Intelligent identification method for abnormal area of step voltage of grounding grid

Abstract

This invention relates to the field of intelligent identification technology for step voltage anomaly areas in grounding grids, specifically disclosing an intelligent identification method for such areas. The method includes: multi-source sensor data acquisition, data sliding filtering, dynamic threshold parameter generation, voltage gradient matrix construction, collaborative judgment result generation, and anomaly identification threshold optimization. This invention utilizes a distributed sensor network to comprehensively acquire multi-source sensor data, effectively removes noise interference through sliding filtering, and then combines historical operating data with real-time environmental parameters to generate dynamic threshold parameters. Simultaneously, a voltage gradient matrix reflecting the spatial characteristics of the potential field is constructed. After accurate identification of anomaly areas through collaborative judgment, the feedback information is used to analyze and optimize the direction, dynamically adjusting the anomaly identification threshold to form a complete intelligent optimization closed loop. This achieves dual verification and accurate identification of the grounding grid status, improving the level of intelligent management.

Description

Technical Field

[0001] This invention relates to the field of intelligent identification technology for step voltage anomaly areas in grounding grids, and more specifically, to an intelligent identification method for step voltage anomaly areas in grounding grids. Background Technology

[0002] The grounding grid is a crucial component in power facilities such as substations and power plants, ensuring the safety of personnel and equipment. Its main function is to provide a low-impedance discharge path for fault current during grounding faults and to control the ground surface potential within a safe range. Step voltage refers to the potential difference between the two feet of a person standing at different positions, and it is one of the important indicators for evaluating the safety performance of the grounding grid. If the performance of the grounding grid deteriorates, it may lead to excessively high step voltages during faults, posing a serious risk of electric shock to workers.

[0003] Currently, there are two main methods for identifying areas with abnormal step voltage in grounding grids. One is regular manual inspection, where maintenance personnel use specialized instruments to conduct sampling or grid-based measurements within the grounding grid coverage area and compare the data on-site to determine if any anomalies exist. The other is an online monitoring system, which deploys sensors at key locations to collect step voltage data in real time. When the measured value exceeds a pre-set fixed alarm threshold, an alarm is triggered, prompting maintenance personnel to pay attention.

[0004] However, the aforementioned existing technologies have obvious technical defects: (i) manual inspection is time-consuming, labor-intensive, and inefficient. Its discreteness and discontinuity make it difficult to capture occasional or dynamic anomalies; (ii) online monitoring systems mostly use fixed threshold judgments, which cannot adapt to normal fluctuations in step voltage caused by environmental factors such as soil moisture and temperature. This easily leads to a large number of false alarms or missed alarms. Moreover, they rely only on a single parameter of voltage amplitude and are not sensitive to local electric field distortions caused by conductor corrosion, loose connection points, etc., and lack the ability to comprehensively assess the physical health status of the grounding grid; (iii) existing technologies have not built an optimized closed-loop feedback mechanism, and cannot optimize their own judgment standards and algorithms based on operation and maintenance conditions and feedback information. They are difficult to adapt to the aging of the grounding grid and long-term environmental changes, which seriously limits the level of intelligence and long-term reliability of grounding grid safety management. Summary of the Invention

[0005] In view of this, in order to solve the problems mentioned in the background technology, an intelligent identification method for abnormal step voltage areas in grounding grids is proposed.

[0006] The objective of this invention can be achieved through the following technical solution: This invention provides an intelligent identification method for abnormal step voltage areas of grounding grids, including the following steps: S1, multi-source sensor data acquisition: acquire step voltage data, soil resistivity data, ambient temperature and humidity data and conduction resistance data of the grounding grid, and generate multi-source sensor data.

[0007] S2. Data sliding filter processing: The multi-source sensor data is subjected to sliding window filtering processing to generate filtered data.

[0008] S3. Dynamic threshold parameter generation: Extract the step voltage data and ambient temperature and humidity data from the filtered data, and combine them with historical data accumulated from long-term system operation to generate dynamic threshold parameters through a weighted moving average algorithm.

[0009] S4. Voltage gradient matrix construction: Extract the step voltage data from the filtered data and calculate its spatial gradient distribution characteristics to generate a voltage gradient matrix.

[0010] S5. Generation of collaborative judgment results: The voltage gradient matrix is ​​correlated with the on-resistance data in the filtered data to generate collaborative judgment results.

[0011] S6. Anomaly Identification Threshold Optimization: Anomaly area identifiers are generated based on the collaborative judgment results, and the generation process of the dynamic threshold parameters is optimized using the received operation and maintenance feedback information.

[0012] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention achieves dual verification of the grounding grid status by forcibly correlating the step voltage gradient characterizing electrical performance with the conduction resistance reflecting the health of the physical structure through the collaborative analysis of multi-source sensor data. This method can accurately distinguish between fundamental anomalies caused by structural defects such as corrosion or loose connection of the grounding body and temporary electrical phenomena caused by external environmental factors such as changes in soil moisture, thereby significantly improving the accuracy of anomaly identification and effectively avoiding false alarms and missed alarms caused by traditional single-parameter detection methods.

[0013] (2) This invention introduces a dynamic threshold generation mechanism based on historical data and real-time environmental parameters, enabling the benchmark for anomaly judgment to be adaptively adjusted in real time according to changes in operating conditions. This overcomes the limitations of traditional fixed threshold schemes in the face of complex environmental changes such as seasonal alternation and wet-dry cycles, ensuring that the identification system can maintain high stability and high sensitivity under various operating conditions, and enhancing the environmental adaptability and robustness of the method.

[0014] (3) This invention constructs a closed-loop learning system that progresses from intelligent diagnosis to visual feedback and then to system self-optimization. By using the on-site confirmation results from maintenance personnel as feedback input, the system can automatically adjust its internal algorithm parameters, achieving continuous iteration and performance improvement of the diagnostic model. This design enables the system to have the ability to self-evolve, continuously adapting to the aging process of the grounding grid itself and long-term environmental changes, significantly improving the intelligence level and long-term reliability of grounding grid safety management. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0017] Figure 2 This is a three-dimensional topology diagram of the step voltage anomaly region of the grounding grid according to the present invention. Detailed Implementation

[0018] The technical solutions of 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.

[0019] Please see Figure 1 The present invention provides an intelligent identification method for abnormal step voltage areas of grounding grid, including: S1, multi-source sensor data acquisition: acquiring step voltage data, soil resistivity data, ambient temperature and humidity data and conduction resistance data of grounding grid, and generating multi-source sensor data.

[0020] In a specific embodiment of the present invention, the generation of multi-source sensing data specifically includes: acquiring the step voltage data, the soil resistivity data, and the environmental temperature and humidity data respectively through voltage sensors, resistivity sensors, and temperature and humidity sensors in a distributed sensor network.

[0021] The on-resistance data is obtained using a continuity tester.

[0022] The step voltage data, soil resistivity data, ambient temperature and humidity data, and on-resistance data are synchronized and aligned in time to generate the multi-source sensing data.

[0023] It should be noted that for the acquisition of step voltage data, multiple voltage sensors are deployed on the surface of the grounding grid coverage area according to a preset matrix topology to measure the potential difference between two points at a predetermined step length, thereby forming step voltage data reflecting the spatial potential distribution. Simultaneously, resistivity sensors are deployed at key locations near the grounding grid by inserting multiple sets of electrodes into the soil to acquire soil resistivity data reflecting the soil's conductivity in real time. In the same area, temperature and humidity sensors are deployed to monitor environmental parameters affecting soil resistivity and output ambient temperature and humidity data. Furthermore, a continuity tester is used to perform connectivity measurements between multiple accessible nodes of the grounding grid, such as between the grounding down conductor and the test well, to obtain conduction resistance data reflecting the health of the grounding grid conductors themselves.

[0024] It should also be noted that, to ensure strong correlation between all data over time, the entire acquisition process is uniformly scheduled by a central control unit or network time synchronization protocol, ensuring that all sensors perform measurements at the same time, or by attaching a high-precision timestamp to each measurement data point. In this way, at any given moment... Both can yield an instantaneous state vector containing multidimensional information. Its composition is ,in, represent Complete state information at any given moment for Step voltage data at each moment for Soil resistivity data at time t, and They are respectively Real-time ambient temperature and humidity data, for The on-resistance data at each moment. By combining the instantaneous state vectors collected continuously at multiple moments, the multi-source sensor data required for subsequent analysis is formed.

[0025] This invention establishes an instantaneous correspondence between electrical state, environmental factors, and the physical health of the grounding grid through synchronous acquisition of multiple physical quantities. This collaborative acquisition method overcomes the problem of misjudgment of causal relationships caused by asynchronous data acquisition time in traditional methods. The generated multi-source sensor data provides a high-fidelity data foundation for subsequent accurate identification of abnormal areas, especially for identifying latent anomalies caused by structural defects, significantly improving the accuracy and reliability of subsequent intelligent analysis.

[0026] S2. Data sliding filter processing: The multi-source sensor data is subjected to sliding window filtering processing to generate filtered data.

[0027] In a specific embodiment of the present invention, the specific steps of performing sliding window filtering on the multi-source sensor data to generate filtered data include: setting the size of the sliding window according to the data sampling frequency of the multi-source sensor data to obtain window parameters.

[0028] The multi-source sensor data is smoothed within a window using the window parameters to generate preliminary filtered data.

[0029] An outlier removal operation is performed on the initial filtered data to generate the filtered data.

[0030] It should be noted that after acquiring multi-source sensor data containing noise and transient interference, sliding window filtering is required to improve data quality. First, a suitable sliding window size is set based on the preset data sampling frequency. This sampling frequency is determined by the configuration of the data acquisition module, while the choice of sliding window size must balance noise suppression effectiveness with real-time signal response. Generally, a higher sampling frequency corresponds to a larger window to include enough data points for smoothing. Then, the sliding window is sequentially slid across the multi-source sensor data in the time series, and data points falling within the window are smoothed to generate preliminary filtered data. A commonly used smoothing method is to calculate the arithmetic mean of the data within the window, as shown in the following formula: ,in, It is a moment A data point in the initial filtered output data. It is a moment A data point from the raw multi-source sensor data collected. This is the set sliding window size. This process is applied independently to all data streams, including step voltage and soil resistivity. After smoothing, to further eliminate extreme outliers caused by equipment failure or sudden strong interference, outlier removal is performed on the generated preliminary filtered data.

[0031] This invention, through the organic combination of smoothing and outlier removal, effectively filters out random noise and sporadic pulse interference in multi-source sensor data. It not only preserves the true trend reflecting changes in the grounding grid state but also avoids misjudgments caused by glitches in the original data in subsequent dynamic threshold generation and spatial gradient analysis. By providing high-quality filtered data as input, it greatly improves the stability of the entire intelligent identification method and the accuracy of the final diagnostic results, laying a solid data foundation for accurately locating abnormal step voltage areas in the grounding grid.

[0032] S3. Dynamic threshold parameter generation: Extract the step voltage data and ambient temperature and humidity data from the filtered data, and combine them with historical data accumulated from long-term system operation to generate dynamic threshold parameters through a weighted moving average algorithm.

[0033] In a specific embodiment of the present invention, the specific steps of generating dynamic threshold parameters by weighted moving average algorithm include: calculating the weighted average of step voltage data in the filtered data and historical step voltage data in the historical data to generate voltage trend parameters.

[0034] It should be noted that, in order to generate dynamic threshold parameters that can adapt to changes in operating conditions, preset historical data is first retrieved from the system's storage unit. This historical data is accumulated over a long period of system operation and includes historical step voltage data and corresponding historical ambient temperature and humidity data under different seasons and weather conditions. These data together constitute the baseline of the grounding grid's normal operation. Next, the system combines the filtered step voltage data acquired at the current moment with the aforementioned historical step voltage data, and generates an initial voltage trend parameter through weighted averaging. This calculation process aims to integrate current measurements with long-term behavioral patterns, and its mathematical expression can be simplified as follows: ,in, It is the current moment. The generated voltage trend parameters, It is the current moment. The filtered step voltage data, It is a statistical average calculated from historical step voltage data. Weighting factor. It is a preset value between 0 and 1, used to adjust the relative importance of current data and historical data when determining trend parameters.

[0035] In a specific embodiment of the present invention, when the weighting factor Only a preset value between 0 and 1 can be selected. When adjusting the relative importance of current data and historical data in determining trend parameters, the specific value needs to be carefully determined by considering multiple factors. First, the characteristics of the data must be considered. If the data is highly real-time, changes frequently, and the current data better reflects the latest trend, then the weighting factor should be a larger value of 0.8 to give the current data a higher weight, allowing the trend parameter to keep up with current changes. Conversely, if the data has long-term stability and historical data has important reference value for trend judgment, then the weighting factor should be a smaller value of 0.2 to highlight the contribution of historical data.

[0036] The ambient temperature and humidity data in the filtered data are obtained, and an environmental correction coefficient is generated accordingly.

[0037] It should be noted that the specific process for generating the environmental correction coefficient is as follows: 1) Determine the correction model: Linear correction model: This assumes that the influence of ambient temperature and humidity on the measurement data is linear. For example, for every 1 degree Celsius increase in temperature, the measurement data will increase or decrease by a fixed value; for every percentage change in humidity, the measurement data will also change linearly. Taking temperature as an example, the correction coefficient is... ,in It is the temperature influence coefficient. This is the current temperature. This is the reference temperature, typically a standard ambient temperature such as 25 degrees Celsius. The correction factor for humidity is similarly calculated. Nonlinear correction models: When the effect of temperature and humidity on the measured data is nonlinear, more complex models may be needed, such as polynomial or exponential models. A polynomial temperature correction model could be... ,in These are coefficients obtained by fitting experimental data. 2) Calculating correction coefficients: Based on experimental data fitting coefficients: If a linear or nonlinear correction model is used, it is necessary to obtain measurement data under different temperature and humidity conditions through experiments, and use the least squares method for data fitting to determine the coefficients in the model. For example, conduct a series of measurement experiments at different temperatures, record the measurement data and actual values, and then use these data to fit a linear model. Where y is the measured data, x is the temperature, and k and b are the coefficients to be fitted. The fitted k and b can then be used to calculate the temperature correction coefficient. 3) Comprehensive temperature and humidity correction coefficient: The effects of temperature and humidity on the measured data are independent of each other. The environmental correction coefficient can be obtained by multiplying the temperature correction coefficient and the humidity correction coefficient, i.e. .

[0038] The voltage trend parameter is corrected using the environmental correction coefficient to generate the dynamic threshold parameter.

[0039] It should be noted that the generation of the dynamic threshold parameter can be expressed as: ,in, It is the current moment. The dynamic threshold parameter.

[0040] This method integrates historical and real-time data and incorporates environmental parameters for dynamic compensation, generating highly adaptive dynamic threshold parameters. This approach fundamentally changes the rigidity of traditional fixed-threshold alarm mechanisms, enabling the judgment criteria to intelligently adjust in real-time according to changes in the environment and operating conditions of the grounding grid. This not only significantly reduces false alarms and missed alarms caused by normal environmental fluctuations such as seasonal changes and alternating wet and dry conditions, but also significantly improves the accuracy and reliability of anomaly identification, ensuring the system can still make accurate judgments under complex and ever-changing operating conditions.

[0041] S4. Voltage gradient matrix construction: Extract the step voltage data from the filtered data and calculate its spatial gradient distribution characteristics to generate a voltage gradient matrix.

[0042] In a specific embodiment of the present invention, the specific steps for calculating its spatial gradient distribution characteristics and generating a voltage gradient matrix include: calculating the rate of change of the step voltage data between adjacent measuring points and generating a voltage change sequence.

[0043] It should be noted that this step aims to transform discrete voltage measurements into structured data that can intuitively reflect the spatial distribution characteristics of the potential field. First, based on the filtered step voltage data, the rate of change of step voltage between adjacent measurement points is calculated. Considering that the sensors are deployed in a grid, for any non-boundary measurement point... The rate of change of voltage between the voltage and its adjacent measuring points, for example, in the x and y directions, can be approximated as a potential gradient component. For example, the rate of change along the x direction... It can be obtained from the following formula: ,in, and Representing the measuring points respectively and Filtered step voltage data at the location, This is the physical distance between the two measuring points. Similarly, the rate of change along the y-direction can be calculated. By combining the voltage change rates calculated from all measuring points, a voltage change sequence is formed.

[0044] The voltage change sequence is matrixed to generate a preliminary gradient matrix.

[0045] It should be noted that the voltage change sequence is matrix-processed. That is, each measuring point... The corresponding voltage rate of change vector Fill the corresponding positions in the matrix. This allows us to construct an initial gradient matrix that is consistent with the physical layout of the sensor.

[0046] The initial gradient matrix is ​​normalized to generate the voltage gradient matrix.

[0047] It should be noted that the specific process of generating the voltage gradient matrix includes: 1) Determining the normalization method: In the scenario of converting the preliminary gradient matrix into a voltage gradient matrix, minimum-maximum normalization is commonly used, and its formula is: ,in, These are the original data in the preliminary gradient matrix. It is the minimum value in the initial gradient matrix. It is the maximum value in the initial gradient matrix. 1) The data is normalized, and its value range is between [0,1]. 2) Obtain the maximum and minimum values: Traverse all elements in the preliminary gradient matrix and find the minimum and maximum values. 3) Perform normalization calculation: According to the minimum-maximum normalization formula, perform normalization calculation on each element in the preliminary gradient matrix to obtain the voltage gradient matrix.

[0048] This invention transforms disordered voltage data points into a structured voltage gradient matrix, achieving a leap from point measurement to field analysis. The core advantage of this approach lies in its focus on the spatial trend of voltage variation, rather than the absolute value of the voltage. A localized high gradient value, even in areas with low overall voltage levels, can be significantly highlighted. This makes the system highly sensitive to latent defects such as conductor breaks in the grounding grid or severe corrosion at connection points that cause drastic changes in local potential. The resulting voltage gradient matrix provides quantified and standardized key feature inputs for subsequent precise location of abnormal areas, greatly improving the accuracy of fault identification and spatial positioning capabilities.

[0049] S5. Generation of collaborative judgment results: The voltage gradient matrix is ​​correlated with the on-resistance data in the filtered data to generate collaborative judgment results.

[0050] In a specific embodiment of the present invention, the specific steps for generating the collaborative determination result include: identifying regions in the voltage gradient matrix that exceed the preset normal gradient range, and generating gradient abnormal regions.

[0051] It should be noted that this method employs a dual verification mechanism for collaborative judgment to accurately identify abnormal regions. First, the voltage gradient matrix generated in the previous step is compared point by point with a preset normal gradient range. When the gradient value at a certain location in the voltage gradient matrix exceeds the upper limit of the preset normal gradient range, it is determined that there is electric field distortion at that location, and its coordinates are marked as an abnormal gradient region. This step mainly aims to preliminarily screen out potential problem points from the perspective of electrical performance.

[0052] In one specific embodiment of the present invention, the preset normal gradient range is determined based on the grounding grid design specifications, historical operating data and simulation analysis, and can be 1-4.5V / m, representing the normal fluctuation range of the potential field under healthy conditions.

[0053] Identify regions in the filtered data where the on-resistance data exceeds a preset resistance threshold, and generate abnormal resistance regions.

[0054] It should be noted that the conduction resistance value at each measuring point is compared with a preset resistance threshold. All measuring point locations exceeding the preset resistance threshold are marked as areas of abnormal resistance. This step assesses the health of the grounding grid's physical structure.

[0055] In one specific embodiment of the present invention, the preset resistance threshold is set according to the material, cross-sectional area and connection process standard of the grounding grid conductor, and can be 0.006Ω. Exceeding this value usually means that the conductor has physical defects such as corrosion, loose connection points or breakage.

[0056] Spatial intersection analysis is performed on the gradient anomaly region and the resistance anomaly region to generate the collaborative determination result.

[0057] In a specific embodiment of the present invention, the specific steps of performing spatial intersection analysis on the gradient anomaly region and the resistance anomaly region to generate the collaborative determination result include: spatially superimposing the gradient anomaly region and the resistance anomaly region to obtain candidate anomaly regions;

[0058] It should be noted that, in order to provide a more refined risk assessment of the identified anomalies, this method introduces an adaptive clustering algorithm for deep analysis based on intersection analysis. First, the system spatially superimposes gradient anomaly regions and resistance anomaly regions. This step identifies all spatially overlapping or adjacent regions, forming a preliminary set of candidate anomaly regions. These regions represent a collection of potential fault points, and their degree of anomaly may vary.

[0059] An adaptive clustering algorithm is applied to the candidate abnormal regions to group them, generating abnormal clustering results.

[0060] It should be noted that, next, an adaptive clustering algorithm is applied to these candidate anomaly regions. Adaptive clustering is a machine learning method that automatically determines the number and boundaries of groups based on the inherent distribution characteristics of the data. Here, each candidate anomaly region is treated as a data point, and its feature vector is composed of the voltage gradient value and on-resistance value of that region. Adaptive clustering algorithms, such as DBSCAN, a density-based spatial clustering algorithm, search for dense regions of data points in this two-dimensional feature space. It automatically groups candidate anomaly regions with similar voltage gradients and on-resistance values ​​into the same group and outputs the anomaly clustering results. This result not only includes the grouping information of the anomaly regions but also automatically identifies and eliminates noise points with isolated feature values, further improving the robustness of the diagnosis.

[0061] Based on the characteristics of the abnormal clustering results, an anomaly level is assigned to each cluster, and the anomaly level is associated with the collaborative judgment result.

[0062] It should be noted that, finally, based on the average voltage gradient value and average on-resistance value of all regions within the group corresponding to the center point of each anomaly clustering result, a predefined anomaly level is assigned to it. These voltage gradient value ranges and on-resistance value ranges are pre-stored in the database based on a large amount of experimental data and actual operating experience, corresponding to severe anomaly, moderate anomaly, and mild anomaly levels, respectively. During comparison, if the average voltage gradient value of the group is within the voltage gradient value range corresponding to a certain anomaly level, and the average on-resistance value is within the on-resistance value range corresponding to that anomaly level, then that anomaly level is taken as the anomaly level of the group. If the average voltage gradient value and average on-resistance value simultaneously meet multiple anomaly level range conditions, or do not meet any anomaly level range conditions, the final anomaly level will be determined according to the preset priority rule, i.e., severe anomaly takes precedence over moderate anomaly takes precedence over mild anomaly. Finally, the system outputs an anomaly region identifier containing precise location, cluster to which it belongs, and specific anomaly level information.

[0063] This invention achieves diagnostic accuracy far exceeding that of single-parameter detection through dual constraints of electrical characterization and physical state. Simple gradient anomalies may be caused by temporary factors such as external electromagnetic interference, while a simple increase in resistance may not yet pose a substantial threat to the grounding grid's current-carrying performance. This method, by performing intersection analysis, ensures that the identified anomaly areas simultaneously possess both physical defect roots and electrical performance degradation consequences. This allows for precise identification of truly hazardous structural defects, significantly reducing false alarm rates and providing a highly reliable basis for subsequent maintenance decisions, achieving accurate tracing from phenomenon to essence.

[0064] S6. Anomaly Identification Threshold Optimization: Anomaly area identifiers are generated based on the collaborative judgment results, and the generation process of the dynamic threshold parameters is optimized using the received operation and maintenance feedback information.

[0065] Please see Figure 2 In a specific embodiment of the present invention, the specific steps for optimizing the generation process of the dynamic threshold parameter using the received operation and maintenance feedback information include: converting the collaborative judgment result into three-dimensional topology map data to generate visualization data.

[0066] It should be noted that this step aims to construct a complete closed loop from intelligent diagnosis to manual intervention and then to system self-optimization. First, the abnormal area identifiers generated in the previous step, which contain the precise coordinates of the abnormal points and quantified anomaly levels, are converted into 3D topology map data. This conversion process is based on a pre-stored 3D digital model of the grounding grid. The system maps the information in the abnormal area identifiers to the corresponding nodes in the model and generates renderingable visualization data through preset visual rules, such as using different colors or icon sizes to represent different anomaly levels.

[0067] The visualized data is sent to the operation and maintenance terminal, and the operation and maintenance feedback information containing the confirmation result is received.

[0068] It should be noted that the visualized data is transmitted in real time to the maintenance personnel's terminals, such as tablets or monitoring center workstations, via a communication network. The software on the maintenance terminal renders this data into an intuitive 3D topology map, allowing maintenance personnel to clearly see the specific location and severity of abnormal areas. After conducting on-site inspections and repairs, maintenance personnel submit maintenance feedback information through the terminal interface. This feedback information clearly records the correctness of the system diagnostic results, such as "confirmed," "false alarm," or "anomaly level corrected."

[0069] The confirmation results in the operation and maintenance feedback information are analyzed, and the weight parameters of the weighted moving average algorithm are adjusted accordingly to optimize the dynamic threshold parameter generation process.

[0070] In a specific embodiment of the present invention, the specific steps of adjusting the weight parameters of the weighted moving average algorithm include: parsing the confirmation result into an optimization indication.

[0071] It should be noted that the system performs in-depth analysis on the received operation and maintenance feedback information, extracting key confirmation result fields. These fields are typically structured information, such as "confirmed," "false alarm," or "anomaly level inconsistent," which the system converts into standardized optimization instructions. For example, a "false alarm" will be parsed as an optimization instruction to "reduce system sensitivity," while a "missed alarm" corresponds to an optimization instruction to "increase system sensitivity."

[0072] The adjustment amount is calculated based on the optimization instructions and the deviation between the current dynamic threshold parameters and the actual situation.

[0073] It should be noted that, based on the optimization instructions, a specific adjustment amount is calculated and generated. This step is crucial for the quantification and correction process. The system traces the dynamic threshold parameter that caused the feedback event and compares it with the actual observed or expected value that may be included in the operational feedback, thereby calculating a deviation value. For example, if the system alarms due to threshold V1 and is confirmed as a false alarm, and the normal value confirmed on-site is V2, then the deviation value is the difference between the two. Then, the system multiplies this deviation value by a preset learning rate factor to generate the final adjustment amount. Its calculation model can be simplified as follows: ,in, It is the final adjustment amount generated for the weight parameters. It is a preset learning rate factor used to control the step size of each optimization. It is the calculated deviation value, and It is the direction coefficient determined by the optimization indication; when the indication is to reduce system sensitivity... It is -1, and +1 when the system sensitivity is increased.

[0074] In one specific embodiment of the present invention, based on preliminary experiments and empirical judgment, the preset learning rate factor can be set to 0.01. This is because 0.01 is a relatively balanced value; it is neither too large, leading to excessively large parameter update steps that cause frequent oscillations and difficulty in convergence during the search for the optimal solution, nor too small, causing excessively slow parameter updates that increase the time and computational cost required for the system to reach the optimal state. Thus, it balances convergence speed and stability in most conventional situations.

[0075] The adjustment amount is applied to update the weight parameters of the weighted moving average algorithm.

[0076] It should be noted that the adjustment amount is applied to update the weight parameters of the weighted moving average algorithm to complete this optimization. The system adds the calculated adjustment amount to the currently used weight parameters to obtain a new weight parameter, which is then stored for use in subsequent dynamic threshold generation.

[0077] This invention achieves intelligent evolution through human-machine collaboration by constructing a closed-loop workflow of "diagnosis-visualization-feedback-optimization". Three-dimensional visualization greatly improves the efficiency of information transmission and the intuitiveness of decision-making, while the feedback learning mechanism endows the system with the ability to self-correct and continuously optimize. This design transforms the system from a static diagnostic tool into an intelligent agent capable of learning from practical experience and becoming increasingly accurate over time. It can continuously adapt to the aging process of the grounding grid itself and long-term environmental changes, ensuring high reliability and accuracy of diagnostic results throughout its entire lifecycle.

[0078] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A method for intelligent identification of abnormal step voltage areas in a grounding grid, characterized in that, Includes the following steps: S1. Multi-source sensor data acquisition: Acquire step voltage data, soil resistivity data, ambient temperature and humidity data, and conduction resistance data of the grounding grid to generate multi-source sensor data; S2. Data sliding filter processing: Perform sliding window filtering processing on the multi-source sensor data to generate filtered data; S3. Dynamic threshold parameter generation: Extract the step voltage data and ambient temperature and humidity data from the filtered data, and combine them with the historical data accumulated from long-term system operation to generate dynamic threshold parameters through a weighted moving average algorithm. The specific steps for generating dynamic threshold parameters using the weighted moving average algorithm include: Calculate the weighted average of the step voltage data in the filtered data and the historical step voltage data in the historical data to generate voltage trend parameters; Obtain the ambient temperature and humidity data from the filtered data, and generate an environmental correction coefficient accordingly; The voltage trend parameter is corrected using the environmental correction coefficient to generate the dynamic threshold parameter; S4. Voltage gradient matrix construction: Extract the step voltage data from the filtered data and calculate its spatial gradient distribution characteristics to generate a voltage gradient matrix; S5. Generation of collaborative judgment results: The voltage gradient matrix is ​​correlated with the on-resistance data in the filtered data to generate collaborative judgment results. S6. Anomaly Identification Threshold Optimization: Anomaly area identifiers are generated based on the collaborative judgment results, and the generation process of the dynamic threshold parameters is optimized using the received operation and maintenance feedback information.

2. The intelligent identification method for abnormal step voltage areas in a grounding grid according to claim 1, characterized in that: The generation of multi-source sensing data specifically includes: The step voltage data, soil resistivity data, and ambient temperature and humidity data are acquired respectively through voltage sensors, resistivity sensors, and temperature and humidity sensors in a distributed sensor network. The on-resistance data is obtained using a continuity tester; The step voltage data, soil resistivity data, ambient temperature and humidity data, and on-resistance data are synchronized and aligned in time to generate the multi-source sensing data.

3. The intelligent identification method for abnormal step voltage areas in a grounding grid according to claim 1, characterized in that: The specific steps for performing sliding window filtering on the multi-source sensor data to generate filtered data include: The size of the sliding window is set according to the data sampling frequency of the multi-source sensor data to obtain the window parameters; The multi-source sensor data is smoothed within a window using the window parameters to generate preliminary filtered data. An outlier removal operation is performed on the initial filtered data to generate the filtered data.

4. The intelligent identification method for abnormal step voltage areas in a grounding grid according to claim 1, characterized in that: The specific steps for calculating its spatial gradient distribution characteristics and generating the voltage gradient matrix include: Calculate the rate of change of the step voltage data between adjacent measuring points to generate a voltage change sequence; The voltage change sequence is matrixed to generate a preliminary gradient matrix; The initial gradient matrix is ​​normalized to generate the voltage gradient matrix.

5. The intelligent identification method for abnormal step voltage areas in a grounding grid according to claim 1, characterized in that: The specific steps for generating the collaborative determination result include: Identify regions in the voltage gradient matrix that exceed the preset normal gradient range and generate gradient anomaly regions; Identify regions in the filtered data where the on-resistance data exceeds a preset resistance threshold, and generate abnormal resistance regions. Spatial intersection analysis is performed on the gradient anomaly region and the resistance anomaly region to generate the collaborative determination result.

6. The intelligent identification method for abnormal step voltage areas in a grounding grid according to claim 5, characterized in that: The specific steps for performing spatial intersection analysis on the gradient anomaly region and the resistance anomaly region to generate the collaborative determination result include: The gradient anomaly region and the resistance anomaly region are spatially superimposed to obtain the candidate anomaly region; An adaptive clustering algorithm is applied to the candidate abnormal regions to group them, generating abnormal clustering results; Based on the characteristics of the abnormal clustering results, an anomaly level is assigned to each cluster, and the anomaly level is associated with the collaborative judgment result.

7. The intelligent identification method for step voltage anomaly areas in a grounding grid according to claim 1, characterized in that: The specific steps for optimizing the generation process of the dynamic threshold parameter using the received operation and maintenance feedback information include: The collaborative determination results are converted into three-dimensional topology map data to generate visualization data; The visualized data is sent to the operation and maintenance terminal, and the operation and maintenance feedback information containing the confirmation result is received. The confirmation results in the operation and maintenance feedback information are analyzed, and the weight parameters of the weighted moving average algorithm are adjusted accordingly to optimize the dynamic threshold parameter generation process.

8. The intelligent identification method for abnormal step voltage areas in a grounding grid according to claim 7, characterized in that: The specific steps for adjusting the weight parameters of the weighted moving average algorithm include: The confirmation result is parsed as an optimization instruction; The adjustment amount is calculated based on the optimization instructions and the deviation between the current dynamic threshold parameters and the actual situation; The adjustment amount is applied to update the weight parameters of the weighted moving average algorithm.

Images