Power distribution network emergency repair path planning method, device and equipment and storage medium

Abstract

This application relates to a method, apparatus, equipment, and storage medium for emergency repair path planning in power distribution networks, and pertains to the field of emergency repair dispatching in power distribution networks. The method includes: constructing a directed graph of the road network based on road network data of the power distribution network area; obtaining a meteorological feature matrix of the power distribution network area within a preset future time period; performing spatial downsampling processing on the matrix; performing threshold segmentation based on the meteorological features of each downscaled spatial grid within each prediction frame in the matrix to obtain a corresponding meteorological obstacle topology mask; obtaining the meteorological obstacle topology mask of the prediction frame corresponding to the current node when the path search reaches the current node, with a preset location as the starting point and the location to be repaired as the ending point; adjusting the directed graph of the road network based on the meteorological obstacle topology mask; and continuing path search on the adjusted directed graph of the road network using a preset path search algorithm to generate a repair path that avoids areas of extreme rainstorms. This achieves proactive spatiotemporal avoidance of moving rainstorm envelopes and plans the safest and most optimal emergency repair path.

Description

Technical Field

[0001] This application relates to the field of emergency repair and dispatching technology for power distribution networks, and in particular to a method, apparatus, equipment, and storage medium for emergency repair path planning in power distribution networks. Background Technology

[0002] With the in-depth construction of new power systems, the 220kV and below distribution network, as the final link in power supply, exhibits characteristics such as numerous and widespread locations, high physical exposure, and relatively weak ability to defend against extreme disasters. During typhoons or sudden severe convective weather events causing extreme torrential rains, the widely distributed distribution network assets are highly susceptible to damage from secondary disasters such as localized heavy rainfall, flash floods, and urban flooding, leading to pole collapses, line breaks, or equipment immersion damage, seriously threatening the reliability of regional power supply. After a disaster, the distribution network dispatch center needs to efficiently dispatch emergency repair teams from the material command center to the sites of the faulty assets to carry out repairs. However, under extreme torrential rain conditions, urban road networks face large-scale flooding and geological disasters, hindering traffic. The spatial distribution and severity of these obstacles change rapidly with the movement of the rainstorm system, exhibiting highly spatiotemporal nonlinear dynamic evolution characteristics.

[0003] Traditional solutions typically employ conventional route planning methods based on static traffic networks or historical traffic flow data, or rely on manual judgment to dispatch emergency repair vehicles. These traditional solutions have the following drawbacks when dealing with extreme, catastrophic rainstorms:

[0004] First, existing weather forecasts suffer from displacement bias. When predicting future precipitation, short-term radar extrapolation models often have a locational deviation of several kilometers between the predicted heavy rainfall area and the actual rainfall area due to atmospheric instability. If conventional route planning directly uses these biased forecast data, it will misjudge the location of heavy rainfall, causing the core rain zone boundaries that should be avoided to be missed, and vehicles may be led to sections of road that have actually been flooded by heavy rain.

[0005] Secondly, weather forecasting and route planning are disconnected. Traditional solutions process static road network maps and dynamic weather forecasts as two separate systems. During route planning, it is impossible to know where the rainstorm will be located at a certain point in the future. When a vehicle actually arrives at a certain intersection, that intersection may happen to be covered by the core area of ​​the rainstorm, but the route planning stage is completely unaware of this and cannot avoid it in advance. Summary of the Invention

[0006] Based on this, it is necessary to provide a method, device, equipment, and storage medium for emergency repair path planning of power distribution networks to address the above-mentioned technical problems, so as to achieve proactive spatiotemporal avoidance of moving rainstorm envelopes and plan the safest and most optimal emergency repair path.

[0007] Firstly, this application provides a method for emergency repair path planning in a power distribution network, including:

[0008] Construct a directed road network graph based on road network data of the distribution network coverage area;

[0009] Obtain the meteorological feature matrix of the distribution network coverage area within a future preset time period; the meteorological feature matrix includes the meteorological features of each spatial grid within multiple prediction frames;

[0010] The meteorological feature matrix is ​​downsampled in terms of spatial dimension to obtain the downsampled feature matrix; the downsampled feature matrix includes meteorological features of each downscaled spatial grid within multiple prediction frames;

[0011] For each prediction frame, threshold segmentation is performed based on the meteorological features of each downscaled spatial grid in the prediction frame in the downsampled feature matrix and the preset extreme rainstorm critical discrimination threshold to obtain the meteorological obstruction topology mask corresponding to the prediction frame.

[0012] Starting from a preset location and ending at the location to be repaired, a preset path search algorithm is used to perform path search on the directed graph of the road network to generate a repair path that avoids areas affected by extreme rainstorms. When the path search reaches the current node, the corresponding meteorological obstruction topology mask is obtained based on the prediction frame corresponding to the current node. The directed graph of the road network is then adjusted based on the obtained meteorological obstruction topology mask. The preset path search algorithm is then used to continue the path search on the adjusted directed graph of the road network until the destination is reached.

[0013] In one embodiment, adjusting the directed graph of the road network based on the acquired meteorological obstruction topology mask includes:

[0014] Determine the target spatial grid region formed by the downscaled spatial grid with the first value in the meteorological obstruction topology mask;

[0015] Temporarily remove all edges covered by the target space grid region in the directed road network graph to obtain the adjusted directed road network graph.

[0016] In one embodiment, adjusting the directed graph of the road network based on the acquired meteorological obstruction topology mask includes:

[0017] For each edge connected to the current node in the directed graph of the road network, if the value of the downscaled spatial grid corresponding to the current edge in the meteorological obstruction topology mask is the second value, then the baseline passage cost of the current edge is calculated as the comprehensive passage cost.

[0018] If the value of the downscaled spatial grid corresponding to the current edge in the meteorological obstruction topology mask is the first value, then the sum of the baseline passage cost and the additional penalty cost of the current edge is calculated as the comprehensive passage cost; the additional penalty cost is determined by the meteorological features of the downscaled spatial grid corresponding to the current edge in the downsampling feature matrix;

[0019] Based on the comprehensive toll cost, the weight information of each edge connected to the current node in the directed graph of the road network is set to obtain the adjusted directed graph of the road network.

[0020] In one embodiment, the preset extreme rainstorm critical discrimination threshold includes a first threshold and a second threshold, wherein the first threshold is greater than the second threshold; the acquired meteorological obstruction topology mask includes a first mask and a second mask, wherein the first mask is generated based on the first threshold and the second mask is generated based on the second threshold; adjusting the directed graph of the road network according to the acquired meteorological obstruction topology mask includes:

[0021] Determine the target spatial grid region formed by the downscaled spatial grid with the first value in the first mask;

[0022] Temporarily remove all edges covered by the target space grid region in the directed road network graph to obtain the remaining directed road network graph;

[0023] For each edge connected to the current node in the remaining road network directed graph, if the value of the downscaled spatial grid corresponding to the current edge in the second mask is the second value, then the baseline passage cost of the current edge is calculated as the comprehensive passage cost.

[0024] If the value of the downscaled spatial grid corresponding to the current edge in the second mask is the first value, then the sum of the baseline passage cost and the additional penalty cost of the current edge is calculated as the comprehensive passage cost; the additional penalty cost is determined by the meteorological features of the downscaled spatial grid corresponding to the current edge in the downsampling feature matrix;

[0025] Based on the comprehensive toll cost, the weight information of each edge connected to the current node in the directed graph of the road network is set to obtain the adjusted directed graph of the road network.

[0026] In one embodiment, the meteorological feature matrix is ​​downsampled in terms of spatial dimension to obtain a downsampled feature matrix, including:

[0027] Using a spatial pooling window of a preset size, perform a sliding window scan on the two-dimensional spatial plane corresponding to each prediction frame in the meteorological feature matrix;

[0028] During the scanning process, the meteorological features of each spatial grid covered by the spatial pooling window are pooled into a pooling value. The pooling value is then used as the meteorological feature of the downscaled spatial grid corresponding to the spatial pooling window to obtain the downsampling feature matrix.

[0029] In one embodiment, obtaining the corresponding meteorological obstruction topology mask based on the prediction frame corresponding to the current node includes:

[0030] Estimate the prediction frame corresponding to the next node of the current node;

[0031] If the prediction frame corresponding to the next node is inconsistent with the prediction frame corresponding to the current node, the meteorological obstruction topology mask obtained based on the prediction frame corresponding to the next node and the meteorological obstruction topology mask obtained based on the prediction frame corresponding to the current node will be merged, and the merged mask will be used as the obtained meteorological obstruction topology mask.

[0032] In one embodiment, the method further includes:

[0033] The updated meteorological feature matrix is ​​continuously acquired at preset time intervals, and the meteorological obstacle topology mask corresponding to each prediction frame is regenerated based on the updated meteorological feature matrix.

[0034] If the updated weather obstruction topology mask indicates that the untraveled section of the repair path is in an extreme rainstorm area in the corresponding forecast frame, then the subsequent path will be replanned based on the current location.

[0035] Secondly, this application also provides a power distribution network emergency repair route planning device, comprising:

[0036] The road network data processing module is used to construct a directed road network graph based on the road network data of the distribution network coverage area;

[0037] The meteorological data processing module is used to obtain the meteorological feature matrix of the distribution network coverage area within a preset time period in the future. The meteorological feature matrix includes the meteorological features of each spatial grid in multiple prediction frames. The meteorological feature matrix is ​​downsampled in spatial dimension to obtain a downsampled feature matrix. The downsampled feature matrix includes the meteorological features of each downscaled spatial grid in multiple prediction frames. For each prediction frame, threshold segmentation is performed based on the meteorological features of each downscaled spatial grid in the prediction frame in the downsampled feature matrix and a preset extreme rainstorm critical discrimination threshold to obtain the meteorological obstruction topology mask corresponding to the prediction frame.

[0038] The path planning module is used to perform path search on the directed graph of the road network, starting from a preset location and ending at the location to be repaired, and generating repair paths that avoid areas affected by extreme rainstorms. Specifically, when the path search reaches the current node, a corresponding meteorological obstruction topology mask is obtained based on the prediction frame corresponding to the current node. The directed graph of the road network is then adjusted based on the obtained meteorological obstruction topology mask. The preset path search algorithm is then used to continue the path search on the adjusted directed graph of the road network until the destination is reached.

[0039] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in the first aspect above.

[0040] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect above.

[0041] The aforementioned distribution network emergency repair route planning method, device, equipment, and storage medium construct a directed road network graph based on road network data of the distribution network coverage area; obtain a meteorological feature matrix of the distribution network coverage area within a preset future time period; the meteorological feature matrix includes meteorological features of each spatial grid within multiple prediction frames; perform spatial dimension downsampling processing on the meteorological feature matrix to obtain a downsampled feature matrix; the downsampled feature matrix includes meteorological features of each downscaled spatial grid within multiple prediction frames; for each prediction frame, based on the meteorological features of each downscaled spatial grid within the prediction frame in the downsampled feature matrix and the predicted... A threshold segmentation method is used to determine the extreme rainstorm threshold, obtaining a meteorological obstacle topology mask corresponding to the predicted frame. Starting from a preset location and ending at the location to be repaired, a preset path search algorithm is used to perform path search on the directed road network graph, generating repair paths that avoid extreme rainstorm areas. When the path search reaches the current node, the corresponding meteorological obstacle topology mask is obtained based on the predicted frame of that node. The directed road network graph is adjusted based on the obtained meteorological obstacle topology mask. The preset path search algorithm continues to perform path search on the adjusted directed road network graph until the endpoint is reached. This method, by introducing spatial dimension downsampling, maps the high-resolution original grid to a larger-scale downscaled spatial grid, overcoming the interference of spatial displacement errors in the meteorological feature matrix on rainstorm area determination. Based on this, threshold segmentation is performed to obtain the meteorological obstacle topology mask, actively tolerating local displacement deviations within the scale of the downscaled grid. This avoids structural underreporting of rainstorm envelope boundaries caused by minor misalignments in traditional grid-by-grid precise alignment, ensuring the integrity and reliability of the rainstorm-restricted area information upon which downstream path planning relies. Since the meteorological feature matrix contains multiple prediction frames, each frame corresponds to the spatial distribution of rainstorms at different future times, the road network is adjusted by dynamically obtaining the mask of the corresponding prediction frame based on the current node during the path search process. This realizes the spatiotemporal linkage between the dynamic evolution of meteorology and the traffic status of the road network, so that the traffic status of the road network is updated synchronously with the movement of the rainstorm system. This allows the planned emergency repair path to match the spatial distribution of rainstorms at each time, thus realizing the spatiotemporal collaborative avoidance of the moving rainstorm envelope. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is an application environment diagram of the distribution network emergency repair path planning method in one embodiment;

[0044] Figure 2 This is a flowchart illustrating a distribution network emergency repair path planning method in one embodiment;

[0045] Figure 3 This is a flowchart illustrating the emergency repair path planning method for a distribution network in another embodiment;

[0046] Figure 4 This is a structural block diagram of a power distribution network emergency repair path planning device in one embodiment;

[0047] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0049] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0050] The emergency repair path planning method for distribution networks provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on another network server.

[0051] For example, the initial position of terminal 102 is a preset position, which can be an emergency supplies command center. Terminal 102 sends a path request to server 104, which includes the coordinate information of the preset position. In response to the path request, server 104 generates a repair path that avoids areas of extreme rainstorms based on the power distribution network emergency repair path planning method provided in this application embodiment, and feeds back the generated repair path to terminal 102 for display. Terminal 102 can navigate based on the repair path.

[0052] Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, and projection equipment. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted displays. Head-mounted displays can be virtual reality (VR) devices, augmented reality (AR) devices, and smart glasses. Server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0053] In one exemplary embodiment, such as Figure 2 As shown, a method for emergency repair path planning in a distribution network is provided, which is then applied to... Figure 1 Taking server 104 as an example, the explanation includes:

[0054] Step 202: Construct a directed road network graph based on the road network data of the distribution network coverage area.

[0055] This involves retrieving regional Geographic Information System (GIS) and digital transportation network layers to obtain road network data within the power distribution network coverage area. For example, the road network data includes basic geographic and traffic attribute information such as road intersection locations, road segment connectivity, road grade, number of lanes, design speed, physical length, and traffic direction. Road intersections are abstracted as nodes to represent the topological connection endpoints in the road network; one-way traffic segments between adjacent nodes are abstracted as directed edges to represent the actual traffic direction and connectivity of roads. The nodes and directed edges together constitute a directed graph of the road network, serving as a static graph topology model describing the spatial connectivity and traffic relationships of roads within the power distribution network coverage area. This directed graph of the road network can be represented as: V represents the set of nodes, and E represents the set of directed edges. In one implementation, for a two-way road, it is split into two directed edges with opposite directions and modeled separately; for a one-way road, a single directed edge with the same direction of travel is constructed.

[0056] For example, the latitude and longitude coordinates of the power distribution network assets affected by the rainstorm are extracted simultaneously from the production management system as the locations to be repaired, and these locations are used as the endpoints for path optimization. Power distribution network assets include, but are not limited to, key equipment such as substations, prefabricated substations, overhead low-voltage lines, and cable branch boxes. Simultaneously, the latitude and longitude coordinates of the emergency material command center are used as a preset location, which is then set as the starting point for path optimization. The starting and ending points correspond to two specific nodes in the directed graph of the road network. Optionally, if there is a positional offset between the starting or ending point and a node in the road network (i.e., it is not at the node's location), a virtual connection edge is used to connect it to the nearest node in the directed graph of the road network. This virtual connection edge is used to establish topological connectivity and does not participate in subsequent travel cost calculations, or is assigned a very small fixed travel cost.

[0057] Optionally, each directed edge is associated with a static attribute tuple, including road grade, physical length, design speed, and number of lanes. The road grade is used to distinguish the traffic priority and basic capacity of different roads, such as national highways, urban arterial roads, secondary arterial roads, and rural roads. In one implementation, different basic traffic speeds can be set according to the road grade, and the basic travel time of each edge can be calculated as the static basic travel cost. Furthermore, each directed edge also reserves dynamic attribute slots for receiving weight information adjusted by the weather mask in subsequent steps.

[0058] Step 204: Obtain the meteorological feature matrix of the distribution network coverage area within a future preset time period; the meteorological feature matrix includes the meteorological features of each spatial grid within multiple prediction frames.

[0059] Specifically, through the emergency disaster reduction digital interface, rolling forecast data for a preset future time period output by the short-term radar extrapolation network model is acquired in real time. This data is then aligned and divided according to the distribution network coverage area to form a meteorological feature matrix. The short-term radar extrapolation network model is a deep learning-based radar echo extrapolation model used to receive historical time-series radar reflectivity factor images and predict the spatial distribution and intensity evolution of future minute-level radar echoes. The meteorological feature matrix is ​​a dataset of meteorological parameters organized in a two-dimensional grid format, with its spatial dimension covering the area where the distribution network is located. It is used to characterize the meteorological parameter values ​​of each spatial grid within this area.

[0060] The meteorological feature matrix can be represented as: Let x represent the horizontal coordinate index of the meteorological feature matrix in the horizontal space, corresponding to the longitude direction in a geographic information system. The meteorological feature matrix is ​​divided into several spatial grids in the horizontal space according to a preset resolution, and x is used to identify the column position of a certain spatial grid in the matrix. Let y represent the vertical coordinate index of the meteorological feature matrix in the horizontal space, corresponding to the latitude direction in a geographic information system, and is used to identify the row position of a certain spatial grid in the matrix. Let t represent the prediction time index, used to identify a certain prediction frame in the meteorological feature matrix. A specific spatial grid in the matrix is ​​determined by (x, y).

[0061] The meteorological feature matrix comprises multiple forecast frames, each corresponding to the spatial distribution of meteorological features at different times within a preset future time period. For example, if the preset future time period is one hour, the meteorological feature matrix contains 12 consecutive rolling forecast frames, with each frame corresponding to a 5-minute forecast lead time. The spatial grid is a basic geographic unit obtained by dividing the power distribution network coverage area horizontally at a preset resolution, used to discretize the continuous meteorological field into a computable raster data structure. The meteorological feature of each spatial grid is the vertically integrated liquid water intensity value corresponding to that grid. The vertically integrated liquid water intensity value characterizes the vertical cumulative intensity of the total amount of liquid water in the atmospheric column above that grid, reflecting the core precipitation intensity of the severe convective weather system.

[0062] In one implementation, the short-term radar extrapolation network model can be implemented using deep learning methods such as convolutional long short-term memory networks, PredRNN, or generative adversarial networks. After training with historical radar echo sequences, it gains the ability to predict future echo evolution trends. The meteorological feature matrix maintains its original pixel resolution (unsampling) in the initial state to fully record the original spatial distribution characteristics of the reflectivity gradient of the severe convective weather system. The spatial distribution characteristics of the reflectivity gradient characterize the spatial rate of change of the intensity within the precipitation system and are a key source of information for subsequent morphological contour preservation and macroscopic envelope extraction. Rolling prediction refers to the synchronous sliding of the entire prediction sequence forward one frame after acquiring a new frame of prediction data at each moment, ensuring that the meteorological feature matrix always maintains its predictive coverage of a fixed-duration window in the future, thereby providing continuously updated meteorological information input for path planning.

[0063] Optionally, the acquired meteorological feature matrix can be spatially cropped as needed, retaining only the spatial subset corresponding to the power distribution network coverage area to reduce the data processing volume in subsequent steps. If the radar data and the geographic information system use different spatial reference systems, spatial coordinate registration needs to be performed after acquisition to ensure that the spatial correspondence between each spatial grid in the meteorological feature matrix and the nodes and directed edges in the directed graph of the road network is clear and consistent. Spatial coordinate registration can use bilinear interpolation or nearest neighbor interpolation methods to resample the radar data to a spatial grid consistent with the geographic coordinates of the road network. In one feasible implementation, the number of prediction frames and the prediction interval contained in the meteorological feature matrix can be configured according to actual emergency response needs. For example, the prediction interval can be shortened to 2 minutes to obtain a finer temporal resolution, or the prediction duration can be extended to 2 hours to obtain longer-term warning information. In another feasible implementation, the meteorological parameters represented by the meteorological feature matrix can be switched according to different disaster types. For example, a combined reflectivity factor can be used to identify hail disasters, or hourly rainfall can be used to identify cumulative flood disasters. Among them, the combined reflectivity factor is the spatial distribution data obtained by projecting the maximum reflectivity of multiple elevation angle scanning layers onto a two-dimensional plane, which is used to characterize the vertical development intensity of the convective system.

[0064] Step 206: Perform spatial dimension downsampling on the meteorological feature matrix to obtain the downsampled feature matrix; the downsampled feature matrix includes meteorological features of each downscaled spatial grid within multiple prediction frames.

[0065] Among them, the meteorological feature matrix is ​​downsampled in spatial dimension to smooth local spatial displacement noise and highlight the macroscopic geometric contour of extreme rainstorm systems, which can be used for subsequent generation of meteorological obstacle topology masks.

[0066] Optionally, a spatial pooling operator is used to perform downsampling smoothing on the two-dimensional spatial plane of each prediction frame in the meteorological feature matrix. The spatial pooling operator is a mathematical operator that performs feature aggregation within a spatial neighborhood, aggregating multiple meteorological feature values ​​within a local neighborhood into a representative value, thereby reducing the spatial resolution of the feature matrix. The preferred spatial pooling operator is a 16×16 average pooling operator, which uses a 16×16 pixel spatial window as the aggregation neighborhood, calculates the arithmetic mean of all meteorological feature values ​​within the window's coverage area, and uses this average value as the output value at the corresponding position of the window.

[0067] In one alternative implementation, the spatial pooling operator can be equivalently replaced by the max pooling operator, that is, taking the maximum value within the window coverage area as the output value. In another alternative implementation, the spatial pooling operator can be equivalently replaced by a Gaussian smoothing filter with spatial Gaussian weights, that is, assigning different weight coefficients according to the distance of the spatial grid from the center position within the window, performing a weighted average on the meteorological feature values ​​within the window coverage area, and using this weighted average as the output value at the corresponding position of the window.

[0068] The downscaled spatial grid is the basic spatial unit in the feature matrix obtained after downsampling. Each downscaled spatial grid corresponds to the spatial extent of all original spatial grids within the coverage area of ​​a pooling window in the meteorological feature matrix. The downsampling feature matrix is ​​a dataset composed of the downscaled spatial grids and their meteorological feature values ​​obtained after downsampling each prediction frame, used to characterize the macroscopic meteorological feature distribution of the rainstorm system at a large spatial scale.

[0069] By performing the aforementioned deep spatial pooling operation, the spatial physical displacement of pixels on the micro-grid is actively tolerated, sacrificing the absolute alignment accuracy of local pixels, thereby smoothing out local spatial drift noise caused by atmospheric chaotic nonlinear extrapolation. After downsampling, the systematic macroscopic dynamic geometric contour and trajectory envelope of the extreme torrential rain cell are preserved and highlighted in the downsampling space. The larger the pooling window, the lower the spatial resolution of the downsampling feature matrix, the greater the tolerance of local displacement deviation, and the smoother the boundary of the macroscopic envelope. The downsampling feature matrix contains the same number of prediction frames as the meteorological feature matrix, and the temporal index relationship between frames remains unchanged, that is, the downsampling feature matrix of each frame still corresponds to the meteorological feature distribution at the same prediction time in the original meteorological feature matrix. The downsampling feature matrix can be represented as: .

[0070] It is understood that this embodiment involves several different spatial data models: the downsampling feature matrix adopts a raster data model, that is, a continuous spatial region is divided into several downscaled spatial grids according to a regular grid, and the meteorological feature values ​​of each grid are stored in the form of a two-dimensional matrix, with each grid corresponding to a specific geographic spatial range; the road network directed graph adopts a vector data model, that is, roads are expressed in the form of directed edges, and each directed edge is composed of a series of geometric line segments defined by geographic coordinate points. In order to achieve joint calculation of meteorological raster data and road vector data under the same spatial reference system, the spatial mapping relationship between each downscaled spatial grid and each road directed edge is pre-established before performing the road network adjustment operation.

[0071] In one implementation, the spatial mapping relationship is established as follows: For each directed edge of a road, its geometric segment is spatially intersected with the geographic boundaries of each downscaled spatial grid. If the geometric segment covered by the directed edge intersects with the spatial extent of a downscaled spatial grid (i.e., at least a portion of the directed edge is located within the downscaled spatial grid), then the directed edge is mapped to that downscaled spatial grid. If a directed edge of a road spans multiple downscaled spatial grids, mapping relationships between the directed edge and multiple downscaled spatial grids are established simultaneously. In a simplified implementation, the overlap determination of the circumscribed rectangle is used instead of the fine intersection calculation; that is, if the circumscribed rectangle of the directed edge of the road overlaps with the circumscribed rectangle of the downscaled spatial grid, a mapping relationship is established to reduce spatial computation overhead.

[0072] In one implementation, the pooling window size can be adjusted according to the spatial scale of the target rainstorm system. A larger pooling window can be used for large-scale typhoon spiral rainbands to highlight the system-level outline, while a smaller pooling window can be used for localized, sudden convective cells to retain necessary boundary details. The spatial pooling operator can perform downsampling simultaneously along both longitude and latitude in a two-dimensional spatial plane, or it can perform non-uniform downsampling in a single direction to adapt to the spatial scale differences of different geographical regions. In another implementation, if there are missing data or outliers in the meteorological feature matrix, data quality control can be performed before downsampling to avoid contamination of the pooling results by outliers. In yet another implementation, downsampling can be optimized by combining spatial interpolation methods, such as using bilinear interpolation or cubic convolutional interpolation to smooth the pooling results, making the downsampled feature matrix more spatially continuous and reducing abrupt changes between grids caused by pooling window boundary effects.

[0073] Step 208: For each prediction frame, threshold segmentation is performed based on the meteorological features of each downscaled spatial grid in the prediction frame in the downsampling feature matrix and the preset extreme rainstorm critical discrimination threshold to obtain the meteorological obstruction topology mask corresponding to the prediction frame.

[0074] In this process, for each prediction frame in the downsampling feature matrix, the meteorological feature values ​​of each downscaled spatial grid within that frame are used as input, and a preset extreme rainstorm critical discrimination threshold is used as the segmentation standard to perform threshold segmentation processing on the entire downscaled spatial grid. The extreme rainstorm critical discrimination threshold is a pre-set meteorological feature threshold value used to determine whether a single downscaled spatial grid is within the absolute blocking envelope of the core disaster of extreme rainstorms. In this embodiment, the extreme rainstorm critical discrimination threshold is fixed at 219.0VIL, meaning that when the vertical integral liquid water intensity value reaches or exceeds this value, the corresponding grid is determined to be within the core disaster area of ​​extreme rainstorms. The meteorological obstruction topology mask is a binary discrete matrix with the same spatial dimension as the downsampling feature matrix, and the values ​​of its elements are used to characterize whether the corresponding downscaled spatial grid is within the core disaster area of ​​extreme rainstorms at the corresponding prediction time. The entire downscaled spatial grid is traversed and scanned, and the meteorological feature values ​​of each grid are compared with the extreme rainstorm critical discrimination threshold. If the meteorological characteristic value of a downscaled spatial grid is greater than or equal to the critical threshold for extreme rainstorm discrimination, the binary value of the corresponding position of that grid in the meteorological obstruction topological mask is set to the first value (e.g., 1), indicating that the spatial physical region is inside the absolute blocking envelope of the core disaster of extreme rainstorm and needs to be marked as blocked. If the meteorological characteristic value is less than the critical threshold for extreme rainstorm discrimination, the binary value of the corresponding position of that grid in the meteorological obstruction topological mask is set to the second value (e.g., 0), indicating that the spatial physical region is in a non-blocked state and its passage attribute is allowed to be retained.

[0075] For example, based on a downsampled, downscaled spatial grid (e.g., a 1km × 1km virtual spatial microgrid). A critical threshold for extreme rainfall is set, for example, 219.0 VIL, corresponding to the core energy level of an exceptionally heavy rainfall event that causes severe ground flooding. The entire microgrid is scanned, and a dynamic binary meteorological barrier topology mask is calculated and generated. The expression for threshold segmentation is as follows:

[0076] ;

[0077] In the formula, For weather-obstructed topological masks, These are meteorological characteristic values.

[0078] In one implementation, the critical threshold for extreme rainstorm discrimination can be differentiated based on the climate characteristics and topographic conditions of different regions. For example, a lower threshold can be used in areas with a history of frequent flooding to increase safety redundancy, while a higher threshold can be used in areas with good drainage conditions to avoid over-conservatism. In another implementation, the meteorological obstruction topology mask can undergo morphological post-processing after threshold segmentation, including but not limited to dilation, erosion, or opening / closing operations, to eliminate isolated noise points caused by downsampling or fill small voids within the mask area, making the boundary of the obstruction area more continuous and smooth. In yet another implementation, for downscaled spatial grids located at the mask edge, if their meteorological feature values ​​fluctuate near the critical threshold, a multi-frame temporal smoothing strategy can be adopted. This involves combining the mask state of the grid in adjacent prediction frames for comprehensive judgment to suppress frequent changes in mask state caused by intensity jitter between prediction frames. It is understood that the meteorological obstruction topology mask records the obstruction state of each downscaled spatial grid at each prediction time. Subsequent steps can determine whether any downscaled spatial grid is in an obstruction state by querying the mask and perform corresponding road network adjustment operations accordingly.

[0079] Step 210: Starting from a preset location and ending at the location to be repaired, a preset path search algorithm is used to perform a path search on the directed graph of the road network to generate a repair path that avoids areas affected by extreme rainstorms. When the path search reaches the current node, the corresponding meteorological obstruction topology mask is obtained based on the prediction frame corresponding to the current node. The directed graph of the road network is adjusted based on the obtained meteorological obstruction topology mask. The preset path search algorithm is then used to continue the path search on the adjusted directed graph of the road network until the destination is reached.

[0080] The process involves using a preset location as the starting point and the location to be repaired as the ending point, and then performing a path search on the directed graph of the road network using a preset path search algorithm. The preset location is the starting node corresponding to the emergency material command center in the directed graph of the road network, and the location to be repaired is the ending node corresponding to the distribution network asset that suffered a rainstorm fault in the directed graph of the road network.

[0081] During the path search process, when the search extends to the current node, the corresponding prediction frame is determined based on the estimated arrival time of the current node, and the corresponding meteorological obstruction topology mask is obtained using this prediction frame as an index. The current node is the node being processed by the path search algorithm during its iteration, and the estimated arrival time is the cumulative travel time required to reach this node from the starting point along the current search path. Optionally, the meteorological obstruction topology mask contains the obstruction status information of each downscaled spatial grid under the prediction frame, used to indicate whether each spatial grid is within the core disaster area of ​​extreme rainstorm at the corresponding prediction time. The directed graph of the road network is adjusted based on the obtained meteorological obstruction topology mask. This adjustment includes at least one of the following methods: temporarily logically removing road segments covered by downscaled spatial grids with a first value from the directed graph, while retaining safe-to-pass road segments in the adjusted graph; or, maintaining the original topology of the directed graph and applying a penalty term to the passage cost of road segments covered by downscaled spatial grids with a first value to adjust the probability of these segments being selected in the path search; or, during the path search process, determining the passage status of the evaluated road segments based on the meteorological obstruction topology mask indexed by the predicted frame corresponding to the arrival time of the current node, and dynamically adjusting the passage cost weight of the road segments based on the determination result. The path search continues on the adjusted directed graph, expanding to the next node, and repeating the process of adjusting the road network based on the corresponding frame mask indexed by the current node until the destination node is reached.

[0082] Optionally, the adjusted directed road network graph is obtained by temporarily logically cutting off road segments covered by downscaled spatial grids with values ​​of the first value based on the original directed road network graph and the meteorological obstruction topology mask corresponding to each prediction frame, retaining road segments that are safe to pass at the current prediction time. During the search process, when expanding from the current node to an adjacent node, it is necessary to determine whether the road segment between them has been cut off from the directed road network graph. If it has been cut off, the edge is rejected from expansion, and other adjacent nodes are evaluated instead; if it has not been cut off, the passage cost of the edge is calculated and the adjacent node is added to the search queue.

[0083] Optionally, the topology of the directed graph of the road network remains unchanged. For road segments covered by the downscaled spatial grid with the first value, an additional penalty cost is added to their baseline travel cost, significantly reducing the probability of these road segments being selected in path search. Optionally, the order of magnitude of the additional penalty cost is much larger than the baseline travel cost, ensuring that the path search algorithm will not prioritize these road segments when any detour paths exist.

[0084] Optionally, when expanding the current node, the path search algorithm determines the value state of the downscaled spatial grid corresponding to the road segment to be evaluated in the meteorological obstruction topology mask of the prediction frame corresponding to the estimated arrival time of the current node, and determines the toll cost weight of the road segment in real time based on the value state to participate in the path search calculation. Finally, all road segments included in the path from the starting point to the ending point together constitute the emergency repair path.

[0085] By using any of the above adjustment methods, the road network of the emergency repair route is adjusted at each node based on the prediction frame corresponding to that node, ensuring that the entire route always keeps in line with the dynamic evolution of the rainstorm system in the time dimension, and achieving spatiotemporal coordinated avoidance of the rainstorm envelope at each moment.

[0086] In one implementation, the preset path search algorithm is the A* algorithm, which uses the basic travel cost of each edge as the baseline weight, combined with a heuristic estimate from the current node to the destination to guide the search direction. In an alternative implementation, the preset path search algorithm can also be replaced by the Dijkstra algorithm, the Floyd-Warshall algorithm, or the dynamic adaptive ant colony algorithm. The basic travel cost of each edge can be calculated based on the road grade, physical length, and design speed, and together with the adjustment applied by the meteorological obstruction topology mask, constitutes the weight basis for the path search. After the search is completed, a trajectory sequence consisting of a series of node latitude and longitude coordinates is output, which is the emergency repair path avoiding extreme rainstorm areas, for downstream terminal equipment to visualize and guide navigation.

[0087] Traditional power distribution network emergency repair vehicle dispatching schemes often attempt to strongly couple complex internal electrical topology states, distributed power flow equations, switch reconfiguration sequences, and other highly dynamic electrical parameters with complex external traffic and logistics paths when establishing mathematical optimization models. However, during extreme disasters, communication links are often damaged and distributed sensor data streams are lost. These highly coupled physical mechanism models often lack complete and accurate real-time electrical parameter support, leading to equation failures. This not only results in highly cumbersome algorithm calculation matrices, significant computational waste, and extremely slow convergence, but also leads to poor timeliness of emergency route planning schemes, making them unsuitable for the high-frequency rolling solution requirements of power distribution network edges or mobile emergency front-ends. To address this issue, this embodiment reduces the computational complexity of the cross-boundary optimization model by converting meteorological information into a meteorological obstacle topology mask and using the mask as a criterion to directly adjust the directed graph of the road network. By reconstructing path planning into a spatiotemporal extended network optimization control that strongly maps weather conditions and spatial road network entities, the dependence on electrical parameters is eliminated, the demand for underlying sensor data is reduced, and the calculation convergence speed is fast. It is suitable for high-frequency rolling deployment in distribution network master stations, material command terminals, and mobile emergency front-ends.

[0088] In the above-mentioned emergency repair path planning method for power distribution networks, a directed graph of the road network is constructed based on the road network data of the power distribution network coverage area; a meteorological feature matrix of the power distribution network coverage area is obtained within a preset time period; the meteorological feature matrix includes meteorological features of each spatial grid within multiple prediction frames; the meteorological feature matrix is ​​downsampled in spatial dimension to obtain a downsampled feature matrix; the downsampled feature matrix includes meteorological features of each downscaled spatial grid within multiple prediction frames; for each prediction frame, threshold segmentation is performed based on the meteorological features of each downscaled spatial grid within the prediction frame in the downsampled feature matrix and a preset extreme rainstorm critical discrimination threshold to obtain a meteorological obstacle topology mask corresponding to the prediction frame; starting from a preset location and ending at the location to be repaired, a preset path search algorithm is used to perform path search on the directed graph of the road network to generate a repair path that avoids the extreme rainstorm area; when the path search reaches the current node, the corresponding meteorological obstacle topology mask is obtained based on the prediction frame corresponding to the current node; the directed graph of the road network is adjusted based on the obtained meteorological obstacle topology mask; the preset path search algorithm is used to continue the path search on the adjusted directed graph of the road network until the destination is reached. By introducing spatial downsampling, the high-resolution original mesh is mapped to a larger-scale downscaled spatial mesh, overcoming the interference of spatial displacement errors in the meteorological feature matrix on the determination of rainstorm areas. Based on this, threshold segmentation is performed to obtain a meteorological obstacle topology mask, which actively tolerates local displacement deviations within the scale of the downscaled mesh. This avoids structural omissions of rainstorm envelope boundaries caused by minor misalignments in traditional grid-by-grid precise alignment, ensuring the integrity and reliability of the rainstorm-restricted area information upon which downstream path planning relies. Since the meteorological feature matrix contains multiple prediction frames, each corresponding to the spatial distribution of rainstorms at different future times, the road network is adjusted by dynamically acquiring the mask of the corresponding prediction frame based on the current node during path search. This achieves spatiotemporal linkage between dynamic meteorological evolution and road network traffic status, ensuring that the road network traffic status updates synchronously with the movement of the rainstorm system. This allows the planned repair routes to match the spatial distribution of rainstorms at each time point, achieving spatiotemporal collaborative avoidance of moving rainstorm envelopes.

[0089] In an exemplary embodiment, adjusting the directed graph of the road network based on the obtained meteorological obstruction topology mask includes: determining a target spatial grid region in the meteorological obstruction topology mask with a first value; temporarily removing all edges covered by the target spatial grid region in the directed graph of the road network to obtain the adjusted directed graph of the road network.

[0090] The meteorological obstruction topology mask is a binary discrete matrix with the same spatial dimension as the downsampling feature matrix. Each element's value characterizes whether the corresponding downscaled spatial grid is within the core disaster area of ​​the extreme rainstorm at the corresponding prediction time. Specifically, downscaled spatial grids with a first value (e.g., 1) in the meteorological obstruction topology mask indicate that the grid is inside the absolute blocking envelope of the core disaster area of ​​the extreme rainstorm at the corresponding prediction time, while downscaled spatial grids with a second value (e.g., 0) indicate that the grid is in a non-blocking state. The target spatial grid region is the set of spatial extents formed by all downscaled spatial grids with a first value (e.g., 1) in the meteorological obstruction topology mask, representing the spatial physical region inside the absolute blocking envelope of the core disaster area of ​​the extreme rainstorm at the corresponding prediction time. After determining the target spatial grid region, all directed edges of roads covered by each downscaled spatial grid within this region are queried from the geographic information system, and all queried directed edges are temporarily removed from the directed road network graph. Temporary removal refers to marking the removed directed edges as impassable during the current path search process, preventing them from participating in subsequent path search expansion calculations. After the current path search ends, the removed directed edges automatically regain their original connectivity, without affecting subsequent path planning tasks. The resulting directed graph of the road network after this processing is the adjusted directed graph, retaining the road segments that are safely passable at the current prediction time for the path search algorithm to continue expanding.

[0091] In one implementation, a spatial index mapping table can be pre-established between each downscaled spatial grid and the directed edges of the roads it covers. The spatial index mapping table is an associated data structure recording the identifiers of all directed edges covered by each downscaled spatial grid. This allows for the rapid retrieval of the set of directed edges to be removed by looking up the table when a downscaled spatial grid has a value of the first value, eliminating the need for real-time spatial lookup calculations and improving the response efficiency of road network adjustments. In one implementation, if a directed edge of a road extends across multiple downscaled spatial grids, and any downscaled spatial grid covered by that edge has a value of the first value, the entire edge is temporarily removed to prevent vehicles from being unable to turn around after partially entering a dangerous area. In another implementation, if a directed edge of a road extends across multiple downscaled spatial grids, the proportion of grids covered by that edge that have a value of the first value can be calculated. When this proportion exceeds a preset threshold, the entire edge is removed to balance safety and road network connectivity. In another implementation, the downscaled spatial grid with the first value in the meteorological obstruction topology mask can be further subjected to connected component labeling, merging multiple spatially adjacent active grids into a single connected obstruction region. This reduces the fragmented road network adjustments caused by scattered active grids, ensuring that the removed road segments form complete obstruction boundaries in space. In yet another implementation, if the meteorological obstruction topology mask is updated, the aforementioned temporary removal operation is re-executed based on the updated target spatial grid region. This allows the adjustment results of the directed road network graph to be dynamically refreshed as meteorological conditions evolve, enabling continuous tracking of the moving rainstorm envelope.

[0092] In this embodiment, by directly mapping the spatial obstruction information of the meteorological obstruction topology mask to the on / off status of road edges in the directed graph of the road network, the possibility of emergency repair vehicles entering the core area of ​​high-risk disasters is blocked at the source of the data flow. Since the temporarily removed edges do not participate in the subsequent extended calculation of path search, the number of candidate edges in the downstream search algorithm is reduced, the search space is significantly reduced, and the search efficiency is significantly improved. Because the adjustment is based on the meteorological obstruction topology mask corresponding to each prediction frame and is performed independently at the prediction time, the road network adjustment results can be dynamically switched with the change of the prediction frame index, realizing a precise temporal match between meteorological spatial obstruction constraints and the path search process.

[0093] In an exemplary embodiment, adjusting the directed graph of the road network based on the obtained meteorological obstruction topology mask includes: for each edge connected to the current node in the directed graph of the road network, if the value of the downscaled spatial grid corresponding to the current edge in the meteorological obstruction topology mask is a second value, then the baseline travel cost of the current edge is calculated as the comprehensive travel cost; if the value of the downscaled spatial grid corresponding to the current edge in the meteorological obstruction topology mask is a first value, then the sum of the baseline travel cost and the additional penalty cost of the current edge is calculated as the comprehensive travel cost; the additional penalty cost is determined by the meteorological features of the downscaled spatial grid corresponding to the current edge in the downsampling feature matrix; based on the comprehensive travel cost, the weight information of each edge connected to the current node in the directed graph of the road network is set to obtain the adjusted directed graph of the road network.

[0094] The road network directed graph is a graph topology containing nodes and directed edges. The current node is the node being expanded during the path search process, and each edge is a directed edge originating from the current node and connecting to adjacent nodes. For each edge connected to the current node, the corresponding downscaled spatial grid in the meteorological obstruction topology mask is determined, and the value state of the downscaled spatial grid in the meteorological obstruction topology mask is obtained. The meteorological obstruction topology mask is a binary discrete matrix with the same spatial dimension as the downsampling feature matrix, and the values ​​of its elements are used to characterize whether the corresponding downscaled spatial grid is located within the core disaster area of ​​extreme rainstorm at the corresponding prediction time. If the downscaled spatial grid corresponding to the current edge takes the second value in the meteorological obstruction topology mask, the baseline travel cost of that edge is calculated as the comprehensive travel cost. The baseline travel cost is an inherent travel cost predetermined based on static traffic parameters such as road grade, physical length, and design speed entered in the geographic information system, used to characterize the basic travel time cost of that edge under conditions without meteorological interference. If the downscaled spatial grid corresponding to the current edge takes the first value in the meteorological obstruction topology mask, then the sum of the baseline passage cost and the additional penalty cost of that edge is calculated as the comprehensive passage cost. The additional penalty cost is the additional passage cost calculated based on the meteorological feature value of the downscaled spatial grid corresponding to the edge in the downsampling feature matrix, used to characterize the nonlinear obstruction effect of heavy precipitation on passage safety and speed. Based on the comprehensive passage cost calculated for each edge, the weight information of each edge connected to the current node in the directed graph of the road network is set, resulting in an adjusted directed graph of the road network. The weight of each edge in the adjusted directed graph of the road network is the comprehensive passage cost of the corresponding edge at the current prediction time, and the path search algorithm uses this weight for subsequent expansion and optimization.

[0095] In one implementation, the additional penalty cost is calculated as follows: based on the pooled meteorological feature value at the downscaled spatial grid, a penalty increment is calculated in combination with a preset penalty adjustment factor. Specifically, the penalty increment can be expressed in logarithmic function form to avoid drastic jumps in the penalty amount due to small fluctuations in the meteorological feature value near the critical threshold. In another implementation, the additional penalty cost is further cascaded with the value of the meteorological obstruction topology mask: when the mask value is the second value, the additional penalty cost is set to zero; when the mask value is the first value, the additional penalty cost is calculated from the meteorological feature value, thereby ensuring that the additional penalty cost is only activated when the corresponding grid is in an obstructed state, achieving precise linkage between the penalty mechanism and the mask determination.

[0096] In one implementation, if the current edge spans multiple downscaled spatial grids, the maximum value or weighted average of the meteorological characteristic values ​​of each grid is taken as the meteorological characteristic value of the edge, which is used to calculate the additional penalty cost to reflect the travel risk of the location with the most unfavorable meteorological conditions in the area traversed by the edge.

[0097] For example, for any downscaled spatial grid g corresponding to a directed graph of a road network, under a specific prediction t, its comprehensive toll cost is... The following formula is used to solve dynamically:

[0098] ;

[0099] In the formula, Let be the dynamic meteorological comprehensive weighted cost function of the road network topology, representing the dynamic comprehensive time cost and safety risk cost of emergency repair vehicles passing through the roadside of the geospatial microgrid g at a specific prediction time node t. This represents the static basic toll cost, which can be statically assigned based on the road grade (such as national highway, urban arterial road, rural road) and the physical length of the road segment entered into the geographic information system. The Dynamic Meteorological Obstacle Penalty Regulating Factor is a positive characteristic constant used to increase the cost of passing through the boundary grid at the algorithm level. In this embodiment, it is statically set to a fixed, extremely large constant (e.g., 10000). As a binary dynamic weather obstruction topology mask, the binary discrete matrix value generated by the aforementioned threshold segmentation step is cascaded into the penalty term as an activation control switch term. At that time, the penalty item is activated; At that time, the penalty item is reset to zero, and the overall toll cost degenerates into a static basic toll cost. It is a nonlinear reflectivity response scaling factor, belonging to the scale characteristic scaling factor, used to smooth the mathematical dimensions of the measured floating-point values ​​of vertically integrated liquid water intensity under different rainstorm energy levels, so that the pooled meteorological characteristic values ​​are within an appropriate order of magnitude range when participating in the penalty term calculation. The multi-scale spatial pooling short-term extrapolation intensity value refers to the macroscopic dynamic vertical integration liquid water intensity value mapped to grid g and prediction time t after smoothing by the spatial pooling operator.

[0100] In this embodiment, by introducing an additional penalty cost based on meteorological obstacle topology masking into the overall traffic cost, the weight of road segments in the core area of ​​rainstorms is significantly increased in the path search. Under the premise that there is an alternative path, the algorithm will autonomously avoid these high-risk road segments, thus achieving a balance between safety and timeliness.

[0101] In an exemplary embodiment, the preset extreme rainstorm critical discrimination threshold includes a first threshold and a second threshold, wherein the first threshold is greater than the second threshold; the acquired meteorological obstruction topology mask includes a first mask and a second mask, wherein the first mask is generated based on the first threshold and the second mask is generated based on the second threshold; the directed graph of the road network is adjusted according to the acquired meteorological obstruction topology mask, including: determining the target spatial grid region composed of downscaled spatial grids with values ​​of a first value in the first mask; temporarily removing all edges covered by the target spatial grid region in the directed graph of the road network to obtain the remaining directed graph of the road network; and adjusting the directed graph of the road network based on the remaining directed graph of the road network. For each edge connected to the current node in the graph, if the value of the downscaled spatial grid corresponding to the current edge in the second mask is the second value, then the baseline travel cost of the current edge is calculated as the comprehensive travel cost; if the value of the downscaled spatial grid corresponding to the current edge in the second mask is the first value, then the sum of the baseline travel cost and the additional penalty cost of the current edge is calculated as the comprehensive travel cost; the additional penalty cost is determined by the meteorological characteristics of the downscaled spatial grid corresponding to the current edge in the downsampling feature matrix; based on the comprehensive travel cost, the weight information of each edge connected to the current node in the directed graph of the road network is set to obtain the adjusted directed graph of the road network.

[0102] The preset extreme rainstorm critical discrimination thresholds include a first threshold and a second threshold, with the first threshold being higher than the second threshold. The first threshold corresponds to the core energy level of extreme torrential rainstorms and is used to delineate the disaster core area where passage is absolutely prohibited; in this embodiment, it is preferably set to 219.0 VIL. The second threshold corresponds to the edge energy level of heavy precipitation and is used to delineate the transition area affected by rainstorms but not yet reaching the level of absolute blockage; in this embodiment, it is preferably set to 133.0 VIL. Based on the above dual thresholds, the meteorological obstruction topology mask obtained in this step includes a first mask and a second mask. The first mask is a binary discrete matrix obtained by performing threshold segmentation on the downsampled feature matrix based on the first threshold, used to identify downscaled spatial grids within the absolute blockage envelope. The second mask is a binary discrete matrix obtained by performing threshold segmentation on the downsampled feature matrix based on the second threshold, used to identify downscaled spatial grids within the rainstorm's influence range. There is a hierarchical relationship between the first mask and the second mask: the set of meshes activated by the first mask is a subset of the set of meshes activated by the second mask, that is, the core area of ​​the absolute disaster is included in the range of rainstorm impact. The first mask is used for hard cut-off, and the second mask is used for soft punishment.

[0103] For example, based on a downsampled, downscaled spatial grid (e.g., a 1km × 1km virtual spatial microgrid). Set a first threshold and a second threshold, for example, 219.0VIL and 133.0VIL. Iterate through all microgrid points, calculate and generate a dynamic binary weather obstruction topology mask, i.e., the weather obstruction topology mask. The expression for threshold segmentation is as follows:

[0104] ;

[0105] ;

[0106] In the formula, As the first mask, For the second mask, These are meteorological characteristic values.

[0107] For example, adjusting the directed graph of the road network based on the obtained meteorological obstruction topology mask specifically includes the following steps:

[0108] The target spatial grid region is determined by downscaling spatial grids with values ​​of the first value (e.g., 1) in the first mask. This target spatial grid region represents the spatial physical region located within the absolute blocking envelope of the core disaster of the extreme rainstorm at the corresponding prediction time. All directed edges of roads covered by each downscaling spatial grid within this region are queried from the geographic information system. All queried directed edges are temporarily removed from the road network directed graph, resulting in the remaining road network directed graph. Temporary removal means that during the current path search process, the removed directed edges are marked as impassable, preventing them from participating in subsequent path search calculations. After this step, road segments within the core disaster area are completely removed from the road network, blocking the possibility of repair vehicles entering the extremely high-risk area from the data flow source. It should be noted that masks corresponding to different extreme rainstorm critical discrimination thresholds can all be called binary masks. Both the first and second masks are binary discrete matrices with values ​​of the first or second value; the only difference between them is the source and purpose of the thresholds.

[0109] For each edge connected to the current node in the remaining road network directed graph, the downscaled spatial grid corresponding to each edge in the second mask is determined, and the value state of the downscaled spatial grid in the second mask is obtained. If the downscaled spatial grid corresponding to the current edge in the second mask has a second value (e.g., 0), the baseline passage cost of that edge is calculated as the comprehensive passage cost. The baseline passage cost is an inherent passage cost predetermined based on static traffic parameters such as road grade, physical length, and design speed entered in the geographic information system, used to characterize the basic passage time cost of that edge under conditions without meteorological interference. If the downscaled spatial grid corresponding to the current edge in the second mask has a first value, the sum of the baseline passage cost and the additional penalty cost of that edge is calculated as the comprehensive passage cost. The additional penalty cost is determined by the meteorological feature value of the downscaled spatial grid corresponding to that edge in the downsampling feature matrix. Specifically, based on the pooled meteorological feature value at the downscaled spatial grid, an additional passage cost is calculated through a preset penalty function, used to characterize the nonlinear hindering effect of heavy precipitation on traffic safety and speed. Based on the comprehensive travel cost calculated for each edge, the weight information of each edge connected to the current node in the remaining directed road network graph is set, resulting in the adjusted directed road network graph. The weight of each edge in the adjusted directed road network graph is the comprehensive travel cost of the corresponding edge at the current prediction time. The path search algorithm uses this weight for subsequent expansion and optimization.

[0110] In this embodiment, a first mask is generated by setting a first threshold to cut off road segments within the core energy level of extreme torrential rain, preserving passable road segments for path search and avoiding over-pruning due to a single low threshold. This hard-cut mechanism ensures zero-tolerance blocking of absolutely high-risk areas, guaranteeing the driving safety of repair vehicles from the data source. A second mask is generated by setting a second threshold, imposing continuous additional penalty costs on road segments in the edge of the rainstorm's impact area but not yet at the level of absolute blockage. This allows the path search algorithm to autonomously avoid these high-risk road segments when alternative routes exist, achieving a balance between safety and timeliness. The hierarchical setting of dual thresholds creates a continuous safety transition zone between the disaster core area and the disaster edge area, ensuring both an absolute safety boundary and preserving the possibility of passability for edge road segments in extreme situations. If all alternative routes are covered by the rainstorm, the penalized road segments remain in the road network and are not physically removed, allowing the search algorithm to potentially traverse these segments at a higher cost, providing a safety guarantee for emergency repairs.

[0111] In an exemplary embodiment, step 206 includes: performing a sliding window scan on the two-dimensional spatial plane corresponding to each prediction frame in the meteorological feature matrix using a spatial pooling window of a preset size; during the scanning process, pooling the meteorological features of each spatial grid covered by the spatial pooling window into a pooling value, and using the pooling value as the meteorological feature of the downscaled spatial grid corresponding to the spatial pooling window to obtain a downsampled feature matrix.

[0112] In this process, a sliding window scan is performed on the two-dimensional spatial plane corresponding to each prediction frame in the meteorological feature matrix using a spatial pooling window of a preset size. The meteorological feature matrix is ​​a multi-frame meteorological parameter dataset organized in the form of a two-dimensional grid, with each prediction frame corresponding to the spatial distribution of meteorological features at different times within a preset future time period. The two-dimensional spatial plane is the two-dimensional spatial distribution corresponding to a single prediction frame in the meteorological feature matrix, formed by arranging all spatial grids covering the power distribution network area according to their horizontal spatial positions. The spatial pooling window is a square spatial neighborhood of a preset size, used to limit the local spatial range covered by each pooling operation. In this embodiment, the preset size of the spatial pooling window is 16×16, that is, a square area of ​​16 spatial grids multiplied by 16 spatial grids is used as a pooling unit. The sliding window scan refers to sliding the spatial pooling window on the two-dimensional spatial plane at a preset step size, sequentially covering each local area of ​​the two-dimensional spatial plane. During the scanning process, the meteorological feature values ​​of each spatial grid currently covered by the spatial pooling window are pooled into a single pooled value. Pooling is a mathematical operation that performs feature aggregation within a spatial neighborhood, used to aggregate multiple input values ​​into a representative value. In this embodiment, pooling is preferably average pooling, which involves taking the arithmetic mean of the meteorological feature values ​​of all spatial grids covered by the window as the pooling value. In an alternative embodiment, pooling can also be equivalently replaced by max pooling, which involves taking the maximum value within the window coverage area as the pooling value. The pooling value is used as the meteorological feature value of the downscaled spatial grid corresponding to the spatial pooling window. The downscaled spatial grid is the basic spatial unit in the feature matrix obtained after downsampling processing, and each downscaled spatial grid corresponds to the spatial range of all original spatial grids within the coverage area of ​​a pooling window in the meteorological feature matrix. After the above processing, the downsampling feature matrix is ​​composed of the downscaled spatial grids and their meteorological feature values ​​corresponding to each prediction frame, which is used for subsequent threshold segmentation to generate the meteorological obstruction topology mask corresponding to each prediction frame.

[0113] In one implementation, the preset size of the spatial pooling window can be adjusted according to the spatial scale of the target rainstorm system: for large-scale typhoon spiral rainbands, a larger pooling window is used to highlight the macroscopic outline of the system; for localized sudden convective cells, a smaller pooling window is used to retain necessary boundary details. The preset step size of the sliding window scan can be configured according to the downsampling compression ratio: when the step size is equal to the pooling window size, the output grids have no overlapping coverage; when the step size is less than the pooling window size, there is overlap between the output grids, which can reduce the influence of window boundary effects. In another implementation, multi-scale pooling can be performed in parallel using pooling windows of multiple sizes for the same meteorological feature matrix, and the pooling results of each scale can be merged into a multi-scale downsampling feature matrix to simultaneously retain meteorological structure information at different scales, allowing downstream systems to adaptively select the optimal scale based on the scale of the rainstorm system. In yet another implementation, the pooling operation can use spatially weighted average pooling, that is, different weight coefficients are assigned according to the position of the spatial grid within the window, so that the grid in the central region of the window has a greater influence on the pooling result than the edge region, in order to better retain the meteorological feature response at the center of the window.

[0114] In this embodiment, the spatial resolution of the downsampling feature matrix is ​​significantly reduced, and the amount of data is greatly reduced, thereby effectively reducing the computational complexity of subsequent threshold segmentation and road network adjustment. Since pooling aggregates multiple original grids within a local neighborhood into a single downscaled grid, the size of a single downscaled grid determines the tolerance range of local displacement deviation. If the displacement deviation of the rainstorm envelope does not exceed the spatial scale of the downscaled grid, this deviation is absorbed within the downscaled grid during threshold segmentation, preventing structural underreporting or misalignment of the rainstorm envelope boundary. The downsampling process itself is a spatial smoothing of the original meteorological features; local micro-fluctuations are aggregated into macro-trends, allowing the systematic dynamic geometric contour and trajectory envelope of the extreme rainstorm system to be preserved in the downsampling space, providing stable and reliable meteorological feature data for the generation of downstream meteorological obstacle topology masks.

[0115] In an exemplary embodiment, obtaining the corresponding meteorological obstruction topology mask based on the prediction frame corresponding to the current node includes: estimating the prediction frame corresponding to the next node of the current node; if the prediction frame corresponding to the next node is inconsistent with the prediction frame corresponding to the current node, the meteorological obstruction topology mask obtained based on the prediction frame corresponding to the next node and the meteorological obstruction topology mask obtained based on the prediction frame corresponding to the current node are merged, and the merged mask is used as the obtained meteorological obstruction topology mask.

[0116] In this process, when the path search reaches the current node and is preparing to expand to the next node, the prediction frame corresponding to the arrival time of the current node is determined, and the prediction frame corresponding to the arrival time of the next node is estimated. The current node is the node being processed by the path search algorithm during its iteration, and the next node refers to the node following the current node in the path sequence, which is the node to be expanded. The arrival time of the current node is the cumulative travel time required to travel from the starting point along the current search path to that node, and the estimated arrival time of the next node is the sum of the arrival time of the current node and the estimated travel time of the segment between the current node and the next node. Optionally, the estimated travel time of the segment between the current node and the next node can be a pre-set fixed value. Each prediction frame corresponds to the spatial distribution of meteorological features at different prediction times in the meteorological feature matrix. If the prediction frame corresponding to the next node is inconsistent with the prediction frame corresponding to the current node, the meteorological obstacle topology mask obtained based on the prediction frame corresponding to the next node and the meteorological obstacle topology mask obtained based on the prediction frame corresponding to the current node are merged, and the merged mask is used as the obtained meteorological obstacle topology mask. The merging process involves combining two meteorological obstruction topology masks using a logical "OR" operation. Specifically, if the same downscaled spatial grid has a first value in either mask, then that grid will also have a first value in the merged mask; if it has a second value in both masks, then that grid will have a second value in the merged mask. The merged mask is the union of downscaled spatial grids located in the core region of extreme rainstorms at any given moment within the time span of two prediction frames. It is used to characterize the spatial extent of obstructions that a vehicle may encounter during the entire time period from the current node to the next node.

[0117] In one implementation, if the estimated travel time between the current node and the next node is long (e.g., spanning multiple prediction frame intervals), the weather obstruction topology masks corresponding to all prediction frames traversed within that travel time period can be merged, instead of just merging the two consecutive frames. This ensures that any section of the road covered by heavy rain at any given moment is marked as blocked throughout the entire travel process. In another implementation, the merging process of the weather obstruction topology masks can be pre-computed offline. That is, the merged mask for each combination of adjacent prediction frames is pre-calculated and cached. When needed during path search, it can be directly retrieved from the table, reducing online computation and improving search efficiency. In yet another implementation, if the prediction frame corresponding to the next node is the same as the prediction frame corresponding to the current node, no merging process is required. The weather obstruction topology mask corresponding to that prediction frame can be directly used as the obtained mask to avoid unnecessary computational overhead.

[0118] In this embodiment, the merged mask provides a safety buffer covering the entire road segment's travel time in the time dimension. The path search algorithm will not select road segments that may be covered by heavy rain during the journey, ensuring that the entire path has sufficient safety redundancy in the time dimension and eliminating the risk of time-related blind spots caused by the prediction frame boundary effect.

[0119] In an exemplary embodiment, the method further includes: continuously acquiring updated meteorological feature matrices at preset time intervals, and regenerating meteorological obstruction topology masks corresponding to each prediction frame based on the updated meteorological feature matrices; if the updated meteorological obstruction topology mask indicates that a section of the repair path that has not yet been traveled is in an extreme rainstorm area in the corresponding prediction frame, then replanning the subsequent path based on the current location.

[0120] After the repair path is output in the preceding steps, the following dynamic update process is executed: The meteorological feature matrix is ​​continuously acquired at preset time intervals. The preset time interval is a pre-defined rolling acquisition cycle used to control the frequency of meteorological data updates. Optionally, the preset time interval is consistent with the rolling prediction cycle of the short-term radar extrapolation network model (e.g., 5 minutes) to ensure that the latest prediction frame sequence is acquired with each update. The updated meteorological feature matrix is ​​a new round of meteorological feature dataset output by the short-term radar extrapolation network model at the new acquisition time, with its prediction start time shifted backward by the preset time interval, thus reflecting the latest prediction results of meteorological features in the future time period. Based on the updated meteorological feature matrix, the meteorological obstacle topology mask corresponding to each prediction frame is regenerated according to the threshold segmentation method of the preceding steps. The regenerated meteorological obstacle topology mask is compared with the mask state of each untraveled segment in the current repair path in the corresponding prediction frame. The untraveled segments are the set of segments in the repair path from the current position to the destination that have not yet been traversed. The corresponding prediction frames are the prediction frames indexed by the estimated arrival time of each untraveled segment. If the updated weather obstruction topology mask indicates that the value of the corresponding grid cell for a road segment that has not yet been traversed changes from the second value to the first value in its corresponding prediction frame—meaning that the road segment will be located within the core area of ​​extreme rainstorms when the vehicle passes through in the updated prediction—then path replanning is triggered. Path replanning involves re-performing a path search on the adjusted directed road network graph using the current location as the starting point and the location to be repaired as the ending point, generating a new subsequent path to replace the untraversed portion of the original path.

[0121] In one implementation, the preset time interval can be dynamically adjusted according to the severity of meteorological changes: when the rainstorm system moves rapidly or its intensity changes drastically, the time interval is shortened (e.g., 2 minutes) to improve response sensitivity; when the rainstorm system is stable and moves little, the time interval is extended (e.g., 10 minutes) to reduce computational overhead. In another implementation, if the replanned path overlaps with the original path after the current position, only the non-overlapping parts are replaced, and the planning results of the overlapping parts are retained to reduce the computational load of replanning.

[0122] In this embodiment, a dynamic update and replanning mechanism enables the repair route to be continuously corrected as meteorological data is updated, ensuring that the route is always based on the latest meteorological forecast information. Even if the direction, speed, or intensity of the rainstorm system changes unexpectedly during the repair vehicle's journey, the system can still detect the change and trigger replanning in the next update, ensuring that the repair route always maintains its spatiotemporal avoidance capability against the dynamic rainstorm envelope. Since replanning only applies to road sections that have not yet been traversed, the planning results for road sections that have already been safely traversed are unaffected, ensuring the continuous safety of the route while avoiding the waste of computational resources caused by full replanning.

[0123] In one exemplary embodiment, refer to Figure 3 The emergency repair route planning method for power distribution networks includes the following steps:

[0124] 1. Construct a directed graph of the road network containing nodes and directed edges;

[0125] 2. Simultaneously acquire the meteorological characteristic matrix of the distribution network coverage area within a future preset time period;

[0126] 3. Determine the isomorphic spatiotemporal data correlation between the directed road network graph and the meteorological feature matrix;

[0127] 4. Call the POOL 16 spatial pooling operator (the preset size of the spatial pooling window is 16×16) to perform downsampling smoothing on the meteorological feature matrix, smoothing local spatial displacement noise while preserving the large-scale systematic macroscopic dynamic geometric contours.

[0128] 5. Traverse the virtual space microgrid based on 1km×1km and perform the determination of the critical threshold for extreme rainstorms;

[0129] 6. Determine whether the water intensity of the vertically integrated pooling system is greater than or equal to the critical discrimination threshold;

[0130] 7. If so, activate the binary graph network model (graph topology pruning logic driven by meteorological obstruction topology mask) and perform spatial feature logic pruning to temporarily remove the road segments covered by the core area of ​​the disaster area from the directed graph of the road network.

[0131] 8. If not, maintain the initial state of the directed graph of the road network and retain the corresponding edge network to enter the optimization matrix;

[0132] 9. Construct a dynamic meteorological integrated weight cost function for the road network topology and solve for each activated edge;

[0133] 10. Use the dynamic integrated passage cost matrix as the real-time adaptive weight of directed edges;

[0134] 11. Call the heuristic shortest path search algorithm to perform a spatiotemporal graph topological optimal search;

[0135] 12. Output the latitude trajectory sequence of the optimal emergency repair route.

[0136] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0137] Based on the same inventive concept, this application also provides a distribution network emergency repair route planning device for implementing the above-mentioned distribution network emergency repair route planning method. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more distribution network emergency repair route planning device embodiments provided below can be found in the limitations of the distribution network emergency repair route planning method above, and will not be repeated here.

[0138] In one exemplary embodiment, such as Figure 4 As shown, a power distribution network emergency repair route planning device is provided, comprising:

[0139] The road network data processing module 402 is used to construct a directed road network graph based on the road network data of the distribution network coverage area;

[0140] The meteorological data processing module 404 is used to obtain the meteorological feature matrix of the distribution network coverage area within a future preset time period; the meteorological feature matrix includes the meteorological features of each spatial grid in multiple prediction frames; the meteorological feature matrix is ​​downsampled in spatial dimension to obtain a downsampled feature matrix; the downsampled feature matrix includes the meteorological features of each downscaled spatial grid in multiple prediction frames; for each prediction frame, threshold segmentation is performed based on the meteorological features of each downscaled spatial grid in the prediction frame in the downsampled feature matrix and a preset extreme rainstorm critical discrimination threshold to obtain the meteorological obstruction topology mask corresponding to the prediction frame;

[0141] The path planning module 406 is used to perform path search on the directed graph of the road network, starting from a preset location and ending at the location to be repaired, and generating a repair path that avoids areas affected by extreme rainstorms. When the path search reaches the current node, the corresponding meteorological obstacle topology mask is obtained based on the prediction frame corresponding to the current node. The directed graph of the road network is adjusted based on the obtained meteorological obstacle topology mask. The path search continues to be performed on the adjusted directed graph of the road network using the preset path search algorithm until the destination is reached.

[0142] In the aforementioned power distribution network emergency repair route planning device, by introducing spatial dimension downsampling processing, the high-resolution original grid is mapped to a larger-scale downscaled spatial grid, overcoming the interference of spatial displacement errors in the meteorological feature matrix on the determination of rainstorm areas. Based on this, threshold segmentation is performed to obtain a meteorological obstacle topology mask, allowing local displacement deviations to be actively tolerated within the scale of the downscaled grid. This avoids structural omissions of rainstorm envelope boundaries caused by minor misalignments in traditional grid-by-grid precise alignment, ensuring the integrity and reliability of the rainstorm-prohibited area information upon which downstream route planning relies. Since the meteorological feature matrix contains multiple prediction frames, each corresponding to the spatial distribution of rainstorms at different future times, the road network is adjusted by dynamically acquiring the corresponding prediction frame mask based on the current node during the route search process. This achieves spatiotemporal linkage between dynamic meteorological evolution and road network traffic status, ensuring that the road network traffic status is updated synchronously with the movement of the rainstorm system. This allows the planned repair route to match the spatial distribution of rainstorms at each moment, achieving spatiotemporal collaborative avoidance of moving rainstorm envelopes.

[0143] In an exemplary embodiment, the path planning module 406 is further configured to determine the target spatial grid region formed by the downscaled spatial grid with a value of the first value in the meteorological obstruction topology mask; and temporarily remove all edges covered by the target spatial grid region in the road network directed graph to obtain an adjusted road network directed graph.

[0144] In an exemplary embodiment, the path planning module 406 is further configured to, for each edge connected to the current node in the directed graph of the road network, calculate the baseline travel cost of the current edge as the comprehensive travel cost if the value of the downscaled spatial grid corresponding to the current edge in the meteorological obstruction topology mask is a second value; if the value of the downscaled spatial grid corresponding to the current edge in the meteorological obstruction topology mask is a first value, calculate the sum of the baseline travel cost and the additional penalty cost of the current edge as the comprehensive travel cost; the additional penalty cost is determined by the meteorological features of the downscaled spatial grid corresponding to the current edge in the downsampling feature matrix; and based on the comprehensive travel cost, set the weight information of each edge connected to the current node in the directed graph of the road network to obtain the adjusted directed graph of the road network.

[0145] In an exemplary embodiment, the preset extreme rainstorm critical discrimination threshold includes a first threshold and a second threshold, wherein the first threshold is greater than the second threshold; the acquired meteorological obstacle topology mask includes a first mask and a second mask, wherein the first mask is generated based on the first threshold and the second mask is generated based on the second threshold; the path planning module 406 is further configured to determine the target spatial grid region composed of downscaled spatial grids with values ​​of a first value in the first mask; temporarily remove all edges covered by the target spatial grid region in the road network directed graph to obtain the remaining road network directed graph; and for the current node in the remaining road network directed graph... For each connected edge, if the value of the downscaled spatial grid corresponding to the current edge in the second mask is the second value, then the baseline travel cost of the current edge is calculated as the comprehensive travel cost; if the value of the downscaled spatial grid corresponding to the current edge in the second mask is the first value, then the sum of the baseline travel cost and the additional penalty cost of the current edge is calculated as the comprehensive travel cost; the additional penalty cost is determined by the meteorological characteristics of the downscaled spatial grid corresponding to the current edge in the downsampling feature matrix; based on the comprehensive travel cost, the weight information of each edge connected to the current node in the directed graph of the road network is set to obtain the adjusted directed graph of the road network.

[0146] In an exemplary embodiment, the meteorological data processing module 404 is further configured to perform a sliding window scan on the two-dimensional spatial plane corresponding to each prediction frame in the meteorological feature matrix using a spatial pooling window of a preset size; during the scanning process, the meteorological features of each spatial grid covered by the spatial pooling window are pooled into a pooling value, and the pooling value is used as the meteorological feature of the downscaled spatial grid corresponding to the spatial pooling window to obtain the downsampled feature matrix.

[0147] In an exemplary embodiment, the path planning module 406 is further configured to estimate the prediction frame corresponding to the next node of the current node; if the prediction frame corresponding to the next node is inconsistent with the prediction frame corresponding to the current node, the meteorological obstruction topology mask obtained based on the prediction frame corresponding to the next node and the meteorological obstruction topology mask obtained based on the prediction frame corresponding to the current node are merged, and the merged mask is used as the obtained meteorological obstruction topology mask.

[0148] In an exemplary embodiment, the meteorological data processing module 404 is further configured to continuously acquire updated meteorological feature matrices at preset time intervals, and regenerate meteorological obstruction topology masks corresponding to each prediction frame based on the updated meteorological feature matrices.

[0149] The path planning module 406 is also used to replan the subsequent path based on the current location if the road segment that has not yet been traveled in the emergency repair path indicated by the updated meteorological obstruction topology mask is in the extreme rainstorm area in the corresponding prediction frame.

[0150] Each module in the aforementioned multi-objective path planning device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0151] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media to run. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a power distribution network emergency repair path planning method.

[0152] Those skilled in the art will understand that Figure 5The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0153] In one exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the distribution network emergency repair path planning method as described in any of the above embodiments.

[0154] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the distribution network emergency repair path planning method as described in any of the above embodiments.

[0155] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the distribution network emergency repair path planning method as described in any of the above embodiments.

[0156] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0157] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0158] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0159] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for emergency repair route planning in a power distribution network, characterized in that, The method includes: Construct a directed road network graph based on road network data of the distribution network coverage area; Obtain the meteorological feature matrix of the distribution network coverage area within a future preset time period; the meteorological feature matrix includes meteorological features of each spatial grid within multiple prediction frames; The meteorological feature matrix is ​​downsampled in spatial dimension to obtain a downsampled feature matrix; the downsampled feature matrix includes meteorological features of each downscaled spatial grid within multiple prediction frames; For each prediction frame, threshold segmentation is performed based on the meteorological features of each downscaled spatial grid in the prediction frame in the downsampling feature matrix and the preset extreme rainstorm critical discrimination threshold to obtain the meteorological obstruction topology mask corresponding to the prediction frame. Starting from a preset location and ending at the location to be repaired, a preset path search algorithm is used to perform path search on the directed graph of the road network to generate a repair path that avoids areas affected by extreme rainstorms. When the path search reaches the current node, a corresponding meteorological obstruction topology mask is obtained based on the prediction frame corresponding to the current node. The directed graph of the road network is then adjusted based on the obtained meteorological obstruction topology mask. The preset path search algorithm is then used to continue the path search on the adjusted directed graph of the road network until the destination is reached.

2. The method according to claim 1, characterized in that, The adjustment of the directed graph of the road network based on the obtained meteorological obstruction topology mask includes: Determine the target spatial grid region formed by the downscaled spatial grid with the first value in the meteorological obstruction topological mask; In the directed road network graph, all edges covered by the target spatial grid region are temporarily removed to obtain the adjusted directed road network graph.

3. The method according to claim 1, characterized in that, The adjustment of the directed graph of the road network based on the obtained meteorological obstruction topology mask includes: For each edge connected to the current node in the directed graph of the road network, if the value of the downscaled spatial grid corresponding to the current edge in the meteorological obstruction topology mask is the second value, then the baseline passage cost of the current edge is calculated as the comprehensive passage cost. If the value of the downscaled spatial grid corresponding to the current edge in the meteorological obstruction topology mask is the first value, then the sum of the baseline passage cost and the additional penalty cost of the current edge is calculated as the comprehensive passage cost; the additional penalty cost is determined by the meteorological features of the downscaled spatial grid corresponding to the current edge in the downsampling feature matrix; Based on the comprehensive toll cost, the weight information of each edge connected to the current node in the directed graph of the road network is set to obtain the adjusted directed graph of the road network.

4. The method according to claim 1, characterized in that, The preset extreme rainstorm critical discrimination threshold includes a first threshold and a second threshold, wherein the first threshold is greater than the second threshold; the obtained meteorological obstruction topology mask includes a first mask and a second mask, wherein the first mask is generated based on the first threshold and the second mask is generated based on the second threshold; The adjustment of the directed graph of the road network based on the obtained meteorological obstruction topology mask includes: Determine the target spatial grid region in the first mask, which is composed of downscaled spatial grids with values ​​of the first value; Temporarily remove all edges covered by the target space grid region in the road network directed graph to obtain the remaining road network directed graph; For each edge connected to the current node in the remaining road network directed graph, if the value of the downscaled spatial grid corresponding to the current edge in the second mask is the second value, then the baseline passage cost of the current edge is calculated as the comprehensive passage cost. If the value of the downscaled spatial grid corresponding to the current edge in the second mask is the first value, then the sum of the baseline passage cost and the additional penalty cost of the current edge is calculated as the comprehensive passage cost; the additional penalty cost is determined by the meteorological features of the downscaled spatial grid corresponding to the current edge in the downsampling feature matrix; Based on the comprehensive toll cost, the weight information of each edge connected to the current node in the directed graph of the road network is set to obtain the adjusted directed graph of the road network.

5. The method according to claim 1, characterized in that, The process of downsampling the meteorological feature matrix in terms of spatial dimension to obtain a downsampled feature matrix includes: A sliding window scan is performed on the two-dimensional spatial plane corresponding to each prediction frame in the meteorological feature matrix using a spatial pooling window of a preset size. During the scanning process, the meteorological features of each spatial grid covered by the spatial pooling window are pooled into a pooling value, and the pooling value is used as the meteorological feature of the downscaled spatial grid corresponding to the spatial pooling window to obtain the downsampling feature matrix.

6. The method according to any one of claims 1 to 5, characterized in that, The step of obtaining the corresponding meteorological obstruction topology mask based on the prediction frame corresponding to the current node includes: Estimate the prediction frame corresponding to the next node of the current node; If the prediction frame corresponding to the next node is inconsistent with the prediction frame corresponding to the current node, the meteorological obstruction topology mask obtained based on the prediction frame corresponding to the next node and the meteorological obstruction topology mask obtained based on the prediction frame corresponding to the current node will be merged, and the merged mask will be used as the obtained meteorological obstruction topology mask.

7. The method according to any one of claims 1 to 5, characterized in that, The method further includes: The updated meteorological feature matrix is ​​continuously acquired at preset time intervals, and the meteorological obstacle topology mask corresponding to each prediction frame is regenerated based on the updated meteorological feature matrix. If the updated weather obstruction topology mask indicates that the untraveled section of the repair path is in an extreme rainstorm area in the corresponding prediction frame, then the subsequent path is replanned based on the current location.

8. A power distribution network emergency repair route planning device, characterized in that, The device includes: The road network data processing module is used to construct a directed road network graph based on the road network data of the distribution network coverage area; A meteorological data processing module is used to acquire a meteorological feature matrix of the distribution network coverage area within a preset future time period; the meteorological feature matrix includes meteorological features of each spatial grid within multiple prediction frames; the meteorological feature matrix is ​​downsampled in spatial dimension to obtain a downsampled feature matrix; the downsampled feature matrix includes meteorological features of each downscaled spatial grid within multiple prediction frames; for each prediction frame, threshold segmentation is performed based on the meteorological features of each downscaled spatial grid within the prediction frame in the downsampled feature matrix and a preset extreme rainstorm critical discrimination threshold to obtain a meteorological obstruction topology mask corresponding to the prediction frame; The path planning module is used to perform path search on the directed graph of the road network, starting from a preset location and ending at the location to be repaired, using a preset path search algorithm to generate a repair path that avoids areas affected by extreme rainstorms. Specifically, when the path search reaches the current node, a corresponding meteorological obstruction topology mask is obtained based on the prediction frame corresponding to the current node. The directed graph of the road network is then adjusted based on the obtained meteorological obstruction topology mask. The preset path search algorithm is then used to continue the path search on the adjusted directed graph of the road network until the destination is reached.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images