Substation grounding grid corrosion state evaluation method and system based on deep learning

Abstract

The invention relates to the related technical field of substations, in particular to a substation grounding grid corrosion state evaluation method and system based on deep learning, and the method comprises the steps: interactively obtaining the structure data of a substation grounding grid, carrying out the graph convolution network modeling according to the structure data, and building a twin model through the learning of the spatial relation of each grounding grid; acquiring operation data of a transformer substation grounding grid; synchronously obtaining soil physicochemical property data; and updating the operation data and the soil physicochemical property data to the twinborn model, and establishing a corrosion state prediction result. The technical problems that static evaluation based on data of a single time point and a single node cannot identify the influence of local corrosion on the overall grounding performance and the evaluation precision of the corrosion state of the grounding grid of the transformer substation is limited are solved, and operation data and soil physicochemical property data are collaboratively updated to a twin model. Mapping the physical structure of the grounding grid and the spatial diffusion effect of local corrosion, and carrying out the precise evaluation of the corrosion state of the grounding grid of the transformer substation.

Description

Technical Field

[0001] This invention relates to the technical field of substations, specifically to a method and system for assessing the corrosion status of substation grounding grids based on deep learning. Background Technology

[0002] Grounding grids, buried underground for extended periods, are susceptible to corrosion due to factors such as soil pH, conductivity, and humidity. This corrosion leads to increased grounding resistance and reduced conductor cross-sections, potentially causing fault current obstruction, equipment burnout, and other accidents that threaten grid stability. However, regular excavation and inspection methods cannot meet early warning requirements. Furthermore, current assessments of grounding grid corrosion status in power plants are mostly static assessments based on data from a single time point and node, failing to identify the impact of localized corrosion on overall grounding performance and making it difficult to predict future corrosion trends. This results in low accuracy and weak predictive capabilities in corrosion status assessments, hindering the early detection of potential risks and forcing reactive measures after grounding grid corrosion faults occur, thus increasing potential grid safety hazards.

[0003] In summary, existing technologies suffer from limitations in the accuracy of substation grounding grid corrosion status assessment due to the inability to identify the impact of localized corrosion on overall grounding performance when using static assessments based on data from a single time point or node. Summary of the Invention

[0004] This application provides a method and system for assessing the corrosion status of substation grounding grids based on deep learning. It aims to solve the technical problem that existing technologies, which rely on static assessments based on data from a single time point and a single node, cannot identify the impact of localized corrosion on overall grounding performance, thus limiting the accuracy of substation grounding grid corrosion status assessment.

[0005] In view of the above problems, the technical solution to achieve the present application is as follows: In a first aspect, this application provides a deep learning-based method for assessing the corrosion status of substation grounding grids. The method includes: interactively acquiring structural data of the substation grounding grid; performing graph convolutional network modeling based on the structural data; constructing a twin model by learning the spatial relationships between each grounding grid; collecting operational data of the substation grounding grid based on a sensor network pre-deployed in the substation grounding grid, the operational data including current data, voltage data, and temperature data; simultaneously acquiring soil physicochemical property data, including soil pH, conductivity, humidity, salinity, and soil temperature; updating the operational data and the soil physicochemical property data to the twin model; and performing spatiotemporal correlation analysis of changes in electrical parameters and soil physicochemical properties through a long short-term memory analysis layer integrated within the twin model to establish a corrosion status prediction result.

[0006] Preferably, the feature extraction channel within the long short-term memory analysis layer is used to extract local features of changes in electrical parameters and soil physicochemical properties, and local feature extraction results are established. The local feature extraction results are sent to the data analysis channel within the long short-term memory analysis layer. After the data analysis channel adaptively configures the memory depth and time window size, the temporal coupling capture of the local feature extraction results is performed. The temporal coupling capture results and the spatial relationship of the grounding grid are used to perform spatiotemporal joint analysis and establish spatiotemporal correlation analysis results.

[0007] Preferably, an adaptive learning mechanism is activated to evaluate the real-time change rate and data stability of the data within the local feature extraction results, and an evaluation result is established. After configuring the corresponding memory depth and time window size according to the evaluation result, a hierarchical time window coupling analysis strategy is used to perform local-global coupling fusion to complete temporal coupling capture.

[0008] Preferably, after extracting the short-term and long-term spatiotemporal features from the spatiotemporal correlation analysis results, an input feature set is established; the deep fusion channel within the long short-term memory analysis layer is activated, and the input feature set is input into the deep fusion channel to perform corrosion state fusion analysis under multi-source data verification. The multi-source data verification includes spatial multi-source data verification and temporal data verification at the same location node, as well as joint location verification at different location nodes; the corrosion state prediction result is output according to the deep fusion channel.

[0009] Preferably, historical operating data of the substation grounding grid is acquired, and time-series regression analysis is performed based on the historical operating data and the corrosion state prediction results to establish authentication prediction identifiers and anomaly prediction identifiers; the authentication prediction identifiers are used to perform prediction enhancement updates on the corrosion state prediction results to establish a first update result; after configuring a verification strategy using the anomaly prediction identifiers, verification detection is performed, and the corrosion state prediction results are verified and updated based on the verification detection results to establish a second update result; corrosion state is assessed based on the first update result and the second update result.

[0010] Preferably, the time-series task dataset of the substation grounding grid is obtained; anomaly warning levels are matched based on the time-series task dataset and the corrosion state prediction results to establish a prediction warning signal; and the prediction warning signal is used for anomaly reporting management.

[0011] Preferably, the local edge node is used to perform data analysis and preprocessing of the running data and the soil physicochemical property data to establish an edge processing dataset; after the edge processing dataset is authenticated, the edge processing dataset is uploaded to the twin model in the cloud.

[0012] In a second aspect, this application provides a deep learning-based substation grounding grid corrosion status assessment system, wherein the system comprises: a first data acquisition module: interactively acquiring structural data of the substation grounding grid, performing graph convolutional network modeling based on the structural data, and constructing a twin model by learning the spatial relationships of each grounding grid; an operational data acquisition module: acquiring operational data of the substation grounding grid based on a sensor network pre-deployed in the substation grounding grid, the operational data including current data, voltage data, and temperature data; a second data acquisition module: synchronously acquiring soil physicochemical property data, the soil physicochemical property data including soil pH, conductivity, humidity, salinity, and soil temperature; and a status assessment module: updating the operational data and the soil physicochemical property data to the twin model, performing spatiotemporal correlation analysis of electrical parameters and soil physicochemical property changes through a long short-term memory analysis layer integrated in the twin model, and establishing corrosion status prediction results.

[0013] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-mentioned deep learning-based substation grounding grid corrosion status assessment method.

[0014] In summary, one or more technical solutions provided in this application achieve the technical effect of collaboratively updating operational data and soil physicochemical property data to a twin model, mapping the physical structure of the grounding grid and the spatial diffusion effect of local corrosion, and conducting accurate assessment of the corrosion status of the substation grounding grid. Attached Figure Description

[0015] Figure 1 This application provides a flowchart illustrating a deep learning-based method for assessing the corrosion status of substation grounding grids.

[0016] Figure 2 This application provides a schematic diagram of the structure of a substation grounding grid corrosion status assessment system based on deep learning.

[0017] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0018] Explanation of reference numerals in the attached drawings: First data acquisition module M100, running data acquisition module M200, second data acquisition module M300, status evaluation module M400, bus 300, receiver 301, processor 302, transmitter 303, memory 304, bus interface 305. Detailed Implementation

[0019] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0020] Example 1: The present application will be described in detail below with reference to the accompanying drawings, as follows... Figure 1 As shown, this application provides a deep learning-based method for assessing the corrosion status of substation grounding grids, wherein the method includes: S1: After interactively acquiring the structural data of the substation grounding network, perform graph convolutional network modeling based on the structural data, and construct a twin model by learning the spatial relationship of each grounding network; S2: Collect the operating data of the substation grounding network based on the sensor network pre-deployed in the substation grounding network, the operating data including current data, voltage data, and temperature data.

[0021] Specifically, interactive acquisition refers to actively acquiring the structural data of the substation grounding network through an interactive system or interface, achieved through network communication, field equipment reading, etc., ensuring the accuracy and real-time nature of the acquired data; Graph convolutional networks are a deep learning model that uses graph convolutional networks to model the structural data of the substation grounding network, learning the spatial relationships between nodes such as grounding electrodes and connection points in the grounding network to construct a model that reflects the topology of the grounding network; Twin models are used to simulate the corrosion state of the substation grounding network, achieving dynamic assessment of the corrosion state through interaction with the structural and operational data of the actual grounding network; Pre-deployed sensor networks refer to the sensor network pre-deployed in the substation grounding network to collect operational data of the grounding network, including current, voltage, and temperature, which is the basis for assessing the corrosion state of the grounding network.

[0022] Execution steps: First, through an interactive system with the substation grounding grid, acquire structural data of the grounding grid, including the location, connection method, and length of the grounding electrodes. This structural data forms the basis for constructing a graph convolutional network model. Second, using the acquired structural data, construct a graph convolutional network model. Third, by learning the spatial relationships between nodes in the grounding grid, construct a twin model that reflects the topology of the grounding grid. These spatial relationships include the distances between nodes. Finally, a sensor network pre-deployed in the substation grounding grid collects real-time operational data, including current, voltage, and temperature. This operational data reflects the actual operating status of the grounding grid and is an important basis for assessing corrosion conditions.

[0023] Preferably, graph convolutional network modeling accurately maps the physical structure of the grounding grid, enabling the analysis of the impact of localized corrosion on overall performance. Specifically, the operational data collected by the sensor network provides real-time evidence for corrosion status assessment. This data includes electrical parameters such as current and voltage, as well as environmental parameters such as temperature, comprehensively reflecting the operating status of the grounding grid. Furthermore, by analyzing changes in current data, it can be determined whether the grounding grid experiences resistance increases due to localized corrosion; by analyzing temperature data, it can be determined whether the soil environment affects grounding grid corrosion; and by combining graph convolutional network modeling with the spatial topology of the grounding grid, accurate assessment of the corrosion status of each grounding electrode in the grounding grid can be achieved. Furthermore, by analyzing abnormal current data of a certain grounding electrode, combined with the location and connection relationship of that grounding electrode, it can be determined that the grounding electrode may have localized corrosion problems, thereby affecting the performance of the entire grounding grid.

[0024] S3: Synchronously acquire soil physicochemical property data, including soil pH, electrical conductivity, humidity, salinity, and soil temperature; S4: Update the running data and the soil physicochemical property data to the twin model, and perform spatiotemporal correlation analysis of electrical parameters and soil physicochemical property changes through the long short-term memory analysis layer integrated in the twin model to establish corrosion state prediction results.

[0025] Specifically, synchronous acquisition of soil physicochemical property data refers to acquiring soil physicochemical property data while collecting substation grounding grid operation data. Soil physicochemical properties include soil pH, electrical conductivity, humidity, salinity, and soil temperature. The long short-term memory analysis layer is integrated into the twin model to analyze the spatiotemporal correlation between electrical parameters and changes in soil physicochemical properties. Spatiotemporal correlation analysis refers to analyzing the correlation between data in both time and space. Through the long short-term memory analysis layer, spatiotemporal correlation analysis is performed to analyze the interaction between electrical parameters and soil physicochemical property data in time and space, in order to establish corrosion state prediction results.

[0026] Execution steps: A pre-deployed sensor network synchronously collects soil physicochemical property data such as pH, conductivity, humidity, salinity, and soil temperature. This data, along with the grounding grid's operational data, provides a comprehensive basis for corrosion status assessment. The collected operational and soil physicochemical property data are then updated into a twin model. By receiving this data in real time, the twin model dynamically reflects the actual operating status of the grounding grid and environmental conditions. A long short-term memory (LSTM) analysis layer integrated within the twin model performs spatiotemporal correlation analysis on the updated data to establish corrosion status prediction results. Preferably, by synchronously acquiring soil physicochemical property data, the impact of the soil environment on grounding grid corrosion can be comprehensively reflected. Furthermore, soil pH and electrical conductivity are important factors affecting metal corrosion. By monitoring these parameters in real time, the corrosion status can be assessed more accurately. Updating operational data and soil physicochemical property data to the twin model ensures that the model can reflect the actual state of the grounding grid in real time. The dynamic update mechanism can capture changes in the corrosion status in a timely manner, improving the accuracy of the assessment. The long short-term memory analysis layer can analyze the temporal and spatial correlation between electrical parameters and soil physicochemical property data. Furthermore, by analyzing the temporal changes in soil moisture and grounding grid current data, the cumulative impact of seasonal fluctuations in soil moisture on corrosion can be discovered. By analyzing the spatial topological relationship of the grounding grid, the spatial diffusion effect of local corrosion can be identified. Spatiotemporal correlation analysis can achieve dynamic prediction of the corrosion status, improving the accuracy and predictability of the assessment.

[0027] Furthermore, by performing spatiotemporal correlation analysis of changes in electrical parameters and soil physicochemical properties through a long short-term memory analysis layer integrated within the twin model, the method of this application includes: Local features of changes in electrical parameters and soil physicochemical properties are extracted using the feature extraction channel within the long short-term memory analysis layer, and local feature extraction results are established. The local feature extraction results are then sent to the data analysis channel within the long short-term memory analysis layer. After the data analysis channel adaptively configures the memory depth and time window size, temporal coupling capture of the local feature extraction results is performed. The temporal coupling capture results and the spatial relationship of the grounding grid are used to perform spatiotemporal joint analysis, and spatiotemporal correlation analysis results are established.

[0028] Specifically, in the Long Short-Term Memory (LSTM) analysis layer, the feature extraction channel is responsible for extracting local features from the input electrical parameters and soil physicochemical properties data. These local features are representative patterns or trends in the data, reflecting local changes. In the LTM analysis layer, the data analysis channel is responsible for further analyzing the extracted local features. By adaptively configuring the memory depth and time window size, it captures the temporal changes of local features, i.e., the evolution of data over time. Temporal coupling capture refers to capturing the coupling relationship of local features over time through the data analysis channel, reflecting the mutual influence and correlation between data at different points in time. Spatiotemporal joint analysis refers to combining the temporal coupling capture results with the spatial relationship of the grounding grid for comprehensive analysis, considering the temporal changes, spatial distribution, and interrelationships of the data, thereby establishing comprehensive spatiotemporal correlation analysis results.

[0029] Execution steps: Utilizing the feature extraction channel within the Long Short-Term Memory (LSTM) analysis layer, local feature extraction is performed on electrical parameters including current, voltage, and temperature, as well as soil physicochemical properties including pH, conductivity, humidity, salinity, and soil temperature. Furthermore, the feature extraction channel identifies abrupt changes or trend variations in the current data, and seasonal fluctuations in the soil moisture data. The extracted local feature results are sent to the data analysis channel, which adaptively configures the memory depth and time window size to perform temporal coupling capture of the local feature extraction results. Further, through adaptive configuration, the correlation between seasonal fluctuations in soil moisture and long-term changes in grounding grid current data can be captured. Spatiotemporal joint analysis is performed using the temporal coupling capture results and the spatial relationship of the grounding grid to establish spatiotemporal correlation analysis results. Furthermore, combining the spatial topology of the grounding grid, the spatial diffusion effect of localized corrosion is analyzed, and how this spatial diffusion effect affects the overall performance of the grounding grid.

[0030] Furthermore, after adaptively configuring the memory depth and time window size using the data analysis channel, the method of this application performs temporal coupling capture of local feature extraction results, including: An adaptive learning mechanism is activated to evaluate the real-time change rate and data stability of data within the local feature extraction results, and an evaluation result is established. After configuring the corresponding memory depth and time window size according to the evaluation result, a hierarchical time window coupling analysis strategy is used to perform local-global coupling fusion to complete temporal coupling capture.

[0031] Specifically, adaptive learning mechanisms refer to mechanisms that allow a system to automatically adjust its parameters and behavior based on the characteristics of the input data. This is used to evaluate the rate of change and stability of data within the local feature extraction results in real time, and to automatically configure memory depth and time window size. Real-time change rate refers to the speed at which data changes over time. This is obtained by calculating the difference or derivative of data at adjacent time points, thus identifying rapid changes or abrupt changes in the data. Data stability evaluation assesses the stability of data over time by calculating statistical measures such as standard deviation, variance, or moving average to determine the degree of data fluctuation and stability. In Long Short-Term Memory (LSTM) networks, memory depth refers to the range of past information the network can remember. Generally, a larger memory depth allows the network to capture longer-term dependencies but also increases computational complexity. Time window size refers to the length of the time period used to analyze data in time series analysis. Generally, a larger time window can capture longer-term trends but may ignore short-term details. Hierarchical time window coupling analysis strategies involve analyzing data through different levels of time windows and coupling and fusing these analysis results. This allows for the simultaneous capture of short-term and long-term changes, achieving comprehensive analysis of both local and global aspects.

[0032] Execution steps: First, activate the adaptive learning mechanism within the Long Short-Term Memory (LSTM) analysis layer to evaluate the real-time rate of change and data stability of the data within the local feature extraction results. Then, assess the real-time rate of change by calculating the difference between data points at adjacent time points and evaluate data stability by calculating the standard deviation of the data, establishing an evaluation result. This evaluation result will serve as the basis for subsequent configuration of memory depth and time window size. The memory depth and time window size are adaptively configured. Furthermore, if the data change rate is fast and stability is low, a smaller time window is needed to capture short-term changes; if the data change rate is slow and stability is high, a larger time window is configured to capture long-term trends. Using the configured memory depth and time window size, a hierarchical time window coupling analysis strategy is executed. Analysis is performed through different levels of time windows, and these analysis results are coupled and fused to achieve comprehensive local and global analysis, completing temporal coupling capture.

[0033] Furthermore, the method used in this application to establish corrosion state prediction results includes: After extracting the short-term and long-term spatiotemporal features from the spatiotemporal correlation analysis results, an input feature set is established. The deep fusion channel within the long short-term memory analysis layer is activated, and the input feature set is input into the deep fusion channel to perform corrosion state fusion analysis under multi-source data verification. The multi-source data verification includes spatial multi-source data verification and temporal data verification at the same location node, as well as joint location verification at different location nodes. The corrosion state prediction result is output based on the deep fusion channel.

[0034] Specifically, short-term spatiotemporal characteristics refer to spatially related data features within a short timeframe, reflecting changes in corrosion status in the short term, specifically the impact of short-term fluctuations in soil moisture on corrosion. Long-term spatiotemporal characteristics refer to spatially related data features within a longer timeframe, reflecting long-term trends, specifically the cumulative impact of seasonal changes in soil moisture on corrosion. The input feature set refers to combining short-term and long-term spatiotemporal characteristics into a single feature set, serving as input data for subsequent analysis. In the Long Short-Term Memory (LSTM) analysis layer, the deep fusion channel is responsible for comprehensively analyzing and fusing multi-source data, ensuring the accuracy of corrosion status prediction results by verifying the reliability of different data sources. Multi-source data verification refers to verifying data from different sensors or locations to ensure data accuracy and reliability, including spatial multi-source data verification at the same location node, time-series data verification, and joint location verification at different locations. The corrosion status prediction result refers to the predicted result regarding the corrosion status of the substation grounding grid obtained after analysis by the deep fusion channel.

[0035] Execution steps: Extract short-term and long-term spatiotemporal features from the spatiotemporal correlation analysis results. Combine the extracted short-term and long-term spatiotemporal features into an input feature set, which contains all the key information required for corrosion state prediction. Activate the deep fusion channel within the Long Short-Term Memory (LSTM) analysis layer and input the input feature set into the deep fusion channel. In the deep fusion channel, validate the input multi-source data, including: spatial multi-source data validation at the same location node, further validating the consistency of multiple sensor data at the same location; time-series data validation, further validating the continuity and consistency of the data over time; joint location validation of different location nodes, further validating the correlation and consistency between sensor data at different locations; analyze the data according to the deep fusion channel and output the corrosion state prediction results.

[0036] Furthermore, to establish corrosion state prediction results, the method of this application also includes: Historical operating data of the substation grounding grid is acquired. Time-series regression analysis is performed based on the historical operating data and the corrosion state prediction results to establish authentication prediction identifiers and anomaly prediction identifiers. The authentication prediction identifiers are used to enhance and update the corrosion state prediction results, establishing a first update result. After configuring a verification strategy using the anomaly prediction identifiers, verification detection is performed. The corrosion state prediction results are then verified and updated based on the verification detection results, establishing a second update result. The corrosion state is assessed based on the first update result and the second update result.

[0037] Specifically, historical operating data refers to the data accumulated by the substation grounding grid during its past operation, including electrical parameters such as current, voltage, and temperature, as well as soil physicochemical properties data, used to analyze the historical trend of corrosion status; time-series regression analysis is used to analyze the relationship between historical operating data and corrosion status prediction results to establish a prediction model; certified prediction identifiers refer to corrosion status prediction results that are consistent with historical data and verified through time-series regression analysis, used to indicate the reliability of the prediction results; abnormal prediction identifiers refer to corrosion status prediction results that are inconsistent with historical data and discovered through time-series regression analysis, used to indicate that the prediction results have errors or anomalies; prediction enhancement and updating refers to optimizing and updating the corrosion status prediction results using certified prediction identifiers to improve the accuracy and reliability of the predictions; verification strategies refer to the detection and verification methods developed for abnormal prediction identifiers to further confirm the accuracy of abnormal prediction results; verification updates refer to adjusting and updating the corrosion status prediction results based on the verification detection results to ensure the accuracy of the prediction results.

[0038] Execution steps: First, retrieve historical operating data from the substation grounding grid database, including electrical parameters and soil physicochemical properties. Second, perform time-series regression analysis on the historical operating data and corrosion state prediction results to establish a prediction model. Third, identify certified and abnormal prediction indicators by analyzing the relationship between historical data and prediction results. Fourth, optimize and update the corrosion state prediction results using the certified indicators to establish a first update result, improving the accuracy and reliability of the prediction results. Fifth, for abnormal prediction indicators, implement verification strategies, such as increasing sensor data acquisition frequency and conducting on-site testing. Based on the verification and testing results, adjust and update the corrosion state prediction results to establish a second update result. Finally, combine the first and second update results to conduct a comprehensive corrosion state assessment, providing a more accurate and reliable corrosion state assessment result.

[0039] Furthermore, after establishing the corrosion state prediction results, the method of this application includes: Obtain the time-series task dataset of the substation grounding grid; perform anomaly warning level matching based on the time-series task dataset and the corrosion state prediction results, and establish a prediction warning signal; use the prediction warning signal for anomaly reporting management.

[0040] Specifically, the time-series task dataset refers to a collection of task-related data arranged in chronological order, used to analyze and predict the corrosion status of substation grounding grids. Specifically, the time-series task dataset includes historical operating data and real-time data collected by sensors. Anomaly warning level matching refers to matching the corrosion status prediction results with preset anomaly warning levels to determine the warning signal level. For example, based on factors such as corrosion rate and corrosion risk, warning signals are divided into low, medium, and high levels. Predictive warning signals are warning signals generated based on the anomaly warning level matching results, used to notify relevant personnel or systems of potential corrosion risks. Anomaly reporting management refers to the management and processing of generated predictive warning signals, including recording, notification, and response operations, to ensure that warning signals are processed in a timely manner.

[0041] Execution steps: Obtain the time-series task dataset from the substation grounding network database or sensor network. The time-series task dataset includes historical operating data and real-time collected electrical parameters, soil physicochemical properties data, etc. Match the corrosion state prediction results with the preset anomaly warning level standards, and generate corresponding prediction warning signals based on the anomaly warning level matching results. The prediction warning signals can be digital signals, alarm information, or other forms of notification. Manage the generated prediction warning signals, including recording warning signals and tracking processing progress.

[0042] Furthermore, before updating the operational data and the soil physicochemical property data to the twin model, the method of this application includes: The local edge node is used to perform data analysis and preprocessing of the running data and the soil physicochemical property data to establish an edge processing dataset; after the edge processing dataset is certified, it is uploaded to the twin model in the cloud.

[0043] Specifically, local edge nodes refer to computing nodes deployed on-site in the substation grounding grid, used for preliminary analysis and preprocessing of collected data to reduce data transmission volume and improve the real-time performance of data processing; data analysis and preprocessing refers to the preliminary processing of collected operational data and soil physicochemical property data, including data cleaning, feature extraction, and normalization, to improve data quality and the efficiency of subsequent processing; edge processing datasets refer to the data set after analysis and preprocessing by local edge nodes, which has a certain structure and quality and is suitable for further analysis and processing; certification means that the edge processing dataset has been verified, confirming the accuracy and integrity of the data and meeting the requirements for uploading to the cloud twin model; the cloud twin model refers to the digital twin model of the substation grounding grid deployed in the cloud, used to receive and process data from local edge nodes for more in-depth analysis and corrosion state prediction.

[0044] Execution steps: At the local edge node of the substation grounding grid, preliminary analysis and preprocessing are performed on the collected operational data and soil physicochemical property data. For example, data cleaning is performed to remove noise and outliers; key features are extracted, such as the mean and variance of current and seasonal variations in soil moisture; the data is normalized to meet the requirements of the model input; the preprocessed data is organized into an edge processing dataset, which already possesses a certain structure and quality suitable for further analysis and processing; the edge processing dataset is certified to verify the accuracy and completeness of the data, and further, to check whether the data meets the preset quality standards and whether there are missing or outliers; after certification, the edge processing dataset is uploaded to the cloud-based twin model for further analysis and corrosion state prediction.

[0045] In summary, the beneficial effects of the embodiments of this application are: This application employs an interactive method and system for assessing the corrosion status of substation grounding grids. After acquiring structural data from the substation grounding grid, a graph convolutional network model is constructed based on this data, and a twin model is built by learning the spatial relationships between the grounding grids. Operational data, including current, voltage, and temperature data, is collected from a pre-deployed sensor network within the substation grounding grid. Simultaneously, soil physicochemical property data, including pH, conductivity, humidity, salinity, and temperature, are acquired. The operational data and soil physicochemical property data are then updated to the twin model. A long short-term memory (LSTM) analysis layer integrated within the twin model performs spatiotemporal correlation analysis of changes in electrical parameters and soil physicochemical properties, establishing a corrosion state prediction result. This application provides a deep learning-based method and system for assessing the corrosion status of substation grounding grids. It achieves the technical effect of collaboratively updating operational data and soil physicochemical property data to the twin model, mapping the spatial diffusion effect of the grounding grid's physical structure and localized corrosion, and enabling accurate assessment of the corrosion status of substation grounding grids.

[0046] Example 2, based on the same inventive concept as the deep learning-based substation grounding grid corrosion status assessment method in the previous examples, such as... Figure 2 As shown in the figure, this application provides a deep learning-based substation grounding grid corrosion status assessment system, wherein the system includes: The first data acquisition module M100: After interactively acquiring the structural data of the substation grounding network, it performs graph convolutional network modeling based on the structural data and constructs a twin model by learning the spatial relationships of each grounding network.

[0047] Operational data acquisition module M200: Based on the sensor network pre-deployed in the substation grounding grid, it collects the operational data of the substation grounding grid, including current data, voltage data, and temperature data.

[0048] The second data acquisition module M300: synchronously acquires soil physicochemical property data, including soil pH, electrical conductivity, humidity, salinity and soil temperature.

[0049] State assessment module M400: Updates the operating data and soil physicochemical property data to the twin model, performs spatiotemporal correlation analysis of electrical parameters and soil physicochemical property changes through the long short-term memory analysis layer integrated in the twin model, and establishes corrosion state prediction results.

[0050] Furthermore, the state assessment module M400 includes: (This is achieved by integrating a long short-term memory analysis layer within the twin model to perform spatiotemporal correlation analysis of changes in electrical parameters and soil physicochemical properties.) Local Feature Extraction Module: Utilizes the feature extraction channel within the Long Short-Term Memory analysis layer to extract local features of changes in electrical parameters and soil physicochemical properties, and establishes the local feature extraction results.

[0051] Temporal coupling capture module: The local feature extraction results are sent to the data analysis channel in the long short-term memory analysis layer. After the memory depth and time window size are adaptively configured using the data analysis channel, the temporal coupling capture of the local feature extraction results is performed.

[0052] Spatiotemporal Joint Analysis Module: Utilizes the temporal coupling capture results and the spatial relationship of the grounding grid to perform spatiotemporal joint analysis and establish spatiotemporal correlation analysis results.

[0053] Furthermore, after adaptively configuring the memory depth and time window size using the data analysis channel, temporal coupling capture of the local feature extraction results is performed. The temporal coupling capture module includes: Evaluation Result Establishment Module: Activate the adaptive learning mechanism to evaluate the real-time change rate and data stability of data within the local feature extraction results, and establish the evaluation results.

[0054] Coupling and fusion module: After configuring the corresponding memory depth and time window size according to the evaluation results, it performs local-global coupling and fusion using a hierarchical time window coupling analysis strategy to complete the temporal coupling capture.

[0055] Furthermore, corrosion state prediction results are established, and the state assessment module M400 includes: Input feature set establishment module: After extracting the short-term and long-term spatiotemporal features from the spatiotemporal correlation analysis results, the input feature set is established.

[0056] Fusion Analysis Module: Activate the deep fusion channel within the Long Short-Term Memory analysis layer, input the input feature set into the deep fusion channel, and perform erosion state fusion analysis under multi-source data verification. The multi-source data verification includes spatial multi-source data verification and temporal data verification at the same location node, as well as joint location verification at different location nodes.

[0057] Corrosion state prediction result output module: Outputs corrosion state prediction results based on the deep fusion channel.

[0058] Furthermore, to establish corrosion state prediction results, the state assessment module M400 also includes: Time-series regression analysis module: acquires historical operating data of the substation grounding grid, performs time-series regression analysis based on the historical operating data and the corrosion state prediction results, and establishes authentication prediction identifiers and anomaly prediction identifiers.

[0059] Prediction Enhancement Update Module: Utilizes the certified prediction identifier to perform prediction enhancement update on the corrosion state prediction results, and establishes a first update result.

[0060] Verification and detection module: After configuring the verification strategy using the anomaly prediction identifier, it performs verification and detection, verifies and updates the corrosion state prediction result based on the verification and detection result, and establishes a second updated result.

[0061] Corrosion status assessment module: performs corrosion status assessment based on the first update result and the second update result.

[0062] Furthermore, after establishing the corrosion state prediction results, the state assessment module M400 includes: Timing task dataset acquisition module: Acquires the timing task dataset of the substation grounding grid.

[0063] Early warning level matching module: Based on the time-series task dataset and the corrosion state prediction results, perform abnormal early warning level matching to establish a predictive early warning signal.

[0064] Anomaly Reporting Management Module: Uses the predicted early warning signals to manage anomaly reporting.

[0065] Furthermore, before updating the operational data and the soil physicochemical property data to the twin model, the state assessment module M400 includes: Edge processing module: Utilizes local edge nodes to perform data analysis and preprocessing of the running data and the soil physicochemical property data, and establishes an edge processing dataset.

[0066] Data upload module: After the edge processing dataset is authenticated, it uploads the edge processing dataset to the twin model in the cloud.

[0067] Example 3: Based on the same inventive concept as the deep learning-based substation grounding grid corrosion status assessment method in Example 1, the present invention also provides an electronic device, including: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the method described in Example 1.

[0068] like Figure 3As shown, the bus architecture is represented by bus 300. Bus 300 may include any number of interconnected buses and bridges, connecting various circuits including one or more processors represented by processor 302 and memory represented by memory 304. Bus 300 may also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. Bus interface 305 provides an interface between bus 300 and receiver 301 and transmitter 303. Receiver 301 and transmitter 303 may be the same element, i.e., a transceiver, providing a unit for communicating with various other devices over a transmission medium. Processor 302 is responsible for managing bus 300 and general processing, while memory 304 can be used to store data used by processor 302 during operation.

[0069] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A substation grounding grid corrosion state evaluation method based on deep learning, characterized in that, The method comprises the steps of: After obtaining the structure data of the grounding grid of the substation, a graph convolution network is modeled according to the structure data, and a twin model is constructed by learning the spatial relationship of each grounding grid; Based on the sensor network pre-deployed in the grounding grid of the substation, the running data of the grounding grid of the substation is collected, and the running data includes current data, voltage data, and temperature data; Synchronously acquire soil physicochemical property data, including soil pH, conductivity, humidity, salinity, and soil temperature; Update the running data and the soil physicochemical property data to the twin model, perform spatiotemporal correlation analysis of electrical parameters and soil physicochemical property changes through the long short-term memory analysis layer integrated in the twin model, and establish a corrosion state prediction result. 2.The substation grounding grid corrosion state evaluation method based on deep learning according to claim 1, wherein, Performing spatiotemporal correlation analysis of electrical parameters and soil physicochemical property changes through the long short-term memory analysis layer integrated in the twin model comprises: Using the feature extraction channel in the long short-term memory analysis layer to extract local features of electrical parameters and soil physicochemical property changes, and establishing a local feature extraction result; After the data analysis channel in the long short-term memory analysis layer is configured with adaptive memory depth and time window size, the time sequence coupling capture of the local feature extraction result is performed; Using the time sequence coupling capture result and the spatial relationship of the grounding grid to perform spatiotemporal joint analysis and establish a spatiotemporal correlation analysis result. 3.The substation grounding grid corrosion state evaluation method based on deep learning according to claim 2, wherein, After the data analysis channel is configured with adaptive memory depth and time window size, the time sequence coupling capture of the local feature extraction result is performed, which comprises: Activate the adaptive learning mechanism to evaluate the real-time change rate and data stability of the data in the local feature extraction result, and establish an evaluation result; According to the evaluation result, configure the corresponding memory depth and time window size, and use the hierarchical time window coupling analysis strategy to perform local-global coupling fusion to complete the time sequence coupling capture. 4.The substation grounding grid corrosion state evaluation method based on deep learning according to claim 1, wherein, Establishing a corrosion state prediction result comprises: After extracting the short-term spatiotemporal features and long-term spatiotemporal features of the spatiotemporal correlation analysis result, an input feature set is established; Activate the deep fusion channel in the long short-term memory analysis layer, input the input feature set into the deep fusion channel, and perform corrosion state fusion analysis under multi-source data verification, which includes spatial multi-source data verification and time sequence data verification of the same position node, and joint position verification of different position nodes; According to the deep fusion channel, a corrosion state prediction result is output. 5.The substation grounding grid corrosion state evaluation method based on deep learning according to claim 1, wherein, Establishing a corrosion state prediction result also comprises: Obtain historical running data of the grounding grid of the substation, perform time sequence regression analysis according to the historical running data and the corrosion state prediction result, and establish authentication prediction identifiers and abnormal prediction identifiers; Use the authentication prediction identifiers to perform prediction enhancement update of the corrosion state prediction result, and establish a first update result; After configuring a verification strategy using the abnormal prediction identifiers, perform verification detection, and perform verification update of the corrosion state prediction result according to the verification detection result, and establish a second update result; According to the first update result and the second update result, corrosion state evaluation is performed. 6.The substation grounding grid corrosion state evaluation method based on deep learning according to claim 1, wherein, After the corrosion state prediction result is established, the method comprises: Obtaining a time series task data set of the grounding grid of the substation; According to the time series task data set, the corrosion state prediction result is matched to establish a prediction warning signal; Using the prediction warning signal to manage the abnormal report. 7.The substation grounding grid corrosion state evaluation method based on deep learning according to claim 1, wherein, Before updating the running data and the soil physicochemical property data to the twin model, the method comprises: Using a local edge node to perform data analysis and preprocessing on the running data and the soil physicochemical property data to establish an edge processing data set; After the edge processing data set is authenticated, the edge processing data set is uploaded to the twin model in the cloud.

8. A substation grounding grid corrosion state evaluation system based on deep learning, characterized in that, The system for implementing the steps of the substation grounding grid corrosion state evaluation method based on deep learning of any one of claims 1-7, the system comprises: A first data acquisition module: after the structure data of the substation grounding grid is interactively acquired, a graph convolution network is modeled according to the structure data, and a twin model is constructed by learning the spatial relationship of each grounding grid; A running data acquisition module: based on the pre-deployed sensor network of the substation grounding grid, the running data of the substation grounding grid is collected, and the running data includes current data, voltage data, and temperature data; A second data acquisition module: synchronously acquiring soil physicochemical property data, the soil physicochemical property data includes soil pH value, conductivity, humidity, salinity, and soil temperature; A state evaluation module: updating the running data and the soil physicochemical property data to the twin model, performing spatio-temporal correlation analysis of electrical parameter and soil physicochemical property changes through a long short-term memory analysis layer integrated in the twin model, and establishing a corrosion state prediction result. 9.The substation grounding grid corrosion state evaluation system based on deep learning of claim 8, wherein, The state evaluation module comprises: A local feature extraction module: using a feature extraction channel in the long short-term memory analysis layer to perform local feature extraction of electrical parameter and soil physicochemical property changes, and establishing a local feature extraction result; A time series coupling capture module: sending the local feature extraction result to a data analysis channel in the long short-term memory analysis layer, and after the data analysis channel is adaptively configured with memory depth and time window size, performing time series coupling capture of the local feature extraction result; A spatio-temporal joint analysis module: using the time series coupling capture result and the spatial relationship of the grounding grid to perform spatio-temporal joint analysis, and establishing a spatio-temporal correlation analysis result. 10.The substation grounding grid corrosion state evaluation system based on deep learning of claim 9, wherein, After the data analysis channel is adaptively configured with memory depth and time window size, the time series coupling capture module comprises: An evaluation result establishment module: activating an adaptive learning mechanism to evaluate the real-time change rate and data stability of the data in the local feature extraction result, and establishing an evaluation result; A coupling fusion module: after the evaluation result is used to configure the corresponding memory depth and time window size, using a hierarchical time window coupling analysis strategy to perform local-global coupling fusion to complete time series coupling capture. 11.The substation grounding grid corrosion state evaluation system based on deep learning of claim 8, wherein, The state evaluation module comprises: The input feature set establishment module: after extracting the short-term spatio-temporal features and long-term spatio-temporal features of the spatio-temporal correlation analysis result, an input feature set is established; The fusion analysis module: a deep fusion channel in the long short-term memory analysis layer is activated, the input feature set is input into the deep fusion channel, and corrosion state fusion analysis under multi-source data verification is performed, the multi-source data verification including spatial multi-source data verification of the same position node, time series data verification, and joint position verification of different position nodes; The corrosion state prediction result output module: corrosion state prediction results are output according to the deep fusion channel. 12.The substation grounding grid corrosion state evaluation system based on deep learning of claim 8, wherein, The corrosion state prediction result is established, and the state evaluation module further includes: The time series regression analysis module: historical operation data of the substation grounding grid are acquired, time series regression analysis is performed according to the historical operation data and the corrosion state prediction result, authentication prediction identifiers and abnormal prediction identifiers are established; The prediction enhancement update module: the corrosion state prediction result is updated by using the authentication prediction identifiers, and a first update result is established; The verification detection module: after a verification strategy is configured by using the abnormal prediction identifiers, verification detection is performed, the corrosion state prediction result is updated according to the verification detection result, a second update result is established; The corrosion state evaluation module: corrosion state evaluation is performed according to the first update result and the second update result. 13.The substation grounding grid corrosion state evaluation system based on deep learning of claim 8, wherein, After the corrosion state prediction result is established, the state evaluation module includes: The time series task data set acquisition module: a time series task data set of the substation grounding grid is acquired; The early warning level matching module: abnormal early warning level matching is performed according to the time series task data set and the corrosion state prediction result, and a prediction early warning signal is established; The abnormal report management module: abnormal report management is performed by using the prediction early warning signal. 14.The substation grounding grid corrosion state evaluation system based on deep learning of claim 8, wherein, Before the operation data and the soil physicochemical property data are updated to the twin model, the state evaluation module includes: The edge processing module: data analysis and preprocessing of the operation data and the soil physicochemical property data are performed by using a local edge node, and an edge processing data set is established; The data uploading module: after the edge processing data set is authenticated, the edge processing data set is uploaded to the twin model in the cloud.

15. An electronic device, comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the steps of the substation grounding grid corrosion state evaluation method based on deep learning in any one of claims 1-7.

Images