Typhoon prediction and power grid maintenance system based on digital twinning and AI cooperation

Abstract

The invention relates to the technical field of intelligent power grid management, in particular to a typhoon prediction and power grid maintenance system based on digital twinning and AI cooperation, which comprises a typhoon path evolution module, a power grid vulnerability mapping module, a risk link aggregation module, a maintenance strategy calibration module and a multi-stage maintenance tracking module. According to the method, through dynamic environment parameter extraction and path deviation trend analysis, continuous evolution judgment of a typhoon influence area can be realized, the problem of insufficient change capture of static prediction is solved, fine description and dynamic identification of a vulnerable part are realized, and through node connection relation induction and load intensity partitioning, a typhoon influence area can be identified. A risk link is identified and a potential fault chain is structurally presented; resource configuration and risk priority dynamic comparative analysis are maintained; a deployment logic can be calibrated; a task sequence is ensured to be consistent with a risk level; and the response efficiency and the scheduling accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of smart grid management technology, and in particular to a typhoon prediction and power grid maintenance system based on collaboration between digital twins and AI. Background Art

[0002] The field of smart grid management technology involves the intelligent operation and maintenance of power systems. Its core areas include power load forecasting, grid equipment status monitoring, automated power dispatching, fault diagnosis and response, and early warning and handling of natural disaster impacts on the grid. This technology integrates sensing, communication, and information processing technologies with power system engineering to achieve real-time perception and intelligent response to grid operating conditions, improving the reliability and efficiency of power systems. This is particularly critical for predicting and maintaining safe grid operation in areas prone to natural disasters. Traditional typhoon prediction and grid maintenance systems use meteorological data to predict typhoon paths and intensities before a typhoon disaster occurs, and develop grid maintenance plans based on empirical data or static historical data. These systems rely on meteorological radar observations for path estimation, followed by manual analysis of vulnerable grid areas and the pre-deployment of emergency personnel and equipment. This approach primarily relies on historical typhoon path distribution, the geographic layout of the grid, and records of areas affected by previous power outages to identify potential risk areas and formulate emergency response measures and maintenance plans.

[0003] Before a typhoon arrives, existing technologies mainly rely on historical path data and manual experience to judge potential risk areas. They lack the ability to judge and dynamically adjust path evolution trends in real time. When dealing with sudden typhoon offsets or path changes, the response is delayed, resulting in a disconnect between resource allocation and the actual affected areas. Maintenance plans formulated based on static data are difficult to identify vulnerabilities caused by real-time load changes in power grid operation, resulting in prediction errors. When historical records do not cover the current typhoon path or the power grid layout has been adjusted, unreasonable emergency deployment is prone to occur. If the scheduling priority relies on a fixed template, the task advancement sequence is prone to confusion in the face of changes in risk levels, ultimately leading to waste of emergency repair resources and delays in power supply restoration. Summary of the Invention

[0004] The purpose of this invention is to solve the shortcomings of the existing technology and propose a typhoon prediction and power grid maintenance system based on the collaboration of digital twins and AI.

[0005] To achieve the above objectives, the present invention adopts the following technical solutions: A typhoon prediction and power grid maintenance system based on digital twin and AI collaboration includes: The typhoon path evolution module is based on the dynamic environmental parameters transmitted by the meteorological sensor network, combined with the geographical layout information of the power grid, to screen the input feature dimensions, analyze the frequency of changes in the path vector in multiple time period samples, and generate a spatiotemporal distribution map of the typhoon's impact range; The power grid vulnerability mapping module calls the spatiotemporal distribution map of the typhoon's impact range, extracts the stress and load change characteristics of key nodes in the typhoon's impact range, combines AI sequence modeling to analyze abnormal fluctuations in power grid operating parameters, compares stress positions through the digital twin node mapping warehouse, and establishes a power grid vulnerability mapping set; The risk link aggregation module calls the power grid vulnerability mapping set, detects potential fault triggering links, records the connection relationship between risk nodes, forms a partition structure according to the node stress strength, and constructs a power grid risk link structure set; The maintenance strategy calibration module calls the power grid risk link structure set, locates the resource configuration chain in the power grid maintenance platform that is consistent with the illustrated link, extracts the current maintenance priority, compares it with the priority benchmark in the standard maintenance strategy template, adjusts the chain with inconsistent priorities, and generates a digital twin maintenance priority sequence table.

[0006] As a further solution of the present invention, the spatiotemporal distribution map of the typhoon impact range includes path deviation trends, impact range boundaries, time series expansion characteristics, spatial coverage parameters, and node stress indicators; the power grid vulnerability mapping set includes key node numbers, load fluctuation labels, abnormal fluctuation sequences, node stress indexes, and vulnerability identification marks; the power grid risk link structure set includes fault trigger link numbers, node stress statistical blocks, risk node groups, link priority identifiers, and aggregated link labels; the digital twin maintenance priority sequence table includes resource configuration chain numbers, current priority labels, standard priority benchmarks, priority deviation values, and calibration classification marks.

[0007] As a further solution of the present invention, the typhoon path evolution module includes: The dynamic environmental parameter extraction submodule identifies the changing trend of environmental variables in a continuous time period based on the dynamic environmental parameters transmitted by the meteorological sensor network, selects the recording intervals that meet the path deviation conditions, and generates a typhoon path evolution sequence; The path deviation trend screening submodule counts the path deviation amplitude in differentiated time periods based on the typhoon path evolution sequence, determines whether it exceeds the path deviation threshold, screens the corresponding path, and records the deviation time period and spatial variation to obtain the high-frequency deviation path change rate; The spatiotemporal distribution map construction submodule calls the high-frequency offset path change rate, identifies the path offset structure, determines the coverage range and offset order, formulates the time series structure and sets the coverage weight according to the offset time period, and establishes the spatiotemporal distribution map of the typhoon impact range.

[0008] As a further solution of the present invention, the power grid vulnerability mapping module includes: The node force identification submodule calls the spatiotemporal distribution map of the typhoon impact range, extracts the node number and load state identifier, screens the numbers with load mutation and state offset, analyzes the load mutation amplitude and state change intensity, and obtains the node force intensity value; The abnormal fluctuation aggregation submodule associates the abnormal fluctuation sequence marked in the power grid operation parameters according to the node stress intensity value, screens the numbers in the number sequence that coincide with the fluctuation sequence, aggregates and classifies them, calculates the number fluctuation aggregation value, merges and classifies them according to the numerical interval, and obtains the fluctuation aggregation value; The vulnerability marker identification submodule compares the number and the fluctuation category label based on the fluctuation aggregation value, extracts the co-occurring node pairs in the mapping sequence, analyzes the overlapping feature intervals and load differences between the numbers, and establishes a power grid vulnerability mapping set.

[0009] As a further solution of the present invention, the risk link aggregation module includes: The fluctuation tag analysis submodule calls the power grid vulnerability mapping set, extracts the fluctuation tag group within the link, detects the fluctuation tag cross structure by combining the link position value and the tag type, calculates the cross frequency and position offset value, divides the link fluctuation clusters according to the offset position, and obtains the link fluctuation cluster frequency data; The fault node mapping submodule calls the link fluctuation cluster frequency data, selects nodes with a stress rate higher than the fluctuation threshold, recodes them according to the node number and frequency, and generates a standard mapping node set; The risk link aggregation submodule maps the node set according to the standard, aggregates the associated link node structure, identifies the node load sequence number, position sequence and trigger signal value, identifies the link structure correlation index, sorts the blocks, counts the number of links and node ratios covered by the blocks, and constructs the power grid risk link structure set.

[0010] As a further solution of the present invention, the maintenance strategy calibration module includes: The risk link identification submodule calls the power grid risk link structure set, compares the node sequence and logic of the illustrated link with the maintenance platform resource configuration chain, identifies the risk link whose structure consistency exceeds the threshold, and generates a power grid risk link set; The priority deviation judgment submodule extracts the link priority label based on the grid risk link set, extracts the standard priority value of the corresponding link in the standard maintenance policy template, determines the corresponding position of the two in the priority sequence and analyzes the difference to obtain the link priority deviation value; The label mapping adjustment submodule selects the link with the priority deviation exceeding the threshold value according to the link priority deviation value, extracts the priority label, node load strength, link coupling strength and the number of fluctuation conflicts, analyzes the maintenance label correction amplitude and reconstructs the priority label mapping, identifies the label mapping relationship, and generates a digital twin maintenance priority sequence table.

[0011] As a further aspect of the present invention, the system further includes a multi-stage maintenance tracking module: The multi-stage maintenance tracking module calls the digital twin maintenance priority sequence table, marks the priority corresponding values ​​of the trigger instructions in the multi-stage maintenance process, identifies the distribution trend of priority changes in the instruction transmission path, screens the chain with continuously increasing priority, and outputs a priority-driven multi-stage maintenance task list; The priority-driven multi-stage maintenance task list includes stage-by-stage maintenance priorities, priority change trajectories, link concentration sections, continuous increment identifiers, and instruction conduction path mapping.

[0012] As a further aspect of the present invention, the multi-stage maintenance tracking module includes: The priority mapping marking submodule calls the digital twin maintenance priority sequence table, identifies the link number, trigger time and response time in the path, extracts the original priority value and compares it with the mapping priority item, selects the corresponding value, and generates a maintenance priority mapping value set; The priority trend identification submodule, based on the maintenance priority mapping value set, sorts the link number sequence and the response time sequence by number, identifies the difference between adjacent values ​​and determines whether it is positive or negative, marks the growth node, extracts the continuous positive difference sequence, records the link number, the total difference and the number of nodes, eliminates the low-frequency change segment, and generates the number of increasing trend sequences; The link screening and aggregation submodule extracts high-frequency increasing path segments according to the number of increasing trend sequences, identifies the start and end numbers, analyzes the cumulative increase and the increase rate, screens the path segments whose rate exceeds the benchmark value, and obtains a priority-driven multi-stage maintenance task list.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, through the extraction of dynamic environmental parameters and analysis of path deviation trends, it is possible to realize the continuous evolution judgment of the typhoon-affected area, effectively breaking through the problem of insufficient capture of actual changes by static predictions, and combining the abnormal fluctuation identification and node force comparison in the power grid operation parameters, it is possible to achieve fine characterization and dynamic identification of vulnerable parts, and further through the induction of the connection relationship between nodes and the load intensity partitioning, it is possible to identify risk links and structured presentation of potential power grid fault chains, and at the same time, through the dynamic comparative analysis of the priority after maintenance resource allocation and risk link mapping, the resource allocation logic can be refined and calibrated to ensure that the priority of maintenance tasks is consistent with the actual risk level, and by tracking the priority distribution trend in the multi-stage task chain, the high-priority continuously increasing links can be screened, thereby efficiently promoting the distributed execution path of the multi-stage emergency repair tasks, and improving the overall response efficiency and the accuracy of power grid restoration scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a system flow chart of the present invention; Figure 2 This is a flow chart of the typhoon path evolution module in the present invention; Figure 3 This is a flow chart of the power grid vulnerability mapping module in the present invention; Figure 4 This is a flow chart of the risk link aggregation module in the present invention; Figure 5 This is a flow chart of the maintenance strategy calibration module in the present invention; Figure 6 This is a flow chart of the multi-stage maintenance tracking module in the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0016] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0017] See also Figure 1 The typhoon prediction and power grid maintenance system based on the collaboration of digital twins and AI includes: The typhoon path evolution module uses dynamic environmental parameters transmitted by the meteorological sensor network and combined with the geographical layout information of the power grid to determine the path deviation trend, filter the input feature dimensions, analyze the change frequency of the path vector in multi-period samples, and generate a spatiotemporal distribution map of the typhoon's impact range; The power grid vulnerability mapping module uses the spatiotemporal distribution map of the typhoon's impact area to extract the stress and load change characteristics of key nodes within the typhoon's impact area. It then uses AI sequence modeling to analyze abnormal fluctuations in power grid operating parameters, compares stress positions through the digital twin node mapping warehouse, and establishes a power grid vulnerability mapping set. The risk link aggregation module calls the power grid vulnerability mapping set, detects potential fault trigger links, records the connection relationship between risk nodes, forms a partition structure according to the node stress strength, and constructs the power grid risk link structure set; The maintenance strategy calibration module calls the power grid risk link structure set, locates the resource configuration chain in the power grid maintenance platform that is consistent with the illustrated link, extracts the current maintenance priority, compares it with the priority benchmark in the standard maintenance strategy template, adjusts the chains with inconsistent priorities, and generates a digital twin maintenance priority sequence table; The multi-stage maintenance tracking module calls the digital twin maintenance priority sequence table, marks the priority corresponding values ​​of the link trigger instructions in the multi-stage maintenance process, identifies the distribution trend of priority changes in the instruction transmission path, screens the priority continuously increasing chain, and outputs the priority-driven multi-stage maintenance task list.

[0018] The spatiotemporal distribution map of the typhoon's impact range includes path deviation trends, impact range boundaries, time series expansion characteristics, spatial coverage parameters, and node stress indicators. The power grid vulnerability mapping set includes key node numbers, load fluctuation labels, abnormal fluctuation sequences, node stress indexes, and vulnerability identification marks. The power grid risk link structure set includes fault trigger link numbers, node stress statistical blocks, risk node groups, link priority identifiers, and aggregated link labels. The digital twin maintenance priority sequence table includes resource configuration chain numbers, current priority labels, standard priority benchmarks, priority deviation values, and calibration classification marks. The priority-driven multi-stage maintenance task list includes stage-by-stage maintenance priorities, priority change trajectories, link concentration sections, continuous incremental identifiers, and instruction conduction path mapping.

[0019] See also Figure 2 , the typhoon path evolution module includes: The dynamic environmental parameter extraction submodule identifies the changing trend of environmental variables in a continuous time period based on the dynamic environmental parameters transmitted by the meteorological sensor network, selects the recording intervals that meet the path deviation conditions, and generates a typhoon path evolution sequence; Based on the typhoon-related dynamic environmental parameter dataset received in real time by the meteorological sensor network, the dataset includes variables such as wind speed, wind direction, air pressure, temperature, and humidity at multiple monitoring points within a continuous time period. The changing trend of each environmental variable within a continuous time period (for example, once every hour in the past 24 hours) is identified and extracted from the dataset to determine whether the wind speed continues to increase (the change trend is positive and the amplitude exceeds the preset enhancement threshold, for example, the wind speed increases by more than 2m / s per hour for 3 consecutive hours) or whether the wind direction has significantly deflected (the change trend is that the direction angle change rate exceeds the preset deflection threshold, for example, the wind direction change rate exceeds 10 degrees / hour for 2 consecutive hours), and at the same time, combined with the air pressure change (for example, the continuous decrease rate of air pressure exceeds the preset threshold, For example, 1hPa / hour), comprehensively judge whether the changes in environmental variables meet the condition set for the typhoon path deviation, and filter out all recorded time intervals that meet the above path deviation conditions (for example, T1 period: [10:00, 12:00], T2 period: [15:00, 18:00]). Based on the meteorological data in the filtered recording intervals, a typhoon path evolution sequence containing key information such as timestamp, center position latitude and longitude, wind speed, and wind direction is generated. This sequence reflects the state changes of the typhoon during the period when path deviation may occur. For example, the sequence contains the typhoon center position and maximum wind speed data points recorded in the T1 and T2 periods. The data points are arranged in chronological order to form a description of the typhoon's path trajectory during a specific period.

[0020] The path deviation trend screening submodule is based on the typhoon path evolution sequence, counts the path deviation amplitude in different time periods, determines whether it exceeds the path deviation threshold, screens the corresponding path, and records the deviation time period and spatial variation to obtain the high-frequency deviation path change rate; Based on the typhoon path evolution sequence, which records the position information of the typhoon in a specific time period, the sequence is first divided according to the preset time granularity (for example, a differentiated time period every 3 hours). For example, the 24-hour evolution sequence is divided into 8 3-hour time periods. For each differentiated time period, the straight-line distance between the starting point and the end point of the time period is counted as the path deviation amplitude. The deviation amplitude is compared with the preset path deviation threshold (the threshold is obtained by analyzing historical typhoon data. For example, the average path deviation of typhoons that caused significant damage to the power grid in similar time periods in the historical records is added with a standard deviation, and the threshold is set to 30 kilometers). If the deviation amplitude of a certain time period exceeds the threshold, it is judged that the path in the time period has significantly deviated, and the typhoons are screened out. All time periods with deviation amplitudes exceeding the threshold and their corresponding typhoon path segments (for example, the path deviation amplitude in the period [15:00, 18:00] is 45 kilometers, which is greater than 30 kilometers) are recorded, and the deviation time period (for example, [15:00, 18:00]) and the spatial change of the typhoon center position during this period are recorded (for example, from (longitude 118.5, latitude 22.3) to (longitude 118.9, latitude 22.7), the spatial change is (0.4 longitude, 0.4 latitude)). By analyzing the ratio of the spatial change of the screened path segments to the length of the corresponding time period (for example, 3 hours), the high-frequency deviation path change rate is calculated. For example, for the period [15:00, 18:00], the path change rate is (0.4 longitude / 3 hours, 0.4 latitude / 3 hours).

[0021] The spatiotemporal distribution map construction submodule calls the high-frequency offset path change rate, identifies the path offset structure, determines the coverage range and offset sequence, formulates the time series structure and sets the coverage weight based on the offset time period, and establishes the spatiotemporal distribution map of the typhoon impact range; A high-frequency offset path change rate set is called, which contains multiple time periods and their corresponding typhoon path spatial change information. Based on the high-frequency offset path change rate, the structure of the typhoon path offset is identified on the geographic information system (GIS) platform (for example, the path segment of each offset time period is drawn as a line with an arrow, and the direction of the arrow indicates the offset direction). The impact coverage of each offset structure is determined (for example, based on the typhoon wind field model, the regional boundary within the impact range of a specific wind speed (for example, the minimum wind speed of 17m / s that reaches the wind resistance level of power grid equipment) is determined in combination with the offset path segment) and the offset sequence (for example, marking according to the order of the offset time periods). The overall typhoon impact is formulated according to each offset time period (for example, [15:00, 18:00], [20:00, 23:00]). Time series structure, and set coverage weights for each time period (the weights are determined according to factors such as offset amplitude, duration, and regional importance. For example, the larger the offset amplitude, the longer the duration, and the affected area contains important power grid facilities, the higher the weight. The weight of [15:00, 18:00] is set to 0.7, and the weight of [20:00, 23:00] is set to 0.5). For example, the weights can be obtained by normalization processing, with the maximum weight set to 1. The weighted sum of the offset amplitude, duration, and regional importance score (for example, 1-5 points) is divided by the maximum possible weighted sum value. Finally, the identified offset structure, coverage range, offset order, time series structure, and coverage weights are superimposed on the geographic information base map to establish a spatiotemporal distribution map of the typhoon's impact range, which intuitively shows the impact range and intensity of the typhoon at different times and spaces.

[0022] See also Figure 3 , the grid vulnerability mapping module includes: The node force identification submodule calls the spatiotemporal distribution map of the typhoon's impact range, extracts the node number and load status identifier, screens the numbers with load mutations and state offsets, analyzes the load mutation amplitude and state change intensity, and obtains the node force intensity value; The spatiotemporal distribution map of the typhoon's impact range is generated by calling the spatiotemporal distribution map construction submodule. The map marks the specific geographical areas and intensities affected by the typhoon at different times. The power grid topology map is superimposed on the spatiotemporal distribution map, and the numbers of the power grid nodes (e.g., substations, transmission towers) located in the typhoon-affected area and the load status identification (e.g., normal, overloaded, light load, the status identification comes from the power grid SCADA system) during the typhoon-affected period are extracted. The node numbers (e.g., node numbers N101, N205) with load mutations (e.g., load change rate exceeds the normal fluctuation threshold, e.g., load change exceeds 10% for 15 consecutive minutes) and state deviations (e.g., from normal state to overload state) during the typhoon are screened out. For each screened node, its load mutation amplitude is analyzed (e.g., the load of N101 suddenly changes from 50MW to 80MW during the period of 16:00-16:15, with a mutation amplitude of 3 0MW) and state change intensity (for example, the state change intensity from normal to overload is marked as 3, and the state change intensity from normal to light load is marked as 1). According to the load mutation amplitude and state change intensity, and combined with the environmental parameters such as wind speed and wind pressure of the typhoon at the node location (for example, the greater the wind speed and the higher the wind pressure, the greater the impact on the node), the node force intensity value of each node is calculated. This value comprehensively reflects the physical and operational pressure that the node bears under the action of the typhoon. For example, it is calculated by weighted summing the load mutation amplitude (normalized), the state change intensity mark and the local wind speed (normalized). For example, the force intensity value of node N101 is: normalized mutation amplitude (30MW / 100MW_max)×0.5+state change intensity mark (3)×0.3+normalized local wind speed (25m / s / 60m / s_max)×0.2=0.15+0.9+0.083=1.133.

[0023] The abnormal fluctuation aggregation submodule associates the abnormal fluctuation sequence marked in the power grid operation parameters according to the node stress intensity value, screens the numbers in the number sequence that coincide with the fluctuation sequence, aggregates and classifies them, calculates the number fluctuation aggregation value, and merges and classifies them according to the numerical interval. The formula is: ; Get the volatility imputed value; in, represents the volatility collection value, Representative The fluctuation intensity change value of the number, Represents the strength of force numbered i, Represents the average strength value of all numbers in the sequence, Indicates the total number of numbers; According to the node stress intensity value set, which contains the stress intensity values ​​of each power grid node under the influence of the typhoon, first associate the abnormal fluctuation sequence marked in the power grid operation parameters. This sequence records the abnormal electrical parameter fluctuations (for example, voltage surges and dips, frequency anomalies, current overloads) that occur at each power grid node within a specific time period. Each fluctuation event has a corresponding node number and fluctuation category label. Filter out the numbers in the number sequence in the node stress intensity value set that coincide with the node numbers in the abnormal fluctuation sequence (for example, node N101 is in both the stress intensity value set and the abnormal fluctuation sequence). Aggregate and classify the overlapping nodes according to the fluctuation category or geographical location (for example, classify the nodes with voltage sags into one category, and classify the nodes in the same substation into one category). For each aggregation category, calculate the fluctuation aggregation value V of all nodes in the category. This value is calculated by the formula Calculated, where represents the volatility collection value, Representative The fluctuation intensity change value of the numbered node in the abnormal fluctuation sequence (for example, the voltage sag amplitude, current overload ratio, this value has been quantized, for example, if the voltage drops from 110kV to 90kV, the fluctuation intensity change value is 20kV, which is quantized to 20). Represents the force strength value of the node in the node force strength value set (for example, the force strength value of node N101 is 1.133), Represents the average strength value of all the coincident numbered nodes in the sequence. For example, if the coincident nodes are N101 (S=1.133) and N205 (S=0.95), then , Indicates the total number of overlapping nodes in the aggregation category. For example, if there are N101 and N205 in the category, and the corresponding fluctuation intensity change value , ; The fluctuation aggregation value of this category is: ; According to the calculated fluctuation aggregation value range (for example, (0, 10) is low, (10, 50) is medium, (50, ) is high) to merge and classify the aggregate categories and obtain the fluctuation aggregation values ​​of different categories. The benefit of this formula is that by introducing the force strength of the node and average stress strength , so that the fluctuation aggregation value not only reflects the intensity of electrical fluctuations, but also combines the impact of external typhoon loads on nodes. The weighting of regions with higher overall stress intensity is adjusted, avoiding the bias inherent in judging by fluctuation intensity alone. This allows for a more accurate assessment of the overall abnormal fluctuation level of the power grid under typhoon conditions. The results indicate that this clustering category experiences moderately high abnormal fluctuations under typhoon conditions, necessitating further analysis of its vulnerability.

[0024] The vulnerability marker identification submodule compares the number and fluctuation category label based on the fluctuation aggregation value, extracts the co-occurring node pairs in the mapping sequence, analyzes the overlapping feature intervals and load differences between the numbers, and establishes a power grid vulnerability mapping set; Based on the fluctuation aggregation value and the corresponding node number and fluctuation category label, the number of each node in the abnormal fluctuation sequence is compared with the associated fluctuation category label (for example, node N101 - voltage sag, N205 - current overload), and the node pairs that co-occur with the fluctuation nodes in the power grid topology mapping sequence (the sequence represents the connection relationship between nodes in the power grid) are extracted, that is, pairs in which at least one node in the directly connected node pair appears in the fluctuation node list (for example, if N101 is connected to N102, and N101 is a fluctuation node, then (N101, N102) is a co-occurring node pair), and the overlapping feature intervals between the co-occurring node pairs are analyzed (for example, the shared line segment , substations in the same area) and load differences (for example, the difference in load levels of connected nodes during the typhoon-affected period, node N101 load 80MW, N102 load 60MW, a difference of 20MW). By analyzing overlapping feature intervals and load differences, the degree of vulnerability correlation between node pairs is evaluated. For example, the more shared line segments and the greater the load difference, the stronger the vulnerability correlation. A power grid vulnerability mapping set is established, which contains co-occurring node pairs, their vulnerability correlation degrees, and corresponding fluctuation category information. This mapping set reveals which node pairs in the power grid have higher vulnerability correlations under the action of typhoons, and what type of fluctuations cause this correlation.

[0025] See also Figure 4 , the risk link aggregation module includes: The fluctuation tag analysis submodule calls the power grid vulnerability mapping set, extracts the fluctuation tag group within the link, detects the fluctuation tag cross structure through the combination of link position value and tag type, calculates the cross frequency and position offset value, and divides the link fluctuation clusters according to the offset position. The formula is: ; Obtain link fluctuation cluster frequency data; in, Represents the link fluctuation cluster frequency data, Representative The location label type of the link label group, Representative The corresponding position value of the link label group, Representative The displacement deviation value between the link label group type and the adjacent label type, Represents the total number of link label groups; The grid vulnerability mapping set is called, which contains node pairs, their vulnerability associations, and fluctuation category information. The fluctuation label group contained in each grid link is extracted from the mapping set (for example, a link contains nodes N101 (voltage sag) and N103 (current overload)). By analyzing the link position value (for example, the order or relative position of the nodes on the link) and the combination of label type (for example, the voltage sag label appears at the beginning of the link, and the current overload label appears in the middle), the cross structure of the fluctuation label on the link is detected (for example, different types of fluctuation labels appear alternately on the link), and the frequency of label crossing (for example, the voltage sag and current overload labels appear alternately 3 times on a link) and the position offset value (for example, the number of nodes between adjacent labels of different types) are calculated. The link fluctuation clusters are divided according to the offset position (for example, labels with an offset position of less than 3 nodes are grouped together) (for example, labels in the same cluster are regarded as related fluctuation events). The formula is used. Calculate the frequency data of link fluctuation clusters, where Represents the link fluctuation cluster frequency data, Representative The location tag type of each link tag group (for example, quantize the voltage dip type as 1 and the current overload type as 2), Representative The corresponding position value of the link label group (for example, the position of the first fluctuation label on the link is set to 1, the second to 2, and so on), Representative The displacement deviation value between the link tag group type (such as voltage sag) and the adjacent tag type (such as current overload) (for example, the difference in the number of nodes between two adjacent tags of different types, if the voltage sag is at position 2 and the current overload is at position 5, the displacement deviation is 3). Represents the total number of label groups on the link. For example, if there are three fluctuating labels on a link, the type quantization values ​​are 1, 2, and 1, and the position values ​​are 1, 3, and 4. but hour, , without a preceding label, the item is considered ; hour, , pre-order label type 1, position 1, displacement deviation is ; hour, , pre-order label type 2, position 3, displacement deviation is 4-3 ; Then the link fluctuation cluster frequency data is: ; This formula is useful for obtaining link fluctuation cluster frequency data. By combining the position of the label type and the positional deviation of adjacent types, it quantifies the degree of clustering and alternation of different types of fluctuations on the link. The smaller the displacement deviation (i.e., the closer the different types of fluctuations are), the greater the contribution to the overall frequency data (because the denominator becomes smaller). This indicates that closely adjacent fluctuation types may point to deeper link vulnerability issues, thereby more accurately identifying fluctuation risk clusters on the link. This result indicates that the frequency of fluctuation risk clusters on this link is high and requires further attention.

[0026] The fault node mapping submodule calls the link fluctuation cluster frequency data, selects nodes with a stress rate higher than the fluctuation threshold, recodes them according to the node number and frequency, and generates a standard mapping node set; The link fluctuation cluster frequency data is called, which quantifies the frequency of fluctuation clusters on the power grid link. The nodes whose stress ratio (the ratio of the node stress intensity value to the node's maximum bearing capacity (for example, rated capacity, physical strength level)) is higher than the fluctuation threshold (the threshold is set according to historical fault data, for example, the lowest stress ratio value of the historical fault nodes plus a buffer value is statistically calculated, and the threshold is set to 0.2) are selected from the power grid node data (for example, the stress ratio of node N101 is 0 .2266>0.2), the node is a node that is subject to greater stress and prone to fluctuations under the influence of typhoons. The number of the screened node (for example, N101) and its corresponding fluctuation cluster frequency (for example, the fluctuation cluster frequency of the link where N101 is located is 3.707) are recoded to generate a standard mapping node set. For example, the node number and frequency can be combined into a new identifier (for example, N101_3.707), or the frequency information can be added to the node set as a node attribute. This mapping node set highlights which nodes in the power grid are not only subject to greater stress under the influence of typhoons, but also have frequent fluctuations on the links related to them.

[0027] The risk link aggregation submodule maps the node set according to the standard, aggregates the associated link node structure, identifies the node load sequence number, position sequence and trigger signal value, identifies the link structure correlation index, sorts the blocks, counts the number of links and node ratios covered by the blocks, and constructs the power grid risk link structure set; According to the standard mapping node set, the set identifies the nodes with large stress and frequent fluctuations on the related links, aggregates the power grid link node structure associated with the standard mapping node, that is, finds all links containing the standard mapping node and the nodes and connection relationships on them, identifies the load sequence number (that is, the identifier of the load change sequence of the standard mapping node during the typhoon-affected period), position sequence (the relative position of the standard mapping node on the link) and trigger signal value (for example, the event signal that causes the node to fluctuate, such as the wind speed reaching a certain threshold) of the standard mapping node contained in the link structure, and identifies the load sequence number, position sequence and trigger signal value. Link structure correlation indicators (for example, analyzing the concentration of standard mapping nodes on the link, the correlation of load changes, and the similarity of trigger signals). For example, if the load change trends of multiple standard mapping nodes on a link are consistent and the trigger signals are the same, the link correlation is high, and block sorting is performed. Link structures with higher correlation are divided into the same risk block and sorted (for example, sorted by correlation level). The number of links covered by each risk block and the proportion of standard mapping nodes included are counted to construct a power grid risk link structure set. This set describes in detail the risk link group in the power grid that fails as a whole under the action of a typhoon and its internal structural characteristics.

[0028] See also Figure 5 , the maintenance strategy calibration module includes: The risk link identification submodule calls the power grid risk link structure set, compares the node sequence and logic of the illustrated link with the maintenance platform resource configuration chain, identifies risk links whose structural consistency exceeds the threshold, and generates a power grid risk link set; The grid risk link structure set is called, which identifies and aggregates high-risk link groups in the grid. The risk links (for example, a link topology consisting of a series of nodes and connections) illustrated in the structure set are compared with the node sequence and logic of the resource configuration chain recorded in the maintenance platform. The resource configuration chain of the maintenance platform describes the link path that resources (such as repair teams and equipment) need to arrive and configure in a specific order and logic when performing maintenance operations. For example, to maintain a line segment, it is necessary to first reach substation A and then go along the line to the fault point. The identification structure consistency exceeds the threshold (the threshold is set based on historical maintenance experience. For example, if the nodes of the risk link and the maintenance resource chain are A risk link with a point order matching degree (e.g., the ratio of the number of matching nodes to the total number of nodes) greater than 80% is considered to have high structural consistency (with a threshold of 0.8). This is a risk link that highly matches the existing maintenance resource chain in terms of topology and logical order (e.g., the node order of risk link L1 is N101-N103-N105, and the node order of maintenance chain M1 is N101-N103-N105, with a matching degree of 100% > 0.8). Links with high consistency indicate that the existing maintenance process can be directly applied to address risks, generating a power grid risk link set that lists high-risk links that are compatible with the existing maintenance resource configuration.

[0029] The priority deviation judgment submodule extracts the link priority label based on the grid risk link set, extracts the standard priority value of the corresponding link in the standard maintenance strategy template, determines the corresponding position of the two in the priority sequence and analyzes the difference, using the formula: ; Get the link priority deviation value; in, Represents the link priority deviation value, Representative The actual priority value of the link, represents the standard priority value of the jth link, Representative The weight of the link, Represents the total number of links; According to the power grid risk link set, which contains a list of high-risk links that are compatible with the existing maintenance resource configuration, the priority label of each risk link is extracted from the power grid risk link set. The label reflects the link risk level (for example, high risk, medium risk, low risk) obtained based on the typhoon impact assessment. At the same time, the standard priority value of the corresponding link in the standard maintenance strategy template is extracted. The standard maintenance strategy template is a preset power grid maintenance plan, in which priority values ​​are preset for different links. For example, the trunk transmission line has a high priority and the branch line has a low priority. The corresponding positions of the two in the priority sequence are judged and the differences are analyzed. For example, the risk priority label of the risk link L1 is "high risk" and it ranks high in the priority sequence (for example, set to 1). The standard priority value of the link L1 in the standard maintenance strategy template is "medium" and it ranks in the middle of the priority sequence (for example, set to 3). The difference in position between the two in the priority sequence is , using the formula Calculate the link priority deviation value, where Represents the link priority deviation value, Representative The actual priority value of each link (i.e., the priority based on typhoon risk assessment, for example, high risk is quantified as 3, medium risk is 2, and low risk is 1), Representative The standard priority value of each link (for example, in the standard maintenance policy, high priority is quantized as 3, medium priority is 2, and low priority is 1), Representative The weight of each link can be set according to the importance of the link (for example, transmission capacity, influence range). For example, the weight of the trunk transmission line is set to 0.8, and the branch line is set to 0.4. Represents the total number of risk links. For example, if there are two links L1 and L2 in the risk link set; L1: actual priority 3, standard priority 2, weight 0.8; L2: actual priority 2, standard priority 3, weight 0.4; Then the link priority deviation value is: ; The link priority deviation value is obtained. The benefit of this formula is that it quantifies the degree of deviation between the link risk priority under the influence of typhoon and the routine maintenance priority by calculating the root mean square of the weighted difference between the actual priority and the standard priority. This makes the priority deviation of important links contribute more to the total deviation value, highlighting the importance of adjusting the priority of key links under extreme weather conditions. This result shows that the overall priority of the risk link set needs to be adjusted according to the typhoon risk.

[0030] The label mapping adjustment submodule selects links with priority deviation exceeding the threshold value based on the link priority deviation value, extracts priority labels, node load strength, link coupling strength, and the number of fluctuation conflicts, analyzes the maintenance label correction amplitude, reconstructs the priority label mapping, identifies the label mapping relationship, and generates a digital twin maintenance priority sequence table; Based on the link priority deviation value, which quantifies the difference between the typhoon risk priority and the standard maintenance priority, links with priority deviation exceeding the threshold (the threshold is set according to management requirements, for example, the threshold is set to 1.0, indicating that links with priority deviation values ​​greater than 1.0 need to adjust their priorities) are selected (for example, the priority deviation value calculated in the previous paragraph is 1.34>1.0, so the risk link set needs to be adjusted). The priority label (for example, high risk), node load intensity (for example, the average force intensity value of the nodes on the link), link coupling strength (for example, the connection density between the nodes on the link), and the number of fluctuation conflicts (for example, the number of different types of fluctuation labels on the link) of the links with priority deviation exceeding the threshold are extracted. The system then reconstructs the priority label mapping. For example, the priority value corresponding to the "high risk" label is increased from 3 to 4 in the standard maintenance strategy, and the priority value corresponding to the "medium risk" label is increased from 2 to 3. The mapping relationship between the adjusted priority label and the original risk label is identified (for example, the original label "high risk" is mapped to the new priority value 4). This generates a digital twin maintenance priority sequence table. This table provides a grid maintenance priority ranking based on typhoon risk calibration, which is used to guide actual maintenance operations.

[0031] See also Figure 6 , the multi-stage maintenance tracking module includes: The priority mapping marking submodule calls the digital twin maintenance priority sequence table, identifies the link number, trigger time and response time in the path, extracts the original priority value and compares it with the mapping priority item, selects the corresponding value, and generates the maintenance priority mapping value set; Call the digital twin maintenance priority sequence list, which provides a grid maintenance priority ranking based on typhoon risk calibration, identify the link number (e.g., node number on the path), trigger time (e.g., the time when the typhoon impact causes the path risk to reach a preset threshold) and response time (e.g., the estimated maintenance start time based on the availability of maintenance resources) in each maintenance path (e.g., one or more associated links) in the sequence list, extract the original priority value (e.g., the priority in the standard maintenance strategy) and compare it with the mapped priority item (i.e., the calibrated priority in the digital twin sequence list), select the corresponding value, and generate a maintenance priority mapping value set, which provides a calibrated priority value for each maintenance path under the typhoon scenario. For example, the original priority of path A is medium (quantized as 2), and the calibrated mapped priority is high (quantized as 4), then the mapping value is 4.

[0032] The priority trend identification submodule is based on the maintenance priority mapping value set, sorts the link number sequence and the response time sequence by number, identifies the difference between adjacent values ​​and judges whether it is positive or negative, marks the growth nodes, extracts the continuous positive difference sequence, records the link number, the total difference and the number of nodes, eliminates the low-frequency change segments, and generates the number of increasing trend sequences; Based on the maintenance priority mapping value set, which contains the calibrated priority values ​​of each maintenance path, the link number sequence of each maintenance path and the corresponding response time sequence are sorted by link number to ensure that the analysis is based on ordered path links. The differences in the calibrated priority mapping values ​​between adjacent links are identified (for example, the priority value of link 1 of path A is 4, and that of link 2 is 3, and the difference is 3-4=-1). The positive or negative difference is determined, and the nodes with increasing priority values ​​are marked, that is, the nodes with higher priority than the current link. The continuous positive difference value sequence, that is, the path segment with continuously increasing priority (for example, the priority value of link 3 is 4, and the difference is 3-4=-1) is extracted. =5, the difference with link 2 is 5-3=2, and there is no direct connection with link 1. If the difference from link 2 to link 3 is positive, a continuous positive difference sequence is formed). The link number range, total difference (for example, the priority of link 2 to link 3 increases from 3 to 5, with a total difference of 2) and the number of nodes included in the increasing trend sequence are recorded. Low-frequency change segments, that is, path segments with small fluctuations in priority values ​​or only a single increase (for example, sequences with a total difference less than or equal to 1 or with fewer than 2 nodes are eliminated), are eliminated to generate the number of increasing trend sequences. This number reflects the number of path segments where the power grid maintenance priority presents a continuous upward trend in space.

[0033] The link screening and aggregation submodule extracts high-frequency increasing path segments based on the number of increasing trend sequences, identifies the start and end numbers, analyzes the cumulative increase and increase rate, screens path segments with rates exceeding the benchmark value, and obtains a priority-driven multi-stage maintenance task list; Based on the number of increasing trend sequences, the number identifies the path segments whose maintenance priority shows a continuously increasing trend. A high-frequency increasing path segment (for example, a path segment containing 5 nodes and a total difference of 8) with a large number (for example, the number exceeds a preset threshold, for example, greater than or equal to 3) or a large total difference (for example, the total difference exceeds a preset threshold, for example, greater than or equal to 5) is extracted. The start and end numbers of the path segment are identified (for example, the path segment starts from node N110 and ends at node N115). The cumulative increase (total difference) and the increase rate (total difference divided by the number of nodes or path length) of the path segment are analyzed. For example, the path segment from N110 to N115 has a total of 5 nodes, a total difference of 8, an average node spacing of 1 km, a path length of 4 km, and an increase rate of 8. / 4=2 (priority increase of 2 per kilometer). Path segments with an increase rate exceeding a benchmark value (the benchmark value is set based on historical maintenance efficiency data. For example, if historical data shows that the maintenance efficiency of path segments with a priority increase rate exceeding 1.5 has the most significant improvement, then the benchmark value is set to 1.5) are screened (for example, an increase rate of 2>1.5). These path segments are areas with the most significant priority increase and the most urgent maintenance needs. A priority-driven, multi-stage maintenance task list is obtained. This list specifies detailed multi-stage maintenance tasks for path segments with high priority increase rates. For example, rapid inspections and temporary reinforcement are performed in the first stage, and equipment replacement and line optimization are performed in the second stage. This ensures that maintenance resources are prioritized in areas with the highest risks and the most significant maintenance priority increases.

[0034] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. Typhoon prediction and power grid maintenance system based on digital twin and AI collaboration, characterized by: The system comprises: The typhoon path evolution module is based on the dynamic environmental parameters transmitted by the meteorological sensor network, combined with the geographical layout information of the power grid, to screen the input feature dimensions, analyze the frequency of changes in the path vector in multiple time period samples, and generate a spatiotemporal distribution map of the typhoon's impact range; The power grid vulnerability mapping module calls the spatiotemporal distribution map of the typhoon's impact range, extracts the stress and load change characteristics of key nodes in the typhoon's impact range, combines AI sequence modeling to analyze abnormal fluctuations in power grid operating parameters, compares stress positions through the digital twin node mapping warehouse, and establishes a power grid vulnerability mapping set; The risk link aggregation module calls the power grid vulnerability mapping set, detects potential fault triggering links, records the connection relationship between risk nodes, forms a partition structure according to the node stress strength, and constructs a power grid risk link structure set; The maintenance strategy calibration module calls the power grid risk link structure set, locates the resource configuration chain in the power grid maintenance platform that is consistent with the illustrated link, extracts the current maintenance priority, compares it with the priority benchmark in the standard maintenance strategy template, adjusts the chain with inconsistent priorities, and generates a digital twin maintenance priority sequence table.

2. The typhoon prediction and power grid maintenance system based on digital twin and AI collaboration according to claim 1 is characterized in that: The spatiotemporal distribution map of the typhoon impact range includes path deviation trends, impact range boundaries, time series expansion characteristics, spatial coverage parameters, and node stress indicators. The power grid vulnerability mapping set includes key node numbers, load fluctuation labels, abnormal fluctuation sequences, node stress indexes, and vulnerability identification marks. The power grid risk link structure set includes fault trigger link numbers, node stress statistical blocks, risk node groups, link priority identifiers, and aggregated link labels. The digital twin maintenance priority sequence table includes resource configuration chain numbers, current priority labels, standard priority benchmarks, priority deviation values, and calibration classification marks.

3. The typhoon prediction and power grid maintenance system based on digital twin and AI collaboration according to claim 1 is characterized in that: The typhoon path evolution module includes: The dynamic environmental parameter extraction submodule identifies the changing trend of environmental variables in a continuous time period based on the dynamic environmental parameters transmitted by the meteorological sensor network, selects the recording intervals that meet the path deviation conditions, and generates a typhoon path evolution sequence; The path deviation trend screening submodule counts the path deviation amplitude in differentiated time periods based on the typhoon path evolution sequence, determines whether it exceeds the path deviation threshold, screens the corresponding path, and records the deviation time period and spatial variation to obtain the high-frequency deviation path change rate; The spatiotemporal distribution map construction submodule calls the high-frequency offset path change rate, identifies the path offset structure, determines the coverage range and offset order, formulates the time series structure and sets the coverage weight according to the offset time period, and establishes the spatiotemporal distribution map of the typhoon impact range.

4. The typhoon prediction and power grid maintenance system based on digital twin and AI collaboration according to claim 3 is characterized in that: The power grid vulnerability mapping module includes: The node force identification submodule calls the spatiotemporal distribution map of the typhoon impact range, extracts the node number and load state identifier, screens the numbers with load mutation and state offset, analyzes the load mutation amplitude and state change intensity, and obtains the node force intensity value; The abnormal fluctuation aggregation submodule associates the abnormal fluctuation sequence marked in the power grid operation parameters according to the node stress intensity value, screens the numbers in the number sequence that coincide with the fluctuation sequence, aggregates and classifies them, calculates the number fluctuation aggregation value, merges and classifies them according to the numerical interval, and obtains the fluctuation aggregation value; The vulnerability marker identification submodule compares the number and the fluctuation category label based on the fluctuation aggregation value, extracts the co-occurring node pairs in the mapping sequence, analyzes the overlapping feature intervals and load differences between the numbers, and establishes a power grid vulnerability mapping set.

5. The typhoon prediction and power grid maintenance system based on digital twin and AI collaboration according to claim 4 is characterized in that: The risk link aggregation module includes: The fluctuation tag analysis submodule calls the power grid vulnerability mapping set, extracts the fluctuation tag group within the link, detects the fluctuation tag cross structure by combining the link position value and the tag type, calculates the cross frequency and position offset value, divides the link fluctuation clusters according to the offset position, and obtains the link fluctuation cluster frequency data; The fault node mapping submodule calls the link fluctuation cluster frequency data, selects nodes with a stress rate higher than the fluctuation threshold, recodes them according to the node number and frequency, and generates a standard mapping node set; The risk link aggregation submodule maps the node set according to the standard, aggregates the associated link node structure, identifies the node load sequence number, position sequence and trigger signal value, identifies the link structure correlation index, sorts the blocks, counts the number of links and node ratios covered by the blocks, and constructs the power grid risk link structure set.

6. The typhoon prediction and power grid maintenance system based on digital twin and AI collaboration according to claim 5 is characterized in that: The maintenance strategy calibration module includes: The risk link identification submodule calls the power grid risk link structure set, compares the node sequence and logic of the illustrated link with the maintenance platform resource configuration chain, identifies the risk link whose structure consistency exceeds the threshold, and generates a power grid risk link set; The priority deviation judgment submodule extracts the link priority label based on the grid risk link set, extracts the standard priority value of the corresponding link in the standard maintenance policy template, determines the corresponding position of the two in the priority sequence and analyzes the difference to obtain the link priority deviation value; The label mapping adjustment submodule selects the link with the priority deviation exceeding the threshold value according to the link priority deviation value, extracts the priority label, node load strength, link coupling strength and the number of fluctuation conflicts, analyzes the maintenance label correction amplitude and reconstructs the priority label mapping, identifies the label mapping relationship, and generates a digital twin maintenance priority sequence table.

7. The typhoon prediction and power grid maintenance system based on digital twin and AI collaboration according to claim 1 is characterized in that: The system also includes a multi-stage maintenance tracking module: The multi-stage maintenance tracking module calls the digital twin maintenance priority sequence table, marks the priority corresponding values ​​of the trigger instructions in the multi-stage maintenance process, identifies the distribution trend of priority changes in the instruction transmission path, screens the chain with continuously increasing priority, and outputs a priority-driven multi-stage maintenance task list; The priority-driven multi-stage maintenance task list includes stage-by-stage maintenance priorities, priority change trajectories, link concentration sections, continuous increment identifiers, and instruction conduction path mapping.

8. The typhoon prediction and power grid maintenance system based on digital twin and AI collaboration according to claim 7 is characterized in that: The multi-stage maintenance tracking module includes: The priority mapping marking submodule calls the digital twin maintenance priority sequence table, identifies the link number, trigger time and response time in the path, extracts the original priority value and compares it with the mapping priority item, selects the corresponding value, and generates a maintenance priority mapping value set; The priority trend identification submodule, based on the maintenance priority mapping value set, sorts the link number sequence and the response time sequence by number, identifies the difference between adjacent values ​​and determines whether it is positive or negative, marks the growth node, extracts the continuous positive difference sequence, records the link number, the total difference and the number of nodes, eliminates the low-frequency change segment, and generates the number of increasing trend sequences; The link screening and aggregation submodule extracts high-frequency increasing path segments according to the number of increasing trend sequences, identifies the start and end numbers, analyzes the cumulative increase and the increase rate, screens the path segments whose rate exceeds the benchmark value, and obtains a priority-driven multi-stage maintenance task list.

Images