An ultra-high voltage insulator full life cycle digital operation and maintenance management method and system

Abstract

This invention relates to the field of data processing technology and discloses a digital operation and maintenance management method and system for the entire life cycle of ultra-high voltage insulators. The method includes: acquiring the digital genetic code of the insulator upon leaving the factory and anchoring it on the blockchain; simultaneously collecting dynamic phenotypic characteristics and micro-meteorological parameters related to grid connection; splicing and fusing the underlying heterogeneous data tensors, calculating edge weights based on wind direction and distance attenuation rules, and constructing a dynamic spatiotemporal map base; using a time-series graph convolutional network to perform deep evolutionary inference on the map, outputting a quantitative degradation risk score; dynamically generating differentiated operation and maintenance strategies and underlying control instructions based on this, and mapping virtual storage capacity demand parameters through trend extrapolation; identifying common defects, and issuing process optimization instructions to the manufacturing source after strict tamper-proof verification using blockchain. This invention completely breaks down cross-domain data silos, realizing a closed-loop leap from passive emergency repairs to predictive operation and maintenance and pre-scheduled supply chain scheduling for insulators.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a digital operation and maintenance management method and system for the entire life cycle of ultra-high voltage insulators. Background Technology

[0002] Ultra-high voltage (UHV) transmission lines are a crucial component of the power grid, and insulators, as key insulation and mechanical support equipment, directly impact the safety and stability of the power grid. Currently, the routine maintenance of UHV insulators mainly relies on periodic manual inspections and fixed threshold alarm mechanisms based on single sensors.

[0003] In practical engineering applications, existing management models have certain limitations. The microscopic parameters generated during the manufacturing process of insulators are usually separate from their dynamic monitoring data during grid connection and operation. Maintenance systems struggle to comprehensively integrate the equipment's factory-set parameters to benchmark and calibrate its aging rate during operation. Furthermore, most existing condition assessment methods analyze individual insulator nodes in isolation, failing to incorporate spatial topological relationships such as environmental stress transmission between adjacent devices, micro-meteorological changes, and terrain and wind direction into the calculation logic, leading to discrepancies between the condition assessment model and the actual physical environment.

[0004] Furthermore, alarm mechanisms based on static thresholds lack the ability to calculate and predict the long-term degradation trajectory of insulators. Because the expected failure time cannot be accurately predicted, power grid maintenance departments struggle to plan material needs in advance, easily leading to delays in spare parts scheduling or excess inventory. When similar high-frequency faults occur on-site, existing mechanisms lack reliable data links and verification methods to support traceability back to the manufacturing source, making it difficult to achieve closed-loop feedback and process improvement from the production end to the operation end. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a digital operation and maintenance management method and system for the entire life cycle of UHV insulators. This method solves the problems in existing UHV insulator operation and maintenance management, such as the disconnect between manufacturing data and grid operation data, the lack of topological correlation in condition assessment due to physical space and environmental constraints, and the difficulty in achieving accurate failure prediction and reliable cross-domain traceability.

[0006] The first aspect of this invention provides a digital operation and maintenance management method for the entire lifecycle of ultra-high voltage (UHV) insulators, comprising: acquiring the microscopic manufacturing parameters of the UHV insulators to generate a factory digital gene; storing the hash value of the factory digital gene and equipment metadata on-chain and storing the original data off-chain; collecting the dynamic phenotypic features and micro-meteorological parameters of the UHV insulators; constructing a dynamic spatiotemporal map of the UHV corridor, concatenating the factory digital gene and the dynamic phenotypic features as feature vectors of the map nodes, and calculating the dynamic edge weights between the map nodes based on the micro-meteorological parameters; extracting spatiotemporal features based on the dynamic spatiotemporal map and outputting a degradation risk score for the UHV insulators; generating an operation and maintenance dispatch strategy based on the degradation risk score, predicting the failure time of the UHV insulators and performing material scheduling, and extracting common features that cause degradation to generate process optimization instructions.

[0007] Further, the step of calculating the dynamic edge weights between map nodes based on the micro-meteorological parameters includes: obtaining the spatial distance between the source node and the target node, and performing a negative exponential decay operation on the spatial distance; extracting the real-time wind direction vector from the micro-meteorological parameters, calculating the spatial angle between the real-time wind direction vector and the direction vector pointing from the source node to the target node, extracting the cosine value of the spatial angle and performing zero-value truncation processing; multiplying the result of the negative exponential decay operation, the cosine value after zero-value truncation processing, and the micro-topography adjustment coefficient assigned based on the meteorological clustering results to obtain the dynamic edge weights.

[0008] Furthermore, the step of extracting spatiotemporal features based on the dynamic spatiotemporal graph and outputting the degradation risk score of the UHV insulator includes: extracting the adjacency matrix of the dynamic spatiotemporal graph; spatially aggregating the feature vectors of the adjacency matrix and graph nodes through a graph convolutional network to obtain spatial topological features; inputting the spatial topological features into a gated recurrent unit, and updating the hidden state of the gated recurrent unit in the current time step by combining it with the hidden state of the previous time step; and inputting the updated hidden state of the gated recurrent unit into a multilayer perceptron to output the degradation risk score.

[0009] Furthermore, the step of generating an operation and maintenance dispatch strategy based on the degradation risk score includes: continuously monitoring the degradation risk score corresponding to the map node; when the degradation risk score of a specific node is greater than or equal to a preset warning sensitivity threshold, generating an upgraded inspection work order for that node; allocating multimodal collaborative inspection tasks according to the upgraded inspection work order, and dynamically increasing the sampling frequency of the corresponding node's sensor channel.

[0010] Furthermore, the process of predicting the failure time of the UHV insulators and scheduling materials, and extracting common features that cause degradation to generate process optimization instructions, includes: predicting the failure time of each UHV insulator that exceeds the absolute failure threshold through a time-series network, generating material demand parameters based on the failure time distribution, and sending a capacity lock instruction to the supplier; extracting the set of UHV insulators that have reached the absolute failure threshold, locating the common digital gene fragments that cause degradation using a frequent itemset association mining algorithm; verifying the original data containing the common digital gene fragments using the hash value stored on the blockchain, and sending the process optimization instruction to the corresponding product lifecycle management system after verification.

[0011] The second aspect of this invention provides a digital operation and maintenance management system for the entire lifecycle of ultra-high voltage (UHV) insulators, comprising: a data acquisition and storage module, used to acquire microscopic manufacturing parameters of UHV insulators to generate a factory digital gene, store the hash value of the factory digital gene and equipment metadata on-chain, and store the original data off-chain, while simultaneously collecting dynamic phenotypic features and micro-meteorological parameters of the UHV insulators; a dynamic map construction module, used to construct a dynamic spatiotemporal map of the UHV corridor, concatenating the factory digital gene and the dynamic phenotypic features as feature vectors of map nodes, and calculating dynamic edge weights between map nodes based on the micro-meteorological parameters; an evolutionary deduction and analysis module, used to extract spatiotemporal features based on the dynamic spatiotemporal map and output a degradation risk score for the UHV insulators; and an operation and maintenance and collaboration module, used to generate an operation and maintenance dispatch strategy based on the degradation risk score, predict the failure time of the UHV insulators and perform material scheduling, and extract common features that cause degradation to generate process optimization instructions.

[0012] Furthermore, when the dynamic map construction module calculates the dynamic edge weights between map nodes based on the micro-meteorological parameters, it specifically performs the following: obtaining the spatial distance between the source node and the target node, and performing a negative exponential decay operation on the spatial distance; extracting the real-time wind direction vector from the micro-meteorological parameters, calculating the spatial angle between the real-time wind direction vector and the direction vector pointing from the source node to the target node, extracting the cosine value of the spatial angle and performing zero-value truncation processing; multiplying the result of the negative exponential decay operation, the cosine value after zero-value truncation processing, and the micro-topography adjustment coefficient assigned based on the meteorological clustering results to obtain the dynamic edge weights.

[0013] Furthermore, when the evolutionary inference and analysis module extracts spatiotemporal features based on the dynamic spatiotemporal graph and outputs the degradation risk score of the UHV insulator, it is specifically used to: extract the adjacency matrix of the dynamic spatiotemporal graph; perform spatial aggregation on the adjacency matrix and the feature vectors of the graph nodes through a graph convolutional network to obtain spatial topological features; input the spatial topological features into the gated recurrent unit, and update the hidden state of the gated recurrent unit in the current time step by combining it with the hidden state of the previous time step; input the updated hidden state of the gated recurrent unit into the multilayer perceptron, and output the degradation risk score.

[0014] Furthermore, when the operation and maintenance and collaboration module generates an operation and maintenance dispatch strategy based on the degradation risk score, it is specifically used to: continuously monitor the degradation risk score corresponding to the map node; when the degradation risk score of a specific node is greater than or equal to a preset warning sensitivity threshold, generate an upgraded inspection work order for that node; allocate multimodal collaborative inspection tasks according to the upgraded inspection work order, and dynamically increase the sampling frequency of the corresponding node's sensor channel.

[0015] Furthermore, when the operation and maintenance and collaboration module predicts the failure time of the UHV insulators and performs material scheduling, extracts common features that cause degradation to generate process optimization instructions, it is specifically used to: predict the failure time of each UHV insulator that exceeds the absolute failure threshold through a time-series network, generate material demand parameters based on the failure time distribution, and send a capacity locking instruction to the supplier; extract the set of UHV insulators that have reached the absolute failure threshold, and use a frequent itemset association mining algorithm to locate the common digital gene fragments that cause degradation; use the hash value stored on the blockchain to verify the original data containing the common digital gene fragments, and after verification, send the process optimization instruction to the corresponding product lifecycle management system.

[0016] This invention provides a digital operation and maintenance management method and system for the entire life cycle of ultra-high voltage insulators. It has the following beneficial effects: 1. Addressing the technical shortcomings of traditional static maps in reflecting the dynamic spread of pollution, this invention extracts real-time wind direction, truncates the spatial angle cosine to zero, and combines this with negative exponential decay calculations based on physical distance to construct dynamic edge weights capable of quantifying the unidirectional spread of environmental stress. This scheme realistically reproduces the objective physical laws governing the accumulation of pollution at specific nodes, accurately pinpointing high-risk, vulnerable insulators downwind. It effectively transforms traditional reactive, blind line inspections into precise dispatching driven by predicted risks, significantly improving the safety and maintenance efficiency of UHV corridors.

[0017] 2. To address the issues of data fragmentation and scheduling lag between power grid operation and maintenance and manufacturing, this invention maps failure predictions to pre-emptive virtual storage capacity through a time-series network, achieving predictive capacity locking and material scheduling. Simultaneously, relying on correlation mining and hash-value blockchain tamper-proof verification, the system accurately traces common manufacturing tolerances from batch failure data and directly feeds back defect characteristics to the manufacturer's PLM system. This cross-domain closed-loop mechanism not only physically blocks the continuous grid connection of insulators with similar defects but also effectively drives iterative optimization of upstream manufacturing processes from the data flow source. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a block diagram of the management system structure of the present invention; Figure 3 This is a schematic diagram of the hardware physical structure of the electronic device of the present invention; Figure 4 This is a schematic diagram of the dynamic spatiotemporal map topology construction of the ultra-high voltage corridor according to the present invention; Figure 5 This is a derivation diagram of the temporal graph convolutional network architecture of the present invention; Figure 6 This is a sequence diagram of the cross-domain closed-loop and blockchain tamper-proof data interaction of the present invention. Detailed Implementation

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see the appendix Figure 1 -Appendix Figure 6 This invention provides a digital operation and maintenance management method and system for the entire life cycle of ultra-high voltage insulators. The digital operation and maintenance management method for the entire life cycle of ultra-high voltage insulators may include the following execution contents: Based on industrial data management standards, the system acquires micro-manufacturing parameters of UHV insulators within the product lifecycle management system through big data collection. These micro-manufacturing parameters objectively record the inherent physical and chemical states of the insulators at the manufacturing stage, specifically covering the mixing ratio of silicone rubber skirts, high-temperature vulcanization time, cross-linking degree test indicators, and mold injection tolerance data. The system vectorizes these multi-dimensional manufacturing parameters to generate initial state characteristics that characterize the natural corrosion resistance and structural weaknesses of individual insulators—the "manufacturing digital gene."

[0021] To address the dimensional differences in multi-source heterogeneous parameters and clarify the specific operational logic of the aforementioned vectorization transformation, this embodiment discloses a feature stitching method based on range standardization. The specific implementation process is as follows: First, the system extracts the micro-manufacturing parameters according to a fixed structured sequence. In this embodiment, the extraction sequence is set to include four core physical dimensions: the first dimension... The compounding ratio for silicone rubber umbrella skirts (dimensions are mass percentages, typically ranging from 60% to 80%); second dimension The third dimension is the high-temperature vulcanization time (measured in minutes, typically ranging from 10 to 20 minutes). Crosslinking degree test index (dimensions are percentages, typical range is 85% to 95%); fourth dimension Mold injection tolerance data (unit: micrometer) Typical range is 10 to 50 Subsequently, the system performs quantization calculations. This involves extracting the raw parameters for each dimension. ( The system calls the theoretical minimum value preset under the engineering physical constraints of this parameter. and maximum value The linear range transform operator is used for normalization, mapping the values ​​to dimensionless floating-point eigenvalues ​​in the interval [0,1]. The conversion formula is defined as follows: ; Finally, the system invokes a tensor concatenation operation in the computer's video memory to combine the four normalized floating-point eigenvalues ​​in sequence, generating a one-dimensional floating-point vector with a fixed length of 4. This fixed-length vector is the factory-issued digital gene. A specific example illustrates this: when a single insulator has a mixed ratio of 70%, a vulcanization time of 15 minutes, a crosslinking degree of 90%, and an injection molding tolerance of 25... At that time, the system transforms it into a feature vector of [0.50, 0.50, 0.50, 0.375] according to the above formula. This mechanism gives "vectorization" a clear numerical algorithm and structural definition, completely eliminating the feature weight offset caused by different units, and enabling those skilled in the art to accurately reproduce the extraction and transformation process of the underlying data through programming.

[0022] To ensure the uniqueness of the insulator's entire lifecycle traceability and the tamper-proof nature of the underlying data, a cryptographic hash operation is performed on the factory-issued digital gene to generate a fixed-length hash value. Specifically, this invention employs an industrial-grade consortium blockchain architecture for data solidification and anchoring. To address the congestion and storage cost issues associated with high-frequency data uploading to the blockchain, the system adopts an isolated storage strategy of "storing the original data off-chain and solidifying the hash value on-chain." Its specific operating mechanism is as follows: Data hashing: Complete micro-manufacturing parameters are stored in an off-chain industrial data storage platform. The system uses a secure one-way hash algorithm (such as the national cryptographic SM3 or SHA-256 algorithm) to process the original data and generate a fixed-length hash value that uniquely represents the data. On-chain data structure construction: The system uses an application programming interface (API) to concatenate the generated hash value with the device identity features to construct a standardized on-chain request message. The structured data within the message includes at least: the unique device identifier of the UHV insulator, the manufacturer's identity code, the on-chain timestamp accurate to the second, the fixed-length hash value of the factory-issued digital gene, and an asymmetric encrypted digital signature for non-repudiation. Smart Contracts and Consensus Endorsement: The structured messages mentioned above trigger on-chain transactions by invoking smart contracts on the consortium blockchain. This consortium blockchain network comprises verification nodes jointly deployed by power grid operation and maintenance companies, insulator manufacturers, and quality inspection and supervision agencies. The underlying consensus mechanism employs a practical Byzantine fault tolerance mechanism or a similar approach. The on-chain process strictly adheres to a pre-defined multi-party endorsement strategy, meaning that transactions must be jointly verified and digitally signed by key nodes from both the power grid and manufacturing ends before being packaged and recorded in a new block.

[0023] This mechanism completely eliminates the possibility of a single interested party forging or tampering with the factory process parameters at the distributed ledger protocol layer and cryptographic level. It establishes an absolutely trustworthy data mapping between physical equipment and the digital twin space, providing authoritative data credentials for subsequent accountability to manufacturers and process feedback. Simultaneously, at the operation site of the UHV corridor, the dynamic phenotypic characteristics and micro-meteorological parameters of the UHV insulators are continuously collected using IoT sensor nodes. The dynamic phenotypic characteristics include the real-time leakage current amplitude, partial discharge pulse frequency, and static contact angle of the insulator surface, which are used to accurately quantify the current apparent degradation degree of the insulation material.

[0024] The micrometeorological parameters include three-dimensional real-time wind direction vector, temperature and humidity, and ultraviolet radiation. The collected dynamic phenotypic features and micrometeorological parameters are saved in real time to the industrial data storage platform via industrial IoT communication protocols, providing continuous high-frequency time-series data support for environmental simulation of the entire corridor.

[0025] In the big data processing stage, a dynamic spatiotemporal graph is constructed using the UHV corridor as the physical space. This spatiotemporal graph maps the various insulator entities distributed on the transmission towers as discrete nodes in a graph structure. Before performing tensor splicing operations, the system first aligns and serializes the heterogeneous data at the underlying level. Specifically, a fixed-length feature code (e.g., the first 16 bytes of the hash value) is extracted from the factory digital gene and converted into a baseline numerical feature using one-hot encoding or hash embedding. Simultaneously, for dynamic phenotypic features (such as hexadecimal sensor feedback messages containing temperature and leakage current), the range transformation method is used to parse and map them to a floating-point range of [0,1]. Subsequently, splicing logic (such as NumPy's concatenate interface) is called in the computer's video memory to fuse the extracted and normalized factory digital gene and the dynamic phenotypic features in the feature dimension, which together serve as the feature vector of the graph node. This feature vector takes into account both the static factory condition and the real-time service appearance of the insulator at the underlying logic.

[0026] In the graph, the edges between nodes are used to characterize the stress transmission effect between adjacent insulators under natural conditions. Based on the micrometeorological parameters, the dynamic edge weights between nodes are calculated according to the environmental stress transmission law. First, the three-dimensional physical straight-line distance between the source node and the target node is obtained, and this distance is then subjected to a negative exponential decay calculation based on a distance constant characterizing the spatial diffusion attenuation of pollution. This calculation objectively reflects the nonlinear decreasing concentration distribution law of suspended particulate matter with increasing spatial distance.

[0027] Furthermore, a real-time wind direction vector is extracted from the micrometeorological parameters, and the spatial angle between this real-time wind direction vector and the direction vector pointing from the source node to the target node is calculated. The cosine value of the spatial angle is extracted and truncated to zero. When the target node is located on the upwind side of the source node, the spatial angle exceeds 90 degrees, the cosine value is negative, and the zero-value truncation operator forces it to zero. This processing method strictly cuts off the physical transmission weight under upwind conditions at the algorithm level, ensuring that the information transmission direction of the map fully conforms to the objective physical principle of unidirectional meteorological diffusion.

[0028] Finally, the result of the negative exponential decay operation, the cosine value after zero-value truncation, and the local micro-topography adjustment coefficient assigned based on historical meteorological clustering results are multiplied to obtain the dynamic edge weights that quantify the meteorological transmission law of pollution and environmental erosion between adjacent UHV insulators. The formula for calculating these dynamic edge weights is defined as follows:

[0029] In the formula, For the calculated source node With the target node Dynamic edge weights between them; For the source node With the target node The three-dimensional physical straight-line distance between them; To characterize the spatial diffusion attenuation of contamination, in actual software code implementation, when the three-dimensional physical straight-line distance... When the unit is meters, the The value range is statically set to be between 0.01 and 0.05. For example, under the calibration conditions of a wind tunnel experiment in a plain, A typical value is 0.025, which means that when the node spacing reaches 40 meters, the baseline physical transfer weight will naturally decay to [value missing]. (Approximately 36.8%), thereby constructing a weight cutoff boundary in computer memory that conforms to the Gaussian diffusion empirical model; Real-time wind direction vector and source node Point to target node The spatial angle between the direction vectors; The operator is used to perform the zero-value truncation process to block the backwind transmission; This describes a topology adjustment coefficient dynamically calculated for local micro-topography within a corridor. To clarify the calibration rules for this topology adjustment coefficient, this embodiment discloses a quantitative mapping method based on historical meteorological clustering. Specifically: First, the system extracts historical meteorological and geographical data of the target node j over the past natural year, constructing a three-dimensional feature vector. This feature vector includes: the daily average maximum wind speed, the variance of the wind speed time series (representing the intensity of gust disturbance), and the spatial elevation gradient between the node and its neighboring nodes. Then, the system calls the K-means clustering algorithm to divide the three-dimensional feature vectors of all nodes along the entire route into a preset number (e.g., K=3) of terrain risk clusters, such as "open plain cluster," "mid-mountain disturbance cluster," and "canyon wind gap cluster," respectively. Finally, the system performs quantitative mapping based on the clustering results. For the baseline node belonging to the "open plain cluster," its adjustment coefficient... Strictly calibrated to 1.0. For nodes belonging to other high-risk clusters, the system extracts the average wind speed at the cluster center. and the average wind speed of the baseline plain Compare the results and calculate the adjustment coefficient based on the following formula for wind speed amplification effect: In the formula, This is a scaling factor for contamination sensitivity, set based on fluid dynamics experience (typically 0.5). For example, when target node j is clustered into the "canyon wind gap cluster" using the K-means algorithm, and the historical average wind speed at its cluster center is... Reaching 12 m / s, while the benchmark wind speed in the plains... When the speed is 8 m / s, the system automatically calculates This mapping logic completely eliminates the arbitrariness of subjective human assignment, allowing the airflow acceleration and severe pollution accumulation caused by the funnel effect to be scientifically reflected in the dynamic edge weights of the map through continuous numerical values. The dynamic spatiotemporal map constructed through the above process fully restores the uneven distribution of environmental pressure, providing a network foundation with strict physical meaning for subsequent in-depth extrapolation of high-dimensional data.

[0030] The dynamic spatiotemporal map constructed through the above process fully restores the uneven distribution of environmental pressure, providing a network foundation with strict physical meaning for subsequent in-depth extrapolation of high-dimensional data.

[0031] In this embodiment, based on the constructed dynamic spatiotemporal map, a deep evolutionary analysis of big data is further performed. This analysis process aims to aggregate susceptibility and environmental stress in the spatial dimension, while capturing the cumulative aging effect in the temporal dimension, and finally output the degradation risk score of the UHV insulator.

[0032] The adjacency matrix composed of dynamic edge weights is extracted from the dynamic spatiotemporal graph. To eliminate scale bias caused by differences in connectivity between different nodes, the adjacency matrix with added self-loops is subjected to degree matrix-based symmetric normalization using a graph convolutional network. The operation of adding self-loops ensures that the inherent digital genes and dynamic phenotypic features carried by the nodes are effectively preserved during information aggregation.

[0033] By combining the input features of the current layer or the hidden state representation of the previous convolutional layer, and the learnable weight parameter matrix set in the network, state propagation is performed between network layers through a non-linear activation function. This propagation mechanism uses a normalized adjacency matrix to guide data flow, spatially weighting and aggregating the natural susceptibility features and real-time dynamic phenotypic features corresponding to the digital genes of adjacent nodes.

[0034] After iterative feature extraction via a graph convolutional network, the final spatial topological feature is obtained. This feature vector not only contains the physical state of the individual insulator but also incorporates the combined erosion pressure from the surrounding environment driven by microclimate dynamics. The specific formula for this state propagation and spatial feature aggregation is expressed as follows:

[0035] In the formula, To increase the adjacency matrix of self-loops, it is defined as follows: , For the above dynamic edge weights Extract the original adjacency matrix. It is the identity matrix; For corresponding A degree matrix whose diagonal elements are equal to The sum of all elements in each row; For the first The hidden state representation matrix of the layer, which is in the first layer The input at that time is a feature vector formed by concatenating the node's factory-issued digital gene and dynamic phenotypic features; For graph convolutional networks in the 1st... The learnable weight parameter matrix of the layer; This represents a non-linear activation function used to introduce mapping complexity; That is, the feature tensor output by the current layer and passed to the next layer, which constitutes the spatial topological features through the output of the last layer.

[0036] To analyze the degradation time series of the erosion resistance of insulator materials, the extracted spatial topological features are used as the input data sequence for the current time step and sent to the gate control loop unit along the time axis.

[0037] The gated loop unit integrates update and reset gate structures, using these two logical gating mechanisms to determine the proportion of historical information forgotten or retained on the timeline. The system combines the hidden state of the gated loop unit left over from the previous time step with the spatial topological features of the current input to perform tensor operations, capturing the material aging effect that gradually accumulates over time under the combined effects of continuous electrical stress and environmental stress.

[0038] After completing gating and state fusion, the network updates and outputs the hidden state of the gated cyclic unit at the current time step. This hidden state vector fully encapsulates the evolution trajectory accumulated under specific spatial topology and micrometeorological environment from the time the insulator entered service to the present moment. The feature update logic of the above-mentioned time-series cumulative effect can be expressed by the following formula:

[0039] In the formula, For the current time step The updated gated loop unit is now hidden. For time steps The input spatial topological features; For the previous time step The passed-in gated loop unit is in a hidden state; This represents a set of state update gating function operations within a gated loop unit.

[0040] Finally, the updated hidden state of the gated recurrent unit is input into a multilayer perceptron for nonlinear mapping processing. The multilayer perceptron uses its internal fully connected layers and activation functions to reduce dimensionality, extracting high-level abstract features, and ultimately generating a continuous quantitative index at the output, namely the degradation risk score. This degradation risk score directly characterizes the probability of transient failure of a specific UHV insulator under the current comprehensive operating conditions. The mapping formula is defined as:

[0041] In the formula, The degradation risk score is the final output of the network; This represents the layer-by-layer fully connected forward propagation and nonlinear activation mapping function performed by the multilayer perceptron.

[0042] In the specific data structure and model training deployment, the initial node feature vectors input to the graph convolutional network are... Configured as a one-dimensional tensor, such as a floating-point vector of length 16 (where the first 8 bits represent manufacturing process features mapped from the hash value, and the last 8 bits represent real-time phenotypic data such as temperature, humidity, and leakage current). The hidden layer feature dimension of the gated recurrent unit (GRU). The dimension is set to 32. In a full-link numerical extrapolation example, assuming that at a certain moment the input feature vector is aggregated in the GCN space, the hidden state obtained by updating the GRU is input. for The state vector is nonlinearly dimension-reduced using a multilayer perceptron (containing one layer of fully connected neurons and a sigmoid activation function). The specific code logic involves performing matrix multiplication. The system is then activated, ultimately generating a continuous scalar in the range [0,1] at the output, for example, an output value of 0.825. The system then multiplies this scalar by 100 to obtain a final degradation risk score of 82.5, which is directly used to characterize the probability of transient failure.

[0043] In this invention, after obtaining quantitative evidence of degradation across all nodes, the system automatically generates differentiated operation and maintenance dispatch strategies based on the degradation risk score. First, a warning sensitivity threshold is set on the management control terminal; for example, within a normalized score range of 0 to 100, this threshold is set to 75 points. A real-time streaming data processing engine is then used to continuously monitor the degradation risk score calculated for each node.

[0044] When the monitoring logic identifies a specific node that carries a weak factory-issued digital gene indicating poor resistance to environmental erosion and is located downwind in a local micro-topography, and after continuously receiving environmental pressure from upstream, causes its calculated deterioration risk score to be greater than or equal to the set warning sensitivity threshold, the system triggers the defense mechanism and automatically generates an upgraded inspection work order for that specific high-risk node.

[0045] Based on the generated upgraded inspection work order, the system adaptively allocates resources. It issues fixed-point verification instructions to UHV line maintenance personnel and simultaneously calls upon the nearest UAV nest to perform multimodal collaborative inspection tasks. Simultaneously, through the downlink control link, it dynamically upgrades the sampling frequency of the IoT sensor channel attached to the specific UHV insulator from the usual low-power polling frequency (e.g., set to once per hour to maximize edge node energy savings) to a high-frequency continuous monitoring mode (e.g., set to 1Hz high-frequency continuous sampling once per second). The specific logic of this dynamic upgrade is as follows: when the degradation risk score exceeds the warning sensitivity threshold, the system's underlying control microprocessor receives a downlink interrupt wake-up command, forcibly activating the sensor's high-speed analog-to-digital conversion channel to capture millisecond-level leakage current transient waveform characteristics and partial discharge pulses, thereby achieving adaptive and precise scheduling of power grid maintenance resources driven by predictive risk.

[0046] In a further embodiment of the invention, in addition to adaptive scheduling of front-end on-site operations and maintenance, the system also simultaneously executes predicted failure times to complete the closed-loop operation of cross-enterprise supply chain management. This mechanism directly penetrates the degradation risk of physical terminals to the upstream manufacturing source.

[0047] The system sets an absolute failure threshold in the management backend. This threshold characterizes the critical state at which the insulator material undergoes irreversible insulation breakdown or mechanical fracture. Combining the aforementioned normalized scoring system from 0 to 100, in the actual system backend configuration, this absolute failure threshold can be specifically set to 95 points, serving as a rigid quantitative benchmark for determining the end of the life cycle of UHV insulators. Using the time-series network constructed by the aforementioned gated cyclic units, the degradation evolution trajectory of UHV insulators that have not yet reached this state is extrapolated.

[0048] Specifically, the system utilizes a trained and converged GRU model, using the hidden state of the current time step as initial conditions, to perform autoregressive prediction. It recursively generates degradation risk scores for future sequences with a preset time step (e.g., each iteration represents the next 7 days). To control the risk of error accumulation during autoregressive iterations and avoid infinite extrapolation, the system introduces a dynamic confidence interval and maximum extrapolation limit mechanism. In each prediction step, the system combines the model's historical prediction variance on the validation set to calculate the upper bound of the 95% confidence interval of the current output score, using this as a conservative risk assessment value. Simultaneously, a maximum extrapolation limit is set (e.g., a preset maximum number of iterations of 156, i.e., extrapolating approximately 3 years). The system continues recursively generating data until the upper bound of the confidence interval of the output score is first greater than or equal to the set 95% absolute failure threshold, or the number of iterations reaches the maximum extrapolation limit. The cumulative step length required for this autoregressive iteration is its expected time point, which is used as the predicted failure time. If the upper bound of the score has not yet exceeded the threshold when the maximum extrapolation limit is triggered, the system will mark the predicted failure time status of the node as 'long-term safe' and will not include it in the current predictive material scheduling.

[0049] Subsequently, the system aggregates and statistically analyzes the predicted failure time distribution of insulators across the entire UHV line, fitting it into a dynamic future material demand function. Through the management interface, this future material demand function is directly mapped to virtual inventory capacity demand parameters for specific insulator models. This mapping logic not only performs absolute quantity mapping but also introduces a supply chain redundancy coefficient (e.g., set to 1.15, an empirical compensation constant derived by weighted statistical summation of the target supplier's historical delivery delay rate and arrival inspection defect rate over the past three calendar years, designed to hedge against supply chain fluctuation risks). Its underlying calculation model is: Virtual inventory capacity demand = (Total number of expected failure nodes within a specific future time period × 1.15) + Basic safety stock baseline value. The specific value of the basic safety stock baseline value is derived from the historical meteorological disaster database of the UHV corridor. The system automatically extracts the maximum statistical number of similar insulators damaged by a single extreme disaster (such as a strong typhoon or widespread freezing) in the past five years (e.g., set as a baseline value of 500 pieces), using this as an emergency backup value to ensure the power grid's safety under extreme conditions. Based on this, the system generates a predictive capacity locking instruction containing timestamps and batch quantities for the corresponding suppliers, in order to eliminate the delays in material dispatch caused by traditional post-disaster emergency repairs.

[0050] While completing material scheduling, the system enters the deep process optimization phase. It extracts a set of all UHV insulators whose calculated state has reached or is predicted to reach the absolute failure threshold. For this failure set, the system invokes a frequent itemset association mining algorithm to traverse and calculate the factory-issued digital gene sequences carried by these insulators.

[0051] This data mining algorithm, by setting minimum support and confidence levels (e.g., 0.15 and 0.80 respectively), filters out feature combinations with abnormally high frequencies from a massive amount of micro-manufacturing parameters. This allows it to pinpoint common manufacturing tolerances that accelerate the deterioration of insulators under specific environments, and extract corresponding common gene feature fragments. For example, the system might discover that a batch of frequently failing insulators commonly exhibits the common feature combination of "5% reduction in the vulcanization time of rubber skirt mixing and lower than the standard value in their factory digital genes," and then extracts this as a gene fragment requiring optimization.

[0052] To ensure the legal validity and authenticity of cross-enterprise feedback data, the system invokes a distributed ledger interface and utilizes the hash value previously anchored on the industrial cloud storage architecture to perform tamper-proof traceability verification of the original digital genetic data containing the common genetic feature fragment. The system rigorously compares the hash value of the currently extracted data with the hash value permanently stored in the cloud.

[0053] After verifying the consistency of the results and confirming that the underlying data has not been truncated or tampered with, the system automatically generates a process optimization instruction containing the common gene characteristic fragment and an environmental stress analysis report. This instruction is then sent directly to the R&D port of the corresponding manufacturer's product lifecycle management system via a dedicated network. This closed-loop operation extends the monitoring of phenotypic degradation to source process control, forcing upstream production lines to iteratively correct specific material ratios or injection molding tolerances.

[0054] In this embodiment, based on the same technical concept as the aforementioned method embodiments, the present invention provides a digital operation and maintenance management system for the entire life cycle of ultra-high voltage insulators. This system, through a modular network architecture with coordinated hardware and software, achieves data interaction and logical deduction across the physical environment and the manufacturing source.

[0055] In this embodiment, the UHV insulator full lifecycle digital operation and maintenance management system includes a data acquisition and storage module. This module, based on industrial data management standards, acquires the micro-manufacturing parameters of the UHV insulator within the product lifecycle management system through a distributed data acquisition interface, generating a digital gene reflecting the insulator's condition upon leaving the factory.

[0056] Simultaneously, the data acquisition and storage module performs cryptographic hash calculations on the aforementioned factory-issued digital genes, employing an isolation and tamper-proof strategy of "off-chain storage of raw data and on-chain solidification of hash values." The generated hash values ​​are uploaded to the distributed ledger of the industrial-grade consortium blockchain, establishing an immutable data baseline. Specifically, this module constructs a standardized message containing a unique device identifier, manufacturer identification code, on-chain timestamp, hash value, and digital signature through an application programming interface (API). It then calls a smart contract to broadcast this structured message to independent verification nodes jointly deployed by the power grid, manufacturer, and other parties. After confirmation through practical Byzantine fault tolerance and multi-party endorsement verification, the message is recorded in the blockchain block, thereby establishing an absolutely trustworthy data mapping between the physical site and the cloud. Furthermore, this module, in collaboration with an IoT communication gateway, continuously collects the dynamic phenotypic characteristics and micro-meteorological parameters of the UHV insulators and persistently saves the dynamic phenotypic characteristics to the industrial data storage platform.

[0057] The system also includes a dynamic graph construction module. This module performs mathematical visualization of spatial topological relationships during the big data processing stage, constructing a dynamic spatiotemporal graph using the UHV corridor as the physical space. This module concatenates the extracted factory-generated digital genes with dynamic phenotypic features using tensors, resulting in feature vectors for the graph nodes.

[0058] When establishing physical connections between nodes, the dynamic map construction module combines micro-meteorological parameters to calculate dynamic edge weights between nodes based on the environmental stress transmission law. Specifically, this module performs a negative exponential decay operation on the three-dimensional physical straight-line distance between nodes, truncates the cosine of the angle between the extracted wind direction vector and the spatial connection direction vector to zero, and multiplies the above calculation results with the local micro-topography adjustment coefficient. The weight calculation formula built into this dynamic map construction module is specifically expressed as follows:

[0059] In the formula, The source node for the output of this module. With the target node Dynamic edge weights between them; This refers to the three-dimensional physical straight-line distance. A distance constant characterizing the rate of pollution diffusion attenuation; The spatial angle between the real-time wind direction vector and the direction vectors of the two nodes; The operator is used to perform zero-value truncation to cut off the weights passed upwind; This refers to the local micro-topography adjustment coefficient preset for the system.

[0060] The system further includes an evolutionary deduction and analysis module. Based on the aforementioned dynamic spatiotemporal graph, this module aggregates susceptibility and environmental pressure across spatial dimensions. It extracts the adjacency matrix from the graph and performs degree-matrix-based symmetric normalization on the adjacency matrix containing self-loops using a built-in graph convolutional network. Its logical operations for state propagation and feature aggregation are as follows:

[0061] In the formula, To increase the adjacency matrix of self-loops; This is the degree matrix corresponding to it; For the network The hidden state representation matrix of the layer; For this module in the The learnable weight parameter matrix initialized for the layer; It is a non-linear activation function; The tensor is used to pass the output spatial topological features to the next layer.

[0062] To clarify the specific network structure, in this embodiment, the total number of layers in the graph convolutional network is set to 2 (i.e., the maximum number of iteration layers). The initial input graph node feature matrix of the system. The feature dimension of the (i.e., the vector concatenated from digital genes and phenotypic features) is configured to be 16-dimensional; after the first layer of graph convolution calculation, the output hidden layer feature matrix is... The dimension mapping is 64-dimensional; after a second layer of graph convolution calculation, the final output is a spatial topological feature matrix. The dimensionality is reduced to 32 dimensions to accurately align with the input dimension requirements of subsequent gated recurrent units. Simultaneously, for state propagation between network layers, the nonlinear activation function... Specifically, the ReLU function is used to effectively alleviate the gradient vanishing problem in deep graph propagation and accelerate model training convergence.

[0063] Subsequently, the evolutionary analysis module inputs the aggregated spatial topological features into the gated loop unit, and combines the hidden state from the previous time step to capture the cumulative aging effect of the time dimension. The hidden state feature update logic is expressed as follows:

[0064] In the formula, and These are the hidden states of the gated loop unit updated at the current and previous time steps, respectively. For time steps The input spatial topological features. Next, this module will... The data is fed into a multilayer perceptron for nonlinear mapping processing, and the final output is a quantified degradation risk score, specifically expressed as:

[0065] In the formula, This refers to the degradation risk scoring index, which characterizes the transient failure probability of a specific insulator. To further disclose the specific network topology of the multilayer perceptron (MLP), this embodiment sets the MLP as a two-layer feedforward neural network with one hidden layer. Specifically, the first layer is a hidden layer with 16 neurons, receiving 32-dimensional hidden states from the output of the GRU. The system employs the ReLU activation function for nonlinear extraction of higher-order features. The second layer is the output layer, configured with one neuron. The Sigmoid activation function is used to forcibly map the hidden layer features to the open interval (0,1). The system then multiplies this output value by 100 in the underlying code to generate a final quantized degradation risk score within the range of 0 to 100. This explicit hierarchical and dimensional setting ensures the transparency and engineering reproducibility of the feature reduction process.

[0066] The system finally includes an operation and maintenance (O&M) module. This module reads the degradation risk score sequence and, based on the warning sensitivity thresholds set in the system backend, automatically generates differentiated O&M dispatch strategies. When it detects that a downwind node is experiencing environmental erosion pressure and exceeds the warning sensitivity threshold, this module automatically distributes an upgraded inspection work order and dynamically increases the sampling frequency of the corresponding node's IoT sensor channel via downlink control commands.

[0067] Meanwhile, the operation and maintenance and collaboration module extrapolates the expected time point when each node will break through the absolute failure threshold through the internal time series analysis network, transforms it into a future material demand function and maps it into virtual warehouse capacity demand parameters, and sends a predictive capacity locking instruction to the supplier data interface.

[0068] In the traceability process, the operation and maintenance and collaboration module uses a frequent itemset association mining algorithm to traverse the data base for insulator sets that have reached the failure threshold, locating common genetic feature fragments that induce accelerated degradation. This module further retrieves hash values ​​fixed in the industrial cloud storage architecture to perform hash comparison verification on the original digital genetic data. After confirming that the underlying data has not been truncated or tampered with, it automatically sends process optimization instructions containing the common genetic feature fragment across the network gateway to the product lifecycle management system, completing the design optimization closed loop from the physical terminal to the manufacturing source.

[0069] In this embodiment, the present invention provides an electronic device designed to provide the underlying computing environment for executing the aforementioned digital operation and maintenance management method for the entire lifecycle of ultra-high voltage insulators. Physically, this electronic device is an industrial-grade server platform deployed in a cloud data center or edge computing node, and its internal hardware topology mainly consists of a processor, memory, communication interface, and communication bus.

[0070] The aforementioned processor, memory, and communication interface are electrically connected and perform low-level data routing via a communication bus. The communication bus provides a standardized broadband data transmission channel, ensuring that under high-concurrency conditions, multimodal tensor data, such as micro-meteorological parameters, dynamic phenotypic characteristics, and factory-generated digital genes, transmitted from the UHV corridor can be exchanged and coordinated at high speed without blocking between various hardware components.

[0071] The memory is used to store various preset characteristic baselines and topology adjustment coefficients required for the operation of computer programs and systems. In actual hardware resource configurations, this memory is typically composed of a combination of high-speed random access memory and non-volatile storage media. It is not only responsible for storing the underlying instruction set for implementing the aforementioned dynamic spatiotemporal map construction and evolutionary analysis, but also serves as the physical entity carrier of the industrial data storage platform, persistently storing the hash values ​​anchored to the chain and massive amounts of time-series state slice data.

[0072] The processor, as the core instruction execution engine of the electronic device, is responsible for reading and running the computer program stored in the memory. When the processor is running at full speed, its built-in arithmetic logic unit and tensor calculation core strictly follow the mathematical principles revealed in the aforementioned embodiments to execute dynamic edge weight extraction logic, including negative exponential decay operations of spatial distance and zero-value truncation of the cosine of the wind direction angle. Relying on its powerful parallel computing capabilities, the processor quickly completes the spatiotemporal cascaded state propagation of graph convolutional networks and gated recurrent units, thereby outputting a precise quantitative degradation risk score at the physical hardware level and generating differentiated operation and maintenance strategies for cross-domain collaboration.

[0073] The communication interface is configured as a multi-protocol compatible network throughput hub, responsible for establishing wide-area connections with external business systems. On one hand, the interface connects downlink to IoT sensor nodes and meteorological monitoring base stations distributed on power transmission towers to receive high-frequency sampled environmental and phenotypic flow data; on the other hand, it connects uplink to cross-enterprise data gateways, establishing encrypted communication tunnels with upstream manufacturers' product lifecycle management systems to ensure that predictive capacity locking instructions and process optimization instructions based on traceability verification can be securely and losslessly delivered across domains.

[0074] In another embodiment of the present invention, a computer-readable storage medium is provided. This computer-readable storage medium contains a set of computer-executable instructions, which can be specifically embodied in an optical disc, magnetic hard disk, solid-state drive, or a cloud-based distributed read-only storage node. When this set of instructions is read by a computer device and loaded into memory for execution, it can completely drive the computer device to reproduce all data processing, network graph deduction, and supply chain closed-loop feedback processes of the aforementioned digital operation and maintenance management method for the entire lifecycle of ultra-high voltage insulators.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that equivalent modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the logical framework and protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. A digital operation and maintenance management method for the entire life cycle of ultra-high voltage insulators, characterized in that, include: The micro-manufacturing parameters of the UHV insulator are obtained to generate a factory digital gene. The hash value of the factory digital gene and the equipment metadata are stored on the blockchain, and the original data is stored off the blockchain. The dynamic phenotypic characteristics and micro-meteorological parameters of the UHV insulators were collected. A dynamic spatiotemporal map of the ultra-high voltage corridor is constructed. The factory-issued digital genes and the dynamic phenotypic features are spliced ​​together as the feature vectors of the map nodes, and the dynamic edge weights between the map nodes are calculated based on the micro-meteorological parameters. Based on the dynamic spatiotemporal map, spatiotemporal features are extracted, and the degradation risk score of the UHV insulator is output. Based on the degradation risk score, an operation and maintenance dispatch strategy is generated, the failure time of the UHV insulator is predicted and materials are dispatched, and common characteristics that cause degradation are extracted to generate process optimization instructions.

2. The method for digital operation and maintenance management of ultra-high voltage insulators throughout their entire life cycle as described in claim 1, characterized in that, The calculation of dynamic edge weights between map nodes based on the micrometeorological parameters includes: Obtain the spatial distance between the source node and the target node, and perform a negative exponential decay operation on the spatial distance; Extract the real-time wind direction vector from the micro-meteorological parameters, calculate the spatial angle between the real-time wind direction vector and the direction vector from the source node to the target node, extract the cosine value of the spatial angle and perform zero-value truncation. The dynamic edge weight is obtained by multiplying the result of the negative exponential decay operation, the cosine value after zero-value truncation, and the micro-topography adjustment coefficient assigned based on the meteorological clustering result.

3. The digital operation and maintenance management method for the entire life cycle of ultra-high voltage insulators according to claim 1, characterized in that, Based on the dynamic spatiotemporal map, spatiotemporal features are extracted, and a degradation risk score for the UHV insulator is output, including: The adjacency matrix of the dynamic spatiotemporal graph is extracted, and the feature vectors of the adjacency matrix and graph nodes are spatially aggregated through a graph convolutional network to obtain spatial topological features. The spatial topological features are input into the gated loop unit, and the hidden state of the gated loop unit at the current time step is updated by combining the hidden state of the previous time step. The updated hidden state of the gated loop unit is input into the multilayer perceptron, and the degradation risk score is output.

4. The digital operation and maintenance management method for the entire life cycle of ultra-high voltage insulators according to claim 1, characterized in that, An operation and maintenance dispatch strategy is generated based on the aforementioned degradation risk score, including: Continuously monitor the degradation risk score corresponding to the map nodes; When the degradation risk score of a specific node is greater than or equal to the preset warning sensitivity threshold, an upgraded inspection work order is generated for that node. Based on the upgraded inspection work order, multimodal collaborative inspection tasks are assigned, and the sampling frequency of the corresponding node sensor channel is dynamically increased.

5. The method for digital operation and maintenance management of ultra-high voltage insulators throughout their entire life cycle as described in claim 1, characterized in that, Predicting the failure time of the ultra-high voltage insulators and scheduling materials, extracting common characteristics that cause degradation to generate process optimization instructions, including: The failure time of each UHV insulator exceeding the absolute failure threshold is predicted by using a time-series network. Based on the failure time distribution, material demand parameters are generated, and a capacity lock instruction is sent to the supplier. Extract the set of UHV insulators that have reached the absolute failure threshold, and use the frequent itemset association mining algorithm to locate the common digital gene fragments that cause the degradation; The hash value stored on the blockchain is used to verify the original data containing the common digital gene fragment. If the verification is successful, the process optimization instruction is sent to the corresponding product lifecycle management system.

6. A digital operation and maintenance management system for the entire life cycle of ultra-high voltage insulators, employing the digital operation and maintenance management method for the entire life cycle of ultra-high voltage insulators as described in any one of claims 1-5, characterized in that, include: The data acquisition and storage module is used to acquire the micro-manufacturing parameters of the UHV insulator to generate the factory digital gene, store the hash value of the factory digital gene and the equipment metadata on the chain, and store the original data off the chain. At the same time, it collects the dynamic phenotypic characteristics and micro-meteorological parameters of the UHV insulator. The dynamic map construction module is used to construct a dynamic spatiotemporal map of the UHV corridor. It splices the factory-issued digital genes with the dynamic phenotypic features as the feature vectors of the map nodes, and calculates the dynamic edge weights between map nodes based on the micro-meteorological parameters. The evolutionary analysis module is used to extract spatiotemporal features based on the dynamic spatiotemporal map and output the degradation risk score of the UHV insulator. The operation and maintenance and collaboration module is used to generate operation and maintenance dispatch strategies based on the degradation risk score, predict the failure time of the UHV insulators and perform material scheduling, and extract common features that cause degradation to generate process optimization instructions.

7. The UHV insulator full life cycle digital operation and maintenance management system according to claim 6, characterized in that, When the dynamic graph construction module calculates the dynamic edge weights between graph nodes based on the micrometeorological parameters, it is specifically used for: Obtain the spatial distance between the source node and the target node, and perform a negative exponential decay operation on the spatial distance; Extract the real-time wind direction vector from the micro-meteorological parameters, calculate the spatial angle between the real-time wind direction vector and the direction vector from the source node to the target node, extract the cosine value of the spatial angle and perform zero-value truncation. The dynamic edge weight is obtained by multiplying the result of the negative exponential decay operation, the cosine value after zero-value truncation, and the micro-topography adjustment coefficient assigned based on the meteorological clustering result.

8. The UHV insulator full life cycle digital operation and maintenance management system according to claim 6, characterized in that, When the evolutionary analysis module extracts spatiotemporal features based on the dynamic spatiotemporal map and outputs the degradation risk score of the UHV insulator, it is specifically used for: The adjacency matrix of the dynamic spatiotemporal graph is extracted, and the feature vectors of the adjacency matrix and graph nodes are spatially aggregated through a graph convolutional network to obtain spatial topological features. The spatial topological features are input into the gated loop unit, and the hidden state of the gated loop unit at the current time step is updated by combining the hidden state of the previous time step. The updated hidden state of the gated loop unit is input into the multilayer perceptron, and the degradation risk score is output.

9. The UHV insulator full life cycle digital operation and maintenance management system according to claim 6, characterized in that, When generating an operation and maintenance dispatch strategy based on the degradation risk score, the operation and maintenance and collaboration module is specifically used for: Continuously monitor the degradation risk score corresponding to the map nodes; When the degradation risk score of a specific node is greater than or equal to the preset warning sensitivity threshold, an upgraded inspection work order is generated for that node. Based on the upgraded inspection work order, multimodal collaborative inspection tasks are assigned, and the sampling frequency of the corresponding node sensor channel is dynamically increased.

10. The UHV insulator full life cycle digital operation and maintenance management system according to claim 6, characterized in that, When the operation and maintenance and coordination module predicts the failure time of the UHV insulators, performs material scheduling, and extracts common characteristics that cause degradation to generate process optimization instructions, it is specifically used for: The failure time of each UHV insulator exceeding the absolute failure threshold is predicted by using a time-series network. Based on the failure time distribution, material demand parameters are generated, and a capacity lock instruction is sent to the supplier. Extract the set of UHV insulators that have reached the absolute failure threshold, and use the frequent itemset association mining algorithm to locate the common digital gene fragments that cause the degradation; The hash value stored on the blockchain is used to verify the original data containing the common digital gene fragment. If the verification is successful, the process optimization instruction is sent to the corresponding product lifecycle management system.

Images