Substation grounding grid facing ground resistance multi-point distribution monitoring and hierarchical early warning method and system
By combining a multi-point distributed sensor network and a regional baseline model, the problems of personalization, predictability, and real-time monitoring of substation grounding grids are solved, enabling efficient and accurate health assessment and early warning of the grounding grid.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BAODING SHANGWEI ELECTRICITY TECH
- Filing Date
- 2026-04-15
- Publication Date
- 2026-07-10
AI Technical Summary
Existing methods for monitoring the grounding resistance of substation grounding grids lack personalized monitoring, resulting in regional false alarms or missed alarms, insufficient predictability, and inadequate real-time performance and coverage. They are difficult to issue early warnings before problems escalate, and traditional methods may not fully consider physical constraints.
By designing a multi-point distributed sensor network, multi-modal data acquisition and processing are carried out, a regional baseline model is constructed, a health index is calculated, health prediction and hierarchical alarm are performed, and combined with digital twin simulation, multi-point distributed monitoring and hierarchical early warning of the grounding grid are realized.
It improves the accuracy and predictive ability of grounding grid monitoring, reduces the false alarm rate, improves the efficiency of operation and maintenance decision-making, and realizes a comprehensive assessment and early warning of the health status of the grounding grid.
Smart Images
Figure CN122361902A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of grounding resistance monitoring technology, and in particular to a method and system for multi-point distribution monitoring and hierarchical early warning of grounding resistance in substation grounding grids. Background Technology
[0002] In substations and their grounding grid systems, a monitoring and alarm mechanism is established by continuously or periodically measuring the resistance (grounding resistance, equivalent impedance, ground potential, etc.) of the grounding electrode and the surrounding grounding grid. When the grounding resistance exceeds the set threshold or shows an abnormal trend, an early warning signal is issued. This ensures the safety of the substation grounding system under various operating conditions, reduces the potential risks of live equipment to people, objects, and the environment, and detects potential hazards such as grounding grid loss, corrosion, loose connections, changes in contact resistance, and grounding electrode disconnection in advance. It also improves the planning of power outage maintenance, shortens fault handling time, and enhances reliability and power supply reliability levels.
[0003] Existing methods for monitoring the grounding resistance of substation grounding networks include, but are not limited to, direct grounding resistance monitoring, online and semi-online monitoring, and monitoring of the equivalent impedance or surface potential of the grounding network. However, existing technologies have the following drawbacks: 1) They lack personalized monitoring for different local areas, which can easily lead to regional false alarms or missed alarms; 2) They lack predictive power, with many systems still relying primarily on threshold alarms, lacking accurate modeling of trends and decay rates, making it difficult to issue early warnings before problems escalate; 3) They lack real-time performance and coverage, as traditional inspections or single-point monitoring cannot cover the complex topology of the entire grounding network and cannot capture local anomalies in real time; 4) Traditional prediction methods may not fully consider physical constraints, causing deviations between predictions and actual physical behavior. Summary of the Invention
[0004] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for multi-point distribution monitoring and hierarchical early warning of grounding resistance in substation grounding networks. By standardizing multimodal data for gridded grounding networks, adaptive thresholding of partitioned baselines, digital twin comparison correction, and multi-stage health prediction and hierarchical alarm, the monitoring accuracy, prediction capability, and operation and maintenance decision-making efficiency of substation grounding networks are significantly improved.
[0005] To achieve the above objectives, the present invention provides the following solution: a method for multi-point distribution monitoring and hierarchical early warning of grounding resistance in substation grounding grids, comprising: Based on the geographical information and grounding grid topology of the substation area, a preliminary topology of the substation area is designed, and a multi-point distributed sensor network is designed based on the preliminary topology of the substation area. Based on the multi-point distributed sensor network, multimodal raw datasets are collected through sensor network and redundancy design, and then preliminary signal processing is performed on the multimodal raw datasets to obtain multimodal cleaned datasets. Based on the multimodal cleaning dataset, data time series unification, data pipeline definition, interface contract design, and environmental variable structuring are performed to obtain standardized data input; Construct a partitioned baseline model, and then use the standardized data input and the partitioned baseline model to output a personalized baseline curve and adaptive threshold sequence for each grid node; Based on the standardized data input and the personalized baseline curve, a hybrid health index is calculated, including site-level health index, grid-level health index, and health trend analysis. Based on the standardized data input and the hybrid health index, health prediction and anomaly detection are performed, and then graded alarms and alarm information are generated according to the anomaly detection results.
[0006] Optionally, a preliminary topology structure for the substation area is designed based on the geographical information and grounding grid topology, and a multi-point distributed sensor network is designed based on the preliminary topology structure, including: Based on the geographic information and grounding grid topology of the substation area, the area is divided into regions, and grid nodes and sensor points are designed for each region to obtain the preliminary topology of the substation area. Based on the preliminary topology of the substation area, the data flow direction, data format, clock alignment requirements and data interface are designed. Based on the preliminary topology of the station area, a set of sensor modules and observation indicators are designed in combination with on-site environmental information. A gridded sensor point layout and sampling strategy are designed for the set of sensor modules to obtain a multi-point distributed sensor network.
[0007] Optionally, the sensor module set includes a grounding status sensor, a soil temperature and humidity sensor, a grounding grid corrosion rate sensor, a potential monitoring sensor, a current sensor, and a voltage sensor. The observed indicators include grounding resistance value, surface potential difference between grid nodes, ambient temperature of grid nodes, soil moisture, and corrosion index.
[0008] Optionally, based on the multi-point distributed sensor network, a multimodal raw dataset is collected through the sensor network and redundant design, and then preliminary signal processing is performed on the multimodal raw dataset to obtain a multimodal cleaned dataset, including: Based on the gridded sensor point layout of the multi-point distributed sensor network, a sensor module combination is assigned to each monitoring point to perform multimodal observation at the same point and output a monitoring point coordinate table. Based on the multi-point distributed sensor network and the monitoring point coordinate table, hardware redundancy strategy, communication redundancy strategy, data redundancy and caching strategy and fault tolerance mechanism are designed. GPS is used to unify the time source of the entire network to align the sensor timestamps and complete the sensor network and redundancy design. Local data is collected using the multi-point distributed sensor network. Noise suppression, filtering, denoising, temporal alignment, and missing data filling are performed on the collected data at the edge nodes to obtain a multimodal raw dataset. Based on the multimodal raw dataset, local equivalent impedance, local surface displacement, potential, temperature, and humidity are calculated to obtain local indicators. Then, the initial local health index is calculated using the local indicators, and a multimodal cleaning dataset is output.
[0009] Optionally, based on the multimodal cleaning dataset, data time series unification, data pipeline definition, interface contract design, and environmental variable structuring are performed to obtain standardized data input, including: Based on the multimodal cleaned dataset, field definitions, observation unit standardization, missing value handling, and environmental variable structuring are performed to obtain unified time series data. Based on the multimodal cleaned dataset, time synchronization and alignment, data cleaning and quality assessment, missing value imputation, and data packaging and transmission format definition are performed to obtain a data pipeline; Based on the unified time-series data, an interface protocol is selected, a message structure is defined, and TLS encryption, device certificates, token authentication, key rotation strategy, and log auditing are used for data security and authentication to obtain an interface contract document. Based on the unified time-series data and environmental data, a mapping relationship between environmental factors and local indicators is constructed to obtain an environmental variable table. Then, the interface contract document and the environmental variable table are organized into a structured input format to obtain standardized data input.
[0010] Optionally, the defined fields include nodes, measurement points, timestamps, observation values, units, sensor status, data quality labels, and environmental metadata. The data pipeline includes data flow direction, preprocessing process, data quality assessment, missing value imputation strategy, and timestamp alignment rules.
[0011] Optionally, a partitioned baseline model is constructed, and then the standardized data input and the partitioned baseline model are used to output a personalized baseline curve and adaptive threshold sequence for each grid node, including: For each grid node, the historical data is decomposed into a time series, and the historical trend, seasonal components and residuals are output. The historical trend is fitted using robust regression to obtain a long-term trend model. The seasonal components are fitted using a periodic sliding window to obtain a seasonal baseline component. The long-term trend model and the seasonal baseline component are combined to obtain a partitioned baseline model, and the partitioned baseline curve is output. Based on the standardized data input and the partitioned baseline curve, an adaptive threshold function is constructed using Bayesian update or adaptive threshold algorithm to self-correct the partitioned baseline model according to environmental factors. For each grid node, the personalized baseline curve and adaptive threshold sequence of each grid node are output using the corrected partitioned baseline model.
[0012] Optionally, based on the standardized data input and the personalized baseline curve, a hybrid health index is calculated, including site-level health index, grid-level health index, and health trend analysis, comprising: Based on the standardized data input and the personalized baseline curve, the impedance normalized value, the displacement potential change normalized value between grid nodes, the environmental variable normalized value, the temperature normalized value, and the corrosion index normalized value are calculated and integrated to obtain the multimodal comprehensive health index. Based on the standardized data input, the equivalent impedance of the grounding grid and the distribution of grounding points are simulated by physical modeling to obtain a digital twin model. The actual observation values are compared and analyzed with the simulation results. The multimodal comprehensive health index is corrected according to the comparison and analysis results, and a hybrid health index including site-level health index, grid-level health index and health trend analysis is output.
[0013] Optionally, based on the standardized data input and the hybrid health index, health prediction and anomaly detection are performed, and then graded alarms and alarm information are generated according to the anomaly detection results, including: Based on the standardized data input and the hybrid health index, a multimodal time series prediction model is obtained by fusing a neural network component with physical constraints and a traditional time series component. The multimodal time series prediction model is then used to predict the local equivalent impedance, the displacement potential change between grid nodes, and the hybrid health index in the future time period to obtain the health prediction results. Based on the health prediction results, anomaly pattern detection is performed using a statistical method based on adaptive thresholds and a time-series anomaly detection algorithm to identify nonlinear anomalies or sudden events, and output anomaly points, as well as the time of the anomaly points, the grid of the anomaly points, and the anomaly intensity, to obtain the anomaly detection results. Set graded alarm levels, define trigger conditions, trigger delay, alarm connection and rectification suggestions, and then generate graded alarms and alarm information based on the health prediction results and the anomaly detection results.
[0014] This invention also provides a multi-point distribution monitoring and hierarchical early warning system for grounding resistance in substation grounding grids, comprising: The substation topology architecture module is used to design the preliminary topology of the substation area based on the geographical information and grounding grid topology of the substation area, and to design a multi-point distributed sensor network based on the preliminary topology of the substation area. The edge computing module is used to collect multimodal raw datasets based on the multi-point distributed sensor network through the sensor network and redundant design, and then perform preliminary signal processing on the multimodal raw datasets to obtain multimodal cleaned datasets. The standardized input module is used to perform data time-series unification, data pipeline definition, interface contract design, and environmental variable structuring based on the multimodal cleaning dataset to obtain standardized data input. The personalized baseline module is used to construct a partitioned baseline model, and then use the standardized data input and the partitioned baseline model to output a personalized baseline curve and adaptive threshold sequence for each grid node. The health trend analysis module is used to calculate a hybrid health index, including site-level health index, grid-level health index and health trend analysis, based on the standardized data input and the personalized baseline curve. The health prediction and alarm module is used to perform health prediction and anomaly detection based on the standardized data input and the hybrid health index, and then perform hierarchical alarms and generate alarm information based on the anomaly detection results. The station area topology module, the edge computing module, the standardized input module, the personalized baseline module, the health trend analysis module, and the health prediction and alarm module are interconnected.
[0015] This invention discloses the following technical effects by providing a method and system for multi-point distribution monitoring and hierarchical early warning of grounding resistance in substation grounding grids: 1. By designing a multi-point distributed sensor network: a structured and object-oriented foundation is laid for subsequent data acquisition and alignment, ensuring consistency and comparability among different monitoring points; and redundancy considerations are used to improve system robustness and reduce the impact of single-point failures.
[0016] 2. By acquiring multimodal data and performing preliminary signal processing, edge processing and redundancy design can improve the quality of raw data and reduce subsequent analysis errors; and provide a consistent and aligned data foundation, which facilitates subsequent unified time series processing and model training.
[0017] 3. By unifying data time series and establishing contracts with pipelines and interfaces, unified time series data is formed, enabling seamless integration of cross-grid and cross-modal data, ensuring that different data sources are comparable and traceable on the same time scale; and strengthening the security and reliability of data transmission, reducing the risk of data leakage and tampering.
[0018] 4. By designing personalized baselines and environmental adaptation, historical data is decomposed for each grid node to obtain long-term trend and seasonal components, forming a partitioned baseline model and partitioned baseline curve. Bayesian updates or adaptive threshold algorithms are applied to generate adaptive threshold functions for self-correction. This allows for dynamic adjustment of thresholds under different environmental factors, improving alarm accuracy and reducing false alarms and missed alarms; it also enables grid-level health assessments, enhancing regional management capabilities.
[0019] 5. By calculating the hybrid health index and digital twin simulation, the health status of the ground network can be comprehensively evaluated, providing more comprehensive monitoring results; and by comparing physical simulation with observation data, the credibility and self-interpretability of the model can be improved.
[0020] 6. Through health prediction and anomaly detection, the system can predict future trends in local impedance, displacement potential, and health index. This enables early warning, risk reduction, and improved operational response speed and accuracy. It also provides actionable alarm strategies and rectification directions to enhance system maintainability.
[0021] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in 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.
[0023] Figure 1 This is a schematic diagram of the method flow provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the multi-point distribution monitoring process for grounding resistance provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the system architecture provided for an embodiment of the present invention. Detailed Implementation
[0024] 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.
[0025] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0026] like Figure 1 As shown, this invention provides a method for multi-point distribution monitoring and hierarchical early warning of grounding resistance in substation grounding grids, including: Step 1, as follows Figure 2 As shown, based on the geographical information and grounding grid topology of the substation area, a preliminary topology structure of the substation area is designed, and a multi-point distributed sensor network is designed based on the preliminary topology structure of the substation area; supplementary step 1 includes: 1.1 Based on the geographic information and grounding grid topology of the substation area, the area is divided into regions, and grid nodes and sensor points are designed for each region to obtain the preliminary topology of the substation area. Based on the preliminary topology of the substation area, the data flow direction, data format, clock alignment requirements and data interface are designed.
[0027] Geographic information and grounding grid topology of substation area: location coordinates of substation area, topography, grounding electrode layout method (grid, pile ground electrode, etc.), grounding grid relationship, name and unique identifier of each grounding point, soil and environmental conditions, etc.
[0028] 1.2 Based on the preliminary topology of the station area, a set of sensor modules and observation indicators are designed in combination with on-site environmental information. A gridded sensor point layout and sampling strategy are designed for the set of sensor modules to obtain a multi-point distributed sensor network.
[0029] The sensor module assembly includes a grounding status sensor, a soil temperature and humidity sensor, a grounding grid corrosion rate sensor, a potential monitoring sensor, a current sensor, and a voltage sensor. The observed indicators include grounding resistance value, surface potential difference between grid nodes, ambient temperature of grid nodes, soil moisture, and corrosion index.
[0030] Step 2, as follows Figure 2 As shown, based on the multi-point distributed sensor network, multimodal raw datasets are collected through sensor network and redundancy design, and then preliminary signal processing is performed on the multimodal raw datasets to obtain multimodal cleaned datasets; Step 2 includes: 2.1 Based on the gridded sensor point layout of the multi-point distributed sensor network, a sensor module combination is assigned to each monitoring point to perform multimodal observation at the same point and output the monitoring point coordinate table.
[0031] Specifically, one or more sensor points are assigned to each grid cell area to ensure minimal coverage blind spots; within each grid cell, key monitoring points are identified based on land conditions and sensor accessibility, such as around the ground pole, key joints, easily corroded areas, wet areas, and intersections of underground pipelines; sensor module combinations are assigned to each monitoring point to ensure that multimodal observations can be achieved at the same point, such as simultaneous observation of grounding impedance, temperature, humidity, and corrosion indicators.
[0032] 2.2 Based on the multi-point distributed sensor network and the monitoring point coordinate table, hardware redundancy strategy, communication redundancy strategy, data redundancy and caching strategy, and fault tolerance mechanism are designed. GPS is used to unify the time source of the entire network to align the sensor timestamps and complete the sensor network and redundancy design.
[0033] Hardware redundancy strategy: Configure dual-mode sensors and dual-mode operation for key grid nodes, that is, two sets of sensors work in parallel at the same monitoring point, and if one set fails, the other set continues to observe.
[0034] Communication redundancy strategy: Multi-channel transmission is used in critical areas, such as redundancy between the main channel and backup channel or short-range wireless links.
[0035] Data redundancy and caching strategy: Edge nodes have local caching, such as data from the last 24 hours, which can be continuously uploaded when network backhaul is interrupted. The caching strategy prioritizes the most recent N data entries.
[0036] Fault tolerance mechanism: Mesh topology allows for hopping between nodes and automatically reroutes when some nodes fail, ensuring uninterrupted data collection.
[0037] 2.3 Local data acquisition is performed using the multi-point distributed sensor network. Noise suppression, filtering, denoising, time alignment, and missing data filling are performed on the acquired data at the edge nodes to obtain a multimodal raw dataset. Based on the multimodal raw dataset, local equivalent impedance, local surface displacement, potential, temperature, and humidity are calculated to obtain local indicators. Then, the initial local health index is calculated using the local indicators, and a multimodal cleaning dataset is output.
[0038] Step 3, as follows Figure 2 As shown, based on the multimodal cleaning dataset, data time series unification, data pipeline definition, interface contract design, and environmental variable structuring are performed to obtain standardized data input; step 3 includes: 3.1 Based on the multimodal cleaned dataset, field definitions, observation indicator unit standardization, missing value handling, and environmental variable structuring are performed to obtain unified time series data. The defined fields include nodes, measurement points, timestamps, observation values, units, sensor status, data quality labels, and environmental metadata. The data pipeline includes data flow direction, preprocessing process, data quality assessment, missing value imputation strategy, and timestamp alignment rules.
[0039] 3.2 Based on the aforementioned multimodal cleaned dataset, time synchronization and alignment, data cleaning and quality assessment, missing value imputation, and data packaging and transmission format definition are performed to obtain a data pipeline. The data pipeline includes data flow direction, processing nodes, data storage location, latency constraints, etc.
[0040] 3.3 Based on the unified time-series data, a connection protocol is selected, a message structure is defined, and data security and authentication are performed using TLS encryption, device certificates, token authentication, key rotation strategies, and log auditing to obtain the interface contract document. Common protocols such as MQTT or OPC UA / REST are selected for the connection protocol, taking into account on-site security policies.
[0041] 3.4 Based on the unified time-series data and environmental data, a mapping relationship between environmental factors and local indicators is constructed to obtain an environmental variable table. Then, the interface contract document and the environmental variable table are organized into a structured input format to obtain standardized data input.
[0042] Step 4, as follows Figure 2 As shown, a partitioned baseline model is constructed, and then the standardized data input and the partitioned baseline model are used to output the personalized baseline curve and adaptive threshold sequence for each grid node; step 4 includes: 4.1 For each grid node, the historical data is decomposed into a time series, and the historical trend, seasonal components and residuals are output. The historical trend is fitted using robust regression to obtain a long-term trend model. The seasonal components are fitted using a periodic sliding window to obtain a seasonal baseline component. The long-term trend model and the seasonal baseline component are combined to obtain a partitioned baseline model. The partitioned baseline curve is output, which includes the trend and seasonal components.
[0043] 4.2 Based on the standardized data input and the partitioned baseline curve, an adaptive threshold function is constructed using Bayesian update or adaptive threshold algorithm to self-correct the partitioned baseline model according to environmental factors.
[0044] 4.3 For each grid node, the personalized baseline curve and adaptive threshold sequence of each grid node are output using the corrected partitioned baseline model.
[0045] Step 5, as follows Figure 2 As shown, based on the standardized data input and the personalized baseline curve, a hybrid health index is calculated, including site-level health index, grid-level health index, and health trend analysis; step 5 includes: 5.1 Based on the standardized data input and the personalized baseline curve, the impedance normalized value, the displacement potential change normalized value between grid nodes, the environmental variable normalized value, the temperature normalized value, and the corrosion index normalized value are calculated and integrated to obtain the multimodal comprehensive health index.
[0046] 5.2 Based on the standardized data input, the equivalent impedance of the grounding grid and the distribution of grounding points are simulated by physical modeling. Boundary conditions, material parameters, contact resistance, etc. are set to obtain a digital twin model. The actual observation values are compared and analyzed with the simulation results. The multimodal comprehensive health index is corrected according to the comparison and analysis results, and a hybrid health index including site-level health index, grid-level health index and health trend analysis is output.
[0047] Step 6: Based on the standardized data input and the hybrid health index, perform health prediction and anomaly detection, and then generate tiered alarms and alarm information based on the anomaly detection results. Step 6 includes: 6.1 Based on the standardized data input and the hybrid health index, a multimodal time series prediction model is obtained by fusing a neural network component with physical constraints and a traditional time series component. The multimodal time series prediction model is used to predict the local equivalent impedance, the displacement potential change between grid nodes and the hybrid health index in the future time period to obtain the health prediction result.
[0048] Physically constrained neural network components, such as LSTM and GRU, have physical loss terms to ensure that the output conforms to physical laws; traditional time series components, such as ARIMA and exponential smoothing, are used to capture linear and short-term trends.
[0049] 6.2 Based on the health prediction results, anomaly pattern detection is performed using a statistical method based on adaptive thresholds and a time-series anomaly detection algorithm to identify nonlinear anomalies or sudden events, and anomaly points, as well as anomaly point time, anomaly point grid, and anomaly intensity, to obtain anomaly detection results; 6.3 Set the graded alarm levels, define the trigger conditions, trigger delay, alarm connection and rectification suggestions, and then generate graded alarms and alarm information based on the health prediction results and the anomaly detection results.
[0050] like Figure 3 As shown, the present invention also provides a multi-point distribution monitoring and hierarchical early warning system for grounding resistance of substation grounding grids, including: The substation topology architecture module is used to design the preliminary topology of the substation area based on the geographical information and grounding grid topology of the substation area, and to design a multi-point distributed sensor network based on the preliminary topology of the substation area. The edge computing module is used to collect multimodal raw datasets based on the multi-point distributed sensor network through the sensor network and redundant design, and then perform preliminary signal processing on the multimodal raw datasets to obtain multimodal cleaned datasets. The standardized input module is used to perform data time-series unification, data pipeline definition, interface contract design, and environmental variable structuring based on the multimodal cleaning dataset to obtain standardized data input. The personalized baseline module is used to construct a partitioned baseline model, and then use the standardized data input and the partitioned baseline model to output a personalized baseline curve and adaptive threshold sequence for each grid node. The health trend analysis module is used to calculate a hybrid health index, including site-level health index, grid-level health index and health trend analysis, based on the standardized data input and the personalized baseline curve. The health prediction and alarm module is used to perform health prediction and anomaly detection based on the standardized data input and the hybrid health index, and then perform hierarchical alarms and generate alarm information based on the anomaly detection results. The station area topology module, the edge computing module, the standardized input module, the personalized baseline module, the health trend analysis module, and the health prediction and alarm module are interconnected.
[0051] Therefore, this invention provides a method and system for multi-point distribution monitoring and hierarchical early warning of grounding resistance in substation grounding networks. Through multi-modal data standardization for gridded grounding networks, adaptive thresholding of partitioned baselines, digital twin comparison correction, and multi-stage health prediction and hierarchical alarm, it significantly improves the monitoring accuracy, prediction capability, and operation and maintenance decision-making efficiency of substation grounding networks.
[0052] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.
[0053] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.
Claims
1. A method for multi-point distribution monitoring and hierarchical early warning of grounding resistance in substation grounding grids, characterized in that, include: Based on the geographical information and grounding grid topology of the substation area, a preliminary topology of the substation area is designed, and a multi-point distributed sensor network is designed based on the preliminary topology of the substation area. Based on the multi-point distributed sensor network, multimodal raw datasets are collected through sensor network and redundancy design, and then preliminary signal processing is performed on the multimodal raw datasets to obtain multimodal cleaned datasets. Based on the multimodal cleaning dataset, data time series unification, data pipeline definition, interface contract design, and environmental variable structuring are performed to obtain standardized data input; Construct a partitioned baseline model, and then use the standardized data input and the partitioned baseline model to output a personalized baseline curve and adaptive threshold sequence for each grid node; Based on the standardized data input and the personalized baseline curve, a hybrid health index is calculated, including site-level health index, grid-level health index, and health trend analysis. Based on the standardized data input and the hybrid health index, health prediction and anomaly detection are performed, and then graded alarms and alarm information are generated according to the anomaly detection results.
2. The method for multi-point distribution monitoring and hierarchical early warning of grounding resistance for substation grounding grids according to claim 1, characterized in that, Based on the geographical information and grounding grid topology of the substation area, a preliminary topology structure of the substation area is designed. Based on this preliminary topology structure, a multi-point distributed sensor network is designed, including: Based on the geographic information and grounding grid topology of the substation area, the area is divided into regions, and grid nodes and sensor points are designed for each region to obtain the preliminary topology of the substation area. Based on the preliminary topology of the substation area, the data flow direction, data format, clock alignment requirements and data interface are designed. Based on the preliminary topology of the station area, a set of sensor modules and observation indicators are designed in combination with on-site environmental information. A gridded sensor point layout and sampling strategy are designed for the set of sensor modules to obtain a multi-point distributed sensor network.
3. The method for multi-point distribution monitoring and hierarchical early warning of grounding resistance for substation grounding grids according to claim 2, characterized in that, The sensor module assembly includes a grounding status sensor, a soil temperature and humidity sensor, a grounding grid corrosion rate sensor, a potential monitoring sensor, a current sensor, and a voltage sensor. The observed indicators include grounding resistance value, surface potential difference between grid nodes, ambient temperature of grid nodes, soil moisture, and corrosion index.
4. The method for multi-point distribution monitoring and hierarchical early warning of grounding resistance for substation grounding grids according to claim 3, characterized in that, Based on the aforementioned multi-point distributed sensor network, multimodal raw datasets are collected through sensor network and redundancy design. Preliminary signal processing is then performed on the multimodal raw datasets to obtain a multimodal cleaned dataset, including: Based on the gridded sensor point layout of the multi-point distributed sensor network, a sensor module combination is assigned to each monitoring point to perform multimodal observation at the same point and output a monitoring point coordinate table. Based on the multi-point distributed sensor network and the monitoring point coordinate table, hardware redundancy strategy, communication redundancy strategy, data redundancy and caching strategy and fault tolerance mechanism are designed. GPS is used to unify the time source of the entire network to align the sensor timestamps and complete the sensor network and redundancy design. Local data is collected using the multi-point distributed sensor network. Noise suppression, filtering, denoising, temporal alignment, and missing data filling are performed on the collected data at the edge nodes to obtain a multimodal raw dataset. Based on the multimodal raw dataset, local equivalent impedance, local surface displacement, potential, temperature, and humidity are calculated to obtain local indicators. Then, the initial local health index is calculated using the local indicators, and a multimodal cleaning dataset is output.
5. The method for multi-point distribution monitoring and hierarchical early warning of grounding resistance for substation grounding grids according to claim 4, characterized in that, Based on the aforementioned multimodal cleaning dataset, data time series unification, data pipeline definition, interface contract design, and environmental variable structuring are performed to obtain standardized data input, including: Based on the multimodal cleaned dataset, field definitions, observation unit standardization, missing value handling, and environmental variable structuring are performed to obtain unified time series data. Based on the multimodal cleaned dataset, time synchronization and alignment, data cleaning and quality assessment, missing value imputation, and data packaging and transmission format definition are performed to obtain a data pipeline; Based on the unified time-series data, an interface protocol is selected, a message structure is defined, and TLS encryption, device certificates, token authentication, key rotation strategy, and log auditing are used for data security and authentication to obtain an interface contract document. Based on the unified time-series data and environmental data, a mapping relationship between environmental factors and local indicators is constructed to obtain an environmental variable table. Then, the interface contract document and the environmental variable table are organized into a structured input format to obtain standardized data input.
6. The method for multi-point distribution monitoring and hierarchical early warning of grounding resistance for substation grounding grids according to claim 5, characterized in that, The defined fields include nodes, measurement points, timestamps, observation values, units, sensor status, data quality labels, and environmental metadata. The data pipeline includes data flow direction, preprocessing process, data quality assessment, missing value imputation strategy, and timestamp alignment rules.
7. The method for multi-point distribution monitoring and hierarchical early warning of grounding resistance for substation grounding grids according to claim 6, characterized in that, Construct a partitioned baseline model, and then use the standardized data input and the partitioned baseline model to output a personalized baseline curve and adaptive threshold sequence for each grid node, including: For each grid node, the historical data is decomposed into a time series, and the historical trend, seasonal components and residuals are output. The historical trend is fitted using robust regression to obtain a long-term trend model. The seasonal components are fitted using a periodic sliding window to obtain a seasonal baseline component. The long-term trend model and the seasonal baseline component are combined to obtain a partitioned baseline model, and the partitioned baseline curve is output. Based on the standardized data input and the partitioned baseline curve, an adaptive threshold function is constructed using Bayesian update or adaptive threshold algorithm to self-correct the partitioned baseline model according to environmental factors. For each grid node, the personalized baseline curve and adaptive threshold sequence of each grid node are output using the corrected partitioned baseline model.
8. The method for multi-point distribution monitoring and hierarchical early warning of grounding resistance for substation grounding grids according to claim 7, characterized in that, Based on the standardized data input and the personalized baseline curve, a hybrid health index is calculated, including site-level health index, grid-level health index, and health trend analysis, comprising: Based on the standardized data input and the personalized baseline curve, the impedance normalized value, the displacement potential change normalized value between grid nodes, the environmental variable normalized value, the temperature normalized value, and the corrosion index normalized value are calculated and integrated to obtain the multimodal comprehensive health index. Based on the standardized data input, the equivalent impedance of the grounding grid and the distribution of grounding points are simulated by physical modeling to obtain a digital twin model. The actual observation values are compared and analyzed with the simulation results. The multimodal comprehensive health index is corrected according to the comparison and analysis results, and a hybrid health index including site-level health index, grid-level health index and health trend analysis is output.
9. The method for multi-point distribution monitoring and hierarchical early warning of grounding resistance for substation grounding grids according to claim 8, characterized in that, Based on the standardized data input and the hybrid health index, health prediction and anomaly detection are performed. Then, based on the anomaly detection results, tiered alarms and alarm information generation are conducted, including: Based on the standardized data input and the hybrid health index, a multimodal time series prediction model is obtained by fusing a neural network component with physical constraints and a traditional time series component. The multimodal time series prediction model is then used to predict the local equivalent impedance, the displacement potential change between grid nodes, and the hybrid health index in the future time period to obtain the health prediction results. Based on the health prediction results, anomaly pattern detection is performed using a statistical method based on adaptive thresholds and a time-series anomaly detection algorithm to identify nonlinear anomalies or sudden events, and output anomaly points, as well as the time of the anomaly points, the grid of the anomaly points, and the anomaly intensity, to obtain the anomaly detection results. Set graded alarm levels, define trigger conditions, trigger delay, alarm connection and rectification suggestions, and then generate graded alarms and alarm information based on the health prediction results and the anomaly detection results.
10. A multi-point distribution monitoring and hierarchical early warning system for grounding resistance of substation grounding grids, characterized in that, include: The substation topology architecture module is used to design the preliminary topology of the substation area based on the geographical information and grounding grid topology of the substation area, and to design a multi-point distributed sensor network based on the preliminary topology of the substation area. The edge computing module is used to collect multimodal raw datasets based on the multi-point distributed sensor network through the sensor network and redundant design, and then perform preliminary signal processing on the multimodal raw datasets to obtain multimodal cleaned datasets. The standardized input module is used to perform data time-series unification, data pipeline definition, interface contract design, and environmental variable structuring based on the multimodal cleaning dataset to obtain standardized data input. The personalized baseline module is used to construct a partitioned baseline model, and then use the standardized data input and the partitioned baseline model to output a personalized baseline curve and adaptive threshold sequence for each grid node. The health trend analysis module is used to calculate a hybrid health index, including site-level health index, grid-level health index and health trend analysis, based on the standardized data input and the personalized baseline curve. The health prediction and alarm module is used to perform health prediction and anomaly detection based on the standardized data input and the hybrid health index, and then perform hierarchical alarms and generate alarm information based on the anomaly detection results. The station area topology module, the edge computing module, the standardized input module, the personalized baseline module, the health trend analysis module, and the health prediction and alarm module are interconnected.