Power grid multi-disaster coupling time sequence evolution data processing modeling method and system

Abstract

The invention provides a power grid multi-disaster coupling time sequence evolution data processing modeling method and system, and relates to the technical field of power grid safety, and the method comprises the steps: obtaining power grid multi-source heterogeneous original data, carrying out the normalization and consistency processing of the multi-source heterogeneous original data, and obtaining standardized basic data; performing multi-scale spatio-temporal feature extraction on the standardized basic data to obtain a multi-dimensional spatio-temporal feature set; constructing a space-time cell system covering a power grid region, mapping and associating the multi-dimensional space-time feature set and the space-time cell system, and generating cellular fusion feature data containing data of each feature observation point; parameterization correction is carried out on the cellular fusion feature data, that is, a steady-state component defined in the data of each feature observation point is mapped into a parameterization plane, and a dynamic component defined in the data of each feature observation point is mapped into a parameterization curved surface. According to the invention, dynamic deduction of the disaster damage process and optimization decision of emergency scheduling can be realized.

Description

Technical Field

[0001] This invention relates to the field of power grid security technology, and in particular to a method and system for processing and modeling time-series evolution data of multiple disasters coupled in power grids. Background Technology

[0002] As the core infrastructure of the energy supply system, the safe and stable operation of the power grid is directly related to social production, daily life, and public safety. In recent years, extreme weather events have occurred frequently around the world, and the probability of multiple disasters such as typhoons, rainstorms, snowstorms, and earthquakes occurring simultaneously and in a chain reaction has increased significantly. The coupling effect between multiple disasters and the temporal evolution of the power grid equipment status and operating parameters form a complex and intertwined relationship, which can easily trigger a chain reaction of power grid failures, leading to serious consequences such as large-scale power outages.

[0003] Existing methods for processing and modeling power grid disaster data suffer from the following technical shortcomings: a lack of quantitative characterization of the temporal correlation patterns between multiple disasters and the power grid system. For example, current solutions often focus on the analysis of the catalytic effects of a single disaster or the monitoring of isolated power grid states, failing to effectively integrate the spatiotemporal dynamic characteristics of multi-source heterogeneous data. They also lack the systematic modeling capability for the entire chain of relationships between multi-disaster coupling evolution, power grid state response, and temporal feedback regulation, resulting in the inability to accurately quantify the dynamic interaction mechanisms between multiple disasters and the power grid. This limitation directly leads to two key problems: First, the accuracy of disaster loss dynamic simulation is insufficient, making it difficult to accurately identify the critical propagation paths and evolution stages of disaster chains in the power grid topology, resulting in significant deviations between the simulation results and the actual disaster development process. Second, the emergency recovery and dispatch strategies generated based on the simulation results lack dynamic adaptability, failing to match the temporal changes in power grid disaster losses, thus affecting the targeting and efficiency of emergency response and hindering the achievement of the goal of rapid power grid restoration. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a method for processing and modeling the time-series evolution data of multiple disasters coupled in power grids, which can realize the accurate characterization of the coupling relationship between multiple disasters and power grids, the dynamic simulation of disaster loss process, and the optimized decision-making of emergency dispatch.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a method for processing and modeling time-series evolution data of multiple disasters coupled in power grids, the method comprising: Obtain multi-source heterogeneous raw data of the power grid, and perform normalization and consistency processing on the multi-source heterogeneous raw data to obtain standardized basic data; Multi-scale spatiotemporal feature extraction is performed on standardized basic data to obtain a multi-dimensional spatiotemporal feature set; Construct a spatiotemporal cellular system covering the power grid area, map and associate multi-dimensional spatiotemporal feature sets with the spatiotemporal cellular system, and generate cellular fusion feature data containing data from each feature observation point; The cellular fusion feature data is parametrically corrected by mapping the steady-state components defined in each feature observation point data to a parametric plane and the dynamic components defined in each feature observation point data to a parametric surface; the intersection of the parametric plane and the parametric surface is solved to obtain the coupling correction parameter set; the cellular fusion feature data is corrected by the coupling correction parameter set to obtain the corrected cellular fusion feature data. Based on the corrected cell-based fusion feature data, a multi-hazard and power grid coupling time-series evolution map model is constructed to characterize the dynamic evolution process of disasters, power grid state changes and their coupling effects. The disaster and power grid data obtained from real-time monitoring are input into the multi-hazard and power grid coupled time-series evolution map model to perform dynamic simulation of power grid disaster damage and comprehensive impact assessment, and obtain the time-series evolution assessment results. Based on the time-series evolution assessment results, power grid emergency recovery optimization scheduling strategies and auxiliary decision-making information are generated.

[0006] Secondly, a power grid multi-hazard coupled time-series evolution data processing and modeling system, which executes the following methods, including: The data acquisition module is used to acquire multi-source heterogeneous raw data of the power grid, and to perform normalization and consistency processing on the multi-source heterogeneous raw data to obtain standardized basic data. The feature extraction module is used to extract multi-scale spatiotemporal features from standardized basic data to obtain a multi-dimensional spatiotemporal feature set. The mapping and association module is used to construct a spatiotemporal cell system covering the power grid area, mapping and associating multi-dimensional spatiotemporal feature sets with the spatiotemporal cell system to generate cell-based fusion feature data containing data from each feature observation point; The optimization and correction module is used to perform parametric correction on the cell-based fusion feature data, that is, to map the steady-state components defined in each feature observation point data to a parametric plane, and to map the dynamic components defined in each feature observation point data to a parametric surface; solve the intersection line of the parametric plane and the parametric surface to obtain the coupling correction parameter set; and correct the cell-based fusion feature data through the coupling correction parameter set to obtain the corrected cell-based fusion feature data. The model building module is used to construct a multi-hazard and power grid coupling time-series evolution map model based on the corrected cell-based fusion feature data, which characterizes the dynamic evolution process of disasters, power grid state changes and their coupling effects. The simulation and evaluation module is used to input disaster and power grid data obtained from real-time monitoring into the multi-hazard and power grid coupled time-series evolution spectrum model to perform dynamic simulation and comprehensive impact assessment of power grid disaster damage, and obtain time-series evolution assessment results. The scheduling decision generation module is used to generate grid emergency recovery optimization scheduling strategies and auxiliary decision-making information based on the time-series evolution evaluation results.

[0007] Thirdly, a computing device including a memory and a processor; The memory stores one or more computer programs, the one or more computer programs including instructions; when the instructions are executed by the processor, the computing device performs the method as described in the first aspect.

[0008] Fourthly, a computer-readable storage medium for storing a computer program for performing the method as described in the first aspect.

[0009] The above-described solution of the present invention has at least the following beneficial effects: Because it employs an integrated approach that combines multi-source data standardization processing, multi-scale spatiotemporal feature extraction, spatiotemporal cell fusion, parameterized correction, and coupled temporal evolution graph modeling, it effectively overcomes the core deficiency in existing technologies where there is a lack of quantitative representation of the coupling temporal correlation between multiple disasters and the power grid system. This allows for a precise characterization of the coupling evolution between multiple disasters and the power grid, effectively improving the accuracy of dynamic simulation of power grid disaster losses. It also enables the generated emergency recovery and optimization scheduling strategies to have dynamic adaptability, enhancing the pertinence and efficiency of emergency response and ensuring the safe and stable operation of the power grid under multiple disaster scenarios. Attached Figure Description

[0010] Figure 1 A schematic diagram of the process for modeling multi-hazard coupled temporal evolution data of power grid; Figure 2 A schematic diagram of a system for modeling and processing time-series evolution data of multiple disasters coupled in a power grid; Figure 3 This is a schematic diagram of a computing device. Detailed Implementation

[0011] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0012] The embodiments of the present invention propose a method for processing and modeling time-series evolution data of multiple disasters coupled in power grids. The figure shows a flowchart of this method, which includes: Step 1: Obtain multi-source heterogeneous raw data of the power grid, and perform normalization and consistency processing on the multi-source heterogeneous raw data to obtain standardized basic data; Step 2: Extract multi-scale spatiotemporal features from the standardized basic data to obtain a multi-dimensional spatiotemporal feature set; Step 3: Construct a spatiotemporal cell system covering the power grid area, map and associate the multi-dimensional spatiotemporal feature set with the spatiotemporal cell system, and generate cell-based fusion feature data containing data from each feature observation point; Step 4: Perform parametric correction on the cell-based fusion feature data, that is, map the steady-state components defined in each feature observation point data to a parametric plane, and map the dynamic components defined in each feature observation point data to a parametric surface; solve for the intersection of the parametric plane and the parametric surface to obtain the coupling correction parameter set; correct the cell-based fusion feature data through the coupling correction parameter set to obtain the corrected cell-based fusion feature data; Step 5: Based on the corrected cell-based fusion feature data, construct a multi-hazard and power grid coupling time-series evolution map model to characterize the dynamic evolution process of disasters, power grid state changes and their coupling effects. Step 6: Input the disaster and power grid data obtained from real-time monitoring into the multi-hazard and power grid coupled time-series evolution spectrum model to perform dynamic simulation and comprehensive impact assessment of power grid disaster damage, and obtain the time-series evolution assessment results; Step 7: Based on the time-series evolution assessment results, generate power grid emergency recovery optimization scheduling strategies and auxiliary decision-making information.

[0013] In this embodiment of the invention, not only is the format of multi-source heterogeneous data standardized, but multi-scale spatiotemporal features are also accurately extracted. Through spatiotemporal cell fusion and parameterized correction, the accuracy and consistency of the feature data are ensured; the constructed coupled temporal evolution map model can clearly present the dynamic correlation between disasters and the power grid. The power grid disaster damage projection results are more realistic, and the emergency recovery and dispatch strategies are more targeted and flexible. This effectively improves emergency response efficiency, helps the power grid quickly restore stable operation under multi-hazard scenarios, and effectively ensures energy supply security.

[0014] In a preferred embodiment of the present invention, step 1 above may include: Step 1.1 involves simultaneously collecting data on multiple hazard-causing factors, basic power grid data, and power grid operation data from multiple sources, including satellite remote sensing, UAV monitoring, power grid sensors, and a pre-defined infrastructure database. This forms a multi-source heterogeneous raw data set. Specifically, satellite remote sensing focuses on collecting data on the spatial distribution, impact intensity, and development trend of multiple hazards, such as typhoon wind speed and direction, rainfall in heavy rain, snow cover thickness, and earthquake magnitude and epicenter location. UAV monitoring focuses on the conductors, insulators, and tower foundations and structures of power grid lines, collecting real-time image data of the surrounding terrain and trace data of disaster impacts. Power grid sensors are installed at key locations such as transmission lines and substation equipment to continuously collect operating parameters such as line current, voltage, and power, as well as status data such as equipment temperature and vibration. Basic information such as model parameters, factory information, installation location, and operation and maintenance records of power grid equipment are retrieved from the pre-defined infrastructure database. Through a multi-source data synchronous acquisition mechanism, these different types and structures of data are integrated and aggregated according to preset time intervals and data transmission protocols to form a multi-source heterogeneous raw data set covering multiple hazards and power grid-related information.

[0015] Step 1.2 involves conducting a data quality assessment on the multi-source heterogeneous raw dataset to identify and remove duplicate records and outliers, generating data that has undergone the quality assessment. This includes: using data duplication verification rules to compare key identifying information and core fields such as collection time, monitoring location, and device number for each data entry; prioritizing deduplication according to data integrity (from highest to lowest), collection time (from newest to oldest), and device reliability (from highest to lowest), retaining complete data entries that meet the priority, and removing the remaining duplicate records; and for outlier detection, first using a box plot method to calculate the first quartile of each data indicator. The outlier thresholds are determined by the quartile, third quartile, and interquartile range. The third quartile plus 1.5 times the interquartile range and the first quartile minus 1.5 times the interquartile range are used as the outlier boundaries. The mean and standard deviation of each indicator are then calculated using the 3σ principle, and the normal data range is determined by the mean plus or minus 3 times the standard deviation. Combining the power grid operation safety threshold, the reasonable range of disaster monitoring data, and the normal operating parameter range of the equipment, the screening results of the two methods are integrated to identify suspected outliers. Professional technicians manually verify and confirm these outliers based on the actual monitoring scenario and equipment operating status to ensure accurate identification before removing them. Finally, the data for the quality investigation is generated.

[0016] Step 1.3 involves converting the data from the completed quality survey into a pre-defined standardized data format. This process includes: pre-establishing a unified structured data format standard, specifying the field names, data types, field lengths, encoding rules, and storage specifications; field names include monitoring time, monitoring location, data type, numerical value, and equipment number; data types are specified as integers, floating-point numbers, and strings; format adaptation and conversion are performed on the data after the quality survey according to this standard; for unstructured image-related attribute data acquired by satellite remote sensing and UAVs, key feature information such as the boundary of the disaster impact range and the degree of equipment damage is extracted using image analysis technology and mapped to the corresponding fields; for streaming data output from power grid sensors, data is extracted and formatted at fixed intervals of minutes or seconds to ensure the uniformity of data records; and for historical data in the database, it is uniformly updated to the pre-defined field definitions and encoding formats to ensure that all data maintains consistency in storage structure and expression, resulting in the converted data.

[0017] Step 1.4: Using a unified spatial coordinate system and time reference, perform spatiotemporal alignment on the format-converted data to obtain spatiotemporally aligned data. Specifically, this includes: selecting the WGS-84 geographic coordinate system as a unified spatial reference standard; using a coordinate transformation algorithm to convert all spatial location information, such as latitude and longitude coordinates, address descriptions, and projected coordinates, from data from different sources to this coordinate system for unified calibration to ensure that spatial location errors are controlled within a preset threshold; using Coordinated Universal Time (UTC) as a unified time reference to verify the collection timestamp of each data entry; and for data with missing timestamps, combining the data acquisition equipment's operation logs. The synchronous acquisition protocol records are supplemented and completed. For data with inconsistent time bases, time zone conversion is first performed according to the international standard time zone offset, and then the time deviation is corrected by linear interpolation. The implementation process of linear interpolation is as follows: select two accurate time points before and after the deviation data as interpolation nodes, calculate the time interval between the two nodes and the data deviation, establish a linear mapping relationship, and calculate the corrected accurate timestamp based on the original time position of the deviation data. Through the above process, it is ensured that all data maintain a high degree of consistency in spatial location and time record, providing a basis for subsequent spatiotemporal correlation analysis and obtaining data with completed spatiotemporal alignment.

[0018] Step 1.5 involves normalizing the numerical range of the spatiotemporally aligned data to eliminate the influence of different dimensions and magnitudes, ultimately generating standardized basic data. This process includes: identifying the indicator types of the spatiotemporally aligned data and clarifying the actual numerical range of each indicator, including electrical parameters such as current and voltage, meteorological parameters such as rainfall and wind speed, and state parameters such as equipment temperature and vibration; calculating the maximum and minimum values ​​of each indicator by traversing all data entries; processing the original values ​​of each indicator using a linear normalization method, i.e., subtracting the minimum value of the indicator from the original value of each data entry, then dividing by the difference between the maximum and minimum values ​​of the indicator, and finally mapping all data to the interval between 0 and 1. This processing method preserves the relative size relationships between data while eliminating interference caused by differences in dimensions and magnitudes between different data indicators, enabling direct comparison and fusion analysis of various data types, ultimately generating standardized basic data.

[0019] In this embodiment of the invention, multi-source data is collected synchronously through multiple channels, enriching the coverage dimensions and completeness of the original data; data quality inspection removes duplicates and outliers, ensuring the reliability of the data foundation. Unified format and spatiotemporal alignment effectively solve the compatibility problem of multi-source heterogeneous data; numerical normalization eliminates interference from differences in units and magnitudes, and the final standardized basic data is accurate, consistent, and usable, laying a solid data foundation for spatiotemporal feature extraction and coupled modeling.

[0020] In a preferred embodiment of the present invention, step 2 above may include: Step 2.1: Based on standardized basic data, analyze the original information of disaster-causing factors and power grid elements to form analyzed original information. Specifically, this includes: regarding disaster-causing factors, analysis is carried out according to a multi-stage refined process; the first step is to accurately identify the type by extracting disaster codes and key feature fields from the standardized basic data, and combining them with preset disaster classification rules to determine the disaster type one by one, clearly distinguishing categories such as typhoons, rainstorms, snowstorms, and earthquakes, while also marking the auxiliary characteristics of each type of disaster, such as the movement path of typhoons and the short-term characteristics of rainstorms. The first step involves labeling heavy rainfall, etc.; the second step is to extract and standardize the impact intensity data. First, the intensity index fields corresponding to the disaster type are selected: wind speed data for typhoons, rainfall data for rainstorms, snow and ice thickness data for snow and ice, and magnitude and intensity data for earthquakes. Then, the original intensity data is cleaned to remove abnormal fluctuation values ​​and invalid records. The units of data from different sources are standardized, such as converting wind speed values ​​in kilometers per hour to meters per second, and converting rainfall statistics from different time periods to hourly rainfall to ensure the consistency and comparability of the intensity data.

[0021] The third step is to calibrate the occurrence time information, extract the original collection time field from the data, check whether the time format is consistent, complete any missing hours, minutes, and seconds, and uniformly calibrate data with inconsistent time zones. Finally, the occurrence time is accurate to the minute and a standard timestamp is added simultaneously to ensure the accuracy of the time information. The fourth step is to delineate the preliminary spatial range of the affected area, integrate information such as geographic coordinates, administrative region names, and topographic descriptions from the data, and combine them with the radiation characteristics of the disaster to determine the approximate latitude and longitude range of the affected area, mark the boundary between the core affected area and the peripheral affected area, and form an intuitive spatial range description.

[0022] Regarding power grid elements, the equipment types are clearly defined, distinguishing between transmission lines, towers, substations, etc.; the precise coordinates of the installation locations are determined to ensure compatibility of the coordinate data with the geographic information system; equipment model parameters are compiled, including core technical indicators such as rated voltage and rated power; key monitoring parameters during operation are extracted, including real-time operating data such as current, voltage, and temperature; the disaster-causing factor information and the compiled power grid element information, after being refined and analyzed, are combined into two main categories to establish structured data catalogs, which are then sorted and archived according to field specifications, forming the original information after analysis that is clearly structured, complete in content, and accurate in data.

[0023] Step 2.2, based on the parsed original information, identifies and extracts the temporal evolution characteristics and spatial distribution characteristics of disaster-causing factors. Specifically, this includes: based on the parsed original information, conducting feature identification and extraction for each disaster-causing factor; when extracting temporal evolution characteristics, using minutes or hours as the time unit, tracking the intensity data of the same disaster at continuous time points, analyzing its overall trend from occurrence, development to weakening and dissipation by comparing the numerical changes at adjacent time points, determining the rate of intensity increase and decrease, statistically analyzing the duration of the disaster, and marking the specific time node where the peak intensity occurs; when extracting spatial distribution characteristics, based on the precise coordinate information of each monitoring point within the disaster-affected area, outlining the boundary contour of the affected area by connecting points one by one, calculating the area of ​​the affected area through coordinates, dividing and identifying the core and peripheral areas of the disaster impact based on the intensity values, statistically analyzing the density distribution pattern of the disaster in space, and clarifying the differences in the degree of disaster impact in different areas.

[0024] Step 2.3, based on the parsed original information, identifies and extracts the spatial layout features and state change features of power grid elements. Specifically, this includes: focusing on feature extraction of power grid elements based on the parsed original information; when extracting spatial layout features, collecting precise coordinate data of all power grid equipment, analyzing the spatial distribution of equipment through coordinate comparison (centralized or dispersed distribution), calculating the straight-line distance between different types of equipment, recording and sorting out the topological connection relationship between equipment based on the connection relationship of the equipment, clarifying the path direction of the starting point, transit points, and ending points of key transmission channels, as well as the location distribution of important hub nodes, such as substations; when extracting state change features, tracking the numerical changes of power grid equipment operating parameters at different time points in fixed time intervals, judging the stable range and fluctuation period of parameters through the range of numerical fluctuations, statistically analyzing the switching frequency of normal operating state and abnormal alarm state of equipment, calculating the difference and proportion of parameters deviating from the rated value under abnormal state, and clarifying the amplitude and trend of equipment state changes.

[0025] Step 2.4 quantifies the temporal evolution characteristics, spatial distribution characteristics, spatial layout characteristics, and state change characteristics to generate corresponding feature quantification indicators. Specifically, this includes: quantifying the four extracted features to form feature quantification indicators that can be directly used for subsequent calculations; for temporal evolution characteristics, calculating the intensity change rate by the ratio of the intensity difference between adjacent time points to the time interval, calculating the total duration from the occurrence of the disaster to its complete dissipation as the duration, recording the maximum intensity value during the disaster as the peak intensity value, and simultaneously calculating the peak lag time by calculating the interval between the peak occurrence time and the start time; for spatial distribution characteristics, calculating the total area of ​​the affected area based on the boundary coordinates as the area of ​​the affected range, and determining the boundary from the core area to the... The average distance from the center point is used as the radius of the core area. The arithmetic mean of the intensity values ​​of all monitoring points within the area is taken to obtain the mean regional influence intensity. The intensity attenuation rate of the edge area is determined by the ratio of the difference between the mean intensity values ​​of the edge area and the core area. For spatial layout characteristics, the arithmetic mean of the distances between similar devices is calculated as the mean device spacing. The number of device connections within a unit space is counted to obtain the topology connection density. The path length of the critical channel is obtained by summing the straight-line distances from the start point to the end point of the channel. For state change characteristics, the difference between the maximum and minimum values ​​of the parameter is used to represent the parameter fluctuation amplitude. The proportion of abnormal states is calculated by the ratio of abnormal running time to total running time. The coefficient of variation of the parameter values ​​within the stable interval is used to characterize the parameter stability.

[0026] Step 2.5: Based on feature quantification indicators, construct spatiotemporal feature vectors corresponding to each disaster-causing factor and each power grid element. Specifically, this includes: constructing spatiotemporal feature vectors using a spatial vector projection algorithm. This algorithm models the spatiotemporal correlation between disaster types and power grid elements through coordinate mapping and vector operations. First, establish a unified spatial rectangular coordinate system, with the geographical center point of the power grid coverage area as the origin, the x-axis pointing east longitude, the y-axis pointing north latitude, and the z-axis corresponding to the time dimension. The time unit is set to hours, and the starting time of the modeling cycle is used as the zero point of the z-axis. Then, convert the spatial distribution feature quantification indicators of the disaster-causing factors, including the area of ​​influence and the radius of the core area, into spatial coordinate vectors under this coordinate system. Simultaneously, quantify the temporal evolution feature indicators, such as intensity variation... The efficiency of peak lag and other parameters are mapped to vector components along the z-axis. Similarly, for power grid elements, spatial layout characteristic quantification indicators, such as average equipment spacing and topology connection density, are transformed into spatial coordinate vectors in the coordinate system. State change characteristic quantification indicators, such as parameter fluctuation amplitude and abnormal state ratio, are mapped to vector components along the z-axis. Next, the projection lengths of the disaster-causing factor's spatiotemporal vector and the power grid element's spatiotemporal vector in the coordinate system are calculated. The spatiotemporal correlation coefficient is calculated by combining the spatial angle between the two, which is the ratio of the projection length to the vector magnitude. Finally, the spatiotemporal correlation coefficient is used as the core component, combined with the specific values ​​of various characteristic quantification indicators, and arranged in a preset order to construct a unique spatiotemporal feature vector for each disaster-causing factor and each power grid element.

[0027] Step 2.6 aggregates all spatiotemporal feature vectors to form a multi-dimensional spatiotemporal feature set. Specifically, this includes: first, systematically classifying all constructed spatiotemporal feature vectors according to disaster type and power grid element type; grouping spatiotemporal feature vectors corresponding to the same disaster-causing factor and different power grid elements into the same major category, distinguished by disaster type number; grouping spatiotemporal feature vectors corresponding to the same power grid element and different disaster-causing factors into the same subcategory, distinguished by equipment number; then, orderly integrating the classified spatiotemporal feature vectors, assigning a unique identifier to each vector, consisting of an 8-digit disaster type number, a 10-digit equipment number, and a 12-digit time period code, where the time period code includes year, month, day, hour, and minute information; next, summarizing all spatiotemporal feature vectors with unique identifiers according to classification hierarchy and entering them into a preset feature database; during the entry process, strictly ensuring that the classification, identifier information, and feature values ​​of each vector correspond one-to-one to avoid information misalignment or omission; finally, through structured storage and integration in the database, forming a multi-dimensional spatiotemporal feature set covering multiple disaster types, multiple power grid elements, and multiple spatiotemporal dimensions.

[0028] In this embodiment of the invention, by systematically analyzing the original information of disaster-causing factors and power grid elements in standardized basic data, the temporal evolution and spatial distribution characteristics of disaster types, as well as the spatial layout and state change characteristics of the power grid, are accurately extracted. Various features are transformed into calculable quantitative indicators, and a unique spatiotemporal feature vector is constructed for each disaster-causing factor and power grid element. These vectors are then aggregated to form a multi-dimensional set, effectively capturing the spatiotemporal dynamic information of multiple disaster types and the power grid. This allows for the structured integration of dispersed feature data, providing an accurate, reasonable, and usable feature foundation for mining the coupling correlation patterns between multiple disaster types and the power grid. This effectively supports the accuracy and reliability of coupled temporal evolution modeling.

[0029] In a preferred embodiment of the present invention, step 3 above may include: Step 3.1: Determine the spatial boundaries of the power grid area and the time period for the modeling process. This includes: referring to power grid planning and design documents, geographic information system data, and administrative boundaries, clarifying the extreme latitude and longitude ranges of the power grid coverage, and delineating the spatial boundaries of the power grid area to ensure complete coverage of all power grid equipment and areas potentially affected by disasters; combining historical disaster occurrence cycles, data collection frequency, and emergency response requirements, determining the time period for modeling. The time period can be set to short-term (e.g., days), medium-term (e.g., months), or long-term (e.g., year) depending on the actual scenario; and simultaneously clarifying the start and end times of the time period to provide a foundation for the unified processing of spatiotemporal dimensions in subsequent steps.

[0030] Step 3.2: After determining the spatial range boundary and time period range, perform spatial clustering analysis on the spatial locations corresponding to each spatiotemporal feature vector in the multi-dimensional spatiotemporal feature set to form multiple feature point clusters; calculate the axial rectangle with the smallest area covering all feature points in each feature point cluster, and take the axial rectangle as a spatial cell; the set of rectangular cells covering the power grid area is composed of all spatial cells, specifically including: after determining the spatial range and time period, first extract the precise latitude and longitude coordinates corresponding to each spatiotemporal feature vector in the multi-dimensional spatiotemporal feature set; then carry out spatial clustering analysis, the distance threshold is set according to the average distribution density of power grid equipment to ensure that the feature vectors of adjacent equipment can be classified into the same cluster, and the minimum number of points threshold is set according to the percentage of the total number of feature vectors to avoid the occurrence of isolated clusters of single points; traverse all spatial coordinates, calculate the distance between each coordinate and the surrounding coordinates, and aggregate the coordinates with a distance less than the threshold and a number that meets the minimum number of points requirement to form multiple Feature point clusters; for each feature point cluster, the minimum enclosing rectangle rotation caliper geometry algorithm is initiated; firstly, the vertices of the convex hull of the feature point cluster are extracted using the Graham scan method, and the vertices are connected in clockwise order to form a convex polygon; then, using each edge of the convex hull as a reference, the convex hull is rotated sequentially until that edge is parallel to the coordinate axis; after each rotation, the axial rectangle that can completely cover all convex hull vertices is calculated, with the left and right boundaries of the rectangle corresponding to the minimum and maximum x-coordinates of the vertices, and the top and bottom boundaries corresponding to the minimum and maximum y-coordinates of the vertices; the length and width of the rectangle after each rotation are recorded, and the area is calculated by multiplying the length and width. The area of ​​the rectangle under all rotation angles is compared, and the axial rectangle with the smallest area that can completely enclose all feature points of the feature point cluster is selected as a spatial cell. This axial rectangle is the minimum area enclosing rectangle that covers all feature points of the feature point cluster; after all feature point clusters have completed spatial cell calculations, all spatial cells are summarized to form a set of rectangular cells that do not overlap and completely cover the power grid area.

[0031] Step 3.3: Based on the time period range, define a unified time reference and time unit to expand each spatial cell in the rectangular cell set in the time dimension, forming a spatiotemporal cell with a unified spatiotemporal identifier. All spatiotemporal cells constitute the spatiotemporal cell system. Specifically, this includes: using Coordinated Universal Time (UTC) as the unified time reference based on a defined time period range to eliminate time differences caused by different time zones; defining a unified time unit according to the time resolution of the spatiotemporal feature vector and the speed of disaster evolution. If the data acquisition frequency is on the minute level and the disaster evolves rapidly, the time unit is set to minutes; if the data acquisition frequency is on the hour level and the evolution cycle is long, the time unit is set to hours. The granularity of the time unit must ensure that it can accurately capture the dynamic changes of disaster and power grid status, while avoiding data redundancy caused by excessively fine granularity. For each spatial cell in the rectangular cell set, it is expanded according to all time units within the time period; a unique spatial identifier is assigned to each spatial cell, which is composed of the latitude and longitude of the lower left and upper right corners of the cell; a unique time identifier is assigned to each time unit, which is obtained by accumulating the time unit number at the start of the time period; through the combination of spatial identifier and time identifier, a unified spatiotemporal identifier is given to each expanded cell, forming a spatiotemporal cell; all spatiotemporal cells are arranged in the order of spatial identifier and time identifier, forming a spatiotemporal cell system covering the entire spatial range and complete time period of the power grid area.

[0032] Step 3.4 establishes a mapping relationship between each spatiotemporal feature vector in the multi-dimensional spatiotemporal feature set and spatiotemporal cells with the same spatiotemporal identifier in the spatiotemporal cell system. Specifically, this includes: extracting the spatial coordinates, timestamp, and corresponding feature identifier of each spatiotemporal feature vector in the multi-dimensional spatiotemporal feature set; parsing the spatiotemporal attributes of each spatiotemporal feature vector to determine its spatial location and time node; traversing each spatiotemporal cell in the spatiotemporal cell system, reading the spatial boundary range of the cell (i.e., the extreme latitude and longitude values ​​of the lower left and upper right corners) and the time unit (i.e., the start and end times); determining whether the spatial coordinates of the spatiotemporal feature vector satisfy the extreme longitude and latitude ranges of the cell, and simultaneously determining whether the timestamp of the vector falls within the start time of the cell's time unit. Between the moment and the end moment; if both conditions are met simultaneously, the spatiotemporal feature vector is determined to match the spatiotemporal cell; when establishing the mapping relationship, a structured mapping table is first created, which contains fixed fields such as vector feature identifier, cell spatiotemporal identifier, and matching verification identifier. The matched vector feature identifier and cell spatiotemporal identifier are entered into the table one by one; if the spatiotemporal attributes of multiple spatiotemporal feature vectors all match the same spatiotemporal cell, the correspondence between each vector and the cell is recorded in the mapping table, and the many-to-one association type is marked in the matching verification identifier field; finally, by traversing and checking, it is ensured that the vector feature identifier and cell spatiotemporal identifier of each mapping record are uniquely associated, and each spatiotemporal feature vector can be accurately associated with the unique corresponding spatiotemporal cell, without omission or mismatch.

[0033] Step 3.5: Based on the mapping relationship, associate and fill the feature information carried by each spatiotemporal feature vector into the corresponding spatiotemporal cell to form a spatiotemporal cell filled with feature information. Specifically, this includes: extracting the complete feature information carried by each spatiotemporal feature vector one by one based on the established mapping relationship; the feature information includes quantitative indicators such as the intensity change rate, duration, peak intensity value, peak lag duration, affected area, and core area radius of disaster-causing factors, as well as quantitative indicators such as the average equipment spacing, topology connection density, critical channel path length, parameter fluctuation amplitude, and abnormal state ratio of power grid elements, and the temporal relationship between the two. The empty correlation coefficient is used to divide the feature storage area for each spatiotemporal cell according to a preset field format, with each field corresponding to a feature indicator. The extracted feature information is then filled into the corresponding spatiotemporal cell one by one in the order of the fields. During the filling process, it is ensured that the value of each feature indicator completely matches the field definition without any misalignment or missing information. After filling, the integrity of the feature information in each spatiotemporal cell is checked to see if there are any empty fields or abnormal values. If any problems are found, the process is traced back to the mapping relationship and feature extraction stage for correction. Finally, a spatiotemporal cell filled with feature information is formed, where each cell carries complete and accurate feature information.

[0034] Step 3.6: Integrate all spatiotemporal cells filled with feature information to generate cell-based fusion feature data containing data from each feature observation point. Specifically, this includes: collecting all spatiotemporal cells filled with feature information, first verifying the uniqueness of the spatiotemporal cells according to their unified spatiotemporal identifiers, eliminating any possible duplicate cells, and then classifying and sorting them according to spatial identifier partitioning and time identifier ascending order to ensure that the spatiotemporal logic of the cells is clear and orderly. The sorted spatiotemporal cells are structurally integrated, with each cell's unified spatiotemporal identifier, spatial boundary latitude and longitude extreme values, and time unit start and end time information linked to the complete data of all feature observation points filled within the cell, including the quantitative indicators of disaster-causing factors and power grid elements, spatiotemporal attributes, and their spatiotemporal correlation coefficients, forming a standardized data structure with standardized fields and clear hierarchy. Redundant information generated during the integration process is removed through data cleaning, and multi-dimensional consistency checks are performed: checking the matching of feature observation point data and mapping relationships within each cell to ensure no mismatches or omissions; checking the spatiotemporal boundary connectivity of adjacent cells to avoid spatial overlap or temporal discontinuities; and checking the consistency of the numerical range of feature data with the previously standardized basic data to prevent abnormal fluctuations. Finally, cell-based fusion feature data with spatiotemporal cells as the basic unit, a regular structure, clear spatiotemporal correlation, and complete feature observation point data are generated.

[0035] In this embodiment of the invention, by accurately defining the spatial boundaries of the power grid area and the modeling time period, and combining spatial clustering analysis to construct a complete set of spatial cells, a unified spatiotemporal cell system is formed through time dimension expansion. A precise mapping between spatiotemporal feature vectors and spatiotemporal cells is established, and feature filling and integration are completed. This effectively solves the problems of scattered multi-dimensional spatiotemporal features and unclear spatiotemporal correlations, enabling the dispersed feature information to be carried in a structured and orderly manner. This not only ensures the spatiotemporal consistency and integrity of feature data, but also provides a regular and efficient data source for subsequent parameterized correction and coupled temporal evolution modeling, significantly improving the accuracy and efficiency of subsequent data processing and modeling.

[0036] In a preferred embodiment of the present invention, step 4 above may include: Step 4.1: Parse the cell-based fusion feature data to extract the feature observation point data covered by the cell-based fusion feature data. Specifically, the cell-based fusion feature data adopts a structured storage format hierarchically based on spatiotemporal identifiers. Each layer corresponds to a time unit, and each time unit is classified and stored according to the spatial identifier of the spatial cell. During parsing, the data is first read in the order of time units, and then all feature observation point data in each spatial cell are extracted one by one. The feature observation point data includes the unique ID of the observation point, the feature type code, the disaster-causing factor or power grid element, the name of the quantitative indicator (such as the intensity change rate, the average equipment spacing, etc.), the observation value, the collection timestamp, and the latitude and longitude coordinates, which need to be accurate to six decimal places. During the parsing process, the validity of each field is checked to ensure that the observation value is of numerical type, the timestamp format is uniform, and the coordinates are within the boundary of the power grid area. Data with missing fields or incorrect formats are removed, and finally a feature observation point dataset sorted by observation point ID is formed.

[0037] Step 4.2: Based on the data from each characteristic observation point, separate and extract the steady-state component from the data at each characteristic observation point. Specifically, this includes: based on the time series characteristics of the data from each characteristic observation point, separating the steady-state component using a sliding window averaging method; firstly, analyzing the patterns of disaster fluctuations and power grid parameter changes in historical data to determine the size of the sliding window. If the short-term fluctuation cycle of disasters such as typhoons and rainstorms is about 30 minutes, and the short-term fluctuation cycle of power grid equipment operating parameters is about 15 minutes, then the window size is set to 20 minutes, covering various fluctuation cycles without obscuring the long-term trend; the sliding step size is set to 5 minutes to ensure the continuity of the trend curve; the time series data of each characteristic observation point is traversed through the sliding window, and extreme values ​​exceeding the median and plus or minus two times the interquartile range are first removed from each window, and then the arithmetic mean of the remaining data is calculated; the average values ​​of all windows are arranged sequentially in chronological order to form a smooth trend curve. This smooth trend curve is the steady-state component, which can accurately characterize the long-term stable trend and baseline state of the data under conditions of no sudden disturbances.

[0038] Step 4.3: Based on the data from each characteristic observation point, separate and extract the dynamic component from the data of each characteristic observation point. Specifically, this includes: using the steady-state component as a benchmark, separating the dynamic component through time-by-time numerical comparison and calculation; for each characteristic observation point, subtracting the original observation value from the steady-state component value corresponding to the same timestamp according to the order of collection timestamps, and obtaining the difference as the dynamic component value at that time; after the calculation, performing a preliminary analysis on the dynamic component sequence, if the absolute value of the difference exceeds 10% of the corresponding steady-state component value, it is marked as significant fluctuation data, which should be given special attention in subsequent processing; at the same time, calculating the standard deviation of the dynamic component sequence to quantify the fluctuation intensity, the larger the standard deviation, the more drastic the real-time change of the data at that observation point, and finally forming a dynamic component sequence, which completely preserves the sudden fluctuation information of the original data that deviates from the long-term benchmark, including dynamic features such as sudden changes in disaster intensity and temporary equipment failures.

[0039] Step 4.4 involves mapping the steady-state components to the parameter space to construct a parameterized plane representing the long-term trend and baseline state of the data. Specifically, this includes: mapping the steady-state components to the parameter space using a three-point plane determination method to construct the parameterized plane; firstly, defining a three-dimensional parameter space with longitude as the x-axis, latitude as the y-axis, and time as the z-axis; where the units for the x-axis and y-axis are degrees, and the unit for the z-axis is hours; the values ​​of the steady-state components are the dependent variable w, which represents the values ​​of the data in the parameter space. Therefore, the actual construction is in a four-dimensional space (x, y... The hyperplane in (x1, y1, z1, w1); the selection of points follows the principle of uniform spatial distribution and complete temporal coverage, selecting three key nodes: the start point, the middle point, and the end point of the time series, which are located in the eastern, central, and western parts of the power grid area, respectively, to ensure that the three points are not collinear; the three feature points are denoted as P1(x1, y1, z1, w1), P2(x2, y2, z2, w2), and P3(x3, y3, z3, w3), where x1, x2, and x3 are the longitude values ​​of each point; y1, y2, and y3 are the latitude values; z1, z2, and z3 are the latitude values; z4, z5, z6, z7, z8, z9 ... 3 represents the time value; w1, w2, and w3 represent the steady-state component values ​​at the corresponding times; construct the plane equation Ax + By + Cz + Dw + E = 0, where A, B, C, and D are the four components of the plane normal vector, and E is the plane constant term; substitute the coordinates and steady-state component values ​​of the three feature points into the equation to obtain three sets of linear equations. The first set is Ax1 + By1 + Cz1 + Dw1 + E = 0, corresponding to the coordinates and steady-state component values ​​of feature point P1; the second set is Ax2 + By2 + Cz2 + Dw2 + E = 0, corresponding to the coordinates and steady-state component values ​​of feature point P... The coordinates and steady-state component values ​​of feature point P3 are given in the first set of equations. The third set is Ax3+By3+Cz3+Dw3+E=0, corresponding to the coordinates and steady-state component values ​​of feature point P3. These three sets of linear equations are organized into a system of linear equations, and the specific values ​​of A, B, C, D, and E in the system of equations are solved by matrix inversion. After the solution is completed, two other feature points are selected and substituted into the plane equations for verification to ensure that the calculation error is within ±0.01. The final plane equation is the parameterized plane, which can accurately characterize the long-term trend and reference state of the steady-state component.

[0040] Step 4.5 involves mapping the dynamic components to the parameter space to construct a parameterized surface representing the real-time fluctuations and evolution paths of the data. Specifically, this includes: mapping the dynamic components to the parameter space using least-squares fitting to construct a parameterized surface; firstly, clarifying the fitting objective, which requires accurately depicting the spatiotemporal fluctuations of the dynamic components with longitude and latitude using the surface; therefore, a bivariate quadratic surface equation is selected as the fitting basis, with the equation form w=a1x 2 +a2y 2The equation +a3xy+a4x+a5y+a6; The meanings of each character in the equation are as follows: x is the longitude coordinate in degrees; y is the latitude coordinate in degrees; w is the actual observed value of the dynamic component; a1, a2, a3, a4, a5, and a6 are the surface coefficients to be solved, corresponding to the weights of the longitude squared term, latitude squared term, longitude-latitude intersection term, first-order longitude term, first-order latitude term, and constant term, respectively. The dynamic component dataset is preprocessed by calculating the mean μ and standard deviation σ of all dynamic component values. The mean μ is the arithmetic mean of all sample points w, and the standard deviation σ is the square root of the average of the squared differences between each sample point w and μ. The dynamic component value of each sample point is compared with μ. If the value exceeds the range of μ ± 3σ, it is determined to be an outlier and directly removed from the dataset to avoid extreme values ​​interfering with the fitting accuracy. The remaining selected sample points form the valid sample dataset, ensuring that the fitting is based on reliable data.

[0041] Construct a design matrix X and an observation vector W. The effective sample dataset contains n sample points, each corresponding to a set of (x, y, w) data. The design matrix X is an n x 6 matrix, with each row corresponding to a sample point, and the elements in each row are in order of the x values ​​of that sample point. 2 y 2 xy, x, y, 1, that is

[0042] The observation vector W is an n x 1 column vector, where each element corresponds to the actual observed value w of the dynamic component of a sample point, i.e., W = [w1; w2; ...; w...]. n According to the core principle of least squares, the goal of fitting is to minimize the sum of squared residuals between the observed values ​​w and the calculated values ​​of the equation for all sample points, where the residuals e are... i , Sum of squared residuals (i from 1 to n); to find the coefficient vector that minimizes S. , By taking the partial derivative of S with respect to each coefficient and setting the partial derivative to 0, the normal equation X is finally derived. T X a =X T W, where X T To design the transpose of matrix X.

[0043] When solving the normal equation, first calculate X. T The product of X and X yields a 6th order square matrix X. T X, then calculate X T The product of W and W yields a 6th-order column vector X. T W; if the square matrix X T If X is invertible, then it can be obtained through matrix inversion. Then combine it with XT W multiplied, that is This yields the optimal solution for the coefficient vector a; if X T If X is irreversible, then add a small regularization term λI (λ is...). A positive number of magnitude, where I is a 6th order identity matrix, modifies the equation to Then, the coefficient solution is obtained by inversion to avoid numerical singularity problems; after fitting, the reliability of the results is ensured by double verification: first, the coefficient of determination R is calculated. 2 , , where μ is the dynamic component w i The mean, R 2 The closer the value is to 1, the better the fit between the fitted surface and the sample data. R must satisfy this condition. 2 ≥0.85; secondly, perform residual analysis to calculate the residuals e for all sample points. i Check whether the residuals are normally distributed and whether there is no obvious trend fluctuation to ensure that there is no systematic error; after verification, substitute the coefficients a1 to a6 obtained by solving into the bivariate quadratic surface equation to obtain the parameterized surface.

[0044] Step 4.6: In the parameter space, the coupling correction parameter set is obtained by solving for the intersection of the parameterized plane and the parameterized surface. Specifically, this includes: solving for the intersection of the parameterized plane and the parameterized surface iteratively; first, the plane equation Ax+By+Cz+Dw+E=0 is compared with the following surface equation: w=a1x 2 +a2y 2 By simultaneously solving the equations +a3xy+a4x+a5y+a6, and substituting w from the surface equation into the plane equation to eliminate the variable w, we obtain the ternary implicit function equation, the formula of which is as follows: F(x, y, z) = Da1x 2 +Da2y 2 +Da3xy+(A+Da4)x+(B+Da5)y+Cz+(Da6+E)=0; The meanings of the characters in the equations are as follows: A, B, C, and D are the coefficients of the plane equation; E is the constant term; a1, a2, a3, a4, a5, and a6 are the coefficients of the surface equation; x is the longitude coordinate of the parameter space, y is the latitude coordinate, and z is the time coordinate; the initial iteration interval is determined, where the interval for x is the minimum and maximum longitude of the power grid area, the interval for y is the minimum and maximum latitude, and the interval for z is the start and end time of the modeling time cycle; within the initial interval, 20 discrete points are uniformly selected in each dimension to form 8000 discrete sample points, and the F value is calculated for each sample point. For each (x, y, z) function value, record the sample point pairs where the function value sign changes, and determine the possible subintervals where the intersection point may exist. For each subinterval, use a bisection method iteratively, dividing the subinterval along the dimension of the function value change, calculating the function value at the midpoint, and determining the half-interval where the intersection point lies. Repeat this process until the length of the subinterval is less than the preset accuracy threshold of 0.001 degrees for longitude and latitude, and 0.01 hours for time. Arrange all intersection points that meet the accuracy requirements in ascending order of z-axis time, and connect them sequentially to form a continuous curve. This curve is the intersection line of the parametric plane and the parametric surface. The curve contains all (x, y, z) points. i y i , z i w i Coordinates and numerical information constitute a set of coupling correction parameters.

[0045] Step 4.7: The cellular fusion feature data is comprehensively corrected and adjusted using the coupling correction parameter set to generate corrected cellular fusion feature data. Specifically, this includes extracting the spatiotemporal coordinates (x, y) of all intersection points in the coupling correction parameter set. i y i , z i ) and correction value w i A calibration parameter index table is established, and the index table is stored in partitions according to the interval ranges of x, y, and z. For each spatiotemporal cell, the extreme longitude and latitude values ​​of its spatial boundary and the start and end times of the time unit are read, and all calibration points falling within that spatiotemporal range are queried in the calibration parameter index table. The w of these calibration points is calculated. i The arithmetic mean of the values ​​is used as the baseline correction value for the spatiotemporal cell. The original data of each feature observation point within the cell are corrected. If the difference between the original observation value and the baseline correction value exceeds twice the standard deviation of the dynamic component, the original value is adjusted to a reasonable range of the baseline correction value plus or minus the standard deviation of the dynamic component. If the difference is within the allowable range, the original value is kept unchanged to avoid overcorrection that could lead to data distortion. After all spatiotemporal cells are corrected, cross-cell consistency is checked, and the deviation rate of the corrected data of adjacent cells is calculated to ensure that the deviation rate is ≤5% to avoid data mutations. Finally, all corrected spatiotemporal cell data are integrated to generate corrected cell-based fusion feature data that is structurally regular, numerically accurate, and spatiotemporally continuous.

[0046] In this embodiment of the invention, feature observation point data is extracted by parsing the cell-based fusion feature data, the steady-state component and the dynamic component are accurately separated and mapped to the parameter space respectively, and a parameterized plane representing the long-term trend and a parameterized surface representing the real-time fluctuation are constructed. The coupling correction parameter set is obtained by solving the intersection of the two, and then the cell-based fusion feature data is corrected as a whole. This process effectively overcomes the trend deviation and fluctuation interference that may exist in the cell-based fusion feature data, improves the accuracy and consistency of the feature data, and ensures that the data can truly reflect the coupling state of multiple disasters and the power grid. It provides a high-quality and highly reliable data foundation for the subsequent construction of the multi-hazard and power grid coupling time series evolution map model, and ensures the accuracy of disaster loss prediction and assessment results.

[0047] In a preferred embodiment of the present invention, step 5 above may include: Step 5.1 involves parsing the corrected cell-based fusion feature data to identify and extract key feature nodes representing the dynamics of disaster-causing factors and the key feature nodes representing the power grid equipment and operating status. Specifically, this includes: parsing the corrected cell-based fusion feature data; extracting key feature nodes one by one from the data according to the spatiotemporal cell order; identifying nodes where the intensity of disaster-causing factors reaches its peak, nodes where the scope of influence begins to expand or contract, and nodes where the intensity mutation exceeds a preset threshold – these nodes accurately represent the key evolution stages of the disaster; and identifying nodes where equipment parameters reach their rated critical values, nodes where the operating status switches from normal to abnormal, and nodes where parameter fluctuations exceed the stable range – these nodes reflect the core state changes of the power grid; assigning a unique identifier to each key feature node, labeling the node type (disaster-causing factor or power grid equipment operating status), and recording the corresponding spatiotemporal coordinates, latitude and longitude, time unit, and core feature values ​​to form a clearly categorized set of key feature nodes.

[0048] Step 5.2: Based on all key feature nodes, analyze and define the correlation and interaction strength between different types of nodes. Specifically, this includes: analyzing the correlation logic of different types of nodes based on the type, spatiotemporal information, and core feature values ​​of the key feature nodes; when determining the correlation between disaster-causing factor nodes and power grid equipment nodes, spatiotemporal overlap is the core criterion. If the time unit of the disaster node is consistent with the state change time unit of the power grid node, and the latitude and longitude coordinates of the power grid node fall within the influence range boundary of the disaster node, then a direct influence relationship is determined; when determining the correlation between power grid equipment nodes, the correlation is based on the topological connection relationship of the equipment, transmission line nodes... There are natural topological relationships between points and the pole nodes at both ends, and between pole nodes and substation nodes. When quantifying the intensity of interaction, the intensity of the direct influence relationship is calculated by dividing the change in the power grid node parameters by the intensity value of the disaster node, and then multiplying by the time decay coefficient and the distance decay coefficient. The closer the time and the closer the distance, the larger the coefficient. The intensity of the topological relationship is assigned according to the importance of the equipment connection. The connection intensity of the key transmission channel is higher than that of ordinary distribution network lines. The connection intensity between the hub substation and the main line is higher than that of the branch line. Finally, a set of relationships with quantitative intensity indicators is formed, that is, the analysis and definition of the relationship and interaction intensity between different types of nodes are completed.

[0049] Step 5.3: Based on the association relationships and interaction strengths, construct a basic topology network with key feature nodes as vertices and association relationships as edges. Specifically, this includes: constructing the basic topology network using a graph structure model, with key feature nodes as the vertices; storing complete node information for each vertex, including a unique identifier, type, latitude and longitude coordinates, time unit, core feature values, state description, and priority, with key equipment nodes having higher priority than ordinary nodes; using the determined association relationships as edges between vertices, directly assigning the edge attributes to the corresponding interaction strength values, with strength values ​​represented in the range of 0 to 1, where above 0.8 indicates strong association, 0.5 to 0.8 indicates medium association, and below 0.5 indicates weak association; grouping vertices according to disaster type and power grid equipment type, with nodes of the same disaster type clustered together; arranging power grid equipment nodes in the same area as close as possible; automatically connecting vertices with association relationships using a topology connection algorithm, verifying the rationality of the associations during the connection process, and eliminating invalid connections with spatiotemporal mismatches or zero strength; ultimately forming a basic topology network with clear vertices, well-defined edge weights, and a regular structure.

[0050] Step 5.4 involves loading the temporal information contained in the corrected cell-based fusion feature data into the basic topology network to assign attribute state sequences that change over time to vertices and edges. Specifically, this includes: extracting temporal information from the corrected cell-based fusion feature data in ascending order of time units; the temporal information includes the core feature values, state identifiers, and trends of change for each key feature node in each time unit (including rising, stable, and falling), as well as the specific values ​​and fluctuation amplitudes of the interaction strength corresponding to the relationships in each time unit; loading the temporal information into the network unit by unit, corresponding one-to-one with the vertices and edges of the basic topology network according to node and edge identifiers; constructing an attribute state sequence for each vertex, with one record corresponding to each time unit in the sequence, containing the feature value, state identifier, trend, and cumulative change at that moment; constructing an attribute state sequence for each edge, with one record corresponding to each time unit in the sequence, containing the interaction strength value, fluctuation amplitude, and strength level at that moment; after loading, verifying the integrity of the temporal data to ensure that each vertex and edge in each time unit has a corresponding state record, thus realizing the dynamic temporal expansion of the basic topology network.

[0051] Step 5.5: Based on the attribute state sequence, time series analysis methods are used to explore and quantify the coupling correlation and transmission patterns between disaster evolution and power grid state changes. Specifically, this includes: using time series processing methods such as sliding window comparison and rate of change analysis based on the attribute state sequences of vertices and edges to explore coupling correlations and transmission patterns; performing a sliding window comparison between the intensity sequence of disaster nodes and the parameter change sequence of power grid nodes, with the window size set to 3 time units, calculating the consistency of the change trends of the two sequences within the window; if the consistency exceeds 80%, it is determined to be a strong coupling correlation; quantifying the degree of coupling correlation by calculating the correlation coefficient of the sequences: an absolute correlation coefficient above 0.7 indicates a strong correlation, 0.4 to 0.7 indicates a moderate correlation, and below 0.4 indicates a weak correlation; tracing the intensity change sequence of edges, analyzing the transmission path of disaster impact, recording the time difference from the starting node to subsequent nodes, and calculating the transmission speed, i.e., the path length divided by the time difference. Statistically analyzing the intensity attenuation ratio of different paths to clarify the energy loss patterns during the transmission process. By combining node priority and correlation strength, the priority ranking of the impact of disasters on the power grid and the transmission order of changes in the state of power grid nodes are determined. These patterns are quantified into specific numerical indicators and rule descriptions, forming a standardized set of coupling correlation and transmission patterns.

[0052] Step 5.6 integrates the basic topology network, attribute state sequences, and coupling correlation and transmission patterns to construct a multi-hazard and power grid coupling time-series evolution graph model, or simply graph model. This includes: In the construction phase, the model strictly adheres to previous results and follows a structured logical progression. The previously constructed basic topology network serves as the core framework. At the node layer, network vertices are the core. Data on extracted hazard-causing factors and key characteristic nodes of power grid equipment operation status are imported in batches. The unique identifier, type, spatiotemporal information, core characteristic values, and priority of each node are verified one by one to ensure complete consistency between node information and previous extraction results, and that attribute fields for nodes of the same type are standardized and uniform. At the edge layer, defined correlation relationships are the core. The vertex identifiers of each edge are entered, and the dynamically changing interaction strength sequence generated over time is correlated. Combined with quantified strength level standards, the edge strength level is labeled, achieving precise binding between edge attributes and node attributes. This ensures that the correlation relationships truly reflect the actual interaction logic between hazard types and the power grid and power grid equipment.

[0053] The temporal layer associates the node and edge layers through a unified time encoding, integrating the attribute state sequences of the generated vertices and edges in ascending order of time units. This establishes a three-dimensional mapping relationship between time units, node states, and edge states, supporting backtracking queries of the characteristic value change trajectory of corresponding nodes and the intensity fluctuation of edges at any time unit. Furthermore, the temporal granularity of the temporal data remains consistent with the corrected cell-based fused feature data. The pattern layer deeply embeds the mining set of coupling and transmission patterns, associating quantitative indicators such as correlation coefficients, transmission speeds, attenuation ratios, and rule descriptions such as disaster impact priority ranking and power grid state transmission order with the corresponding node types, edge association types, and temporal data change ranges. This ensures that each pattern accurately matches the specific multi-hazard and power grid coupling scenario.

[0054] The training phase relies on the core data resources of this invention, focusing on improving the model's ability to accurately simulate the coupled evolution process. First, corrected cellular fusion feature data covering different disaster types such as typhoons, rainstorms, snow accumulation, and earthquakes, as well as different scenarios such as normal power grid operation, heavy load operation, and maintenance status, are selected from the past five years. This data is divided into training and validation sets in a 7:3 ratio to ensure that the scenario coverage and data volume of the training data meet the model's generalization requirements. The training set data is then input into the initially constructed model in batches according to time units, driving the node layer, edge layer, and time-series layer to simulate coupled evolution. The deviations of the model's output node state changes, edge strength fluctuations, and coupling association triggering results are compared with historical actual evolution results. To address the issue of node state prediction deviation exceeding 5%, the weight allocation of core feature values ​​for the corresponding node type is adjusted, and the time mapping rules of the temporal layer are optimized. For the problem of edge interaction strength simulation error exceeding 0.05, the interaction strength quantification logic is revisited, and the calculation parameters of the association relationship are corrected. To address deviations in the application of coupling rules, such as discrepancies between the disaster impact transmission path and reality, the association matching conditions between the rule layer and nodes, edges, and temporal data are optimized, and rule adaptation rules for different disaster scenarios are supplemented. After each round of adjustments, the effect is verified using validation set data. The accuracy of the overall evolution simulation, the fit of the coupling rule matching, and the synchronization rate of temporal state updates are calculated. Training stops when the evolution simulation accuracy on the validation set reaches over 90%, and the fit and synchronization rates both reach over 85%.

[0055] During integration and training, multi-dimensional logical verification is carried out simultaneously to check the adaptability of node types and associated edges. For example, disaster nodes only establish direct influence edges with power grid equipment nodes, and power grid equipment nodes only establish connection edges based on topological relationships. The integrity of the correspondence between time series data and time units is checked, i.e., there are no missing state records of time units. The matching degree between the rules of the rule layer and the actual coupling scenario is checked, i.e., the coupling rules corresponding to strong association edges have higher priority than those of weak association edges. It is ensured that the information of the node layer, edge layer, time series layer, and rule layer are mutually matched and without contradictions. Finally, a multi-hazard coupling time series evolution graph model is formed, or graph model for short. It not only supports dynamic query in the spatiotemporal dimension, visualization of coupling relationships, and inference of evolution trends, but also accurately reproduces the chain change process of power grid state under the action of different disasters. It fully presents the internal logic of the coupling evolution of multi-hazards and power grid, and provides comprehensive and accurate model support for dynamic inference of power grid disaster damage, comprehensive impact assessment, and emergency recovery and dispatch strategy formulation, realizing a closed loop of the whole process from feature extraction, correlation analysis to model application.

[0056] In a preferred embodiment of the present invention, step 6 above may include: Step 6.1: Receive and preprocess disaster-causing factor data and power grid operation status data acquired through real-time monitoring to form a real-time input data stream matching the input format of the graph model. Specifically, this includes: receiving disaster-causing factor data and power grid operation status data acquired through real-time monitoring; disaster-causing factor data including real-time observed values ​​such as wind speed, rainfall, snow thickness, and magnitude; power grid operation status data including operating parameters such as current, voltage, and temperature of transmission lines, towers, and substations, as well as equipment status indicators; preprocessing the received data, first removing abnormal data with values ​​exceeding reasonable ranges, and then using linear interpolation to complete missing short-term data; standardizing the units of data from different sources, for example, unifying wind speed to meters per second and rainfall to millimeters per hour; and organizing the data types, spatiotemporal coordinates, and numerical information according to the field order required by the multi-hazard and power grid coupling time-series evolution graph model to form a real-time input data stream with standardized fields and matching formats.

[0057] Step 6.2 involves inputting the real-time input data stream into the multi-hazard and power grid coupling time-series evolution model. Specifically, this includes: pushing the real-time input data stream to the multi-hazard and power grid coupling time-series evolution model in real time according to time units through a preset data interface; each pushed data corresponds to a fixed time unit, and the data acquisition timestamp and integrity check code are attached during the push process; after receiving the data, the multi-hazard and power grid coupling time-series evolution model first verifies the check code to ensure that the data transmission is not lost or tampered with, and then checks the consistency between the data fields and the model input format; after confirming that there are no errors, the data is temporarily stored in the real-time data cache area of ​​the multi-hazard and power grid coupling time-series evolution model.

[0058] Step 6.3, based on the relationships and evolutionary rules defined in the graph model, drives the state attributes of corresponding vertices and edges in the graph model, and performs time-series updates according to the real-time input data stream to obtain the updated graph model. Specifically, this includes: reading the predefined relationships and evolutionary rules in the multi-hazard and power grid coupling time-series evolution graph model; matching key feature nodes in the model that have overlapping spatial locations and consistent time dimensions according to the latitude and longitude coordinates and time units of the real-time input data stream; classifying and matching by data type, with hazard-causing factor data corresponding to hazard nodes in the model, and power grid operation parameter data corresponding to equipment nodes in the model; and using real-time data to perform time-series updates. Driven by changes in real-time data, the core feature values ​​and status identifiers of corresponding vertices are updated; disaster nodes update their intensity values ​​and influence range boundaries, while power grid nodes update their operating parameters such as current, voltage, and temperature, as well as their status identifiers for normal, abnormal, alarm, and failure. The interaction strength of associated edges is updated synchronously. If the intensity of a disaster node increases compared to the previous unit, the intensity of the associated edge increases synchronously in a linear proportion; if the intensity of a disaster node decreases, the intensity of the associated edge gradually decreases according to the attenuation coefficient. After the update is completed, the state of each node and edge is verified to ensure that it conforms to the built-in evolution law of the model and that there are no logical contradictions, thus forming a graph model after the state update.

[0059] Step 6.4: Based on the updated graph model, simulate and calculate the propagation process of the disaster chain on the power grid topology, as well as the chain change process of the power grid state under the action of the disaster, to complete the dynamic simulation of power grid disaster loss including critical path and evolution stage, and generate dynamic simulation results simultaneously. Specifically, this includes: starting the dynamic simulation process of disaster loss based on the updated graph model; when simulating the propagation of the disaster chain, starting from the disaster node triggered by real-time data, sorting the nodes according to the interaction strength of the associated edges from high to low; calculating the time of disaster impact transmission based on the edge strength level and the spatial distance between nodes, with the transmission speed determined according to the medium propagation coefficient preset by the model; and sequentially determining the time point and parameter change magnitude of each associated power grid node affected. When simulating the cascading changes in the power grid state, the system tracks the changes in operating parameters and state transitions of each power grid node. When a parameter exceeds a safety threshold for 10 minutes, the node state switches from normal to abnormal. If the parameter exceeds the threshold for 20 minutes, it switches to alarm mode. If the parameter exceeds the threshold and the associated edge strength reaches 0.8 or higher, it switches to failure mode. When identifying critical paths, the system counts the number of power grid nodes and load scale covered by each propagation path, and selects the path with the widest impact and the greatest load loss as the critical path. The system divides the evolution stages according to time sequence and the degree of disaster impact: the early warning stage is when the disaster triggers the node but does not affect critical equipment; the mild impact stage is when less than 10% of equipment is abnormal; the severe impact stage is when 10% to 30% of equipment fails; and the recovery stage is when the equipment state gradually returns to normal. The system simultaneously generates dynamic simulation results, including the start time of each stage, the core affected nodes, the propagation path, and details of state changes.

[0060] Step 6.5: Based on the dynamic simulation process and results, assess the power grid load loss, critical equipment failure risk, and power supply reliability degradation within the simulation scope to complete the comprehensive impact assessment. Simultaneously, obtain quantitative indicators for the comprehensive impact assessment. Specifically, this includes: conducting a comprehensive impact assessment based on node state changes and path impact range recorded during the dynamic simulation process, according to time units and evolution stages; when calculating the load loss rate, statistically analyze the real-time load loss of all affected users within the simulation scope and sum them to obtain the total load loss; retrieve the design rated total load of the power grid in the area, divide the total load loss by the rated total load to obtain the load loss rate; when assessing the failure risk of critical equipment, consider the deviation of current equipment parameters... The threshold level, parameter matching degree of historical failure cases, and vulnerability coefficient corresponding to the service life of the equipment are comprehensively calculated to obtain a failure risk value in the range of 0 to 1. The closer the value is to 1, the higher the probability of equipment failure. When quantifying the degree of power supply reliability degradation, the total outage time of affected users within the simulation range is statistically analyzed and divided by the total number of affected users to obtain the average outage time. The area of ​​the normal power supply area is divided by the total area of ​​the simulation area to obtain the power supply area coverage rate. Combined with the expected fault recovery time, the power supply reliability index is calculated, and finally, comprehensive impact assessment quantitative indicators such as load loss rate, equipment failure risk value, average outage time, and power supply area coverage rate are formed for each time unit and each evolution stage.

[0061] Step 6.6 integrates the critical path, evolution stages, and quantitative indicators of the comprehensive impact assessment from the dynamic simulation to generate the time-series evolution assessment results. Specifically, this includes: integrating the dynamic simulation results and quantitative indicators of the comprehensive impact assessment in chronological order; first, identifying the spatiotemporal coordinates and impact nodes of the critical path corresponding to each time unit, clarifying the start time, core characteristics, and scope of impact of each evolution stage; classifying the quantitative indicators of the comprehensive impact assessment according to load loss, equipment failure risk, and power supply reliability, and mapping them to each time unit and evolution stage; and finally, organizing and integrating the information using a structured format to ensure that the critical path, evolution stage, and quantitative indicators correspond one-to-one and the time sequence is clear, ultimately generating the time-series evolution assessment results.

[0062] In a preferred embodiment of the present invention, step 7 above may include: Step 7.1 involves analyzing the time-series evolution assessment results to identify the set of critical damaged power grid equipment, affected load areas, and estimated recovery time windows. Specifically, this includes: analyzing the quantitative indicators and dynamic projection details in the time-series evolution assessment results; extracting power grid equipment with a failure risk value higher than 0.7; combining the importance level of the equipment to screen transmission lines, towers, substations, etc., that play a key supporting role in regional power supply, forming the set of critical damaged power grid equipment; classifying affected load areas according to load loss rate data: areas with a load loss rate above 30% are severely affected; areas with a load loss rate between 10% and 30% are moderately affected; and areas with a load loss rate below 10% are slightly affected; referencing historical repair data, equipment damage severity, and resource allocation cycles, estimating the time interval from the occurrence of the fault to the restoration of normal power supply for each affected area, clarifying the estimated recovery time windows for different areas, and ensuring clear information classification and accurate data.

[0063] Step 7.2: Based on the set of damaged critical equipment and the affected load area, and in conjunction with the pre-established emergency resource database, determine the emergency resource demand list for personnel, equipment, vehicles, and materials. Specifically, this includes: identifying the repair needs of each piece of equipment in the damaged critical equipment set, clarifying the skill types of the required maintenance personnel (e.g., high-voltage line maintenance, substation equipment maintenance), and the number of personnel; determining the required specialized equipment models based on the type of equipment damage (e.g., line repair vehicles, insulation tools, testing instruments), and the quantity of such equipment; matching the required emergency vehicle types (e.g., freight vehicles, repair vehicles, emergency power supply vehicles), and the vehicle capacity and quantity based on the location, scope, and estimated recovery time window of the affected load area; listing the required materials (e.g., cables, fuses, emergency generators, lighting equipment), and the specifications and quantities of such materials based on equipment repair consumption and the temporary power supply needs of the affected population; and retrieving the pre-established emergency resource database to verify the existing inventory and distribution of various resources, ultimately forming a precise emergency resource demand list that matches the needs.

[0064] Step 7.3: Based on the emergency resource demand list and the estimated recovery time window, perform multi-objective optimization calculations to generate a preliminary emergency recovery scheduling plan that includes resource allocation paths, task execution sequences, and time nodes. Specifically, this includes: optimizing for the shortest recovery time, lowest resource consumption, and highest power restoration efficiency, setting weight coefficients for each objective, with recovery time having the highest weight, followed by power restoration efficiency, and finally resource consumption; integrating the distribution location and availability status of various resources in the emergency resource database based on the emergency resource demand list; planning resource allocation paths by combining the estimated recovery time window and the geographical information of the affected areas, avoiding road interruptions and traffic congestion areas caused by the disaster, and selecting the optimal travel route; determining the task execution sequence by sorting by the importance of damaged equipment and the level of affected load areas, prioritizing the repair of critical equipment and severely affected areas; and assigning specific time nodes to each task based on path distance, task complexity, and resource allocation time, specifying the resource departure time, arrival time, repair start time, and estimated completion time, thus generating a preliminary emergency recovery scheduling plan that includes resource allocation paths, task execution sequences, and time nodes.

[0065] Step 7.4: Based on the time-series evolution assessment results, conduct feasibility verification and dynamic adaptive adjustments to the preliminary emergency recovery dispatch plan to form the final optimized dispatch strategy for power grid emergency recovery. Specifically, this includes: verifying the feasibility of the preliminary emergency recovery dispatch plan based on the real-time disaster evolution and power grid status changes in the time-series evolution assessment results; verifying whether resource allocation paths are unobstructed and whether there are any new road blockages; verifying whether the resource quantity can meet actual needs and whether there are resource gaps; verifying whether task time nodes are within the estimated recovery time window and whether there is a risk of timeout; if paths are not unobstructed, replanning alternative routes; if there are resource gaps, adjusting the resource allocation plan and supplementing from other areas; if time nodes are exceeded, optimizing the task execution sequence and shortening the time consumed by non-critical links; and dynamically adapting the plan based on the disaster development trend, potential new damaged equipment, or expanded affected areas, supplementing new tasks, adjusting resource allocation priorities and time nodes to ensure the plan matches the actual situation, thus forming the final optimized dispatch strategy for power grid emergency recovery.

[0066] Step 7.5 integrates the final power grid emergency recovery optimization scheduling strategy with the core conclusions defined in the time-series evolution assessment results to obtain auxiliary decision-making information to support emergency command. Specifically, this includes: extracting the core conclusions from the time-series evolution assessment results, including the specific locations of damaged critical equipment in the power grid, the level and scope of affected load areas, key nodes in the estimated recovery time window, and the main trends in disaster evolution; integrating the final power grid emergency recovery optimization scheduling strategy with these core conclusions to clarify the damaged equipment, affected areas, and time requirements corresponding to each scheduling task; and organizing structured auxiliary decision-making information according to the needs of emergency command, including resource allocation instructions, task allocation tables, time node schedules, and risk warning prompts. The auxiliary decision-making information must clearly indicate the responsible entity, execution requirements, and expected goals for each stage to ensure that emergency command personnel can quickly grasp key information and directly use it to guide emergency recovery work.

[0067] In this embodiment of the invention, the set of critical damaged equipment, affected load areas, and estimated recovery time windows of the power grid are identified by analyzing the time-series evolution assessment results. Combined with a pre-established emergency resource database, a precise list of personnel, equipment, vehicles, and materials needs is determined. Based on resource needs and recovery time windows, a preliminary emergency recovery scheduling plan, including resource allocation paths, task execution sequences, and time nodes, is generated through multi-objective optimization. After feasibility verification and dynamic adaptive adjustments, a final optimized scheduling strategy is formed. Emergency command auxiliary decision-making information is output by integrating core conclusions. This process achieves precise matching between emergency needs and resource supply, ensuring that the scheduling plan balances efficient resource utilization, orderly task execution, and time window constraints. The dynamic adjustment mechanism guarantees the strategy's adaptability to the evolving disaster situation. The auxiliary decision-making information provides clear and actionable action guidelines for emergency command, effectively improving the targeting and timeliness of power grid emergency recovery, minimizing power outage duration and load losses, and accelerating the restoration of normal power grid operation.

[0068] Embodiments of the present invention also propose a power grid multi-hazard coupled time-series evolution data processing and modeling system, such as... Figure 2 The diagram shown is a schematic of the system, which includes: The data acquisition module is used to acquire multi-source heterogeneous raw data of the power grid, and to perform normalization and consistency processing on the multi-source heterogeneous raw data to obtain standardized basic data. The feature extraction module is used to extract multi-scale spatiotemporal features from standardized basic data to obtain a multi-dimensional spatiotemporal feature set. The mapping and association module is used to construct a spatiotemporal cell system covering the power grid area, mapping and associating multi-dimensional spatiotemporal feature sets with the spatiotemporal cell system to generate cell-based fusion feature data containing data from each feature observation point; The optimization and correction module is used to perform parametric correction on the cell-based fusion feature data, that is, to map the steady-state components defined in each feature observation point data to a parametric plane, and to map the dynamic components defined in each feature observation point data to a parametric surface; solve the intersection line of the parametric plane and the parametric surface to obtain the coupling correction parameter set; and correct the cell-based fusion feature data through the coupling correction parameter set to obtain the corrected cell-based fusion feature data. The model building module is used to construct a multi-hazard and power grid coupling time-series evolution map model based on the corrected cell-based fusion feature data, which characterizes the dynamic evolution process of disasters, power grid state changes and their coupling effects. The simulation and evaluation module is used to input disaster and power grid data obtained from real-time monitoring into the multi-hazard and power grid coupled time-series evolution spectrum model to perform dynamic simulation and comprehensive impact assessment of power grid disaster damage, and obtain time-series evolution assessment results. The scheduling decision generation module is used to generate grid emergency recovery optimization scheduling strategies and auxiliary decision-making information based on the time-series evolution evaluation results.

[0069] The data processing modeling system according to embodiments of the present invention can correspond to the execution of the methods described in the embodiments of the present invention, and the above and other operations and / or functions of each module of the data processing modeling system are respectively for implementing Figure 1 The corresponding process of the method in the illustrated embodiment will not be described in detail here for the sake of brevity.

[0070] This application also provides a computing device. This computing device can utilize a server.

[0071] like Figure 3 As shown in the figure, this is a schematic diagram of a computing device provided in an embodiment of this application. The computing device 700 includes a bus 701, a processor 702, a communication interface 703, and a memory 704. The processor 702, the memory 704, and the communication interface 703 communicate with each other via the bus 701.

[0072] The 701 bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0073] The processor 702 can be any one or more of the following processors: central processing unit (CPU), graphics processing unit (GPU), microprocessor (MP), or digital signal processor (DSP).

[0074] Communication interface 703 is used for external communication. Memory 704 may include volatile memory, such as random access memory (RAM). Memory 704 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD). Memory 704 stores executable code, which processor 702 executes to perform the aforementioned power grid multi-hazard coupling time-series evolution data processing and modeling method.

[0075] Specifically, in implementing the power grid multi-hazard coupling time-series evolution data processing and modeling system described in the above embodiments, and where each module or unit of the power grid multi-hazard coupling time-series evolution data processing and modeling system described in the above embodiments is implemented by software, the software or program code required to execute the functions of each module / unit in the power grid multi-hazard coupling time-series evolution data processing and modeling system described in the above embodiments can be partially or entirely stored in the memory 704. The processor 702 executes the program code corresponding to each unit stored in the memory 704 to execute the aforementioned power grid multi-hazard coupling time-series evolution data processing and modeling method.

[0076] This application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that a computing device can store, or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive). The computer-readable storage medium includes instructions that instruct the computing device to execute the above-described power grid multi-hazard coupling time-series evolution data processing and modeling method.

[0077] This application also provides a computer program product comprising one or more computer instructions. When the computer instructions are loaded and executed on a computing device, all or part of the processes or functions described in this application are generated.

[0078] The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, or data center to another website, computer, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0079] When the computer program product is executed by a computer, the computer executes any of the methods described in the aforementioned method for processing and modeling multi-hazard coupled time-series evolution data of the power grid. The computer program product can be a software installation package; when any of the methods described in the aforementioned method for processing and modeling multi-hazard coupled time-series evolution data of the power grid needs to be used, the computer program product can be downloaded and executed on the computer.

[0080] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for processing and modeling time-series evolution data of multiple disasters coupled in power grids, characterized in that, The method includes: Obtain multi-source heterogeneous raw data of the power grid, and perform normalization and consistency processing on the multi-source heterogeneous raw data to obtain standardized basic data; Multi-scale spatiotemporal feature extraction is performed on standardized basic data to obtain a multi-dimensional spatiotemporal feature set; A spatiotemporal cellular system covering the power grid area is constructed. By mapping and associating multi-dimensional spatiotemporal feature sets with the spatiotemporal cellular system, cellular fusion feature data containing data from each feature observation point is generated. The cellular fusion feature data is parametrically corrected by mapping the steady-state components defined in each feature observation point data to a parametric plane and the dynamic components defined in each feature observation point data to a parametric surface; the intersection of the parametric plane and the parametric surface is solved to obtain the coupling correction parameter set; the cellular fusion feature data is corrected by the coupling correction parameter set to obtain the corrected cellular fusion feature data. Based on the corrected cell-based fusion feature data, a multi-hazard and power grid coupling time-series evolution map model is constructed to characterize the dynamic evolution process of disasters, power grid state changes and their coupling effects. The disaster and power grid data obtained from real-time monitoring are input into the multi-hazard and power grid coupled time-series evolution map model to perform dynamic simulation of power grid disaster damage and comprehensive impact assessment, and obtain the time-series evolution assessment results. Based on the time-series evolution assessment results, power grid emergency recovery optimization scheduling strategies and auxiliary decision-making information are generated.

2. The power grid multi-hazard coupled time-series evolution data processing and modeling method according to claim 1, characterized in that, Multi-scale spatiotemporal feature extraction is performed on standardized basic data to obtain a multi-dimensional spatiotemporal feature set, including: Based on standardized basic data, the original information of disaster-causing factors and power grid elements are analyzed to form the analyzed original information. Based on the parsed raw information, the temporal evolution characteristics and spatial distribution characteristics of disaster-causing factors are identified and extracted. Based on the parsed raw information, spatial layout features and state change features of power grid elements are identified and extracted. The temporal evolution characteristics, spatial distribution characteristics, spatial layout characteristics, and state change characteristics are quantitatively expressed to generate corresponding feature quantification indicators. Based on feature quantification indicators, a spatiotemporal feature vector corresponding to each disaster-causing factor and each power grid element is constructed. All spatiotemporal feature vectors are aggregated to form a multi-dimensional spatiotemporal feature set.

3. The power grid multi-hazard coupled time-series evolution data processing and modeling method according to claim 2, characterized in that, A spatiotemporal cellular system covering the power grid area is constructed. Multi-dimensional spatiotemporal feature sets are mapped and associated with the spatiotemporal cellular system to generate cellular fusion feature data containing data from each feature observation point, including: Determine the spatial boundaries of the power grid area and the time period range of the modeling process; After the spatial range boundary and time period range are determined, spatial clustering analysis is performed on the spatial locations corresponding to each spatiotemporal feature vector in the multidimensional spatiotemporal feature set to form multiple feature point clusters; a compact axial rectangle covering all feature points of each feature point cluster is calculated, and the axial rectangle is taken as a spatial cell; all spatial cells constitute a rectangular cell set covering the power grid area. Based on the time period range, a unified time reference and time unit are defined to expand each spatial cell in the rectangular cell set in the time dimension, forming a spatiotemporal cell with a unified spatiotemporal identifier. All spatiotemporal cells constitute the spatiotemporal cell system. Establish a mapping relationship between each spatiotemporal feature vector in the multidimensional spatiotemporal feature set and spatiotemporal cells with the same spatiotemporal identifier in the spatiotemporal cell system; Based on the mapping relationship, the feature information carried by each spatiotemporal feature vector is associated and filled into the corresponding spatiotemporal cell to form a spatiotemporal cell filled with feature information. Integrate all spatiotemporal cells filled with feature information to generate cell-based fused feature data containing data from each feature observation point.

4. The power grid multi-hazard coupled time-series evolution data processing and modeling method according to claim 3, characterized in that, The cellular fusion feature data is parametrically corrected by mapping the steady-state components defined in each feature observation point data to a parametric plane and the dynamic components defined in each feature observation point data to a parametric surface. The intersection of the parametric plane and the parametric surface is solved to obtain a set of coupling correction parameters. The cellular fusion feature data is then corrected using this set of coupling correction parameters to obtain the corrected cellular fusion feature data, including: Analyze the cell-based fusion feature data to extract the data of each feature observation point covered by the cell-based fusion feature data; Based on the data from each feature observation point, the steady-state component in the data from each feature observation point is separated and extracted; Based on the data from each feature observation point, separate and extract the dynamic component from the data of each feature observation point; The steady-state components are mapped to the parameter space to construct a parameterized plane that characterizes the long-term trend and baseline state of the data; Dynamic components are mapped to parameter space to construct parameterized surfaces that characterize the real-time fluctuations and evolution paths of data. In the parameter space, the set of coupling correction parameters is obtained by solving the intersection of the parameterized plane and the parameterized surface; By coupling the calibration parameter set, the cellular fusion feature data is calibrated and adjusted as a whole to generate the calibrated cellular fusion feature data.

5. The power grid multi-hazard coupled time-series evolution data processing and modeling method according to claim 4, characterized in that, Based on the corrected cell-based fusion feature data, a multi-hazard and power grid coupling time-series evolution model is constructed to characterize the dynamic evolution process of disasters, power grid state changes, and the temporal correlation of their coupling effects. This model includes: The corrected cell-based fusion feature data is analyzed to identify and extract key feature nodes that characterize the dynamics of disaster-causing factors and key feature nodes that characterize the power grid equipment and its operating status. Based on all key feature nodes, analyze and define the association relationships and interaction strengths between different categories of nodes; Based on the relationships and interaction strength, a basic topological network is constructed with key feature nodes as vertices and relationships as edges; The temporal information contained in the corrected cell-based fusion feature data is loaded into the basic topology network to assign a time-varying sequence of attribute states to vertices and edges. Based on the attribute state sequence, time series analysis is used to explore and quantify the coupling correlation and transmission pattern between disaster evolution and power grid state changes; By integrating the basic topology network, attribute state sequence, and coupling correlation and transmission rules, a multi-hazard and power grid coupling time-series evolution graph model is constructed, which is referred to as the graph model.

6. The power grid multi-hazard coupled time-series evolution data processing and modeling method according to claim 5, characterized in that, Real-time monitoring data on disasters and power grids is input into a multi-hazard and power grid coupled time-series evolution model to perform dynamic simulation and comprehensive impact assessment of power grid disaster losses, yielding time-series evolution assessment results, including: Receive and preprocess disaster-causing factor data and power grid operation status data acquired in real time, and form a real-time input data stream that matches the input format of the graph model; The real-time input data stream will be used to input the time-series evolution graph model of the coupling between multiple disasters and the power grid; Based on the relationships and evolution rules defined in the graph model, the state attributes of the corresponding vertices and edges in the graph model are driven and updated in time according to the real-time input data stream to obtain the graph model with updated state. Based on the updated graph model, the propagation process of disaster chains on the power grid topology and the chain change process of power grid state under the action of disaster are simulated and calculated to complete the dynamic simulation of power grid disaster loss including critical path and evolution stage, and generate dynamic simulation results simultaneously. Based on the dynamic simulation process and results, the power grid load loss, key equipment failure risk and power supply reliability degradation within the simulation scope are assessed to complete the comprehensive impact assessment and obtain the quantitative indicators of the comprehensive impact assessment simultaneously. By integrating the critical paths, evolution stages, and quantitative indicators of comprehensive impact assessment in dynamic simulation, a time-series evolution assessment result is generated.

7. The power grid multi-hazard coupled time-series evolution data processing and modeling method according to claim 6, characterized in that, Based on the time-series evolution assessment results, power grid emergency recovery optimization scheduling strategies and auxiliary decision-making information are generated, including: Analyze the time-series evolution assessment results to identify the set of critical power grid equipment damaged, the affected load areas, and the estimated recovery time windows; Based on the collection of damaged critical equipment and the affected load areas, and in conjunction with the pre-established emergency resource database, a list of emergency resource requirements for personnel, equipment, vehicles and materials is determined. Based on the emergency resource demand list and the estimated recovery time window, multi-objective optimization calculations are performed to generate a preliminary emergency recovery scheduling plan that includes resource allocation paths, task execution sequences and time nodes. Based on the time-series evolution assessment results, the feasibility of the preliminary emergency recovery dispatch plan is verified and dynamically adjusted to form the final power grid emergency recovery optimized dispatch strategy. By integrating the final power grid emergency recovery optimization scheduling strategy with the core conclusions defined in the time-series evolution assessment results, auxiliary decision-making information for supporting emergency command is obtained.

8. A power grid multi-hazard coupled time-series evolution data processing and modeling system, characterized in that, The system performs the method as described in any one of claims 1 to 7, comprising: The data acquisition module is used to acquire multi-source heterogeneous raw data of the power grid, and to perform normalization and consistency processing on the multi-source heterogeneous raw data to obtain standardized basic data. The feature extraction module is used to extract multi-scale spatiotemporal features from standardized basic data to obtain a multi-dimensional spatiotemporal feature set. The mapping and association module is used to construct a spatiotemporal cell system covering the power grid area, mapping and associating multi-dimensional spatiotemporal feature sets with the spatiotemporal cell system to generate cell-based fusion feature data containing data from each feature observation point; The optimization and correction module is used to perform parametric correction on the cell-based fusion feature data, that is, to map the steady-state components defined in each feature observation point data to a parametric plane, and to map the dynamic components defined in each feature observation point data to a parametric surface; solve the intersection line of the parametric plane and the parametric surface to obtain the coupling correction parameter set; and correct the cell-based fusion feature data through the coupling correction parameter set to obtain the corrected cell-based fusion feature data. The model building module is used to construct a multi-hazard and power grid coupling time-series evolution map model based on the corrected cell-based fusion feature data, which characterizes the dynamic evolution process of disasters, power grid state changes and their coupling effects. The simulation and evaluation module is used to input disaster and power grid data obtained from real-time monitoring into the multi-hazard and power grid coupled time-series evolution spectrum model to perform dynamic simulation and comprehensive impact assessment of power grid disaster damage, and obtain time-series evolution assessment results. The scheduling decision generation module is used to generate grid emergency recovery optimization scheduling strategies and auxiliary decision-making information based on the time-series evolution evaluation results.

9. A computing device, characterized in that, Including memory and processor; The memory stores one or more computer programs, the one or more computer programs including instructions; when the instructions are executed by the processor, the computing device performs the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program for performing the method as described in any one of claims 1 to 7.

Images