A method and system for evaluating the grid structure of low-voltage power grids

By integrating multi-source data and extracting features, constructing spatiotemporal sequences, dividing dynamic graphs, encoding topology structures, and weighted evaluation, this method solves the problem of early warning in existing low-voltage power grid evaluation technologies. It enables the diagnosis of the health of the power grid structure and the location of weak links, thereby improving the foresight of power grid transformation decisions and the efficiency of resource allocation.

CN123069291APending Publication Date: 2026-11-10STATE GRID ZHEJIANG ELECTRIC POWER CO LTD YUNHE COUNTY POWER SUPPLY CO +1
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202611569040.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-12
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing low-voltage power grid evaluation technologies are insufficient for early warning and quantitative assessment before structural hazards cause significant deterioration in operational indicators, resulting in delayed and untargeted decisions regarding power grid transformation and upgrading.

Method used

By receiving topological and operational data, multi-source heterogeneous data acquisition and fusion are performed to generate a multi-dimensional feature dataset, construct a spatiotemporal feature coordinate system, extract key performance index sequences, generate a quantized structure spatiotemporal graph, use an adaptive segmentation algorithm to divide the graph, perform graph structure feature extraction and pattern recognition, establish spatial correlation between structural topological data and grid-transformed data, define evaluation weight vectors for weighted fusion calculation, and output structural anomaly statistics.

Benefits of technology

It enables early warning and quantitative assessment of the physical structure of low-voltage power grids, enhances the foresight and pertinence of power grid upgrade decisions, and provides a direct quantitative reflection of the power grid's health status and risk level.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123069291A_ABST
    Figure CN123069291A_ABST
Patent Text Reader

Abstract

This invention relates to the field of power grid safety technology, and particularly to a method and system for assessing the grid structure of low-voltage power grids. The method includes: receiving topology data and operational data of a target low-voltage power grid to generate a multi-dimensional feature dataset; performing feature engineering on the multi-dimensional feature dataset to extract key performance indicator sequences as grid conversion data; dividing the quantified structure spatiotemporal map to output multiple quantified monitoring segments; encoding the structural topology data according to a predefined pattern library to generate quantified structural values; establishing a spatial correlation between the structural topology data and the grid conversion data; calculating the structural matching quantification value to output the structural anomaly statistics. This approach solves the technical problem that existing evaluation technologies struggle to achieve early warning and quantified assessment of structures, improving the foresight, targeting, and resource allocation efficiency of power grid upgrade decisions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of power grid safety technology, and in particular to a method and system for evaluating the grid structure of a low-voltage power grid. Background Technology

[0002] As a crucial link connecting the main grid and end users, the low-voltage power grid's operational status directly determines power supply reliability and power quality. With the widespread integration of new power sources and loads, the operating characteristics of the low-voltage power grid are undergoing profound changes, shifting from a traditional unidirectional power flow model to a complex bidirectional interaction model between power sources and loads. This places higher demands on the grid's carrying capacity, flexible regulation, and safe and stable operation. To address these challenges and promote the evolution of the low-voltage power grid towards robustness, flexibility, and intelligence, a scientific evaluation of the rationality of its physical structure has become an important prerequisite for guiding precise investment and intelligent upgrading of the power grid.

[0003] In the field of power grid evaluation and management, some explorations have been made in existing technologies. One existing technology (application publication number CN115965258A) discloses a grid-based platform management method, which divides the grid into target grid units and calculates and evaluates grid data based on operational indicators such as grid line loss rate, power supply reliability rate, and electricity bill collection rate, thereby identifying weak grid units. This method grids the management area and uses operational result data to evaluate grid performance and trace problems, which plays a certain role in improving the operational management efficiency of the distribution network. However, this method mainly focuses on the statistical analysis and management response of historical and current power grid operational results, and its evaluation system relies on ex-post statistical indicators such as "grid line loss rate" and "power supply reliability rate." Although these indicators can reflect the operational effect over a period of time, they are essentially a comprehensive result of the combined effects of multiple factors such as power grid structure, equipment status, operating mode, and external load, rather than a direct diagnosis of the health and rationality of the inherent "structural" characteristics of the power grid, such as physical topology, equipment connection relationships, and line parameter configuration. Therefore, existing grid-based evaluation technologies based on operational indicators are insufficient to provide early warning and quantitative assessment of structural weaknesses, capacity bottlenecks, and adaptability issues to future new businesses (such as high-proportion distributed energy access and electric vehicle charging) before potential structural hazards in the power grid cause significant deterioration in operational indicators. This results in power grid upgrade decisions often lagging behind the emergence of problems, and investments lacking specificity, making it difficult to achieve the transformation from "passive repair" to "proactive defense" and from "solving existing problems" to "preventing potential risks." Summary of the Invention

[0004] In view of the above-mentioned shortcomings or disadvantages, the present invention provides a method and system for evaluating the grid structure of low-voltage power grids, which can solve the technical problem that existing evaluation technologies are unable to achieve early warning and quantitative evaluation of the structure.

[0005] This invention provides a method for evaluating the grid structure of a low-voltage power grid, comprising: The system receives topology and operational data from the target low-voltage power grid, performs multi-source heterogeneous data acquisition and fusion, and generates a multi-dimensional feature dataset. A spatiotemporal feature coordinate system is constructed, and feature engineering is performed on the multi-dimensional feature dataset to extract key performance indicator sequences as grid conversion data. Based on the grid conversion data, a quantized structure spatiotemporal map is generated in the spatiotemporal feature coordinate system, and an adaptive segmentation algorithm is used to divide the quantized structure spatiotemporal map, outputting multiple quantized monitoring segments. Graph structure feature extraction and pattern recognition are performed on the multi-dimensional feature dataset to obtain structural topology data, which is then encoded according to a predefined pattern library to generate structural quantization assignments. Spatial correlation is established between the structural topology data and the grid conversion data, and structural matching quantization values ​​are calculated. An evaluation weight vector is defined, and a weighted fusion calculation is performed on the structural quantization assignments and structural matching quantization values ​​based on the evaluation weight vector, outputting the structural anomaly statistics.

[0006] According to a second aspect, the present invention provides a grid structure evaluation system for low-voltage power grids, comprising: The data fusion and feature extraction module is used to receive topology data and operational data of the target low-voltage power grid, perform multi-source heterogeneous data acquisition and fusion, and generate a multi-dimensional feature dataset.

[0007] The spatiotemporal sequence construction module is used to construct a spatiotemporal feature coordinate system, perform feature engineering on multidimensional feature datasets, and extract key performance index sequences as grid-transformed data.

[0008] The dynamic map segmentation module is used to generate a quantized structure spatiotemporal map in a spatiotemporal feature coordinate system based on grid-converted data, and to divide the quantized structure spatiotemporal map using an adaptive segmentation algorithm to output multiple quantized monitoring segments.

[0009] The topology encoding module is used to extract graph structure features and recognize patterns from a multidimensional feature dataset to obtain structural topology data. It then encodes the structural topology data according to a predefined pattern library to generate structural quantization values.

[0010] The spatiotemporal correlation analysis module is used to establish the spatial correlation between structural topology data and grid-converted data, and to calculate the structural matching quantization value.

[0011] The weighted evaluation and output module is used to define the evaluation weight vector, perform weighted fusion calculation on the structure quantization assignment and structure matching quantization value based on the evaluation weight vector, and output the structural anomaly statistics.

[0012] According to a third aspect, the present invention provides a computer device comprising: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform any low-voltage power grid grid structure evaluation method in the embodiments of the present invention.

[0013] According to another aspect of the present invention, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute a grid structure evaluation method for any low-voltage power grid in the embodiments of the present invention.

[0014] The present invention provides a method for evaluating the grid structure of a low-voltage power grid. This method is achieved through six core steps: data fusion and feature extraction, spatiotemporal sequence construction, dynamic graph partitioning, topology encoding, spatiotemporal correlation analysis, and weighted evaluation and output. Specifically, it receives topology and operational data of the target low-voltage power grid to perform multi-source heterogeneous data acquisition and fusion, generating a multi-dimensional feature dataset covering static connectivity and dynamic operational status, providing a complete data foundation for subsequent analysis. A spatiotemporal feature coordinate system is constructed, and feature engineering is performed on the multi-dimensional feature dataset to extract key performance indicator sequences as grid transformation data, achieving dimensionality reduction and quantifiable transformation of massive heterogeneous operational data. Based on the grid transformation data, a quantized structure spatiotemporal graph is generated in the spatiotemporal feature coordinate system, and an adaptive segmentation algorithm is used to divide it into multiple quantifiable monitoring segments, transforming the continuous operational data stream into a discrete, independently divisible graph. The system analyzes the state monitoring unit; it extracts graph structure features and performs pattern recognition on a multi-dimensional feature dataset to obtain structural topology data, and encodes it according to a predefined pattern library to generate structural quantification values, thus completing the abstraction and standardization scoring of the physical connection form of the power grid; it establishes spatial correlation between structural topology data and grid transformation data to calculate structural matching quantification values, realizing the correlation measurement between specific topology forms and their corresponding regional operational anomalies; it defines an evaluation weight vector, and performs weighted fusion calculation on structural quantification values ​​and structural matching quantification values ​​based on the vector to output structural anomaly statistics, ultimately generating a comprehensive quantitative index characterizing the rationality of the power grid grid structure.

[0015] In this technical solution, the present invention addresses the technical problem described in the background section: existing evaluation technologies rely on ex-post statistical operational indicators such as "grid line loss rate" and "power supply reliability rate," making it difficult to provide early warning and quantitative assessment of the rationality of the power grid's physical structure before structural hazards cause operational problems. By introducing steps for extracting "structural topology data" and generating "structural quantitative values," as well as calculating "structural matching quantitative values," the invention creatively shifts the evaluation focus from "operational results" to "the structure itself." The "structural quantitative values" directly score the health of the power grid's physical topology, while the "structural matching quantitative values" reveal the correlation between a specific topology and the historical anomalies in its associated regions. Together, they constitute a forward-looking diagnostic basis for the inherent vulnerability of the structure. Finally, the "structural anomaly statistics" output through weighted fusion calculation is a scalar value that is independent of traditional operational result indicators and directly quantifies the health status and risk level of the power grid's physical structure. Therefore, the technical solution of this invention solves the technical problem that existing evaluation technologies struggle to provide early warning and quantitative assessment of structures, improving the foresight, targeting, and resource allocation efficiency of power grid upgrade decisions. Attached Figure Description

[0016] Figure 1 This is a flowchart of a low-voltage power grid grid structure evaluation method according to an embodiment of the present invention; Figure 2 This diagram illustrates the complete application process of the grid structure evaluation method of another embodiment of the present invention in the pre-modification evaluation of a specific area; Figure 3 This is a schematic diagram of the structure of a low-voltage power grid grid structure evaluation system according to an embodiment of the present invention; Figure 4 This is a block diagram of a computer device for implementing embodiments of the present invention. Detailed Implementation

[0017] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0018] During the development of this invention, researchers, through extensive experiments and data analysis, discovered a significant intrinsic correlation between the physical topology characteristics of low-voltage power grids and the abnormal patterns of their corresponding regional operational data: specific power grid topologies (such as radial or single-ring networks) exhibit a stable mapping pattern with the historical frequency and spatial distribution of abnormal operational indicators such as voltage fluctuations and load anomalies within their power supply areas. Based on this relationship, researchers innovatively proposed this technical solution, which uses structural topology data to encode and quantify the physical morphology, captures the spatiotemporal changes in operational status through grid-based data conversion, and combines the spatial correlation analysis between the two to demonstrate the establishment of a statistical correlation between power grid topology and historical operational anomalies, thereby realizing the analytical logic of tracing structural rationality from abnormal manifestations.

[0019] Specifically, through comparative experiments, the invention team discovered that traditional lagging evaluation methods based on operational indicators (such as line loss rate and voltage qualification rate) have a technical deficiency: they cannot provide early warnings before structural hazards evolve into operational accidents. Their evaluation conclusions are "diagnoses" after operational anomalies occur, rather than "health check reports" of structural risks. However, the assessment method based on structural quantification and spatiotemporal correlation proposed in this invention can improve the ability to proactively perceive the vulnerability of power grid structures. By constructing a spatiotemporal feature coordinate system and extracting grid-transformed data, unified spatiotemporal modeling and key feature extraction of massive and heterogeneous operational data can be achieved. By extracting graph structure features and generating structural quantification values, standardized and digital scoring of abstract power grid topology can be achieved, overcoming the limitations of traditional methods that can only process numerical operational data and cannot directly quantify and evaluate non-numerical structural features. By establishing spatial correlations and calculating structural matching quantification values, the "static score" of a specific topology can be scientifically bound to the "dynamic anomaly history" of its associated region, forming a joint evaluation basis of "inherent structural attributes - historical operational performance". Through the weighted fusion calculation and structural anomaly statistical output mechanism formed by the above steps, a comprehensive early warning indicator that is independent of a single operational indicator and directly quantifies the health status and potential risk level of the power grid's physical structure can be generated, thus providing direct data support for preventive and precise transformation decisions of the power grid.

[0020] Therefore, this invention provides a method for evaluating the grid structure of a low-voltage power grid, which can be applied to an intelligent evaluation and decision support system for low-voltage power grids (hereinafter referred to as the "system"). This system integrates IoT sensing, multi-source data fusion, intelligent analysis models, and visualization interaction modules to quantitatively evaluate the rationality of the physical topology of a low-voltage power grid, diagnose its health, and locate weak points. Specifically, this system can be deployed in various hardware environments, including but not limited to: cloud server clusters, local data center servers of power grid companies, or edge computing devices deployed in substations / power supply stations. This flexible deployment architecture allows the system to meet the large-scale analysis needs of provincial or municipal power grid companies for centralized, batch-based periodic structural surveys and ratings of low-voltage distribution areas across the entire region, while also adapting to the lightweight, low-latency field application requirements of specific power supply stations or maintenance teams for real-time monitoring, fault tracing, or instant simulation of transformation plans for key areas of the power grid.

[0021] like Figure 1 As shown, the method may include: Step S110: Receive topology data and operation data of the target low-voltage power grid, perform multi-source heterogeneous data acquisition and fusion, and generate a multi-dimensional feature dataset.

[0022] Among them, topology data refers to structured data describing the physical connection relationships of equipment (such as transformers, switches, and lines) in the power grid, including equipment identification, electrical parameters, and connection matrices; operational data refers to data collected in real time by monitoring equipment that reflects the dynamic operating status of the power grid, including electrical quantities such as voltage, current, power, and frequency; multi-source heterogeneous data acquisition and fusion refers to the process of accessing and integrating raw data from different manufacturers, different types of sensors, or management systems with different formats and sampling rates through a unified Internet of Things (IoT) interface protocol; and multi-dimensional feature datasets refer to standardized data sets containing multi-dimensional information such as spatial, temporal, and electrical attributes, formed after cleaning, aligning, and initially structuring the fused raw data.

[0023] Specifically, the system can periodically or triggerively collect topology and operational data through various sensors deployed in distribution transformer areas, branch boxes, smart meters, and other locations, as well as by connecting to the distribution automation system and the electricity consumption information collection system. The data is then transmitted to the data platform via message queues (such as the MQTT protocol) for time synchronization, unit unification, and associated storage.

[0024] For example, when the system evaluates a low-voltage distribution area in an old residential community, the topology data collected includes the connection relationship and wire diameter parameters of one distribution transformer, 12 outgoing lines, and 45 meter boxes; the operational data includes the effective values ​​of voltage and current collected every 15 minutes during the evaluation period of the above nodes; the final multidimensional feature dataset contains tens of thousands of time-space-electrical quantity correlation records.

[0025] Step S120: Construct a spatiotemporal feature coordinate system, perform feature engineering on the multidimensional feature dataset, and extract the key performance index sequence as grid transformation data.

[0026] Among them, the spatiotemporal feature coordinate system is a two-dimensional rectangular coordinate system with the data acquisition timestamp as the horizontal axis (T-axis) and the data value as the vertical axis (V-axis), used to map the data to a unified spatiotemporal framework; feature engineering refers to the operations of preprocessing, filtering and transforming the data in the coordinate system, aiming to extract features that are sensitive to structural assessment and quantifiable; the key performance index sequence refers to the data sequence formed by arranging the core quantitative indicators (such as voltage deviation rate and load rate) that can characterize the health status of the power grid operation in chronological order after screening; grid-converted data specifically refers to the set of key performance index sequences that can be converted from the multidimensional feature dataset and used for subsequent gridded structural analysis.

[0027] Specifically, the system can map data values ​​to corresponding coordinate points in a spatiotemporal feature coordinate system by parsing the device location identifier and collection timestamp associated with each data point in the multidimensional feature dataset. Subsequently, it performs feature selection operations to automatically filter out indicators with numerical data types and clear physical meanings (such as voltage, current, and power) and filter out unstructured data such as text-based device nameplate information.

[0028] For example, the system selects "phase A voltage deviation rate" as the key performance indicator for the data of the old residential area mentioned above. It generates a continuously fluctuating curve in the coordinate system by taking the voltage data of all monitoring points within the evaluation period and arranging them in the order of their timestamps. This curve is a "phase A voltage deviation rate" indicator sequence. The set of all such sequences constitutes grid-converted data.

[0029] Step S130: Based on the grid-converted data, generate a quantized structure spatiotemporal map in the spatiotemporal feature coordinate system, and use an adaptive segmentation algorithm to divide the quantized structure spatiotemporal map to output multiple quantized monitoring segments.

[0030] Among them, the quantized structure spatiotemporal graph refers to a dynamic graph that can intuitively reflect the evolution of indicators over time by connecting one or more key performance indicator sequences into a smooth curve through interpolation algorithms in a spatiotemporal feature coordinate system; the adaptive segmentation algorithm is an algorithm that automatically divides a continuous curve into several sub-intervals based on a preset step size shift decision boundary; the quantized monitoring segment refers to a continuous time period and its corresponding data sub-sequence that are divided from the curve of the quantized structure spatiotemporal graph through the adaptive segmentation algorithm, and each segment is used for independent analysis.

[0031] Specifically, the system can connect the discrete coordinate points of the key performance index sequence into a continuous grid quantization curve using linear interpolation or spline interpolation algorithms. Then, a straight line parallel to the time axis (T-axis) and movable in the direction of the numerical axis (V-axis) is introduced into the graph as a monitoring and screening axis. A fixed offset interval (such as 0.5 percentage points for voltage deviation) is set. Starting from the minimum point of the curve, the monitoring and screening axis is driven to gradually shift towards the maximum point at this interval. Each time it moves, the area where the axis intersects with the curve is divided into an independent interval.

[0032] For example, the system plots the "A-phase voltage deviation rate" curve on the quantization structure spatiotemporal diagram, setting the offset interval to 0.5%. Starting from a deviation rate of -1.0%, the monitoring and screening axis moves up by 0.5% each time, successively passing through -0.5%, 0%, +0.5%, etc., thereby dividing the entire curve into multiple deviation intervals, such as "the period of deviation between -1.0% and -0.5%", "the period of deviation between -0.5% and 0%", etc., with each period being a quantization monitoring segment.

[0033] Step S140: Extract graph structure features and perform pattern recognition on the multidimensional feature dataset to obtain structural topology data, and encode the structural topology data according to the predefined pattern library to generate structural quantization values.

[0034] Among them, graph structure feature extraction and pattern recognition refers to the process of identifying typical physical connection patterns of the power grid (such as radial, ring, and grid) from topological data describing the connection relationships of devices; structural topology data is an abstract representation describing the actual physical connection patterns of the power grid; the predefined pattern library is a pre-established data information containing various standard power grid topology patterns (such as radial, single-ring, double-ring, and grid) and their corresponding health scores; structural quantification assignment refers to the basic score that characterizes the inherent rationality of the topology structure obtained from the predefined pattern library based on the standard patterns matched by the structural topology data.

[0035] Specifically, the system can automatically identify whether a low-voltage power grid belongs to a "radial," "dual-power-supply chain," or "simple ring network" pattern by analyzing the device connection matrix, and then match the identification results with a predefined pattern library. The predefined pattern library has set empirical scores for each standard pattern, and the scoring criteria include power supply reliability, power transfer capacity, and operational flexibility.

[0036] For example, if the system identifies the target area as a purely "radial" structure, the baseline score for matching this pattern from the predefined pattern library is B1 (e.g., 60 points); if it is identified as a "single-ring network" structure, the baseline score is B2 (e.g., 75 points). This score is the structure quantification assignment. For example, after the system matches the "radial" paradigm, its standard score read from the knowledge base is 60 (out of 100). To facilitate subsequent weighted fusion with the structure matching quantification value, this score needs to be normalized. An exemplary linear normalization method is: divide the standard score by 100 and map it to... Interval. In this example, the normalized structure quantization assignment is 0.6.

[0037] Step S150: Establish the spatial correlation between the structural topology data and the mesh transformation data, and calculate the structural matching quantization value.

[0038] Spatial correlation refers to establishing a mapping relationship between the physical topological area of ​​the power grid (such as a radial branch) and the operational data collected by the monitoring equipment deployed in that area; the structural matching quantification value is a statistic used to quantify the frequency or density of historical operational data anomalies within the area covered by a specific topological structure.

[0039] Specifically, the system first constructs a mapping table to clarify which monitoring devices (whose data constitute grid-converted data) are associated with each line and each distribution area. Then, for each quantified monitoring segment divided in step S130, based on preset anomaly judgment conditions (such as voltage exceeding limits), the proportion of abnormal data points within that segment is calculated, referred to as the anomaly percentage. Finally, the anomaly percentages of all quantified monitoring segments belonging to the same topology are aggregated (e.g., averaged), and the result is the quantified value of the structure matching for that topology.

[0040] For example, the system identifies five monitoring points associated with the "radial" main line. Analysis of the "A-phase voltage deviation rate" of these points across all quantified monitoring sections over the past month reveals that 10% of the sections experienced abnormally low voltage. Therefore, the structural matching quantification value for this "radial" structure is recorded as 0.1 (or 10%).

[0041] Step S160: Define the evaluation weight vector, perform weighted fusion calculation on the structure quantization assignment and structure matching quantization value based on the evaluation weight vector, and output the structure anomaly statistics.

[0042] Among them, the evaluation weight vector is a set of multiple weight coefficients, each weight coefficient corresponding to the importance ratio of a key performance indicator (or its derived structural matching quantification value) in the final evaluation; the weighted fusion calculation refers to the linear weighted summation of the structural quantification assignment and one or more structural matching quantification values ​​according to the proportion specified by the evaluation weight vector; the structural anomaly statistics is a scalar value, which is the result of the weighted fusion calculation. Its value directly and comprehensively reflects the rationality level of the target low-voltage power grid grid structure. The lower the value, the healthier the structure usually is.

[0043] Specifically, the system assigns weights to various indicators based on a pre-configured evaluation strategy (such as prioritizing safety or reliability). For example, in the evaluation strategy for older residential communities, a higher weight is assigned to the structural matching quantification value related to "voltage deviation". Assign weights to "line load" Weights are assigned to the basic score of the "topology" itself (i.e., the structure quantification assignment). Subsequently, the following calculation was performed: Structural anomaly statistics = (Structural matching quantization value voltage * ... ) + (Structure matching quantization value load* ) + (structural quantization assignment* Here, the structural quantization assignment needs to be pre-normalized to a dimension similar to the anomaly degree.

[0044] For example, in the aforementioned "radial" case of an old residential area, let the structural quantification value be 0.6 (after normalization), the voltage anomaly matching value be 0.1, and the load anomaly matching value be 0.05. Then, the structural anomaly statistical degree = 0.1*0.4 + 0.05*0.3 + 0.6*0.3 = 0.04 + 0.015 + 0.18 = 0.235. This value can be used as a quantitative basis for the structural health of this transformer substation area, and can be used as a reference for prioritizing subsequent renovations.

[0045] In another embodiment, such as Figure 2 This diagram illustrates the complete application process of the low-voltage power grid structure assessment method of the present invention in the pre-renovation assessment of a specific area (e.g., "Guangming Community" in a certain city). The diagram uses four core step boxes and connecting arrows to depict the end-to-end processing logic from raw data to evaluation conclusions.

[0046] The first step corresponds to "constructing a low-voltage network layout map and collecting comprehensive grid data." In this step, the system retrieves the power distribution geographic information system map and equipment ledger of "Guangming Community" and automatically generates a digital low-voltage network layout map for the community. The map clearly marks the location and connection relationships of distribution transformers, outgoing switches, branch boxes, and user meter boxes. Simultaneously, the system connects to a total of 85 smart meters and 3 power distribution monitoring terminals within the community through an IoT gateway, collecting one month's worth of historical voltage and current data. This spatially distributed operational data is then correlated with the equipment locations in the layout map, merging them to generate "comprehensive grid data" containing both topology and operational information.

[0047] The second step corresponds to "constructing a two-dimensional rectangular coordinate system based on the integrated grid data and extracting grid-converted data." In this step, the system analyzes the integrated grid data, constructing a unified two-dimensional rectangular coordinate system with data acquisition time as the horizontal axis (T-axis) and data value (such as voltage value) as the vertical axis (V-axis). The system maps the voltage time series data of all monitoring points to this coordinate system, forming a large number of discrete points. Subsequently, the system performs feature selection, specifically extracting the key indicator "phase A voltage deviation rate," and organizes the data of this indicator from all monitoring points in chronological order to generate a regular time series, i.e., "grid-converted data."

[0048] The third step corresponds to "constructing a dynamic quantitative structure graph based on grid conversion data and dividing the data into segments." In this step, the system selects the "Phase A voltage deviation rate" sequence from the main meter at the community entrance as a representative. In the coordinate system constructed in the second step, a "grid quantitative curve" that fluctuates continuously from the beginning to the end of the month is generated through curve fitting. This curve and the coordinate system together constitute the "dynamic quantitative structure graph." The system sets a monitoring screening axis in this graph, starting from the minimum deviation rate (-2.0%), and gradually shifts the axis upwards to the maximum value (+1.5%) with a fixed step size of 0.5%. Each shift extracts the portion of the curve between adjacent axis positions as an independent time period, ultimately dividing the continuous curve for the entire month into multiple independent "quantitative monitoring segments," such as "the period with a deviation between -2.0% and -1.5%" and "the period with a deviation between -1.5% and -1.0%."

[0049] The fourth step corresponds to "extracting and assigning values ​​to the integrated grid data structure, evaluating the low-voltage network layout structure, and obtaining anomaly statistics." This step processes two tasks in parallel: First, it performs "structure extraction" on the connection relationships in the integrated grid data generated in the first step, identifying the community as a typical "radial" topology, and matching its "structure quantification assignment" as B1 (e.g., 60 points) from a predefined pattern library. Second, based on the grid transformation data from the second step and the quantified monitoring sections from the third step, it calculates the "anomaly proportion" of voltage anomalies in each branch within the community during historical monitoring, and merges them according to topological relationships to obtain a "structure matching quantification value" of F1 (e.g., 0.15) reflecting the historical anomaly frequency of the "radial" structure. Finally, the system performs a weighted fusion calculation of B1 and F1 according to the weight configuration of the "old community" scenario, outputting a quantified "structural anomaly statistics" JY (e.g., 0.25). This value intuitively indicates that the community's power grid structure has certain irrationalities, providing a clear quantitative decision-making basis for subsequent planning to transform the "radial" structure into a "ring network."

[0050] In summary, based on the above implementation methods, the system achieves its goals through six core steps: data fusion and feature extraction, spatiotemporal sequence construction, dynamic graph partitioning, topology encoding, spatiotemporal correlation analysis, and weighted evaluation and output. Specifically, it receives topology and operational data from the target low-voltage power grid to perform multi-source heterogeneous data acquisition and fusion, generating a multi-dimensional feature dataset covering static connectivity and dynamic operational status, providing a complete data foundation for subsequent analysis. It constructs a spatiotemporal feature coordinate system and performs feature engineering on the multi-dimensional feature dataset, extracting key performance indicator sequences as grid-converted data, achieving dimensionality reduction and quantifiable transformation of massive heterogeneous operational data. Based on the grid-converted data, it generates a quantized spatiotemporal graph in the spatiotemporal feature coordinate system and uses an adaptive segmentation algorithm to divide it into multiple quantifiable monitoring segments, transforming the continuous operational data stream into a discrete, independently divisible graph. The system analyzes the state monitoring unit; it extracts graph structure features and performs pattern recognition on a multi-dimensional feature dataset to obtain structural topology data, and encodes it according to a predefined pattern library to generate structural quantification values, thus completing the abstraction and standardization scoring of the physical connection form of the power grid; it establishes spatial correlation between structural topology data and grid transformation data to calculate structural matching quantification values, realizing the correlation measurement between specific topology forms and their corresponding regional operational anomalies; it defines an evaluation weight vector, and performs weighted fusion calculation on structural quantification values ​​and structural matching quantification values ​​based on the vector to output structural anomaly statistics, ultimately generating a comprehensive quantitative index characterizing the rationality of the power grid grid structure.

[0051] Specifically, in the technical solution of this embodiment, addressing the technical problem described in the background art that existing evaluation technologies rely on ex-post statistical operational indicators such as "grid line loss rate" and "power supply reliability rate," making it difficult to provide early warning and quantitative assessment of the rationality of the power grid's physical structure before structural hazards cause operational problems, this invention creatively shifts the evaluation focus from "operational results" to "the structure itself" by introducing steps for "structural topology data" extraction and "structural quantitative assignment," as well as the calculation of "structural matching quantitative values." "Structural quantitative assignment" directly scores the health of the power grid's physical topology, while "structural matching quantitative values" reveal the correlation between a specific topology and the historical anomalies of its associated regions. Together, they constitute a forward-looking diagnostic basis for the inherent vulnerability of the structure. Finally, the "structural anomaly statistics" output through weighted fusion calculation is a scalar value that is independent of traditional operational result indicators and directly quantifies the health status and risk level of the power grid's physical structure. Therefore, the technical solution of this invention solves the technical problem that existing evaluation technologies struggle to provide early warning and quantitative assessment of structures, improving the foresight, targeting, and resource allocation efficiency of power grid upgrade decisions.

[0052] In some embodiments, a spatiotemporal feature coordinate system is constructed, feature engineering is performed on the multidimensional feature dataset, and key performance indicator sequences are extracted as grid-transformed data, including: From the multidimensional feature dataset, extract the spatial identifier and corresponding timestamp metadata of IoT sensing nodes.

[0053] Among them, IoT sensing nodes refer to intelligent devices deployed in low-voltage power grid sites to collect various operational data, such as smart meters, temperature sensors, and fault indicators; spatial identifiers are data used to uniquely identify the location of the sensing node in the power grid topology, usually the device code or logical address; timestamp metadata refers to the information generated when each piece of operational data is collected, recording the precise time point, usually in the format of "year-month-day hour:minute:second".

[0054] Specifically, the system can extract metadata by parsing each record in the multidimensional feature dataset and reading the embedded device identifier field and data acquisition time field. For example, from a record {"deviceID": "Meter_A_Zone1_Node5", "time": "2023-10-26 14:30:00", "Ua": 220.5}, the extracted spatial identifier is "Meter_A_Zone1_Node5" and the timestamp metadata is "2023-10-26 14:30:00".

[0055] Based on spatial identifiers and timestamp metadata, a two-dimensional Cartesian coordinate system with time as the first dimension and data values ​​as the second dimension is instantiated, and the data values ​​in the multidimensional feature dataset are mapped to coordinate points in this coordinate system.

[0056] Instantiation refers to creating a specific data structure object in a computer program; the two-dimensional rectangular coordinate system is a plane coordinate system in mathematics consisting of two mutually perpendicular straight lines (horizontal axis and vertical axis), which is used in this method to uniformly organize spatiotemporal data.

[0057] Specifically, the system can programmatically create a coordinate system that converts timestamps into relative seconds or sequence numbers starting from a certain start time (such as the start time of the evaluation cycle) as the horizontal axis (T), and data values ​​(such as the voltage value 220.5V) as the vertical axis (V). Then, for each data record, a unique point is determined in the aforementioned coordinate system based on its timestamp and data value. For example, assuming the evaluation start time is "2023-10-26 14:00:00", then the x-axis T corresponding to the timestamp "14:30:00" is 1800 (seconds). The data value 220.5 (V) is used as the y-axis V. This data point is then mapped to a coordinate point in the coordinate system. .

[0058] Perform feature selection on the coordinate points mapped to the coordinate system to filter out unstructured and unmeasurable data dimensions.

[0059] Feature selection is a preprocessing step in machine learning and data analysis. It refers to the process of selecting the most valuable subset of features for subsequent analysis tasks from all data features (dimensions) according to specific rules (such as data type and coefficient of variation). Unstructured and unmeasurable data dimensions refer to data categories that cannot be directly represented by numerical values ​​or whose numerical values ​​are not continuously comparable, such as equipment manufacturers (text type), equipment photos (binary stream), and switch status (only enumerated values ​​of "open / closed").

[0060] Specifically, the system can iterate through all data categories (i.e., "columns") of a multidimensional feature dataset, examining the data type and numerical distribution of each category. It automatically filters out categories whose data type is integer or floating-point number and whose values ​​exhibit meaningful fluctuations within the evaluation period. Categories whose data type is string, boolean, or whose values ​​remain constant are filtered out. For example, in a dataset containing "voltage (V)," "current (A)," "device model," and "switch status," the system will retain the continuously changing numerical dimensions of "voltage" and "current," while filtering out the dimensions of "device model" (text type) and "switch status" (enumerated type, changing only during faults).

[0061] The data dimensions retained after feature selection are processed by time-series serialization based on timestamp metadata to generate a sequence of key performance indicators (KPIs), and the KPI sequence is marked as grid transformation data.

[0062] Among them, time-series serialization refers to the process of strictly sorting the filtered data according to the order of timestamps and organizing it into a data stream arranged in chronological order.

[0063] Specifically, the system can process each retained data dimension (such as "voltage") individually: First, it sorts the data points for that dimension across all time points according to their timestamps from earliest to latest; then, it stores the sorted data values ​​sequentially into a list or array, thus forming a time series for that indicator. After processing all dimensions, a set of time-synchronized key performance indicator sequences is obtained, and this set of sequences constitutes the grid-converted data. For example, for the "Phase A voltage" dimension, after time-series serialization, a list might be generated: Each value in the list is arranged sequentially, one point every 15 minutes. This list is a sequence of key performance indicators for "Phase A voltage". Sequences such as "Phase A current" and "total active power" together constitute the grid-converted data.

[0064] Therefore, according to the above implementation method, the system can automatically and accurately extract a regular time series data set suitable for structural evaluation from the original messy multi-source heterogeneous data, laying a reliable data foundation for subsequent quantitative analysis and dynamic map construction.

[0065] In some embodiments, based on grid-transformed data, a quantization structure spatiotemporal map is generated in a spatiotemporal feature coordinate system, and an adaptive segmentation algorithm is used to divide the quantization structure spatiotemporal map, outputting multiple quantization monitoring segments, including: In the spatiotemporal characteristic coordinate system, temporal interpolation and curve fitting are performed on the grid-transformed data to generate continuous grid quantization curves, and the coordinate system integrating the grid quantization curves is instantiated as a dynamic graph of quantization structure.

[0066] Among them, temporal interpolation is a data processing method used to estimate the value of intermediate points between known discrete data points according to certain mathematical rules (such as linear or cubic splines), thereby transforming a discrete sequence into a continuous curve; curve fitting refers to finding a continuous and smooth curve that best approximates the overall trend of change of a given data point through a mathematical function model; grid quantization curve refers to a smooth curve that represents the continuous change of a key performance indicator over time after the key performance indicator sequence has been processed by interpolation and fitting; and quantization structure dynamic graph refers to a graphical or data representation that reflects the dynamic evolution of an indicator after grid quantization curves are drawn and integrated in a visualization view or data structure that carries a spatiotemporal coordinate system.

[0067] Specifically, the system can employ a cubic spline interpolation algorithm to connect the discrete coordinate points of the key performance indicator sequence in a spatiotemporal characteristic coordinate system, generating a smooth, continuous curve. This curve, along with the coordinate system background, is then saved as a graphical file or a data structure containing curve function parameters. For example, the system interpolates the "voltage deviation rate" indicator sequence (discrete points) to generate a smooth, fluctuating curve from time point T0 (x-axis 0) to Tn (x-axis 86400, representing one day). This view, containing the coordinate axes and the curve, is instantiated as a dynamic graph of a quantized structure.

[0068] In the dynamic graph of the quantitative structure, a linear decision boundary parallel to the time dimension and capable of being translated along the data value dimension is introduced as the monitoring and screening axis.

[0069] In this method, the linear decision boundary specifically refers to a straight line defined in the graph or data space, which is used to divide or classify data according to the position of data points (above or below the line); the monitoring and filtering axis is the specific application name of the linear decision boundary in this invention, and its function is to move in the numerical direction (V axis) like a "ruler" to intercept segments of the curve in different numerical ranges.

[0070] Specifically, the system can define a straight line with the function expression V=C (where C is a constant) in the data structure corresponding to the quantized structure dynamic graph. This line is parallel to the time axis (T-axis), and its position is determined by the value of the constant C. By changing the value of C, the line can be shifted up or down along the data value dimension (V-axis). For example, in the quantized structure dynamic graph of voltage deviation rate, the system defines a line with an initial position at... The horizontal straight line representing the deviation rate as -1.0% is the monitoring and screening axis.

[0071] The monitoring and screening axis is initially positioned at a reference position and recorded as the original positioning point.

[0072] The reference position is the starting reference position for executing the adaptive segmentation algorithm, which is usually selected as a boundary point of the data value range (minimum to maximum value) of the grid quantization curve within the evaluation period; the original positioning point is a data point that records the reference position of the monitoring and screening axis, which is usually represented by a coordinate value (such as the V coordinate value).

[0073] Specifically, the system can first calculate the minimum value of all data points on the grid quantization curve ( Subsequently, the monitoring and screening axis was positioned to... The location, and this The numerical values ​​are recorded as the original location points. For example, the minimum value of the voltage deviation rate curve is... The system will place the monitoring and screening axis in At %, and record the original positioning point as %. .

[0074] Starting from the original positioning point, the monitoring and filtering axis is driven to perform iterative translation along the data value dimension according to a preset fixed step size.

[0075] Among them, fixed step size refers to the fixed distance that the monitoring and filtering axis moves in the data value dimension (V axis) each time it is translated. This distance is a parameter that is preset according to the evaluation accuracy requirements; iterative translation refers to the repeated execution of the loop process of "moving the monitoring and filtering axis → processing data".

[0076] Specifically, the system can be set with a step size (e.g., 0.5%). From the initial positioning point... Initially, in a loop, the position of the monitored filter axis is updated each time. ,until Exceeding the maximum value of the curve Until then. For example, setting the step size to 0.5%, the initial positioning point is... In the first iteration, the monitoring filter axis is located at... In the second iteration, the position is updated to... In the third iteration, the position is updated to... And so on.

[0077] Based on the continuous data intervals generated by the monitoring and screening axis segmenting the grid quantization curve after each iteration of translation, the grid quantization curve is divided into multiple discretized monitoring subsequences, and each monitoring subsequence is output as a quantization monitoring segment.

[0078] Among them, the continuous data interval refers to the time period corresponding to all data points on the grid quantization curve between two adjacent monitoring and screening axis positions; the monitoring subsequence refers to the data segment corresponding to a specific continuous data interval extracted from the original grid quantization curve data.

[0079] Specifically, in each iteration, when the monitoring filter axis is located at position At that time, the system identifies that the value V in the grid quantization curve satisfies All consecutive time periods ( (This refers to the position of the previous axis). The corresponding raw time-series data within these time periods are extracted to form an independent data segment, i.e., a monitoring sub-sequence. For example, the monitoring filtering axis starts from... %Translate to After %, the curve is at arrive The portion between % is "cut off". Assuming this portion corresponds to the time period from 8:00 AM to 9:30 AM, the system will output all raw voltage deviation rate data within this timeframe as a monitoring subsequence; this subsequence is a quantified monitoring segment. The axis moves to... At that time, then intercept %arrive The time period between % (such as data from 9:30 to 11:00) is used as the next quantitative monitoring segment.

[0080] Therefore, according to the above implementation method, the system can automatically and objectively divide the continuous change curve representing the power grid operating status into multiple analysis periods with clear numerical range boundaries, thereby achieving fine separation of different operating conditions (such as normal, minor abnormality, and severe abnormality), and providing a standardized input unit for subsequent analysis of the performance of different structures under different operating conditions.

[0081] In some embodiments, graph structure feature extraction and pattern recognition are performed on a multidimensional feature dataset to obtain structural topology data, and the structural topology data is encoded according to a predefined pattern library to generate structural quantization values, including: Perform feature filtering on the multidimensional feature dataset for non-numerical structured data, and extract graph structure features that represent physical connectivity relationships as structural topology data.

[0082] Non-numerical structured data refers to data in multidimensional feature datasets that describe the logical connection relationships and organizational structure between power grid devices, rather than directly measured electrical operating values. Graph structure features refer to features obtained by abstracting the power grid into a graph model, where power grid devices (such as transformers and switches) are regarded as "nodes" and electrical connection lines between devices are regarded as "edges". This feature fully describes the physical topology of the power grid.

[0083] Specifically, the system can parse the equipment ledger and connection relationship table in the multidimensional feature dataset to filter fields such as "upper-level equipment identifier," "lower-level equipment identifier," and "connection type." Based on these fields, an adjacency matrix or edge list is constructed with equipment as nodes and connection relationships as edges. This data structure represents the graph structure features of physical connection relationships. For example, the following connection relationship fragment can be extracted from the dataset: transformer T1 is connected to switch cabinet S1; switch cabinet S1 is connected to outgoing line L1 and outgoing line L2; outgoing line L1 is connected to meter box B1. The system transforms this relationship into a graph structure containing 5 nodes and 4 edges, which is the structural topology data.

[0084] It calls a pre-built topology pattern knowledge base, which pre-stores various standard topology paradigms and their corresponding standard scores.

[0085] Among them, the standard topology paradigm refers to a typical and universal power grid connection mode abstracted from power grid planning and operation experience; the standard score is a benchmark score set in advance by domain experts based on the theoretical power supply reliability, operation flexibility, investment economy and other dimensions of each paradigm, used to characterize the rationality of the inherent structure of the paradigm.

[0086] Specifically, the system can access a topology pattern knowledge base stored in a database or configuration file. This knowledge base is organized in key-value pairs or a similar structure, where the key is the name or unique code of a standard topology paradigm, and the value contains the graphical definition of that paradigm (such as a standard adjacency matrix) and its corresponding standard score. For example, one record in the topology pattern knowledge base could be: {"Paradigm Name": "Radial", "Standard Graphic": "[Matrix Data]", "Standard Score": 60}; another record could be: {"Paradigm Name": "Single Ring", "Standard Graphic": "[Matrix Data]", "Standard Score": 75}.

[0087] The pattern matching engine matches similar standard topological paradigms for structural topological data in the topological pattern knowledge base.

[0088] The pattern matching engine is a computer program or algorithm module used to calculate the similarity between two graph structures.

[0089] Specifically, the system can compare the structural topology data (a graph structure) obtained in step S140 with the standard graph corresponding to each standard topology paradigm in the topology pattern knowledge base. The comparison algorithm can be based on the approximate judgment of graph isomorphism, or compare the differences in graph feature indicators such as node degree distribution and connected components. The system selects the standard topology paradigm with the highest similarity as the matching result. For example, after calculation by the matching engine, the graph structure composed of transformer T1, switch cabinet S1, outgoing line L1, outgoing line L2, and meter box B1 has a 90% similarity (out of 100) with the standard graph of the "radial" paradigm in the knowledge base, while the similarity with the "single ring" paradigm is only 30%. Therefore, the system determines that the structural topology data matches the "radial" standard topology paradigm.

[0090] Based on the matching results, the corresponding standard score is obtained, and the standard score is used as the encoding result to assign to the structural topology data, generating structural quantization assignment.

[0091] The encoding result refers to the process of transforming an abstract topological structure into a specific numerical value; the structural quantization assignment refers to the final quantization score assigned to the structural topological data, which represents the rationality of its basic structure.

[0092] Specifically, after completing pattern matching, the system reads the "standard score" field associated with the matched standard topology paradigm from the topology pattern knowledge base. Subsequently, the system establishes a binding relationship between this score value and the structural topology data; this score value is the structural quantification assignment for the structural topology data. For example, after matching the "radial" paradigm, the system reads its standard score of 60 from the knowledge base. The system then uses the value 60 as the encoding result to generate the structural quantification assignment for the power grid area. This means that, from the perspective of static topology morphology alone, the basic rationality score of the area's structure is 60 points (assuming a percentage system).

[0093] Therefore, according to the above implementation method, the system can automatically and accurately classify the actual and complex physical connection relationships of the power grid into known standard structure types, and assign them an objective benchmark rationality score based on domain knowledge. This realizes the standardization and digital transformation of the non-numerical structural characteristics of the power grid, and provides a unified quantitative benchmark for subsequent correlation analysis with operational data.

[0094] In some embodiments, establishing a spatial correlation between structural topology data and mesh transformation data, and calculating structural matching quantization values, includes: Construct a mapping table between the topological patterns represented by the structural topology data and the spatial coordinates of the grid-transformed data.

[0095] Among them, the mapping relationship table is a data structure (such as a database table or key-value dictionary) used to explicitly record the subordinate relationship between each physical topology area of ​​the power grid (defined by structural topology data) and one or more IoT sensing nodes deployed in that area (whose collected data constitutes grid transformation data).

[0096] Specifically, the system can establish a "line-equipment-monitoring point" attribution relationship based on the power grid's Geographic Information System (GIS) ledger or construction drawings. For each identified topology pattern (such as "radial trunk line L1"), a record is created in the mapping table, associating the pattern with the device identifiers (i.e., the source of spatial coordinates) of all smart meters and monitoring terminals connected to this line. For example, suppose the power grid is identified as having a "radial" topology pattern, named "feeder F1". Through a ledger query, it is determined that there are 5 smart meters (numbered...) The feeder is installed within the area it supplies power to. The system then creates a record in the mapping table: Topology mode: "Feeder F1 - Radial", associated monitoring point: .

[0097] Based on the preset comparison conditions, anomaly detection is performed on each quantitative monitoring segment associated with the mapping relationship table.

[0098] Anomaly detection refers to the process of determining whether a data point is in an abnormal state based on pre-set threshold rules. The preset comparison conditions are used to define specific numerical judgment standards for data anomalies, which are usually set based on national standards, industry specifications or operational experience.

[0099] Specifically, after mapping, the system reads all the raw data points within each quantitative monitoring segment generated for each monitoring point in the mapping table. The value of each data point is compared with a preset threshold; points meeting the threshold are marked as anomalies. Common comparison conditions include "voltage value below 198V", "voltage value above 242V", and "load rate exceeding 80%". For example, for "phase A voltage" data, the preset comparison condition is "voltage < 198V or voltage > 242V". For a quantitative monitoring segment from 8:00 AM to 9:30 AM, this monitoring point has one data point every 15 minutes within this interval, for a total of 6 points: Upon testing, the data point with a value of 197 (V) met the condition of "voltage < 198V" and was marked as an outlier.

[0100] Calculate the density of outliers in each quantitative monitoring segment within the total data points of that segment, and use this density as the proportion of outliers in that segment.

[0101] In this context, density specifically refers to the proportion of the number of abnormal data points to the total number of all valid data points in the segment; the anomaly percentage is a value between 0 and 1 (or between 0 and 100%) used to quantify the severity of anomalies in the operating status within the specific monitoring segment.

[0102] Specifically, the system can use the following formula to calculate the abnormality ratio (Abnormal_Ratio) of a single quantitative monitoring segment: Anomaly percentage = (number of anomalies) / (total number of data points in the segment); After calculation, the value is uniquely associated with the quantitative monitoring segment (e.g., associated with its monitoring point identifier and time interval). For example, continuing the previous example, the total number of data points in the quantitative monitoring segment is 6, and the number of abnormal points is 1. Then the abnormality rate of this segment is 1 / 6 ≈ 0.1667, or recorded as 16.67%.

[0103] Based on the mapping relationship table, the anomaly proportion of each quantitative monitoring segment is merged into the corresponding topological pattern in the mapping relationship table, and the merging result is output as the structure matching quantitative value.

[0104] Merging is a data aggregation operation that combines multiple scattered data points belonging to the same category into a single data point that represents the overall characteristics of that category, according to specific rules (such as averaging or taking the maximum value).

[0105] Specifically, the system iterates through the mapping table. For each topology pattern in the table, the system collects the anomaly percentage calculated for all associated monitoring points across all quantified monitoring segments. Then, it performs aggregation calculations (such as calculating the arithmetic mean) on these percentages, and the result is the structural matching quantification value for that topology pattern. This value characterizes the average frequency of anomalies in the data generated during the historical evaluation period for the area covered by this topology. For example, the "feeder F1-radial" topology pattern is associated with 5 monitoring points (M1 to M5). During a one-week evaluation, each point generated multiple quantified monitoring segments. Assuming a total of 100 related anomaly percentages are calculated, with an arithmetic mean of 0.12 (i.e., 12%), the system outputs a structural matching quantification value of 0.12 for this topology pattern. This means that, based on historical data, the power supply area of ​​this "radial" structure experiences voltage anomalies for an average of 12% of the time.

[0106] Therefore, according to the above implementation method, the system can scientifically link the static topological characteristics of the power grid with its dynamic operation performance. By using the quantified "structural matching quantification value", it can intuitively reveal the frequency of vulnerability exposed by a specific type of power grid structure in historical operation, thereby realizing the correlation diagnosis of "inferring structural rationality from abnormal operation results", and providing key real-world performance dimension data for comprehensive evaluation.

[0107] In some embodiments, an evaluation weight vector is defined, and a weighted fusion calculation is performed on the structure quantization assignment and the structure matching quantization value based on the evaluation weight vector to output the structure anomaly statistics, including: Based on the pre-configured evaluation strategy, corresponding weight coefficients are assigned to different types of grid transformation data to generate evaluation weight vectors.

[0108] Among them, the pre-configured evaluation strategy refers to a set of evaluation focuses and rules that the system pre-sets based on the characteristics of the target low-voltage power grid application scenario (such as renovation of old residential areas, acceptance of new residential areas, evaluation of high reliability guarantee areas, etc.). This strategy determines the importance ranking of each indicator in the final evaluation. The weight coefficient is a decimal between 0 and 1, used to represent the proportion of a certain indicator in the comprehensive evaluation. The sum of all weight coefficients is 1.

[0109] Specifically, the system maintains an evaluation strategy configuration library, defining weight coefficient templates for different grid transformation data (such as voltage, current, load, and power quality) for each typical application scenario. At the start of the evaluation, the system matches the corresponding evaluation strategy based on the target grid's labels or characteristics and loads the weight coefficient template for that strategy, thereby instantiating an evaluation weight vector composed of specific numerical values. For example, for the "old residential area renovation" scenario, its evaluation strategy focuses on power supply safety and overload prevention; the preset weight coefficient template might be: voltage-related index weights. Weighting of load rate related metrics Topological structure basic scoring weight The generated evaluation weight vector is then... .

[0110] Based on the evaluation weight vector, a linear weighted summation operation is performed on the structural quantization assignment and the structural matching quantization value to obtain a standardized comprehensive evaluation value.

[0111] Among them, linear weighted summation is a mathematical operation with the formula: comprehensive evaluation value = (weight 1 × value 1) + (weight 2 × value 2) + ... + (weight n × value n); scalarization refers to the process of merging multiple input values ​​into a single numerical output after calculation, which is convenient for comparison and sorting.

[0112] Specifically, the system first ensures that the structural quantification assignment and the matching quantification values ​​for each structure have been normalized to the same numerical range (e.g., between 0 and 1, with 1 representing the worst). Then, according to the coefficients specified in the evaluation weight vector, each value is multiplied by its corresponding quantification value. Finally, all products are summed to obtain the final, single comprehensive evaluation value. For example, assuming the above-mentioned old residential area is considered, the calculated normalized structural quantification assignment (corresponding to weights) is: )for ; Structural matching quantization value of voltage anomaly (corresponding weight) )for Quantization value of structure matching for load anomalies (corresponding weights) )for Then perform a linear weighted summation: comprehensive evaluation value. .

[0113] The comprehensive evaluation value is output as the statistical degree of structural anomaly.

[0114] The structural anomaly statistic is a direct result of the aforementioned linear weighted summation operation. It is a scalarized comprehensive evaluation value used to ultimately characterize the overall rationality level of the target low-voltage power grid structure. A higher value generally indicates a more irrational structure or higher potential risks.

[0115] Specifically, after completing the calculations, the system encapsulates the resulting comprehensive evaluation value into an output parameter called "structural anomaly statistics." This value can be intuitively used for horizontal comparisons between different power grid areas within the same scenario, or for vertical trend analysis of the same power grid area at different times, providing core data support for prioritizing power grid upgrades and making investment decisions. For example, if the system calculates a structural anomaly statistics of 0.32 for one transformer substation and 0.45 for another, decision-makers can initially determine that the grid structure of the substation with a structural anomaly statistics of 0.45 may be more unreasonable, and the need for upgrades is more urgent.

[0116] Therefore, according to the above implementation method, the system can flexibly configure the evaluation focus according to different actual evaluation needs, and through a rigorous mathematical fusion model, organically combine the static score reflecting the inherent attributes of the structure with the dynamic performance data reflecting historical operational problems, and finally output a single quantitative health index that is intuitive, comparable and has strong decision support, namely the structural anomaly statistics, thereby realizing the intelligent and standardized comprehensive evaluation of the rationality of the low-voltage power grid structure.

[0117] In some embodiments, the pre-configured evaluation strategy includes weight configuration templates bound to various typical power grid scenarios; and assigns corresponding weight coefficients to different types of grid transformation data, including: Based on the static topology and geographical planning characteristics of the target low-voltage power grid, the scenario category to which the target low-voltage power grid belongs is identified through a scenario classification model.

[0118] The scenario classification model is a set of computer program or algorithm rules. Its function is to automatically determine which predefined typical application scenario the power grid best fits based on the input target power grid feature data. Scenario categories refer to common power grid types divided according to factors such as the construction year of the power grid, user type, functional positioning, and geographical environment, such as "old residential communities", "newly built commercial residential areas", "rural public transformer substations", "urban core business districts", and "industrial parks".

[0119] Specifically, the system can input static topology data (such as whether it is a ring network, the number of line levels) and geographical planning data (such as the administrative region it belongs to, user density) of the target low-voltage power grid into a pre-trained classification model (such as a model built based on machine learning algorithms such as decision trees and support vector machines) or a rule base formulated by expert experience. This model or rule base outputs a label representing the scenario category. For example, the system input features might include: "radial topology," "average service life of the lines exceeds 20 years," and "users are mainly residents." Based on these features, the scenario classification model would output the scenario category as "old residential community."

[0120] Based on the scenario category, load the corresponding weight configuration template from the weight configuration template library.

[0121] The weight configuration template library is a data structure or configuration file that stores various weight configuration templates. The weight configuration template is a data structure that predefines the weight ratio coefficients that various grid transformation data (such as voltage, current, load, and power quality) should occupy in the final evaluation for a specific scenario category. These coefficients are placeholders or parameter names without specific values.

[0122] Specifically, after identifying the scenario category, the system queries the weight configuration template library using the category name as the key. It retrieves a weight configuration template record that perfectly matches the scenario. This template specifies which indicators should be considered when evaluating this type of scenario, as well as the corresponding weight parameter names for each indicator. For example, for the "old residential community" scenario, the weight configuration template loaded from the template library might be: {"Voltage Deviation Weight": g_1, "Line Load Weight": g_2, "Topology Foundation Weight": g_3}, where g_1, g_2, and g_3 are the parameter names to be assigned values.

[0123] Based on the loaded weight configuration template, specific weight coefficients are instantiated for each type of grid transformation data to generate an evaluation weight vector.

[0124] Instantiation refers to the process in a computer program of creating an object or variable containing specific numerical values ​​based on the definition of a template or class.

[0125] Specifically, the system reads pre-defined values ​​from the configuration items bound to the scenario based on the parameter names defined in the loaded weight configuration template, and assigns these values ​​to the corresponding parameters. After all weight coefficients are assigned, these specific values ​​are organized into an ordered list or array according to the order defined in the template; this data structure is the evaluation weight vector. For example, for the "old residential community" scenario, the configuration items specify: g_1=0.4, g_2=0.3, g_3=0.3 (and satisfy 0.4+0.3+0.3=1). The system generates the evaluation weight vector based on the template {g_1, g_2, g_3} and these specific values. This means that in the evaluation of this scenario, the voltage deviation index is the most important (weight 40%), followed by line load (30%) and topology (30%).

[0126] Therefore, according to the above implementation method, the system can intelligently identify the application scenario of the target power grid and dynamically load a weight configuration that matches the evaluation focus of the scenario, so that the final generated structural anomaly statistics can more accurately reflect the main contradictions and structural rationality shortcomings faced by the specific type of power grid in operation, thereby making the evaluation results more targeted and practically guiding.

[0127] In some embodiments, topology data and operational data of a target low-voltage power grid are received, multi-source heterogeneous data acquisition and fusion are performed, and a multi-dimensional feature dataset is generated, including: Based on the historical operating data and topological connections of the target low-voltage power grid, a spatial probability density field of line load distribution and historical fault frequency is instantiated.

[0128] Among them, the spatial probability density field is a two-dimensional mathematical function field used to describe the probability density of a certain physical quantity (such as load value, number of faults) at each point in the geographic space of the power grid; instantiation refers to creating a specific data structure object containing numerical values ​​in a computer program based on input data and algorithms.

[0129] Specifically, the system can read the historical maximum load rate data of all line segments of the target power grid over a typical period (e.g., one year), as well as the geographical locations of all recorded fault events. For load distribution, the system uses a kernel density estimation algorithm, taking the center point of each line segment as a sample and the load rate as the weight, to calculate the load probability density of each grid point on the geographic grid map of the entire power supply area. For fault frequency, the system uses each fault point as a sample to calculate the fault event probability density of each geographic grid point. Finally, two spatial probability density fields in the form of numerical matrices bound to geographic coordinates are generated. For example, for an old residential area with an area of ​​approximately 0.5 square kilometers, the system divides it into 50-meter by 50-meter grids. Based on the operating data of the past three years, the system calculates the "load rate probability density field" matrix L and the "annual fault frequency probability density field" matrix F, both with a size of 20 rows by 20 columns. A location in matrix L... A value of 0.85 indicates that the grid area is likely to be a high-load dense area.

[0130] By integrating the spatial probability density field with the topological connectivity, and through a pre-set multi-objective optimization solution engine, the Pareto optimal set of monitoring points is calculated.

[0131] In this context, the Pareto optimal monitoring point set refers to a set of monitoring point deployment schemes where it is impossible to improve either the "coverage effect in high-load / high-failure areas" or the "deployment cost / number of monitoring points" without worsening the other by adjusting the monitoring point locations. Multi-objective optimization engine is an algorithmic program used to solve such problems with multiple conflicting optimization objectives, such as the Non-Dominated Sorting Genetic Algorithm (NSGA-II).

[0132] Specifically, the system takes the load probability density field, the fault probability density field, and the power grid topology connections (line routes, equipment locations) as inputs. The optimization objective is defined as: (1) Maximize the total coverage of high probability density areas by monitoring points; (2) Minimize the total number of monitoring points. The constraints are defined as follows: monitoring points must be deployed at existing power equipment locations (such as utility poles and distribution boxes), and network connectivity must be guaranteed. Subsequently, a multi-objective optimization engine is invoked to search the solution space and output a set of Pareto optimal solutions, each of which represents a candidate set of monitoring points. For example, the optimization engine ultimately outputs three Pareto optimal solutions: Solution A contains 8 monitoring points with an estimated coverage of 92%; Solution B contains 12 monitoring points with an estimated coverage of 96%; and Solution C contains 15 monitoring points with an estimated coverage of 98%. The system selects Solution B based on the cost budget; the set of coordinates of the 12 monitoring points included in this solution is the Pareto optimal monitoring point set.

[0133] Based on the Pareto optimal set of monitoring points, the physical deployment of IoT sensing terminals is driven, and a distributed sensor network is instantiated.

[0134] Among them, driving physical deployment refers to generating a construction guidance list containing information such as equipment model, installation location, and network configuration based on the calculated optimization scheme, and issuing it to on-site operators for execution; instantiating the distributed sensor network refers to registering, networking, and going online these actually installed sensor terminals in the software system after the physical deployment is completed, making it an online data acquisition system that can be uniformly managed and called.

[0135] Specifically, the system converts the Pareto optimal monitoring point set into work orders, specifying the intelligent monitoring terminal (such as a fusion terminal with voltage, current, and power measurement functions) to be installed at each point. After on-site personnel complete the installation and communication debugging, each terminal connects to the IoT management platform via wireless public network or power line carrier. The platform performs identity authentication, protocol adaptation, and data channel configuration for the terminals, ultimately forming a logically complete distributed sensor network covering the target power grid. For example, according to the above scheme B which includes 12 monitoring points, construction personnel installed intelligent fusion terminals in 12 designated transformer area poles or meter boxes. After all terminals are online, they are displayed as 12 online nodes on the IoT management platform, forming a distributed network named "XX Community Structure Assessment Sensor Network," which can receive unified data collection commands from the platform.

[0136] The time-series running data stream collected by the distributed sensor network is combined with the static topology data obtained from the production management system, and spatiotemporal alignment and pattern fusion are performed to generate a multidimensional feature dataset.

[0137] Among them, time-series operational data stream refers to the sequence of operational parameters continuously and periodically reported from various nodes of the distributed sensor network, arranged in chronological order; static topology data refers to data obtained from the Production Management System (PMS) or Geographic Information System (GIS) that describes the attributes and connection relationships of power grid equipment and does not change rapidly over time; spatiotemporal alignment refers to associating and matching dynamic time-series data with static spatial topology data based on a unified time reference and unique equipment identifier; pattern fusion refers to organizing the associated data according to a preset format and dimensions to form a structured data set containing multiple dimensions such as time, space, electrical attributes, and equipment attributes.

[0138] Specifically, the system uses a distributed sensor network to collect voltage, current, and power data from 12 monitoring nodes every 15 minutes, forming a continuous data stream. Simultaneously, it retrieves the complete equipment ledger, electrical parameters, and connection relationship table for the community from the PMS system. Using the unique asset code of each device as the key, the system associates the real-time collected electrical quantity data with the static parameters of the devices in the PMS (such as rated capacity, model, and associated line). Then, within each time slice (e.g., a 15-minute period), all successfully associated device data is organized into a record containing fields such as "timestamp," "device ID," "electrical quantity," "device parameters," and "topology relationship." The collection of records for all time slices constitutes a multidimensional feature dataset. For example, at 14:30:00 on October 26, 2023, the system generates a record for this community. This record contains the voltage and current values ​​of the 12 monitoring points at that moment, as well as information such as the device model, associated branch, and upstream power supply for each of these 12 points. By summarizing the records of all moments within the evaluation period (e.g., one week), a multidimensional feature data table with tens of thousands of rows and dozens of columns is obtained, which is the final multidimensional feature dataset used for evaluation.

[0139] Therefore, according to the above implementation method, the system can achieve accurate and economical deployment of monitoring resources by taking data-driven and optimization algorithms as the guide, and automatically complete the collection, alignment and fusion of massive multi-source data on this basis, providing a high-quality and highly correlated standardized data foundation for the subsequent evaluation process, and ensuring the accuracy and reliability of the evaluation results from the source.

[0140] In some embodiments, the method for generating a weight configuration template includes: Knowledge extraction and feature engineering are performed on the construction features and risk profiles of scene categories to deduce the priority vector of the assessment targets.

[0141] Among them, the construction features refer to the set of attribute data describing the physical structure and equipment status of the power grid scenario, such as topology type distribution, average line diameter, and average service life of equipment; the risk profile refers to the set of data describing the faults and anomalies exposed in the historical operation of the power grid scenario, such as historical failure rate, overload frequency, and voltage limit exceedance probability; knowledge extraction and feature engineering refer to the process of extracting quantitative indicators that can characterize the core characteristics and weak links of the scenario from the above raw data through statistical analysis, rule induction and other methods; the priority vector is a mathematical vector in which each element represents the relative importance of a predefined evaluation target (such as power supply safety, power supply reliability, operating efficiency, and economy) in the specific scenario, and is usually normalized so that the sum of all elements is 1.

[0142] Specifically, the system can retrieve detailed historical data of the target scenario from the ledger and operational event database of the power grid management system. Through statistical analysis and association rule mining of this data, the key structural features and risk indicators of the scenario are quantified. Then, based on rules set in the expert experience database (e.g., "If the average wire diameter is too thin, then the priority of the 'power supply safety' target is increased"), these quantified indicators are converted into importance scores for each assessment target and normalized to form a priority vector. For example, for the scenario of "old residential communities," the system extracts its structural features as "radial topology accounts for 90%", "thin wire diameter accounts for 60%", and "average service life exceeds 20 years"; the risk profile is "annual failure rate of 0.15 times / km" and "overload period accounts for 8%". Through feature engineering and rule calculation, the priority order of the assessment targets in this scenario is deduced as: power supply safety > power supply reliability > operational efficiency. The final generated priority vector may be [0.5, 0.3, 0.2], corresponding to the weights of the three dimensions of "power supply safety", "power supply reliability", and "operational efficiency", respectively.

[0143] Based on the priority vector, establish a correlation mapping between different types of grid transformation data and predefined evaluation dimensions.

[0144] Among them, the predefined evaluation dimensions are several aspects that are pre-set for evaluating the power grid structure from multiple perspectives, and usually correspond one-to-one with the elements in the priority vector; the association mapping refers to establishing a logical correspondence between "a certain type of grid transformation data (such as voltage data) mainly serves which evaluation dimensions (such as power supply safety and power supply reliability)".

[0145] Specifically, the system maintains a "Data-Dimension Relationship Table." This table defines which predefined evaluation dimensions ("Power Supply Safety," "Power Supply Reliability," and "Operating Efficiency") are strongly correlated with each quantifiable grid transformation data type (such as "Voltage Deviation Rate Sequence," "Current RMS Value Sequence," "Load Rate Sequence," and "Power Outage Record"). When generating weight configuration templates for specific scenarios, the system uses this table to categorize all grid transformation data types involved in the current scenario under their associated evaluation dimensions. For example, according to the "Data-Dimension Relationship Table," "Voltage Deviation Rate Sequence" is mapped to the "Power Supply Safety" and "Power Supply Reliability" dimensions; "Power Outage Record" is mapped to the "Power Supply Reliability" dimension; and "Average Load Rate Sequence" is mapped to the "Operating Efficiency" dimension. Based on this mapping relationship, the system establishes specific association mappings for the old residential community scenario.

[0146] Based on the association mapping, differentiated parameter initialization is performed on the grid transformation data under each evaluation dimension, thereby generating a weight configuration template bound to the scene category.

[0147] Parameter initialization is a commonly used term in deep learning, which here refers to assigning initial values ​​to each weight coefficient in the weight configuration template to be generated; differentiated parameter initialization refers to assigning different initial weight coefficients to the grid-transformed data according to the evaluation dimension and its priority.

[0148] Specifically, after completing the mapping between data and dimensions, the system first assigns the scenario's priority vector value (representing the basic importance of each dimension) to the corresponding evaluation dimension. Then, within the same evaluation dimension, if multiple types of grid transformation data are associated, the total weight assigned to that dimension is further distributed to the various types of data under it according to sub-rules (such as equal division, or fine-tuning based on data quality and sampling density). Finally, a specific key-value pair structure is formed, where the key is the grid transformation data type and the value is its corresponding specific weight coefficient. This structure serves as the weight configuration template for that scenario.

[0149] For example, continuing from the previous example, the priority vector for old residential areas is [0.5 (safety), 0.3 (reliability), 0.2 (efficiency)].

[0150] The "voltage deviation rate sequence" is mapped to the "power supply security" dimension (weight 0.5). Assuming that this is the only type of data in this dimension, the weight of "voltage deviation rate" is initialized to 0.5.

[0151] The "voltage deviation rate sequence" and "power interruption record" are mapped to the "power supply reliability" dimension (weight 0.3). The system allocates the weight of 0.3 according to sub-rules (e.g., each accounting for half), so the weight of "voltage deviation rate" increases by 0.15, and the weight of "power interruption record" is 0.15. Note that the weight of "voltage deviation rate" needs to be accumulated.

[0152] The “average load rate sequence” is mapped to the “operational efficiency” dimension (weight 0.2), with its weight initialized to 0.2.

[0153] The final weight configuration template is: {"Voltage Deviation Rate": 0.5 + 0.15 = 0.65, "Power Interruption Record": 0.15, "Average Load Rate": 0.2}. The sum of all weight coefficients is 1.

[0154] Therefore, according to the above implementation method, the system can dynamically generate a scientific and quantitative weight allocation scheme that is highly matched with the evaluation needs of a specific power grid scenario based on in-depth feature analysis and risk assessment of the specific power grid scenario in an automated and rule-based manner. This enables the final structural anomaly statistics to accurately reflect the key issues that the power grid structure needs to address most urgently in that scenario, thereby improving the pertinence and decision-making guidance value of the evaluation results.

[0155] Figure 3 This is a structural block diagram of a low-voltage power grid grid structure evaluation system according to an embodiment of the present invention.

[0156] like Figure 3 As shown, the grid structure evaluation system for this low-voltage power grid includes: The data fusion and feature extraction module 210 is used to receive topology data and operation data of the target low-voltage power grid, perform multi-source heterogeneous data acquisition and fusion, and generate a multi-dimensional feature dataset.

[0157] The spatiotemporal sequence construction module 220 is used to construct a spatiotemporal feature coordinate system, perform feature engineering processing on the multidimensional feature dataset, and extract key performance index sequences as grid transformation data.

[0158] The dynamic map segmentation module 230 is used to generate a quantized structure spatiotemporal map in a spatiotemporal feature coordinate system based on grid-converted data, and to divide the quantized structure spatiotemporal map using an adaptive segmentation algorithm to output multiple quantized monitoring segments.

[0159] The topology encoding module 240 is used to extract graph structure features and recognize patterns from a multidimensional feature dataset to obtain structural topology data, and to encode the structural topology data according to a predefined pattern library to generate structural quantization values.

[0160] The spatiotemporal correlation analysis module 250 is used to establish the spatial correlation between structural topology data and grid-converted data, and to calculate the structural matching quantization value.

[0161] The weighted evaluation and output module 260 is used to define the evaluation weight vector, perform weighted fusion calculation on the structure quantization assignment and structure matching quantization value based on the evaluation weight vector, and output the structural anomaly statistics.

[0162] The specific functions and examples of each module and submodule of the device in this embodiment can be found in the relevant descriptions of the corresponding steps in the above method embodiments, and will not be repeated here.

[0163] According to embodiments of the present invention, the above-described method of the present invention can be applied to a computer device and a readable storage medium.

[0164] Figure 4 A schematic block diagram of an example computer device 600 that can be used to implement embodiments of the present invention is shown. The computer device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The computer device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0165] like Figure 4 As shown, the computer device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. The RAM 603 may also store various programs and data required for the operation of the computer device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0166] Multiple components in computer device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of monitors, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows computer device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0167] The computing unit 601 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as a method for evaluating the grid structure of a low-voltage power grid. For example, in some embodiments, a method for evaluating the grid structure of a low-voltage power grid can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed on the computer device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the method for evaluating the grid structure of a low-voltage power grid described above can be performed. Alternatively, in other embodiments, computing unit 601 may be configured by any other suitable means (e.g., by means of firmware) to perform a grid structure evaluation method for a low-voltage power grid.

[0168] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0169] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0170] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0171] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0172] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0173] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0174] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0175] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for evaluating the grid structure of a low-voltage power grid, characterized in that, include: Receive topology and operational data of the target low-voltage power grid, perform multi-source heterogeneous data acquisition and fusion, and generate a multi-dimensional feature dataset; A spatiotemporal feature coordinate system is constructed, and feature engineering processing is performed on the multidimensional feature dataset to extract the key performance index sequence as grid transformation data; Based on the grid transformation data, a quantized structure spatiotemporal map is generated in the spatiotemporal feature coordinate system, and an adaptive segmentation algorithm is used to divide the quantized structure spatiotemporal map to output multiple quantized monitoring segments; Graph structure feature extraction and pattern recognition are performed on the multidimensional feature dataset to obtain structural topology data, and the structural topology data is encoded according to a predefined pattern library to generate structural quantization values. Establish the spatial correlation between the structural topology data and the mesh transformation data, and calculate the structural matching quantization value; Define an evaluation weight vector, and perform a weighted fusion calculation on the structure quantization assignment and the structure matching quantization value based on the evaluation weight vector to output the structure anomaly statistics.

2. The method according to claim 1, characterized in that, The process of constructing a spatiotemporal feature coordinate system, performing feature engineering on the multidimensional feature dataset, and extracting key performance indicator sequences as grid-transformed data includes: From the multidimensional feature dataset, extract the spatial identifier and corresponding timestamp metadata of the IoT sensing nodes; Based on the spatial identifier and the timestamp metadata, a two-dimensional Cartesian coordinate system with time as the first dimension and data value as the second dimension is instantiated, and the data values ​​in the multidimensional feature dataset are mapped to coordinate points in the two-dimensional Cartesian coordinate system; A feature selection operation is performed on the coordinate points mapped to the two-dimensional Cartesian coordinate system to filter out unstructured and unmeasurable data dimensions; The data dimensions retained after feature selection are subjected to time-series serialization processing based on the timestamp metadata to generate the key performance indicator sequence, and the key performance indicator sequence is marked as the grid transformation data.

3. The method according to claim 1, characterized in that, Based on the grid-transformed data, a quantized structure spatiotemporal map is generated in the spatiotemporal feature coordinate system, and an adaptive segmentation algorithm is used to divide the quantized structure spatiotemporal map, outputting multiple quantization monitoring segments, including: In the spatiotemporal feature coordinate system, temporal interpolation and curve fitting are performed on the grid transformation data to generate continuous grid quantization curves, and the coordinate system integrating the grid quantization curves is instantiated as a quantization structure dynamic graph. In the dynamic graph of the quantized structure, a linear decision boundary parallel to the time dimension and capable of being translated along the data value dimension is introduced as the monitoring and screening axis; The monitoring and screening axis is initially positioned at a reference position and recorded as the original positioning point; Starting from the original positioning point, the monitoring and screening axis is driven to perform iterative translation along the data value dimension according to a preset fixed step size; Based on the continuous data intervals generated by the monitoring and filtering axis dividing the grid quantization curve after each iteration of translation, the grid quantization curve is divided into multiple discrete monitoring subsequences, and each monitoring subsequence is output as a quantization monitoring segment.

4. The method according to claim 1, characterized in that, The step of extracting graph structure features and performing pattern recognition on the multidimensional feature dataset to obtain structural topology data, and encoding the structural topology data according to a predefined pattern library to generate structural quantization values, includes: The non-numerical structured data feature filtering is performed on the multidimensional feature dataset to extract graph structure features that represent physical connection relationships, which are used as the structure topology data; The pre-built topology pattern knowledge base is invoked, which pre-stores a variety of standard topology paradigms and the standard scores corresponding to the standard topology paradigms. Using a pattern matching engine, similar standard topological paradigms are matched for the structural topological data in the topological pattern knowledge base; Based on the matching results, the corresponding standard score is obtained, and the standard score is used as the encoding result to assign to the structural topology data, thereby generating the structural quantization assignment.

5. The method according to claim 1, characterized in that, The step of establishing the spatial association between the structural topology data and the mesh transformation data, and calculating the structural matching quantization value, includes: Construct a mapping table between the topological patterns represented by the structural topology data and the spatial coordinates of the grid-transformed data; Based on preset comparison conditions, anomaly detection is performed on each quantized monitoring segment associated with the mapping relationship table; Calculate the density of abnormal points in each of the quantitative monitoring segments within the total data points of that segment, and use it as the abnormality percentage of that segment; Based on the mapping table, the abnormal proportion of each quantitative monitoring segment is merged into the corresponding topological pattern in the mapping table, and the merging result is output as the structure matching quantitative value.

6. The method according to claim 1, characterized in that, The defined evaluation weight vector is used to perform a weighted fusion calculation on the structure quantization assignment and the structure matching quantization value based on the evaluation weight vector, and outputs the structure anomaly statistics, including: According to the pre-configured evaluation strategy, corresponding weight coefficients are assigned to different types of grid transformation data to generate the evaluation weight vector; Based on the evaluation weight vector, a linear weighted summation operation is performed on the structure quantization assignment and the structure matching quantization value to obtain a standardized comprehensive evaluation value; The comprehensive evaluation value is output as the structural anomaly statistical degree.

7. The method according to claim 6, characterized in that, The pre-configured evaluation strategy includes weight configuration templates bound to various typical power grid scenarios; the allocation of corresponding weight coefficients to different types of grid transformation data includes: Based on the static topology and geographical planning characteristics of the target low-voltage power grid, the scene category to which the target low-voltage power grid belongs is identified through a scene classification model; Based on the scenario category, load the corresponding weight configuration template from the weight configuration template library; Based on the loaded weight configuration template, specific weight coefficients are instantiated for each type of grid transformation data to generate the evaluation weight vector.

8. The method according to claim 1, characterized in that, The process involves receiving topology and operational data from the target low-voltage power grid, performing multi-source heterogeneous data acquisition and fusion, and generating a multi-dimensional feature dataset, including: Based on the historical operating data and topological connection relationship of the target low-voltage power grid, the spatial probability density field of line load distribution and historical fault frequency is instantiated. By integrating the spatial probability density field with the topological connectivity, the Pareto optimal set of monitoring points is calculated using a preset multi-objective optimization solution engine; Based on the Pareto optimal monitoring point set, drive the physical deployment of IoT sensing terminals and instantiate a distributed sensor network; The time-series running data stream collected by the distributed sensor network is combined with the static topology data obtained from the production management system, and spatiotemporal alignment and pattern fusion are performed to generate the multidimensional feature dataset.

9. The method according to claim 7, characterized in that, The method for generating the weight configuration template includes: Knowledge extraction and feature engineering are performed on the construction features and risk profiles of the aforementioned scenario categories to deduce the priority vector of the evaluation targets; Based on the priority vector, establish a mapping between different types of grid transformation data and predefined evaluation dimensions; Based on the aforementioned association mapping, differentiated parameter initialization is performed on the grid transformation data under each evaluation dimension, thereby generating the weight configuration template bound to the scenario category.

10. A grid structure evaluation system for low-voltage power grids, characterized in that, include: The data fusion and feature extraction module is used to receive topology data and operational data of the target low-voltage power grid, perform multi-source heterogeneous data acquisition and fusion, and generate a multi-dimensional feature dataset. The spatiotemporal sequence construction module is used to construct a spatiotemporal feature coordinate system, perform feature engineering processing on the multidimensional feature dataset, and extract key performance index sequences as grid transformation data. The dynamic map segmentation module is used to generate a quantized structure spatiotemporal map in the spatiotemporal feature coordinate system based on the grid transformation data, and to divide the quantized structure spatiotemporal map using an adaptive segmentation algorithm to output multiple quantized monitoring segments. The topology encoding module is used to extract graph structure features and recognize patterns from the multidimensional feature dataset to obtain structural topology data, and to encode the structural topology data according to a predefined pattern library to generate structural quantization values. The spatiotemporal correlation analysis module is used to establish the spatial correlation between the structural topology data and the grid transformation data, and to calculate the structural matching quantization value. The weighted evaluation and output module is used to define an evaluation weight vector, perform weighted fusion calculation on the structure quantization assignment and the structure matching quantization value based on the evaluation weight vector, and output the structural anomaly statistics.

Citation Information

Patent Citations

  • Gridding platform management method and system, terminal equipment and storage medium

    CN115965258A