Power grid mountain fire spreading prediction method and device, electronic equipment and storage medium

Abstract

One or more embodiments of the invention provide a power grid mountain fire spreading prediction method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring multi-source observation data of a target area; performing cross-modal fusion on the multi-source observation data to obtain a high-dimensional feature vector; the target area is divided into a plurality of grid units, a power grid structure chart corresponding to the target area is constructed, and the power grid structure chart focuses on a power grid related area of the target area; the power grid structure diagram and the vertex features of the power grid structure diagram in the historical time period are input into a pre-trained space-time attention enhancement diagram convolutional network to obtain a predicted fire behavior state of the target area, and the space-time attention enhancement diagram convolutional network adaptively adjusts the weight of the input data for different positions and time steps; and performing risk assessment on the power equipment in the target area according to the predicted fire behavior state.

Description

Technical Field

[0001] This disclosure relates to the field of power grid disaster prevention technology, and in particular to a method, apparatus, electronic device and storage medium for predicting the spread of wildfires in power grids. Background Technology

[0002] It should be noted that the above description of the technical background is only for the purpose of providing a clear and complete explanation of the technical solutions of the present invention and facilitating understanding by those skilled in the art. It should not be assumed that the above technical solutions are known to those skilled in the art simply because they have been described in the background section of this invention.

[0003] In recent years, global warming has led to more frequent extreme weather events, significantly increasing the risk of wildfires along power grids and seriously threatening the safe and stable operation of the power grid. How to identify fire points and predict risks has become a research hotspot. Summary of the Invention

[0004] In view of the above, the purpose of one or more embodiments of this disclosure is to provide a method, apparatus, electronic device and storage medium for predicting the spread of wildfires in power grids, so as to solve the problems raised in the background art.

[0005] To achieve the above objectives, one or more embodiments of this disclosure provide a method for predicting the spread of wildfires in power grids, including: Acquire multi-source observation data for the target area; Cross-modal fusion of the multi-source observation data yields a high-dimensional feature vector; The target region is divided into multiple grid cells, and a power grid structure diagram corresponding to the target region is constructed. The power grid structure diagram focuses on the power grid-related areas of the target region. The vertex features of the power grid structure map and the power grid structure map in historical time periods are input into a pre-trained spatiotemporal attention-enhanced graph convolutional network to obtain the predicted fire status of the target area. The spatiotemporal attention-enhanced graph convolutional network adaptively adjusts the weights of the input data for different locations and time steps. Based on the predicted fire situation, a risk assessment is conducted on the power equipment in the target area.

[0006] Optionally, after acquiring the multi-source observation data, the method further includes: Acquire optimized multi-source observation data at historical moments; Based on the optimized multi-source observation data of the historical time and the preset inference model, the predicted data for the current time is obtained; Based on the multi-source observation data and the preset credibility assessment model, the weight coefficients of the multi-source observation data are obtained; The multi-source observation data is optimized based on the predicted data and the weighting coefficients to obtain the optimized multi-source observation data at the current time.

[0007] Optionally, the multi-source observation data includes synthetic aperture radar satellite data, multispectral satellite data, digital elevation data, power grid vector data, meteorological data, and real-time monitoring data; Cross-modal fusion of the multi-source observation data yields a high-dimensional feature vector, including: Extract terrain features from the synthetic aperture radar satellite data; Based on the terrain features, guide the extraction of vegetation categories from the multispectral satellite data and the extraction of humidity features from the meteorological data; Based on the tower locations identified in the power grid vector data, weighting coefficients are assigned to the features extracted from the multi-source observation data and / or the multi-source observation data, with the tower locations as anchor points; The high-dimensional feature vector is obtained based on the multi-source observation data and the cross-modal fusion model; ; Where p represents the coordinates of the tower location within the target area, and t represents the acquisition time of the multi-source observation data. This represents the cross-domain attention-gated fusion function. Let S represent the environmental feature extraction function, D represent the synthetic aperture radar satellite data, and M represent the meteorological data. This represents the fire feature extraction function. O represents multispectral satellite data, P represents real-time monitoring data, and O represents power grid vector data.

[0008] Optionally, the target region is divided into multiple grid cells, and a power grid structure diagram corresponding to the target region is constructed, including: The target region is discretized into multiple grid cells; Based on the grid cells, the power grid structure diagram is constructed. The nodes of the power grid structure diagram represent the grid cells, the node values ​​represent the high-dimensional feature vectors corresponding to the grid cells, and the edges represent the connection relationships of the nodes. The nodes are assigned weights based on the location of the power grid equipment, and the edges are assigned weights based on the spatial distance between the connected nodes, the influence of wind direction, and the continuity of combustibles.

[0009] Optionally, the spatiotemporal attention-enhanced graph convolutional network is trained based on a preset loss function, which includes data fitting loss and physical loss.

[0010] Optionally, the physical loss is obtained based on the degree of violation of physical laws by the predicted fire state using multi-source observation data for training.

[0011] Optionally, based on the predicted fire situation, a risk assessment is conducted on the electrical equipment in the target area, including: Based on the predicted fire situation, the individual foundation risks of power equipment within the target area are obtained; Based on the single-point basic risk and the topological relationship of the power equipment in the target area, the cascading risk of the power equipment is obtained. The cascading risk represents the probability of the impact on adjacent equipment after any power equipment is damaged. A risk assessment is conducted on the power equipment in the target area based on the single-point basic risk and the cascading risk.

[0012] Based on the same inventive concept, one or more embodiments of this disclosure also provide a power grid wildfire spread prediction device, comprising: The data acquisition module is configured to acquire multi-source observation data of the target area; The feature extraction module is configured to perform cross-modal fusion on the multi-source observation data to obtain a high-dimensional feature vector; The first calculation module is configured to divide the target region into multiple grid cells and construct a power grid structure diagram corresponding to the target region, wherein the power grid structure diagram focuses on the power grid-related areas of the target region; The second calculation module is configured to input the power grid structure map and the vertex features of the power grid structure map in historical time periods into a pre-trained spatiotemporal attention-enhanced graph convolutional network to obtain the predicted fire status of the target area. The spatiotemporal attention-enhanced graph convolutional network adaptively adjusts the weights of the input data for different locations and time steps. The risk assessment module is configured to perform a risk assessment on the electrical equipment in the target area based on the predicted fire status.

[0013] Based on the same inventive concept, one or more embodiments of this disclosure also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the power grid wildfire spread prediction method as described in any of the above.

[0014] Based on the same inventive concept, one or more embodiments of this disclosure also provide a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute any of the above-described power grid wildfire spread prediction methods.

[0015] As can be seen from the above, the power grid wildfire spread prediction method provided in one or more embodiments of this disclosure first extracts high-dimensional feature vectors from multi-source observation data. Then, by constructing a power grid structure graph and assigning weights to nodes and edges related to power grid equipment in the graph, it focuses on high-dimensional feature vectors related to high-risk areas. Furthermore, it obtains a more accurate prediction of the fire state through a spatiotemporal attention-enhanced graph convolutional network that can adaptively adjust the weights of the input data for different locations and time steps. This disclosure improves the accuracy of fire state prediction, thereby enhancing the accuracy of power grid wildfire spread assessment.

[0016] The power grid wildfire spread prediction device, electronic device, and computer-readable storage medium disclosed herein can all implement the steps of the above-mentioned power grid wildfire spread prediction method, and therefore also have the beneficial effects of the above-mentioned power grid wildfire spread prediction method. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in one or more embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only one or more embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating one or more embodiments of the power grid wildfire spread prediction method disclosed herein; Figure 2 This is a schematic diagram of the structure of a power grid wildfire spread prediction device according to one or more embodiments of the present disclosure; Figure 3 This is a schematic diagram of the hardware structure of an electronic device according to one or more embodiments of this disclosure. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0020] It should be noted that, unless otherwise defined, the technical or scientific terms used in one or more embodiments of this disclosure should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar words used in one or more embodiments of this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0021] As described in the background section, how to identify fire points and predict risks has become a research hotspot.

[0022] Traditional technologies use wildfire spread models (such as the Rothermel model and the FARSITE system) to predict wildfire risks. However, these models are mainly based on physical equations and semi-empirical formulas. Their deduction is highly dependent on the accuracy and consistency of the input parameters and is computationally complex, making it difficult to meet the dual requirements of power grid emergency response for deduction speed and accuracy.

[0023] With the development of remote sensing and deep learning technologies, some researchers have proposed applying these technologies to this problem. However, these technologies have not focused on critical power grid equipment, resulting in poor adaptability to power grid scenarios and low coupling with the power grid.

[0024] refer to Figure 1 This disclosure includes one or more embodiments of a power grid wildfire spread prediction method, comprising the following steps: Step S101: Acquire multi-source observation data of the target area; Step S102: Perform cross-modal fusion on multi-source observation data to obtain high-dimensional feature vectors; Step S103: Divide the target area into multiple grid cells and construct the power grid structure diagram corresponding to the target area. The power grid structure diagram focuses on the power grid-related areas of the target area. Step S104: Input the vertex features of the power grid structure map and the power grid structure map in the historical period into the pre-trained spatiotemporal attention-enhanced graph convolutional network to obtain the predicted fire status of the target area. The spatiotemporal attention-enhanced graph convolutional network adaptively adjusts the weights of the input data for different locations and time steps. Step S105: Based on the predicted fire situation, conduct a risk assessment of the electrical equipment in the target area.

[0025] In some embodiments, multi-source observation data may include synthetic aperture radar (SAR) satellite data, multispectral satellite data, digital elevation data (DEM), power grid vector data, meteorological data, and real-time monitoring data. SAR satellite data can be used to extract high-precision topographic data of the target area. Multispectral satellite data can be used to extract features such as vegetation, humidity, and combustion zones. DEM can be used to obtain the slope and aspect characteristics of the target area. Power grid vector data can be used to extract the geographical location and topological relationships of power equipment (e.g., poles, power lines, etc.). Meteorological data can be used to extract features such as wind speed, wind direction, temperature, humidity, and precipitation. Real-time monitoring data can be used to provide real-time data on power grid equipment.

[0026] In some embodiments, real-time monitoring data may include drone patrol video data and data from ground-based smart monitoring devices deployed on power grid equipment.

[0027] In some embodiments, land use data may also be included to determine the distribution of combustible material types.

[0028] In some embodiments, multi-source observation data can also be weighted to ensure that subsequent data processing focuses on data with higher reliability.

[0029] In some embodiments, the weighting coefficients are determined based on indicators such as the type of observable data, the accuracy of the data acquisition equipment, and real-time performance.

[0030] In some embodiments, to improve the reliability of the weighting coefficients, the weighting coefficients are dynamically adjusted in each calculation.

[0031] In some embodiments, the process of calculating dynamic weights and updating multi-source observation data may include: acquiring optimized multi-source observation data at historical time points; obtaining predicted data at the current time point based on the optimized multi-source observation data at historical time points and a preset inference model; obtaining weight coefficients of the multi-source observation data based on the multi-source observation data and a preset credibility assessment model; and optimizing the multi-source observation data based on the predicted data and weight coefficients to obtain optimized multi-source observation data at the current time point.

[0032] In some embodiments, the predicted data for the current moment can be obtained through a deductive model: ; Where i represents the prediction step of dynamic data assimilation, and a represents the optimal state of the system obtained after data assimilation. Optimized multi-source observation data representing historical moments. Indicates model noise. This indicates the number of set members in dynamic data assimilation.

[0033] In some embodiments, the optimized multi-source observation data is obtained through the following calculation: ; .

[0034] It is understandable that the obtained multi-source observation data can be used to optimize the observation data for the next time step.

[0035] By dynamically adjusting the weighting coefficients of multi-source observation data, the differences in real-time reliability of different data sources are fully considered.

[0036] In some embodiments, before fusing multi-source observation data across modalities, the multi-source observation data is first standardized and aligned to unify the multi-source observation data into the same spatiotemporal grid coordinate system.

[0037] In some embodiments, cross-domain guided feature extraction is achieved by leveraging the association between different data sources.

[0038] In some embodiments, the extraction of vegetation type and humidity features from multispectral satellites is optimized by using high-precision terrain features from synthetic aperture radar as a guide; and the data feature weights of the surrounding area are strengthened by using the tower locations in the power grid vector data as anchor points.

[0039] In some embodiments, data feature extraction can be represented as: ; Where p represents the coordinates of the tower location within the target area, and t represents the acquisition time of the multi-source observation data. This represents the cross-domain attention-gated fusion function. Let S represent the environmental feature extraction function, D represent the synthetic aperture radar satellite data, and M represent the meteorological data. This represents the fire feature extraction function. O represents multispectral satellite data, P represents real-time monitoring data, and O represents power grid vector data.

[0040] In some embodiments, p can also represent the spatial location of all geographically defined landmarks within the target area. Examples include the geographic coordinates of geographically defined landmarks such as power grid poles, fire monitoring stations, and fire detectors.

[0041] In some embodiments, environmental features and fire features can be extracted by a dual-branch U-Net encoder network, and then the two types of features can be dynamically fused through a cross-domain attention gating mechanism to generate a multimodal fusion feature tensor rich in power grid scenario-specific features.

[0042] The above embodiments fully integrate cross-domain guidance information with power grid scenario requirements, providing high-quality input for subsequent simulations.

[0043] In some embodiments, the power grid structure diagram can be represented as G=(V,E,A), where V represents a node, E represents an edge, and A represents a weight. ; in, This represents the power grid adaptation weight factor. If node i or j is the location of power grid equipment or its surrounding area, then... =1.5, otherwise =1.0, enabling focused attention on areas related to the power grid.

[0044] In some embodiments, the spatiotemporal attention-enhanced graph convolutional network is responsible for capturing the spatiotemporal dependencies of fire spread. It adaptively adjusts the feature weights of different regions and time steps through a spatiotemporal attention mechanism, focusing on the fire spread patterns in high-risk areas along the power grid.

[0045] The process of a spatiotemporal attention-enhanced graph convolutional network predicting fire status can be represented as: .

[0046] In some embodiments, The loss function of a spatiotemporal attention-enhanced graph convolutional network can be expressed as: ; in, This represents the data fitting loss, indicating the error between the predicted fire state and the actual fire state. The physical loss is represented so that the model learning process can both fit the data patterns and follow the specific physical mechanisms of wildfire propagation in the power grid corridor area.

[0047] In some embodiments, .

[0048] By using spatiotemporal attention-enhanced graph convolutional networks, high-risk areas (such as areas around power poles and densely packed combustible materials) can be adaptively focused on, allowing for optimization of specific power grid areas such as power grid corridors. The aforementioned constraints can also be combined with the specific physical mechanisms of wildfire spread and the needs of power grid scenarios to improve the model's generalization ability and reliability.

[0049] In some embodiments, the risk assessment process may include: obtaining the single-point basic risk of power equipment in the target area based on the predicted fire status; obtaining the cascading risk of power equipment based on the single-point basic risk and the topological relationship of power equipment in the target area, where the cascading risk represents the probability of impact on adjacent equipment after any power equipment is damaged; and conducting a risk assessment of the power equipment in the target area based on the single-point basic risk and the cascading risk.

[0050] In some embodiments, single-point basic risk includes multiple dimensions: the time of arrival of the fire approaching electrical equipment (e.g., lines, towers), the fire intensity, and the importance of the equipment. Fire intensity indicators may include temperature, spread rate, etc., while equipment importance indicators may include voltage level and power supply range.

[0051] In some embodiments, a comprehensive risk index can be obtained based on the single-point basic risk and the cascading risk. The comprehensive risk index can be expressed as: ; in, Let k represent the set of adjacent devices of facility k. This represents the fire propagation intensity between k and m. f( represents the topological affinity between k and m) ) is the chain risk transmission function.

[0052] In some embodiments, to further improve the visualization of risk assessment results, the target area can be divided according to risk level and represented by different colors.

[0053] In some embodiments, red indicates an emergency risk that requires immediate power outage to avoid danger; orange indicates a high risk that requires enhanced monitoring and preparation for a power outage; yellow indicates a warning risk that requires continuous monitoring; and blue indicates a low risk that requires normal monitoring.

[0054] It is understandable that this method can be executed by any device, equipment, platform, or cluster of devices with computing and processing capabilities.

[0055] It should be noted that the methods of one or more embodiments of this disclosure can be executed by a single device, such as a computer or server. The methods of this embodiment can also be applied in a distributed scenario, where multiple devices cooperate to complete the process. In such a distributed scenario, one of these devices may execute only one or more steps of the methods of one or more embodiments of this disclosure, and the multiple devices will interact with each other to complete the method described.

[0056] It should be noted that the above description pertains to specific embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than those shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0057] Based on the same inventive concept, and corresponding to any of the methods in the above embodiments, this disclosure also provides a power grid wildfire spread prediction device. For example... Figure 2 As shown, the device includes: Data acquisition module 11 is configured to acquire multi-source observation data of the target area; Feature extraction module 12 is configured to perform cross-modal fusion of multi-source observation data to obtain high-dimensional feature vectors; The first calculation module 13 is configured to divide the target area into multiple grid cells and construct a power grid structure diagram corresponding to the target area, the power grid structure diagram focusing on the power grid-related areas of the target area; The second calculation module 14 is configured to input the vertex features of the power grid structure map and the power grid structure map in historical time periods into a pre-trained spatiotemporal attention-enhanced graph convolutional network to obtain the predicted fire status of the target area. The spatiotemporal attention-enhanced graph convolutional network adaptively adjusts the weights of the input data for different locations and time steps. Risk assessment module 15 is configured to conduct risk assessments on electrical equipment in the target area based on the predicted fire status.

[0058] In some embodiments, the above-described apparatus is further configured to: Acquire optimized multi-source observation data at historical moments; Based on the optimized multi-source observation data of the historical time and the preset inference model, the predicted data for the current time is obtained; Based on the multi-source observation data and the preset credibility assessment model, the weight coefficients of the multi-source observation data are obtained; The multi-source observation data is optimized based on the predicted data and the weighting coefficients to obtain the optimized multi-source observation data at the current time.

[0059] In some embodiments, multi-source observation data includes synthetic aperture radar satellite data, multispectral satellite data, digital elevation data, power grid vector data, meteorological data, and real-time monitoring data; Feature extraction module 12 is specifically configured as follows: Extract terrain features from the synthetic aperture radar satellite data; Based on the terrain features, guide the extraction of vegetation categories from the multispectral satellite data and the extraction of humidity features from the meteorological data; Based on the tower locations identified in the power grid vector data, weighting coefficients are assigned to the features extracted from the multi-source observation data and / or the multi-source observation data, with the tower locations as anchor points; The high-dimensional feature vector is obtained based on the multi-source observation data and the cross-modal fusion model; ; Where p represents the coordinates of the tower location within the target area, and t represents the acquisition time of the multi-source observation data. This represents the cross-domain attention-gated fusion function. Let S represent the environmental feature extraction function, D represent the synthetic aperture radar satellite data, and M represent the meteorological data. This represents the fire feature extraction function. O represents multispectral satellite data, P represents real-time monitoring data, and O represents power grid vector data.

[0060] In some embodiments, the first computing module 13 is specifically configured as follows: The target region is discretized into multiple grid cells; Based on the grid cells, the power grid structure diagram is constructed. The nodes of the power grid structure diagram represent the grid cells, the node values ​​represent the high-dimensional feature vectors corresponding to the grid cells, and the edges represent the connection relationships of the nodes. The nodes are assigned weights based on the location of the power grid equipment, and the edges are assigned weights based on the spatial distance between the connected nodes, the influence of wind direction, and the continuity of combustibles.

[0061] In some embodiments, the spatiotemporal attention-enhanced graph convolutional network is trained based on a preset loss function, which includes data fitting loss and physical loss.

[0062] In some embodiments, the physical loss is obtained based on the degree of violation of physical laws by the predicted fire state trained using multi-source observation data.

[0063] In some embodiments, the evaluation module 15 is specifically configured as follows: Based on the predicted fire situation, the individual foundation risks of power equipment within the target area are obtained; Based on the single-point basic risk and the topological relationship of the power equipment in the target area, the cascading risk of the power equipment is obtained. The cascading risk represents the probability of the impact on adjacent equipment after any power equipment is damaged. A risk assessment is conducted on the power equipment in the target area based on the single-point basic risk and the cascading risk.

[0064] For ease of description, the above apparatus is described in terms of its functions, divided into various modules. Of course, when implementing one or more embodiments of this disclosure, the functions of each module can be implemented in one or more software and / or hardware.

[0065] The apparatus described above is used to implement the corresponding methods in the foregoing embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0066] Figure 3This embodiment illustrates a more specific hardware structure of an electronic device. The device may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0067] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this disclosure.

[0068] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this disclosure are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0069] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0070] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0071] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0072] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this disclosure, and not necessarily all the components shown in the figures.

[0073] The electronic devices described above are used to implement the corresponding methods in the foregoing embodiments and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0074] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0075] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure (including the claims) is limited to these examples; within the framework of this disclosure, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of this disclosure as described above, which are not provided in detail for the sake of brevity.

[0076] Additionally, to simplify the description and discussion, and to avoid obscuring one or more embodiments of this disclosure, the provided drawings may or may not show well-known power / ground connections to integrated circuit (IC) chips and other components. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring one or more embodiments of this disclosure, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which one or more embodiments of this disclosure will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuitry) are set forth to describe exemplary embodiments of this disclosure, it will be apparent to those skilled in the art that one or more embodiments of this disclosure may be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0077] Although this disclosure has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0078] This disclosure includes one or more embodiments intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for predicting the spread of wildfires over power grids, characterized in that, include: Acquire multi-source observation data for the target area; Cross-modal fusion of the multi-source observation data yields a high-dimensional feature vector; The target region is divided into multiple grid cells, and a power grid structure diagram corresponding to the target region is constructed. The power grid structure diagram focuses on the power grid-related areas of the target region. The vertex features of the power grid structure map and the power grid structure map in historical time periods are input into a pre-trained spatiotemporal attention-enhanced graph convolutional network to obtain the predicted fire status of the target area. The spatiotemporal attention-enhanced graph convolutional network adaptively adjusts the weights of the input data for different locations and time steps. Based on the predicted fire situation, a risk assessment is conducted on the power equipment in the target area.

2. The method according to claim 1, characterized in that, After acquiring the multi-source observation data, the method further includes: Acquire optimized multi-source observation data at historical moments; Based on the optimized multi-source observation data of the historical time and the preset inference model, the predicted data for the current time is obtained; Based on the multi-source observation data and the preset credibility assessment model, the weight coefficients of the multi-source observation data are obtained; The multi-source observation data is optimized based on the predicted data and the weighting coefficients to obtain the optimized multi-source observation data at the current time.

3. The method according to claim 1, characterized in that, The multi-source observation data includes synthetic aperture radar satellite data, multispectral satellite data, digital elevation data, power grid vector data, meteorological data, and real-time monitoring data; Cross-modal fusion of the multi-source observation data yields a high-dimensional feature vector, including: Extract terrain features from the synthetic aperture radar satellite data; Based on the terrain features, guide the extraction of vegetation categories from the multispectral satellite data and the extraction of humidity features from the meteorological data; Based on the tower locations identified in the power grid vector data, weighting coefficients are assigned to the features extracted from the multi-source observation data and / or the multi-source observation data, with the tower locations as anchor points; The high-dimensional feature vector is obtained based on the multi-source observation data and the cross-modal fusion model; ； Where p represents the coordinates of the tower location within the target area, and t represents the acquisition time of the multi-source observation data. This represents the cross-domain attention-gated fusion function. Let S represent the environmental feature extraction function, D represent the synthetic aperture radar satellite data, and M represent the meteorological data. This represents the fire feature extraction function. O represents multispectral satellite data, P represents real-time monitoring data, and O represents power grid vector data.

4. The method according to claim 1, characterized in that, The target region is divided into multiple grid cells, and a power grid structure diagram corresponding to the target region is constructed, including: The target region is discretized into multiple grid cells; Based on the grid cells, the power grid structure diagram is constructed. The nodes of the power grid structure diagram represent the grid cells, the node values ​​represent the high-dimensional feature vectors corresponding to the grid cells, and the edges represent the connection relationships of the nodes. The nodes are assigned weights based on the location of the power grid equipment, and the edges are assigned weights based on the spatial distance between the connected nodes, the influence of wind direction, and the continuity of combustibles.

5. The method according to claim 1, characterized in that, The spatiotemporal attention-enhanced graph convolutional network is trained based on a preset loss function, which includes data fitting loss and physical loss.

6. The method according to claim 5, characterized in that, The physical loss is obtained based on the degree of violation of physical laws by the predicted fire state from the multi-source observation data used in training.

7. The method according to claim 1, characterized in that, Based on the predicted fire situation, a risk assessment is conducted on the power equipment in the target area, including: Based on the predicted fire situation, the individual foundation risks of power equipment within the target area are obtained; Based on the single-point basic risk and the topological relationship of the power equipment in the target area, the cascading risk of the power equipment is obtained. The cascading risk represents the probability of the impact on adjacent equipment after any power equipment is damaged. Based on the single-point basic risk and the cascading risk, a risk assessment is conducted on the power equipment in the target area.

8. A power grid wildfire spread prediction device, characterized in that, include: The data acquisition module is configured to acquire multi-source observation data of the target area; The feature extraction module is configured to perform cross-modal fusion on the multi-source observation data to obtain a high-dimensional feature vector; The first calculation module is configured to divide the target region into multiple grid cells and construct a power grid structure diagram corresponding to the target region, wherein the power grid structure diagram focuses on the power grid-related areas of the target region; The second calculation module is configured to input the power grid structure map and the vertex features of the power grid structure map in historical time periods into a pre-trained spatiotemporal attention-enhanced graph convolutional network to obtain the predicted fire status of the target area. The spatiotemporal attention-enhanced graph convolutional network adaptively adjusts the weights of the input data for different locations and time steps. The risk assessment module is configured to perform a risk assessment on the electrical equipment in the target area based on the predicted fire status.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executed by the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing the computer to perform the method of any one of claims 1 to 7.

Images