debugging strategy parameters in the operation and maintenance decision matrix into the digital twin model to simulate and generate a predicted status dataset;

[0050] Calculate the geodesic distance between the predicted state dataset and the measured data of the physical device in the high-dimensional state space as a quantitative indicator of decision effectiveness;

[0051] When the geodetic distance exceeds the preset threshold, the density distribution of historical fault cases in the high-dimensional state space is recalculated, and the fault warning threshold surface of key parameters is adjusted;

[0052] By comparing the topological feature changes in the predicted status dataset with the topological features of historical fault cases, a verification report containing the fault pattern matching degree is generated, and the parameter weights of the spatiotemporal feature extraction algorithm are adjusted based on the verification report.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] This invention systematically addresses the core pain points of traditional electric field monitoring systems through the technical approach of "multimodal data fusion - spatiotemporal feature mining - dynamic risk assessment - intelligent decision-making optimization." The combination of multimodal data acquisition and tensor decomposition technology significantly enhances early fault warning capabilities. A fault propagation prediction model based on a graph neural network improves the efficiency of inspection route planning and reduces unnecessary manual inspections. A digital twin verification mechanism ensures the reliability of operation and maintenance decisions and increases the accuracy of maintenance window predictions. This provides an intelligent solution for the entire process of electric field operation and maintenance, from monitoring and early warning to decision-making, significantly reducing the risk of equipment failure, shortening power outages, and improving the safety and economy of grid operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a schematic diagram of the system framework structure of the present invention. DETAILED DESCRIPTION

[0056] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all 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.

[0057] In the field of electric field operation, maintenance, commissioning and monitoring, traditional systems face problems such as difficulty in integrating multi-dimensional data, delayed fault warning, and lack of global optimization of operation and maintenance decisions. On the one hand, the operating status of electric field equipment (such as transformers, circuit breakers, etc.) is affected by multi-dimensional factors such as electrical parameters, mechanical vibrations, environmental factors and chemical gas components, but traditional monitoring systems can only analyze a single type of data in isolation and have difficulty capturing the spatiotemporal coupling relationship between different parameters. For example, in the early stages of partial discharge of a transformer, the electrical parameters change slightly, but the chemical gas components (such as hydrogen and acetylene concentrations) may have increased abnormally. Traditional systems are unable to correlate these cross-dimensional information, resulting in the failure of early fault warning. On the other hand, when a device fails, traditional systems are unable to accurately predict the fault propagation path, which often leads to blind inspections by operation and maintenance personnel, missing the optimal maintenance window, and increasing power outage time and maintenance costs.

[0058] See also Figure 1 , this application provides an operation, maintenance, debugging and monitoring system applicable to electric fields, including:

[0059] The acquisition unit collects the operating data of the electric field equipment at multi-dimensional monitoring points. The operating data includes electrical parameters, mechanical vibration parameters, environmental parameters and chemical gas composition data to construct a multimodal monitoring data set;

[0060] The feature analysis module extracts spatiotemporal features from multimodal monitoring datasets, generates a multidimensional feature tensor containing time domain features, frequency domain features, and spatial distribution features, and obtains the spatiotemporal coupling relationship between features through tensor decomposition technology;

[0061] The acquisition unit is in communication with the feature analysis module and is used to transmit the multimodal monitoring data set to the feature analysis module.

[0062] The calculation module constructs the equipment health state space based on the spatiotemporal coupling relationship, calculates the density distribution of historical failure cases in the state space, and determines the fault warning threshold surface of key parameters;

[0063] The risk assessment module maps real-time monitoring data to the equipment health status space, measures the deviation between the real-time status point and the warning threshold surface using Mahalanobis distance, and generates a fault risk heat map.

[0064] The decision-making module combines the fault risk heat map with the equipment topology, uses a graph neural network to predict the fault propagation path, and generates an operation and maintenance decision matrix that includes inspection priorities and maintenance windows;

[0065] The verification unit verifies the effectiveness of the operation and maintenance decision matrix through digital twin technology, and dynamically adjusts the parameter weights of the spatiotemporal feature extraction algorithm based on the verification results.

[0066] As an optional embodiment, the electrical parameters include at least three of partial discharge amount, insulation resistance, dielectric loss factor, leakage current, dynamic capacitance and capacitance change rate, ultra-high frequency discharge signal and pulse current signal; the mechanical vibration parameters include acceleration amplitude, vibration frequency distribution, vibration phase characteristics; the environmental parameters include temperature, humidity, air pressure, wind speed; and the chemical gas composition data include SF6 decomposition product concentration and dissolved gas content in transformer oil.

[0067] Specifically, a multimodal monitoring data set is constructed through the acquisition unit, which comprehensively covers the electrical parameters of the electric field equipment (such as voltage and current), mechanical vibration parameters (such as vibration frequency and amplitude), environmental parameters (such as temperature and humidity), and chemical gas component data (such as the concentration of SF6 decomposition products), etc.

[0068] The feature analysis module extracts spatiotemporal features from the dataset, generating a multidimensional feature tensor. Combined with the spatial layout of the equipment, it analyzes the spatial distribution characteristics of data from different monitoring points. Using tensor decomposition techniques (such as CANDECOMP / PARAFAC decomposition), it reveals the spatiotemporal coupling relationship between electrical parameter fluctuations, mechanical vibration, and ambient temperature changes. For example, it was found that rising temperature can exacerbate mechanical vibrations at certain frequencies, thereby affecting electrical connection stability.

[0069] Among them, the equipment health status space uses historical failure cases as samples, and determines the fault warning threshold surface of key parameters through density distribution analysis. For example, historical transformer insulation breakdown cases are analyzed to form a high-risk area in the state space, and an early warning is triggered when the real-time monitoring data is mapped near this area. The risk assessment module uses the Mahalanobis distance to measure the deviation between the real-time status point and the early warning threshold surface to generate a fault risk heat map. Compared with the Euclidean distance, the Mahalanobis distance takes into account the correlation between parameters and can more accurately reflect the actual risk of the equipment. For example, when voltage fluctuations, abnormal increases in vibration frequency and simultaneous increases in temperature are detected, the Mahalanobis distance will increase significantly, and the corresponding area will be displayed as a high-risk red area in the heat map;

[0070] Among them, taking the substation equipment network as an example, the nodes represent the equipment and the edges represent the electrical connection relationship. By training the graph neural network model, the propagation law of the fault in the network is learned. When a high-risk warning occurs on a circuit breaker, the model can predict that the fault may propagate along the bus to the adjacent transformer or other load equipment, and generate an operation and maintenance decision matrix that includes inspection priorities (such as prioritizing the inspection of components directly connected to the faulty equipment) and maintenance windows (such as predicting that the probability of failure in the next 24 hours exceeds 80%). The verification unit uses digital twin technology to simulate the execution of operation and maintenance decisions in a virtual environment and evaluate their effectiveness, thereby providing a full-process intelligent solution for power field operation and maintenance from monitoring, warning to decision-making, which is conducive to significantly reducing the risk of equipment failure, shortening power outage time, and improving the safety and economy of power grid operation.

[0071] As an optional embodiment, spatiotemporal feature extraction is performed on the multimodal monitoring dataset, specifically in the following manner:

[0072] Build a hybrid neural network model consisting of a two-dimensional convolutional neural network layer, a long short-term memory network layer, and a fully connected layer;

[0073] Among them, the hybrid neural network model combines the advantages of convolutional neural networks in spatial feature extraction and recurrent neural networks in time series processing, and at the same time realizes deep fusion of features through fully connected layers, providing efficient network structure support for the generation of multi-dimensional feature tensors.

[0074] The two-dimensional convolutional neural network layer sets up multiple convolution layers and pooling layers with different convolution kernel sizes to extract data spatial features;

[0075] Among them, convolution kernels of different sizes (such as 3×3 and 5×5) can capture spatial correlation information in different ranges. For example, small-sized convolution kernels are suitable for extracting signal details of local monitoring points (such as the vibration peak of a certain sensor), while large-sized convolution kernels can capture the spatial distribution patterns of multiple adjacent monitoring points (such as the temperature gradient in different parts of a transformer). The pooling layer reduces the data dimension through downsampling operations, retains key features, and avoids overfitting. For example, after performing a convolution operation on the temperature monitoring data of various areas of the substation, the spatial clustering pattern of temperature can be extracted, and the regional distribution characteristics of poor equipment heat dissipation can be identified.

[0076] Use the gating mechanism of the long short-term memory network layer to process the time series characteristics of the data;

[0077] Gating, which includes input, forget, and output gates, helps address the vanishing or exploding gradient issues of traditional recurrent neural networks and is well-suited for capturing dynamic patterns in long-term time series data. For example, for voltage fluctuation data on electrical equipment, a long short-term memory network can remember abnormal voltage fluctuations from hours ago and correlate them with current vibration signal changes, identifying the cumulative impact of long-term voltage instability on the equipment's mechanical structure. Through this gating mechanism, the network automatically ignores irrelevant short-term noise and focuses on long-term trend characteristics that have a significant impact on the health of the equipment.

[0078] Finally, the spatiotemporal features are fused through the fully connected layer to output a multidimensional feature tensor containing time domain features, frequency domain features and spatial distribution features.

[0079] The fully connected layer nonlinearly combines the spatial features extracted by the convolutional neural network with the temporal features extracted by the long short-term memory network to explore the deep connections between the two. For example, the transformer's vibration frequency characteristics (the proportion of energy at a specific frequency) are combined with the temperature variation characteristics over time (the amplitude of temperature fluctuations caused by the diurnal temperature difference) to form a comprehensive feature that reflects the thermal-mechanical coupling effect of the equipment. The output multidimensional feature tensor provides structured feature data for subsequent tensor decomposition and spatiotemporal coupling relationship analysis, ensuring that the correlation between features is fully preserved.

[0080] Specifically, the hybrid neural network model breaks through the limitations of a single network structure and improves the accuracy of extracting spatial and temporal features. The combination of multi-size convolution kernels and gating mechanisms enables accurate capture of multi-scale, long-term features, helping to avoid feature loss or redundancy. The fusion processing of the fully connected layer establishes a deep correlation between spatiotemporal features, providing a high-quality feature foundation for subsequent equipment health status assessment and fault warning. This helps improve the feature utilization of multimodal data, and in conjunction with tensor decomposition technology, enhances the ability to perceive the operating status of electric field equipment, providing a strong guarantee for the accuracy of operation and maintenance decisions.

[0081] As an optional embodiment, the device health state space is constructed based on the spatiotemporal coupling relationship, specifically:

[0082] Use Riemannian manifolds to construct high-dimensional state space;

[0083] First, whiten the multidimensional feature tensor to eliminate the correlation between features;

[0084] Then the processed data is mapped to the Riemannian manifold space through exponential mapping;

[0085] In the Riemannian manifold space, the kernel density estimation method is used to calculate the probability density distribution of historical fault cases, and then the fault warning threshold surface of key parameters is determined.

[0086] Among them, in the electric field operation, maintenance, commissioning and monitoring, on the one hand, there are complex correlations between the multi-dimensional characteristics of equipment operation, such as the mutual influence between electrical parameters and temperature, vibration and other characteristics. Traditional Euclidean space is difficult to accurately describe this nonlinear relationship, which easily leads to fuzzy state division; on the other hand, the distribution of historical fault cases in high-dimensional feature space is non-uniform, and traditional threshold setting methods (such as single thresholds or hyperplanes) cannot adapt to this distribution characteristic, and often cause false alarms or missed alarms, affecting the accuracy of the judgment of the equipment health status. These problems make the equipment health status assessment lack of reliable spatial model support, which restricts the effectiveness of fault warning;

[0087] Among them, Riemannian manifold is selected as the carrier of the equipment health state space. Compared with the traditional Euclidean space, Riemannian manifold can better characterize the nonlinear correlation and geometric structure between high-dimensional features. The operating characteristics of electric field equipment (such as voltage fluctuation, vibration frequency, gas concentration, etc.) show a complex curved distribution in high-dimensional space. Riemannian manifold can adapt to this distribution through its intrinsic geometric properties, which is conducive to avoiding the deviation of state description caused by unreasonable spatial assumptions. For example, the combination of health characteristics of transformers under different loads and environments can form a more realistic state area division in Riemannian manifold, providing a more accurate spatial basis for subsequent fault probability calculations;

[0088] Among them, in the multidimensional feature tensor, the features of different monitoring parameters are often coupled. For example, a temperature increase may cause a change in resistance and an increase in vibration at the same time. This correlation will interfere with the construction of the state space. The whitening process converts the features into an uncorrelated form through linear transformation, normalizing the variance of each feature. For example, after processing the temperature feature and vibration feature, which originally had a correlation coefficient of 0.8, the correlation coefficient is reduced to below 0.1. This process simplifies the relationship between features, ensuring that the data subsequently mapped to the Riemannian manifold space can more realistically reflect the independent contribution of each feature, thereby improving the discrimination of the state space.

[0089] Among them, exponential mapping is an important tool in Riemannian geometry, which can transform data points in linear space into the curved structure of manifold space, preserving the local geometric characteristics of the data. For example, after whitening the characteristic data of the transformer (including voltage, temperature, vibration, etc.), after exponential mapping, it forms a point set corresponding to the actual health status of the equipment on the Riemann manifold. The points in the normal operating state are concentrated in one area of ​​the manifold, while the points in the fault state are distributed in another area. The point sets of different types of faults have clear geometric separation.

[0090] Among them, density estimation fits the fault probability distribution of the entire space by assigning a kernel function to each historical fault point, and can accurately capture the clustering characteristics of fault cases on the manifold. For example, for transformer insulation aging failure, historical cases show a specific high-density distribution area on the Riemann manifold, and the boundary of this area can be clearly outlined through kernel density estimation. Based on this distribution, the fault warning threshold surface of key parameters is determined. This surface can adaptively fit the contour line of the fault probability. When the real-time monitoring data is mapped to the point on the manifold and enters the inner side of the surface (high probability area), the fault warning is triggered, effectively avoiding the problem that traditional planar thresholds cannot adapt to non-uniform distributions, thereby providing an accurate spatial model for equipment health status assessment. Collaborating with the feature analysis module and the risk assessment module, it is beneficial to improve the reliability of electric field equipment fault warning and provide a solid status judgment basis for operation and maintenance decisions.

[0091] As an optional embodiment, when mapping the processed data to the Riemannian manifold space through exponential mapping, the method further includes:

[0092] Calculate the geodesic distance between the current data point and the origin of the Riemannian manifold space;

[0093] The scale parameter of the exponential mapping is dynamically adjusted according to the geodetic distance. When the distance is greater than a preset threshold, a contraction-type exponential mapping is used to suppress the influence of outliers.

[0094] When the distance is less than the preset threshold, the dilated exponential mapping is used to enhance the discrimination of normal state data;

[0095] The geometric relationship between data points is kept unchanged through parallel movement operations on the manifold.

[0096] Specifically, in the electric field operation, maintenance, debugging and monitoring system, when mapping data to the Riemannian manifold space, on the one hand, there may be outliers in the data (such as extreme values ​​caused by abnormal sensor fluctuations). If these points are directly mapped to the manifold space, the distribution structure of the normal state data will be distorted, resulting in a shift in the fault warning threshold surface. On the other hand, subtle data differences under normal conditions (such as the difference in characteristics between slight aging of equipment and normal operation) may be compressed in the manifold space, making it difficult to form an effective distinction, affecting the accuracy of health status assessment. In addition, the geometric relationship between data points in the traditional mapping process is easily affected by scale changes, resulting in the distortion of the relative positions of originally related state points on the manifold, which cannot accurately reflect the correlation of the actual operation of the equipment.

[0097] The geodesic distance is the shortest path length between two points on a Riemannian manifold. Unlike the straight-line distance in Euclidean space, it more accurately describes the actual distance between data points in the curved structure of the manifold. For example, in the health state space of electric field equipment, the origin can be set as the characteristic point of the equipment's ideal operating state. The geodesic distance between the real-time monitoring data point of a transformer and the origin is 5.2 (in manifold space units of measurement). This distance intuitively reflects the degree of deviation between the current state of the equipment and the ideal state, providing a quantitative basis for subsequent mapping scale adjustments.

[0098] Among them, when the distance is greater than the preset threshold (such as the threshold is set to 8.0), the data point is determined to be an outlier, and a contraction-type exponential mapping is used. The contraction-type mapping compresses the outlier toward the origin of the manifold by reducing the scale parameter, thereby reducing its interference with the overall distribution. For example, the extreme vibration data point generated by a sensor failure has a geodesic distance of 12.0. After the contraction-type mapping, the distance is reduced to 9.5, avoiding the distortion of the clustering structure of the normal state data. When the distance is less than the preset threshold, the expansion-type exponential mapping is used to enlarge the distance between the normal state data points by amplifying the scale parameter, thereby enhancing the distinction between subtle differences. For example, when the equipment is operating normally, the geodesic distance of the temperature feature points under different loads is only 1.0. After expansion, the distance increases to 1.5, so that the "light load normal" and "heavy load slightly overheating" states are clearly separated on the manifold;

[0099] Among them, parallel translation is a transformation method that maintains the direction of vectors in Riemannian geometry. During the data mapping process, it ensures that the relative position and angular relationship between data points are consistent with the original space. For example, in the topology of electric field equipment, the state points of three adjacent devices A, B, and C form a specific triangular relationship in the original space. After parallel translation, this relationship is completely preserved on the Riemann manifold, which helps to avoid geometric distortion caused by mapping scale adjustment. This operation ensures that the correlation between device states (such as the causal relationship between the failure of device A and the abnormality of device B) is not destroyed in the manifold space, providing a reliable geometric basis for subsequent kernel density estimation and fault probability distribution calculation, thereby providing key support for the precise construction of the Riemann manifold space, so that the equipment health state space can more realistically reflect the operating characteristics of the electric field equipment, and lay the foundation for the determination of the subsequent fault warning threshold surface and the accuracy of risk assessment, which is conducive to significantly improving the adaptability of the electric field operation and maintenance debugging monitoring system to complex data distribution.

[0100] As an optional embodiment, when determining the fault warning threshold surface of key parameters, a topological data analysis method is used to extract the fault feature topological structure in the Riemannian manifold space, specifically:

[0101] Construct the Vietoris-Rips complex of historical fault cases to capture the characteristics of holes and connected branches in the manifold space;

[0102] Compute persistent homology groups at different scales and identify fault signatures with topological persistence;

[0103] Correlate and map the persistent coherence characteristics with the physical failure modes of the equipment to determine the topological critical points of key parameters;

[0104] Generate geometrically invariant threshold surfaces in Riemannian manifold space via topological critical points.

[0105] Specifically, in electric field operation, maintenance, commissioning and monitoring, when determining the fault warning threshold surface of key parameters, on the one hand, relying solely on statistical methods or empirical values ​​to set thresholds makes it difficult to capture the complex topological structure of fault characteristics in high-dimensional space. For example, different equipment failure modes may form mutually nested characteristic regions on the Riemann manifold, and traditional threshold surfaces cannot accurately outline their boundaries. On the other hand, fault characteristics may show dynamic topological changes with equipment operating status and environmental changes. Traditional methods lack the ability to identify such topological persistence features, resulting in poor adaptability of threshold surfaces and prone to warning lags or false alarms, affecting the timely judgment and handling of equipment failures.

[0106] Among them, the Vietoris-Rips complex is an important tool in topological data analysis. It connects historical fault case data points into a polyhedron structure by setting different distance parameters. For example, in the Riemannian manifold health state space of electric field equipment, for the insulation fault case data points of the transformer, when the distance parameter is set to 2, a complex structure composed of multiple triangles can be formed. The holes may correspond to the characteristic blank areas under a certain specific insulation fault mode, while the connected branches represent different types of insulation fault groups. In this way, abstract fault data can be converted into a topological structure with geometric meaning, and the distribution of fault characteristics in the manifold space can be intuitively presented;

[0107] Persistent homology groups are used to measure the duration of existence of topological features (such as points, edges, holes, etc.) at different distance parameter scales. The longer the topological features exist, the stronger their topological persistence is, and the more they represent the essential fault characteristics. For example, when analyzing circuit breaker failure cases, as the distance parameter gradually increases from 1 to 10, some hole features persist at multiple scales, while some temporary edge features quickly disappear. Those persistent hole features are the key fault characteristics with topological persistence. This process can filter out non-essential features caused by noise or accidental factors, focusing on the core topological structure that is important for fault warning;

[0108] Among them, the persistent coherent feature is an abstract topological description, which needs to be combined with the actual physical failure mode of the equipment to be meaningful. For example, the identified hole feature with strong persistence is correlated with specific physical failure modes such as excessive dissolved gas concentration in transformer oil and excessive local discharge intensity. It is found that the hole feature corresponds to the typical characteristics of transformer insulation aging failure. Then, under this correlation, the topological critical points of key parameters (such as gas concentration and discharge intensity) are determined. These critical points are the key numerical points where the fault feature changes from one topological state to another. For example, when the concentration of a certain gas reaches this critical point, the topological structure of the fault feature changes significantly.

[0109] Among them, topological critical points have stable positional properties in manifold space. The threshold surfaces generated based on these points will not change drastically due to slight deformations of the manifold or small fluctuations in the data, that is, they have geometric invariance. For example, after determining multiple topological critical points for transformer insulation faults, the threshold surface formed by connecting these points can stably enclose the characteristic region of the insulation fault in the Riemann manifold space. Regardless of any small-scale state fluctuations during normal operation of the equipment, this threshold surface can maintain an accurate definition of the fault area, ensuring the stability and reliability of fault warning.

[0110] Specifically, by constructing the Vietoris-Rips complex and calculating the persistent homology group, the limitations of traditional methods in capturing complex fault characteristics in high-dimensional space have been overcome, the accuracy of fault feature identification has been improved, and the correlation mapping between persistent homology features and physical fault modes has transformed abstract topological features into critical points with practical physical significance, solving the problem of the disconnect between topological features and actual faults. The generated threshold surface with geometric invariance helps to avoid warning deviations caused by data fluctuations or manifold deformation, and improves the stability of fault warning. This solution provides a precise and stable threshold standard for electric field equipment fault warning. In conjunction with the Riemannian manifold state space construction and risk assessment modules, it significantly improves the electric field operation and maintenance debugging monitoring system's ability to predict equipment failures and its reliability.

[0111] As an optional embodiment, a graph neural network is used to predict the fault propagation path by combining the fault risk heat map with the device topology. Specifically:

[0112] Build a graph neural network consisting of graph convolutional layer, graph attention layer and output layer;

[0113] Use graph convolutional layers to aggregate neighborhood information of device nodes to update node features;

[0114] The attention coefficient between nodes is calculated through the graph attention layer to enhance the ability to identify key nodes in the device topology;

[0115] Finally, the output layer predicts the fault propagation path in the device topology based on the updated node features and outputs the path probability distribution.

[0116] Specifically, in the operation, maintenance, commissioning and monitoring of electric fields, on the one hand, electric field equipment forms an interconnected system through complex topological structures (such as busbar connections and cable laying). A single device failure may spread to other devices along the topological link, and traditional prediction methods (such as rule-based reasoning) are difficult to capture this nonlinear and dynamic propagation law. On the other hand, the fault propagation influence of key nodes in the equipment topology (such as hub substations and main transformers) is far greater than that of ordinary nodes, but traditional methods are insufficient in distinguishing the importance of nodes, resulting in a lack of priority in the predicted fault path and an increased probability of invalid inspections. These problems make it difficult for operation and maintenance personnel to accurately locate the direction of fault propagation, delaying the best time for maintenance and even expanding the scope of fault impact;

[0117] Graph neural networks use devices as nodes and electrical connections as edges, converting device topologies into graph data, which naturally matches the relationships between devices. Compared to traditional neural networks, graph neural networks do not rely on the grid structure of data and can directly handle irregular device topologies. For example, they can effectively model the star or ring connection structure of "transformer-busbar-circuit breaker" in substations, providing an adaptive network architecture for fault propagation path prediction.

[0118] The convolutional layer captures information interactions within the local topology by weighted summing the features of each device node and its directly connected neighboring nodes. For example, when a disconnector is at risk of failure, the graph convolutional layer fuses its features with those of adjacent busbars and transformers, updating the feature vectors of these nodes to reflect the initial impact of the potential transmission of the fault. This aggregation mechanism simulates the physical propagation of faults between devices. For example, if current anomalies affect the voltage or temperature of adjacent devices through electrical connections, the graph convolutional layer quantifies this impact as changes in node features through feature updates, providing a foundation for subsequent path prediction.

[0119] Among them, the attention coefficient reflects the influence weight of the neighboring nodes on the current node. Key nodes (such as main transformers) are in a core position in the topology and have denser connections with other nodes. Their attention coefficients will be significantly higher than those of ordinary nodes. For example, in the substation topology, the main transformer is connected to multiple feeders and circuit breakers. When the graph attention layer calculates its attention coefficient with the neighboring nodes, the weight of the main transformer may reach 0.6 (the average for ordinary nodes is 0.2), which makes the network pay more attention to the state changes of the main transformer when predicting fault propagation. This mechanism solves the problem of insufficient distinction of node importance in traditional methods and ensures that the predicted fault path focuses on key propagation links;

[0120] The output layer uses a softmax function to convert node features into the probability of fault propagation to neighboring nodes. For example, when a circuit breaker fault is detected, the output layer may predict a 70% probability of fault propagation to the busbar, a 20% probability of propagation to adjacent load switches, and a 10% probability of propagation to other equipment. This probability distribution provides a quantitative basis for O&M decisions, such as prioritizing inspections of buses with the highest propagation probabilities and allocating appropriate inspection resources to different paths. The output fault propagation path clearly illustrates the link from the fault source to potentially affected equipment, such as "circuit breaker A → busbar B → transformer C." This helps O&M personnel quickly identify the risk area and accurately predict the fault propagation path within the power plant equipment topology. This helps O&M personnel focus on high-risk links and optimize inspection resource allocation, significantly shortening fault location time and reducing the risk of fault expansion, providing strong support for the scientific and timely nature of power plant O&M decisions.

[0121] As an optional embodiment, the deviation between the real-time state point and the warning threshold surface is measured by Mahalanobis distance to generate a fault risk heat map, specifically:

[0122] First, the real-time monitoring data is standardized to have zero mean and unit variance;

[0123] The local weighted regression method is used to perform local fitting on the warning threshold surface;

[0124] Calculate the Mahalanobis distance from the real-time state point to the fitted surface, and smooth the Mahalanobis distance calculation result by combining Gaussian kernel density estimation;

[0125] Using the geographical location of the equipment as the coordinate reference, a fault risk heat map is ultimately generated.

[0126] Specifically, in electric field operation, maintenance, and debugging monitoring, firstly, direct use of real-time monitoring data for distance calculations can lead to distorted deviation assessments due to dimensional differences and uneven distribution. Secondly, the warning threshold surfaces are mostly global fits, which are difficult to adapt to the local nonlinear characteristics of the equipment health status space, resulting in insufficient risk assessment accuracy in some areas.

[0127] Real-time monitoring data comes from different sensors, and the numerical ranges of electrical parameters (such as voltage kV) and environmental parameters (such as humidity%) vary significantly. Direct calculation will result in large numerical parameters dominating the distance results. Standardization is done through the formula (where μ is the mean and σ is the standard deviation) to eliminate dimensionality effects. For example, the temperature data (mean 25°C, standard deviation 5°C) and vibration data (mean 0.1mm / s, standard deviation 0.02mm / s) of a certain area are uniformly converted to standardized data with zero mean and unit variance. This process ensures that all parameters are weighted equally in the distance calculation, helps avoid deviation assessment bias caused by dimensional differences, and provides a reliable data foundation for subsequent Mahalanobis distance calculations.

[0128] Globally fitted threshold surfaces struggle to adapt to complex nonlinear distributions in Riemannian manifold space. For example, in the equipment health state space, the fault signature distributions in the transformer region and the circuit breaker region may exhibit different curvatures. Local weighted regression constructs a local polynomial model by assigning a weight to each point to be evaluated (higher weights for nearby points and lower weights for distant points). For example, near a transformer health point, a surface is fitted using only the 50 historical fault points surrounding it, making the surface more consistent with local characteristics.

[0129] Among them, the Mahalanobis distance takes into account the correlation between parameters, such as the positive correlation between temperature and resistance, and can more truly reflect the degree of state deviation than the Euclidean distance. For example, the Mahalanobis distance between a real-time state point and the threshold surface is 2.5, indicating that its degree of deviation from the normal range is 2.5 times the average level. Since real-time data may contain noise, the Gaussian kernel density estimation smoothes local fluctuations by weighted averaging the distance values ​​(such as with a bandwidth of 0.3) to make the risk value transition more continuous. For example, the distance value that originally fluctuated between 2.3-2.7 is smoothed to 2.5±0.1 to avoid misjudgment of risks due to noise;

[0130] Among them, an effort is made to map the risk value corresponding to the Mahalanobis distance into a color gradient (such as red for high risk and blue for low risk), which is superimposed on the topological map of the electric field equipment. For example, a transformer area in a substation is displayed in dark red (risk value 0.85), and its adjacent busbar area is yellow (risk value 0.6), which intuitively presents the spatial distribution of risks. Operation and maintenance personnel can quickly identify the high-risk link of "transformer A-busbar B" through the heat map and prioritize inspections, solving the problem of poor readability of traditional numerical lists and improving the efficiency of risk positioning. This provides an accurate and readable risk distribution map for electric field operation and maintenance, and cooperates with the fault propagation prediction of the graph neural network to improve the efficiency of identifying high-risk areas and the rationality of maintenance resource allocation, which is conducive to significantly enhancing the risk perception and decision-making support capabilities of the electric field operation and maintenance debugging monitoring system.

[0131] As an optional embodiment, the effectiveness of the operation and maintenance decision matrix is ​​verified by digital twin technology, specifically:

[0132] Build a digital twin that includes equipment geometry, material properties, operational data, and environmental impacts;

[0133] Use 3D modeling software to build a geometric model of the equipment and set parameters according to material characteristics to form a material property model;

[0134] Synchronize equipment operation data to establish an operation data model, and simulate environmental factors to build an environmental impact model;

[0135] The operation and maintenance decision matrix is ​​simulated and executed in the digital twin, and the results including at least the operating status and failure probability are collected.

[0136] Specifically, in electric field operation, maintenance, commissioning, and monitoring, directly executing unverified decisions on physical equipment poses security risks. Furthermore, relying solely on historical data or simulation models for offline verification fails to reflect the current state of the equipment and the impact of complex environments. These issues make it difficult to ensure the effectiveness of operation and maintenance decisions, potentially leading to "over-maintenance" or "under-maintenance," increasing operation and maintenance costs and reducing equipment reliability.

[0137] Among them, a digital twin that includes equipment geometry, material properties, operating data and environmental impact is established to fully reproduce the real characteristics of the physical equipment. Use 3D modeling software (such as SolidWorks) to build a geometric model of the equipment, accurate to millimeter-level dimensions, such as the number of winding layers and the shape of the core of the transformer. Set parameters based on material properties to form a material property model, such as setting the resistivity of copper windings, the magnetic permeability of silicon steel sheets, etc. These parameters directly affect the electrical and thermal performance of the equipment. Synchronize equipment operation data to establish an operation data model, and obtain parameters such as voltage, current, and temperature through a real-time interface so that the digital twin can reflect the immediate status of the equipment. Simulate environmental factors to build an environmental impact model, such as setting the temperature change range, humidity fluctuation pattern, pollution level, etc. in different seasons. Environmental factors have a significant impact on equipment aging and failure probability. For example, a high temperature environment will accelerate the aging of insulation materials;

[0138] The operation and maintenance decision matrix includes inspection priorities, maintenance windows, parameter adjustment plans, and other content. For example, when the decision matrix proposes "to perform cooling system maintenance on a transformer in the next 24 hours", the digital twin will simulate this operation and update the equipment operation status in real time. The temperature changes before and after maintenance are monitored by running data models, such as whether the hot spot temperature drops from 85°C to 75°C after cooling system maintenance; the fault probability model is used to evaluate the improvement of equipment reliability caused by maintenance, such as whether the probability of insulation aging failure drops from 15% to 8%. For complex decisions (such as those involving the coordinated operation of multiple devices), the digital twin can simulate the chain reaction during the decision execution process. For example, adjusting the opening and closing time of a circuit breaker may affect the voltage stability of adjacent equipment. The digital twin can quantify this impact and output the corresponding operating parameter change curve;

[0139] In summary, this provides a virtual verification environment for operation and maintenance decisions, enabling operation and maintenance personnel to evaluate the effectiveness of decisions before implementation, optimize maintenance plans, reduce unnecessary maintenance work, and lower operation and maintenance costs, which is conducive to significantly improving the decision-making reliability and economy of the electric field operation and maintenance debugging monitoring system.

[0140] As an optional embodiment, when verifying the validity of the operation and maintenance decision matrix using digital twin technology, the following steps may also be included:

[0141] Build a digital twin model that is consistent with the physical electric field equipment parameters. The digital twin model includes the equipment electrical parameter model, mechanical structure model, and environmental impact model.

[0142] Input the debugging strategy parameters in the operation and maintenance decision matrix into the digital twin model to simulate and generate a predicted status dataset;

[0143] Calculate the geodesic distance between the predicted state dataset and the measured data of the physical device in the high-dimensional state space as a quantitative indicator of decision effectiveness;

[0144] When the geodetic distance exceeds the preset threshold, the density distribution of historical fault cases in the high-dimensional state space is recalculated, and the fault warning threshold surface of key parameters is adjusted;

[0145] By comparing the topological feature changes in the predicted status dataset with the topological features of historical fault cases, a verification report containing the fault pattern matching degree is generated, and the parameter weights of the spatiotemporal feature extraction algorithm are adjusted based on the verification report.

[0146] Among them, the electrical parameter model is accurate to the volt-ampere characteristics, impedance-frequency characteristics, etc. of each component, such as the contact resistance of the circuit breaker contacts, the eddy current loss coefficient of the transformer winding, etc. The mechanical structure model uses the finite element analysis method to simulate the structural response of the equipment under vibration and stress, such as the fatigue life prediction of the disconnector operating mechanism. The environmental impact model comprehensively considers the impact of factors such as temperature, humidity, and pollution level on equipment performance, such as establishing a nonlinear relationship model between temperature and insulation aging rate. The model parameters are continuously calibrated through field measured data to ensure that the error between the digital twin model and the physical equipment is less than 5%, meeting the industry's highest accuracy standards;

[0147] Among them, when the decision matrix proposes "increasing the reactive compensation capacity of a certain line by 20%", the digital twin model will output the changes in parameters such as voltage distribution and current density of the equipment after the operation. The geodesic distance between the predicted state data set and the measured data of the physical equipment in the high-dimensional state space is calculated as a quantitative indicator of the effectiveness of the decision. The geodesic distance takes into account the geometric characteristics of the state space and can more accurately reflect the degree of difference between the two data sets. For example, after a circuit breaker implements the opening and closing time adjustment, the geodesic distance between the contact wear rate predicted by the digital twin model and the measured value of the physical equipment in the high-dimensional state space is 3.2, which intuitively shows the degree of deviation between the decision implementation effect and the expectation;

[0148] If the distance between a transformer's predicted and measured states continues to increase, the system automatically analyzes the density distribution of historical insulation fault cases and adaptively adjusts the warning threshold surfaces for key parameters such as temperature and partial discharge to better align the warnings with the equipment's actual operating status. By comparing the topological feature changes in the predicted state dataset with the topological features of historical fault cases, a verification report is generated, including the degree of fault pattern matching. For example, the hole features appearing in the predicted state dataset match the topological features of historical insulation aging faults at an 85% match, indicating the risk of insulation aging. Based on this verification report, the parameter weights of the spatiotemporal feature extraction algorithm are adjusted to enhance the ability to identify these fault features. For example, the weight of the temperature change rate feature is increased from 0.3 to 0.5, enhancing sensitivity to abnormal temperature changes. This improves the intelligence and reliability of the electric field operation, maintenance, and commissioning monitoring system.

[0149] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. Applicable to the operation, maintenance, debugging and monitoring system of the electric field, characterized by: include: An acquisition unit collects operating data of the electric field equipment at multi-dimensional monitoring points, wherein the operating data includes electrical parameters, mechanical vibration parameters, environmental parameters, and chemical gas composition data, and constructs a multimodal monitoring data set; A feature analysis module extracts spatiotemporal features from the multimodal monitoring dataset, generates a multidimensional feature tensor containing time domain features, frequency domain features, and spatial distribution features, and obtains the spatiotemporal coupling relationship between features through tensor decomposition technology; a calculation module, which constructs a device health state space based on the spatiotemporal coupling relationship, calculates the density distribution of historical failure cases in the state space, and determines a fault warning threshold surface of key parameters; The risk assessment module maps the real-time monitoring data to the equipment health status space, measures the deviation between the real-time status point and the warning threshold surface using Mahalanobis distance, and generates a fault risk heat map; The decision module combines the fault risk heat map with the equipment topology, uses a graph neural network to predict the fault propagation path, and generates an operation and maintenance decision matrix that includes inspection priorities and maintenance windows; The verification unit verifies the validity of the operation and maintenance decision matrix through digital twin technology, and dynamically adjusts the parameter weights of the spatiotemporal feature extraction algorithm according to the verification results.

2. The operation, maintenance, debugging and monitoring system for electric fields according to claim 1, characterized in that: The spatiotemporal feature extraction of the multimodal monitoring data set is specifically performed in the following manner: Build a hybrid neural network model consisting of a two-dimensional convolutional neural network layer, a long short-term memory network layer, and a fully connected layer; The two-dimensional convolutional neural network layer is provided with multiple convolution layers and pooling layers with different convolution kernel sizes to extract data spatial features; Use the gating mechanism of the long short-term memory network layer to process the time series characteristics of the data; Finally, the spatiotemporal features are fused through the fully connected layer to output a multidimensional feature tensor containing time domain features, frequency domain features and spatial distribution features.

3. The operation, maintenance, debugging and monitoring system for electric fields according to claim 1, characterized in that: The device health status space is constructed based on the spatiotemporal coupling relationship, specifically: Use Riemannian manifolds to construct high-dimensional state space; First, whiten the multidimensional feature tensor to eliminate the correlation between features; Then the processed data is mapped to the Riemannian manifold space through exponential mapping; In the Riemannian manifold space, the kernel density estimation method is used to calculate the probability density distribution of historical fault cases, and then the fault warning threshold surface of key parameters is determined.

4. The operation, maintenance, debugging and monitoring system for electric fields according to claim 1, characterized in that: The fault risk heat map and the equipment topology are combined to predict the fault propagation path using a graph neural network, specifically: Build a graph neural network consisting of graph convolutional layer, graph attention layer and output layer; Use graph convolutional layers to aggregate neighborhood information of device nodes to update node features; The attention coefficient between nodes is calculated through the graph attention layer to enhance the ability to identify key nodes in the device topology; Finally, the output layer predicts the fault propagation path in the device topology based on the updated node features and outputs the path probability distribution.

5. The operation, maintenance, debugging and monitoring system for electric fields according to claim 1, characterized in that: The Mahalanobis distance is used to measure the deviation between the real-time state point and the warning threshold surface to generate a fault risk heat map, specifically: First, the real-time monitoring data is standardized to have zero mean and unit variance; The local weighted regression method is used to perform local fitting on the warning threshold surface; Calculate the Mahalanobis distance from the real-time state point to the fitted surface, and smooth the Mahalanobis distance calculation result by combining Gaussian kernel density estimation; Using the geographical location of the equipment as the coordinate reference, a fault risk heat map is ultimately generated.

6. The operation, maintenance, debugging and monitoring system for electric fields according to claim 1, characterized in that: The effectiveness of the operation and maintenance decision matrix is ​​verified by digital twin technology, specifically: Build a digital twin that includes equipment geometry, material properties, operational data, and environmental impacts; Use 3D modeling software to build a geometric model of the equipment and set parameters according to material characteristics to form a material property model; Synchronize equipment operation data to establish an operation data model, and simulate environmental factors to build an environmental impact model; The operation and maintenance decision matrix is ​​simulated and executed in the digital twin, and the results including at least the operating status and failure probability are collected.

7. The operation, maintenance, debugging and monitoring system for electric fields according to claim 1, characterized in that: The electrical parameters include at least three of partial discharge, insulation resistance, dielectric loss factor, leakage current, dynamic capacitance and capacitance change rate, ultra-high frequency discharge signal and pulse current signal; the mechanical vibration parameters include acceleration amplitude, vibration frequency distribution, and vibration phase characteristics; the environmental parameters include temperature, humidity, air pressure, and wind speed; and the chemical gas component data include SF6 decomposition product concentration and dissolved gas content in transformer oil.

8. The operation, maintenance, debugging and monitoring system for electric fields according to claim 3, characterized in that: When the processed data is mapped to the Riemannian manifold space by exponential mapping, the method further includes: Calculate the geodesic distance between the current data point and the origin of the Riemannian manifold space; The scale parameter of the exponential mapping is dynamically adjusted according to the geodetic distance. When the distance is greater than a preset threshold, a contraction-type exponential mapping is used to suppress the influence of outliers. When the distance is less than the preset threshold, the dilated exponential mapping is used to enhance the discrimination of normal state data; The geometric relationship between data points is kept unchanged through parallel movement operations on the manifold.

9. The operation, maintenance, debugging and monitoring system for electric fields according to claim 8, characterized in that: When determining the fault warning threshold surface of the key parameters, a topological data analysis method is used to extract the fault feature topological structure in the Riemannian manifold space, specifically: Construct the Vietoris-Rips complex of historical fault cases to capture the characteristics of holes and connected branches in the manifold space; Compute persistent homology groups at different scales and identify fault signatures with topological persistence; Correlate and map the persistent coherence characteristics with the physical failure modes of the equipment to determine the topological critical points of key parameters; Generate geometrically invariant threshold surfaces in Riemannian manifold space via topological critical points.

10. The operation, maintenance, debugging and monitoring system for electric fields according to claim 1, characterized in that: When the digital twin technology verifies the validity of the operation and maintenance decision matrix, it also includes: Constructing a digital twin model consistent with the physical electric field device parameters, wherein the digital twin model includes a device electrical parameter model, a mechanical structure model, and an environmental impact model; Input the debugging strategy parameters in the operation and maintenance decision matrix into the digital twin model to simulate and generate a predicted status dataset; Calculate the geodesic distance between the predicted state dataset and the measured data of the physical device in the high-dimensional state space as a quantitative indicator of decision effectiveness; When the geodetic distance exceeds the preset threshold, the density distribution of historical fault cases in the high-dimensional state space is recalculated, and the fault warning threshold surface of key parameters is adjusted; By comparing the topological feature changes in the predicted status dataset with the topological features of historical fault cases, a verification report containing the fault pattern matching degree is generated, and the parameter weights of the spatiotemporal feature extraction algorithm are adjusted based on the verification report.

Images