A GIS-based intelligent monitoring and early warning method and system for power grid meteorological disasters

Abstract

The present application provides a method and system for intelligent monitoring and early warning of meteorological disasters in power grids based on GIS, the method comprising: obtaining historical multi-source data including power grid equipment data, meteorological data and equipment operation data; standardizing the historical multi-source data to obtain a four-dimensional data model including longitude, latitude, altitude and time dimensions; constructing a disaster risk assessment model based on the four-dimensional data model using a deep learning method; inputting real-time multi-source data into the disaster risk assessment model to obtain disaster risk assessment results; making early warning decisions based on the disaster risk assessment results to generate early warning information; and generating emergency response plans based on the early warning information. The system comprises a data processing module, a model building module, a risk assessment module, an early warning decision module and an emergency response module. The technical solution of the present application realizes the effective fusion of multi-source heterogeneous data, improves the accuracy and real-time performance of early warnings, and improves the efficiency of emergency response.

Description

Technical Field

[0001] The present application relates to the technical field of safe operation of power systems, and in particular to a GIS-based intelligent monitoring and early warning method and system for meteorological disasters in power grids. Background Art

[0002] As an important infrastructure, the safe and stable operation of the power grid system is crucial to social and economic development and people's lives. Meteorological disasters are one of the important factors threatening the safe operation of the power grid. Natural disasters such as lightning, ice and snow, and strong winds may cause serious consequences such as damage to power grid equipment and system failures.

[0003] At present, the existing technology will adopt a power grid meteorological disaster early warning system based on statistical learning methods. The system mainly uses traditional machine learning algorithms (such as support vector machines, random forests, etc.) to analyze historical meteorological and power grid operation data and establish a mapping relationship between meteorological disasters and power grid failures. The system sets a fixed threshold for early warning and generates disposal suggestions using a predefined rule base. However, this method has the following limitations: First, it is difficult to effectively process multi-source heterogeneous data, and various types of data are often processed separately, failing to fully explore the correlation characteristics between data; second, the early warning model lacks the ability to model complex spatiotemporal features, and the accuracy and timeliness of the early warning results are insufficient; third, the generation of disposal plans is too dependent on expert experience and lacks intelligence and adaptive capabilities. These problems have seriously affected the early warning effect and practical value of the system in practical applications. Summary of the invention

[0004] In view of this, the present application provides a GIS-based intelligent monitoring and early warning method and system for power grid meteorological disasters, which solves the problems of data fusion difficulty, insufficient early warning accuracy and lack of intelligence in the formulation of emergency response plans in the power grid meteorological disaster monitoring and early warning system in the prior art.

[0005] The present application embodiment provides a GIS-based intelligent monitoring and early warning method for power grid meteorological disasters, including:

[0006] Acquiring historical multi-source data, wherein the historical multi-source data includes power grid equipment data, meteorological data, and equipment operation data;

[0007] Standardizing the historical multi-source data to obtain a four-dimensional data model including longitude, latitude, altitude and time dimensions;

[0008] Based on the four-dimensional data model, a disaster risk assessment model is constructed using a deep learning method, wherein the disaster risk assessment model includes a meteorological disaster type identification model, a power grid vulnerability assessment model, and a disaster impact prediction model;

[0009] Inputting real-time multi-source data into the disaster risk assessment model, identifying the type of meteorological disaster based on the meteorological disaster type identification model, assessing the vulnerability of equipment based on the power grid vulnerability assessment model, and predicting the scope and extent of disaster impact based on the meteorological disaster type, the equipment vulnerability assessment results and the disaster impact prediction model to obtain a disaster risk assessment result;

[0010] Make early warning decisions based on the disaster risk assessment results and generate early warning information;

[0011] An emergency response plan is generated based on the warning information.

[0012] Optionally, the standardization of historical multi-source data includes:

[0013] Standardize the formats of data from different sources and establish a unified data description standard;

[0014] Use statistically based outlier detection methods and time series interpolation algorithms for data cleaning and correction;

[0015] Based on spatiotemporal indexing technology, the alignment and fusion of multi-source data are achieved, and a four-dimensional data model including longitude, latitude, altitude and time dimensions is constructed.

[0016] Optionally, constructing a disaster risk assessment model based on the four-dimensional data model using a deep learning method includes:

[0017] Using a convolutional neural network to build a meteorological disaster type recognition model, the input of the meteorological disaster type recognition model is the meteorological data, and the output is the meteorological disaster type;

[0018] A power grid vulnerability assessment model is established by using a graph neural network, wherein the input of the power grid vulnerability assessment model is the power grid equipment data and equipment operation data, and the output is the equipment vulnerability assessment result;

[0019] A Lipschitz neural network is used to construct a disaster impact prediction model, the input of which is the type of meteorological disaster and the equipment vulnerability assessment results, and the output is the disaster risk assessment results including the scope and degree of disaster impact.

[0020] Optionally, the use of a Lipschitz neural network to construct a disaster impact prediction model includes:

[0021] Design a multi-layer Lipschitz neural network structure and constrain the weight matrix of each layer of the network through spectral normalization technology;

[0022] Quantify the sensitivity of multi-layer Lipschitz neural network structure to different input variables through a quality assessment mechanism based on the extended Jacobian matrix;

[0023] Design an adaptive loss function including prediction error term, Lipschitz constraint term and regularization term;

[0024] The disaster impact prediction model is optimized based on the optimization strategy of adaptive loss function and gradient clipping.

[0025] Optionally, the establishing of a power grid vulnerability assessment model using a graph neural network includes:

[0026] The power grid topology structure in the power grid equipment data is modeled as a graph structure, where nodes represent equipment and edges represent line connection relationships;

[0027] Capture the interaction relationship between nodes through graph convolutional layers and message passing mechanism;

[0028] Introducing a temporal attention layer to model the temporal dependencies of device states;

[0029] The vulnerability score of each node is output through a multi-layer perceptron.

[0030] Optionally, making a warning decision based on the disaster risk assessment result and generating warning information includes:

[0031] Based on the disaster risk assessment results, a multi-level fuzzy comprehensive evaluation method is used to calculate the risk level;

[0032] Based on the risk level, a dynamic programming algorithm is used to optimize the warning triggering and release conditions;

[0033] The risk level, warning trigger and release conditions are input into a visual warning display system based on WebGIS to generate warning information.

[0034] Optionally, the using a dynamic programming algorithm to optimize the warning triggering and release conditions according to the risk level includes:

[0035] Modeling the early warning process as a multi-stage decision-making problem;

[0036] Constructing a state transfer equation, calculating the optimal warning time and warning level corresponding to different risk levels, and constructing warning triggering and release conditions for different risk levels according to the optimal warning time and warning level;

[0037] Through a reinforcement learning-based strategy, the warning triggering and release conditions are continuously optimized by learning historical warning effect feedback.

[0038] Optionally, generating an emergency response plan based on the warning information includes:

[0039] Construct a knowledge graph of emergency plans and establish the relationship between event types, disposal measures and resource requirements;

[0040] Based on the risk level and warning triggering and release conditions in the warning information, a step-by-step emergency response process is generated using graph reasoning and case matching algorithms;

[0041] According to the emergency response process, an optimal scheduling scheme for emergency resources is calculated based on a heuristic algorithm;

[0042] The emergency response process and the optimal scheduling plan are combined to generate a final emergency response plan.

[0043] Optionally, constructing the emergency plan knowledge graph includes:

[0044] Use multi-level semantic network structure to build knowledge graph;

[0045] Extract key knowledge from historical emergency plan documents, incident handling reports and expert experience.

[0046] The present application also provides a GIS-based intelligent monitoring and early warning device for power grid meteorological disasters, including:

[0047] A data processing module is used to obtain historical multi-source data and perform standardization processing to obtain a four-dimensional data model including longitude, latitude, altitude and time dimensions, wherein the historical multi-source data includes power grid equipment data, meteorological data and equipment operation data;

[0048] A model building module, used to build a disaster risk assessment model based on the four-dimensional data model using a deep learning method; the disaster risk assessment model includes a meteorological disaster type identification model, a power grid vulnerability assessment model and a disaster impact prediction model;

[0049] A risk assessment module, used to input real-time multi-source data into the disaster risk assessment model, identify the type of meteorological disaster based on the meteorological disaster type identification model, assess the equipment vulnerability based on the power grid vulnerability assessment model, predict the scope and extent of disaster impact based on the meteorological disaster type, the equipment vulnerability assessment result and the disaster impact prediction model, and obtain a disaster risk assessment result;

[0050] An early warning decision module is used to make early warning decisions based on the disaster risk assessment results and generate early warning information;

[0051] The emergency response module is used to generate an emergency response plan based on the warning information.

[0052] The present application also provides a computer device, the computer device comprising:

[0053] at least one processor; and,

[0054] a memory communicatively connected to the at least one processor; wherein,

[0055] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned GIS-based power grid meteorological disaster intelligent monitoring and early warning method.

[0056] An embodiment of the present application also provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the above-mentioned GIS-based power grid meteorological disaster intelligent monitoring and early warning method.

[0057] An embodiment of the present application also provides a computer program product, including computer instructions, characterized in that when the computer instructions are executed by a processor, the steps of the above-mentioned GIS-based power grid meteorological disaster intelligent monitoring and early warning method are implemented.

[0058] This application has the following technical effects:

[0059] 1. By building a four-dimensional data model, the effective integration of multi-source heterogeneous data is achieved, and the efficiency of data utilization is improved;

[0060] 2. Use deep learning methods to build a disaster risk assessment model, which improves the accuracy and real-time nature of early warning;

[0061] 3. Generate emergency response plans based on knowledge graphs and intelligent algorithms to improve the efficiency and pertinence of emergency response. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for describing the embodiments are briefly introduced below.

[0063] Figure 1 A flowchart of a GIS-based intelligent monitoring and early warning method for power grid meteorological disasters provided in an embodiment of the present application;

[0064] Figure 2 A schematic diagram of the structure of a four-dimensional data model provided in an embodiment of the present application;

[0065] Figure 3 A schematic diagram of the architecture of a disaster risk assessment model provided in an embodiment of the present application;

[0066] Figure 4 A schematic diagram of the emergency response plan generation process provided in the embodiment of the present application;

[0067] Figure 5A schematic diagram of the structure of a GIS-based intelligent monitoring and early warning device for power grid meteorological disasters provided in an embodiment of the present application. DETAILED DESCRIPTION

[0068] The technical solution of the present application is described in detail below through specific embodiments, but the protection scope of the present application is not limited to the following description.

[0069] Figure 1 A flow chart of a GIS-based intelligent monitoring and early warning method for power grid meteorological disasters provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the embodiment of the present application includes the following steps:

[0070] S1: Acquire historical multi-source data, where the historical multi-source data includes power grid equipment data, meteorological data, and equipment operation data;

[0071] Historical multi-source data is obtained mainly through three channels: power grid equipment data can be collected through the power grid SCADA system and EMS system, including basic attribute information of equipment (such as equipment type, installation location, operating years, etc.), real-time operating status (such as voltage, current, active power, reactive power, etc.) and historical fault records; meteorological data is obtained through cooperation with the meteorological department, including historical observation data and forecast data of meteorological elements such as temperature, humidity, wind speed, wind direction, precipitation, lightning intensity, etc.; equipment operation data comes from the equipment status monitoring system, including equipment vibration, temperature, noise and other state quantities, as well as equipment inspection and maintenance records and other information. These data are transmitted using standard data interface protocols, such as Modbus protocol, IEC 61850 protocol, etc.

[0072] S2: Standardize the historical multi-source data to obtain a four-dimensional data model including longitude, latitude, altitude and time dimensions;

[0073] This is a step to standardize historical multi-source data. Data standardization is mainly divided into three sub-steps:

[0074] S2.1: Standardize the formats of data from different sources and establish a unified data description standard;

[0075] First, the format is unified. For heterogeneous data from different sources, a unified data description standard is established, including defining data attributes such as type, format, unit, and precision. For example, all time data is uniformly converted to UTC time format, temperature data is uniformly converted to Celsius, and coordinate information is uniformly converted to WGS84 coordinate system. At the same time, a unified naming mapping mechanism is established to ensure the semantic consistency of data items in view of the differences in naming specifications of different data sources.

[0076] S2.2: Use statistically based outlier detection methods and time series interpolation algorithms for data cleaning and correction;

[0077] Secondly, data cleaning and correction are carried out. A statistically based outlier detection method is used, such as calculating the mean and standard deviation of the data and using the 3σ criterion to identify outliers; for missing values ​​in time series data, a time series interpolation algorithm is used to repair them, such as linear interpolation, spline interpolation and other methods. Specifically, when an abnormal jump in the temperature data of a certain monitoring point is detected, the data point will be automatically marked, and reasonable interpolation will be performed based on the historical data before and after the monitoring point to ensure the continuity and reliability of the data.

[0078] S2.3: Based on the spatiotemporal indexing technology, the alignment and fusion of multi-source data are realized, and a four-dimensional data model including longitude, latitude, altitude and time dimensions is constructed (such as Figure 2 );

[0079] Finally, the alignment and fusion of multi-source data are achieved. Based on the spatiotemporal indexing technology, a four-dimensional data model is constructed. In the spatial dimension, the geospatial grid indexing technology is used to divide the study area into several grid cells, each of which contains information in three dimensions: longitude, latitude, and altitude; in the temporal dimension, a multi-scale time index structure is used to support data organization and query at different time granularities such as year, month, day, and hour. By establishing a spatiotemporal index, multi-source data at a specific spatiotemporal location can be quickly located and associated to achieve effective data fusion. For example, the equipment operation status and corresponding meteorological conditions of a substation in a specific time period can be easily queried.

[0080] S3: Based on the four-dimensional data model, a disaster risk assessment model is constructed using deep learning methods;

[0081] Based on the four-dimensional data model, a deep learning method is used to construct a disaster risk assessment model. Figure 3 ,The model adopts a modular design and contains three core sub-models:

[0082] S3.1: Use convolutional neural network to build a meteorological disaster type recognition model;

[0083] The meteorological disaster type recognition model adopts a convolutional neural network structure. The model organizes meteorological data into a spatiotemporal data cube, which contains multiple meteorological element channels. A three-dimensional convolution operation is used to extract spatiotemporal features, in which the convolution kernel slides simultaneously in the spatial and temporal dimensions, which can effectively capture the spatiotemporal evolution pattern of meteorological elements. The input of the model is meteorological data, including meteorological elements such as temperature, humidity, and wind speed, and the output is the probability distribution of different types of meteorological disasters (such as lightning, typhoons, heavy rains, etc.). For example, when the meteorological element data such as air pressure, wind speed, and temperature at multiple consecutive moments are input, the model can identify whether a typhoon disaster is likely to occur.

[0084] S3.2: Establish a power grid vulnerability assessment model using graph neural network;

[0085] The grid vulnerability assessment model uses a graph neural network structure. The assessment process is divided into the following stages:

[0086] S3.2.1 First, the grid topology is modeled as a graph structure, where substations, transmission lines and other equipment are nodes of the graph, and the physical connection relationship between the equipment is the edge of the graph. Each node contains the attribute characteristics of the equipment (such as equipment type, operating years, etc.) and state characteristics (such as load rate, temperature, etc.).

[0087] S3.2.2 then captures the interaction between nodes through the graph convolution layer and message passing mechanism. In specific implementation, each node updates its own state based on the feature information of its adjacent nodes, thereby realizing the propagation of information in the graph structure. This mechanism enables the model to consider the mutual influence and cascade effect between devices.

[0088] S3.2.3 then introduces the temporal attention layer to model the temporal dependencies of device states. This layer can adaptively learn the importance weights of state information at different time points and effectively capture the dynamic changes in device states. For example, it can identify abnormal state patterns of devices in different time periods.

[0089] S3.2.4 finally outputs the vulnerability score of each node through a multi-layer perceptron. This score comprehensively considers the characteristics of the device itself, the network topology relationship and the time series evolution characteristics, and can more accurately reflect the degree of vulnerability of the device in the face of meteorological disasters.

[0090] S3.3: Use Lipschitz neural network to build a disaster impact prediction model;

[0091] The disaster impact prediction model uses the Lipschitz neural network structure. The construction process of the model includes the following key steps:

[0092] S3.3.1: Design a multi-layer Lipschitz neural network structure and constrain the weight matrix of each layer of the network through spectral normalization technology;

[0093] Design a multi-layer Lipschitz neural network structure. Constrain the weight matrix of each layer of the network through spectral normalization technology to ensure that the network meets the Lipschitz continuity condition. Specifically, for each layer of the network, calculate the spectral norm of the weight matrix and control it within the preset threshold through normalization operation. This constraint mechanism can improve the robustness and generalization ability of the model, and make the prediction results have good stability to input disturbances. For example, even if there is slight noise in the input data, the prediction results of the model will not change drastically.

[0094] S3.3.2: Quantify the sensitivity of the network structure to different input variables through a quality assessment mechanism based on the extended Jacobian matrix;

[0095] The sensitivity of the network structure to different input variables is quantified through a quality assessment mechanism based on the extended Jacobian matrix. This mechanism calculates the partial derivatives of the model output with respect to each input variable and constructs an extended Jacobian matrix to analyze the response characteristics of the model to different input factors. This analysis can help understand the contribution of different meteorological factors and equipment characteristics to the impact of disasters, and provide a basis for the optimization and interpretation of the model.

[0096] S3.3.3: Design an adaptive loss function that includes a prediction error term, a Lipschitz constraint term, and a regularization term.

[0097] Design an adaptive loss function with multiple components. The loss function includes prediction error term, Lipschitz constraint term and regularization term. The prediction error term is used to measure the difference between the model prediction result and the actual observation value; the Lipschitz constraint term is used to ensure the Lipschitz characteristic of the model; and the regularization term is used to prevent the model from overfitting. These three terms are balanced by adaptive weight coefficients to ensure that the model achieves a good balance between prediction accuracy and generalization performance.

[0098] S3.3.4: Optimize the model based on the optimization strategy of adaptive loss function and gradient clipping;

[0099] The model is optimized using an optimization strategy based on adaptive loss function and gradient clipping. During the training process, the learning rate and gradient clipping threshold are dynamically adjusted to ensure that the model converges to the optimal solution. At the same time, an early stopping mechanism is introduced to prevent the model from overfitting the training data. Through this optimization strategy, the model can better learn the nonlinear characteristics of disaster impacts and provide more accurate prediction results.

[0100] Take the disaster impact prediction of a 330kV transmission line in typhoon weather as an example. The line crosses the coastal area and is vulnerable to typhoons. The Lipschitz neural network model receives the typhoon warning information output by the meteorological disaster type identification model (including the wind speed prediction value of 35m / s and the rainfall prediction value of 150mm / h) and the equipment vulnerability score output by the power grid vulnerability assessment model (the transmission tower structure stability score is 0.7 and the line insulation strength score is 0.8). The model first processes these input data through a multi-layer Lipschitz network structure, in which the weight matrix of each layer of the network is spectrally normalized to keep the Lipschitz constant less than 1. Then, the model calculates the possible affected range of the line (the predicted affected line length is 15 kilometers) and the degree of damage (the predicted trip probability is 85% and the equipment damage probability is 35%). These prediction results have good stability. Even if the input data fluctuates by ±5%, the change in the prediction results will not exceed 8%.

[0101] S4: Input real-time multi-source data into disaster risk assessment models for evaluation;

[0102] After the model is built, the real-time multi-source data is input into the disaster risk assessment model for evaluation. Monitoring data from various data sources is received in real time, including the latest meteorological observation data, equipment operation status data, etc. These real-time data are input into the trained disaster risk assessment model after going through the same standardized processing process as historical data. The model first identifies the potential types of meteorological disasters, then evaluates the vulnerability of power grid equipment, and finally predicts the possible scope and extent of disaster impact, thereby obtaining a comprehensive disaster risk assessment result.

[0103] Take the lightning disaster warning in a certain area as an example. Before a thunderstorm arrives, the following data are received in real time: the atmospheric electric field intensity monitored by the meteorological station reaches 4kV / m and continues to rise, the radar echo shows that the strong convective cloud is approaching, and there have been 5 cloud-to-ground lightning records in the surrounding area; at the same time, the power grid monitoring system shows that the operating current of the lightning arrester of 10 towers is abnormal, and the grounding resistance value of 2 substations has increased. These real-time data are input into the model after standardized processing. The meteorological disaster type identification model determines it as a strong thunderstorm (confidence 92%); the power grid vulnerability assessment model assesses 5 towers and 1 substation in the high-risk area based on the equipment status; the disaster impact prediction model gives a probability of 75% that the equipment in the target area will be struck by lightning in the next 2 hours, and the possible impact range covers about 20 square kilometers. These assessment results provide an important basis for subsequent warning decisions and emergency response.

[0104] S5: Make early warning decisions based on disaster risk assessment results and generate early warning information;

[0105] S5.1: Based on the disaster risk assessment results, a multi-level fuzzy comprehensive evaluation method is used to calculate the risk level;

[0106] S5.2: Use dynamic programming algorithms to optimize warning triggering and release conditions based on risk levels;

[0107] S5.2.1: Model the warning process as a multi-stage decision problem;

[0108] S5.2.2: Construct the state transition equation to calculate the optimal warning time and warning level corresponding to different risk levels;

[0109] S5.2.3: Through a reinforcement learning-based strategy, continuously optimize conditions by learning historical warning effect feedback;

[0110] S5.3: Input the risk level, warning trigger and release conditions into the WebGIS-based visual warning display system;

[0111] Specifically, in the above process, early warning decisions are made based on the disaster risk assessment results and early warning information is generated. This process includes the following main links:

[0112] The risk level is calculated using a multi-level fuzzy comprehensive evaluation method. This method establishes a multi-level evaluation index system, including factors such as the intensity of meteorological disasters, the degree of equipment vulnerability, and the possible impact range. Each indicator is quantified using a fuzzy membership function, and then the weight of each indicator is determined by the hierarchical analysis method. Finally, through fuzzy comprehensive operations, the evaluation results of each indicator are integrated into the final risk level. For example, the risk level is divided into four levels: low risk (green), medium risk (yellow), high risk (orange), and extremely high risk (red).

[0113] Use dynamic programming algorithms to optimize the warning triggering and release conditions. The process is divided into the following steps:

[0114] The warning process is modeled as a multi-stage decision-making problem. Each stage corresponds to a time point, at which point it is necessary to decide whether to issue a warning or cancel it. Multiple factors can be considered, such as the current risk level, historical warning effects, and warning costs. For example, before a typhoon arrives, it is necessary to decide whether to issue a warning and what level of warning to issue at different time points.

[0115] By constructing a state transition equation, the optimal warning time and warning level corresponding to different risk levels are calculated. The state transition equation describes the law of the system's transition from the current state to the next state, including factors such as the evolution trend of the risk level and the impact of the warning decision. By solving this equation, the optimal warning strategy under different risk levels can be obtained.

[0116] Introduce a strategy based on reinforcement learning to continuously optimize the warning conditions by analyzing the feedback of historical warning effects. Record the effect of each warning, including the accuracy and timeliness of the warning, and the actual losses after the warning. Through the reinforcement learning algorithm, it is possible to learn from these historical experiences, continuously adjust and optimize the warning strategy, and improve the accuracy and efficiency of the warning.

[0117] The risk level, warning trigger and release conditions are input into the visual warning display system based on WebGIS. The system adopts a layered design, with the bottom layer being the GIS map engine, the middle layer being the data processing and analysis module, and the top layer being the user interaction interface. The system can intuitively display the spatial distribution of warning information, support multiple layer overlay displays, and provide rich interactive operation functions. For example, different colors can be used to identify areas with different risk levels, flashing can be used to identify the location of the warning, and detailed warning information can be viewed by clicking.

[0118] S6: Generate emergency response plans based on early warning information, mainly including:

[0119] S6.1: Construct emergency plan knowledge graph;

[0120] S6.1.1: Use a multi-level semantic network structure to construct a knowledge graph;

[0121] S6.1.2: Extract key knowledge from historical emergency plan documents, accident handling reports, and expert experience;

[0122] S6.2: Generate emergency response procedures based on early warning information using graph reasoning and case matching algorithms;

[0123] S6.3: Calculate the optimal scheduling plan for emergency resources based on heuristic algorithms;

[0124] S6.4: Combine the emergency response process and the optimal scheduling plan to generate the final emergency response plan.

[0125] like Figure 4 , generating emergency response plans based on early warning information. This is a comprehensive decision-making process that includes the following key links:

[0126] The first step is to build a knowledge graph of emergency plans. This process includes two main steps:

[0127] A multi-level semantic network structure is used to construct the knowledge graph. The network structure includes a concept layer, a relationship layer, and an instance layer. The concept layer defines basic concepts such as disaster type, equipment type, and disposal measures; the relationship layer describes various semantic associations between concepts, such as "cause", "need", "include" and other relationships; the instance layer stores specific emergency plan cases. For example, the concept of "lightning disaster" is defined in the concept layer, the relationship "lightning disaster-cause-equipment failure" is defined in the relationship layer, and specific lightning disaster emergency disposal cases are stored in the instance layer.

[0128] Extract key knowledge from historical emergency plan documents, accident handling reports and expert experience. Use natural language processing technology to perform word segmentation, entity recognition and relationship extraction on text to convert unstructured text information into structured knowledge. At the same time, through expert interviews and experience summaries, supplement implicit knowledge that is difficult to obtain directly from documents. For example, you can extract empirical rules such as "When encountering a lightning disaster, the first thing to do is to check the grounding of the equipment" from historical handling reports.

[0129] Based on the warning information, the emergency response process is generated using graph reasoning and case matching algorithms. First, according to the disaster type and risk level in the warning information, a path search is performed in the knowledge graph to find out the relevant response measures and resource requirements. Then, the case matching algorithm is used to retrieve similar cases from the historical case library and refer to their response processes. Finally, through rule reasoning and case adaptation, a specific response process for the current situation is generated.

[0130] The optimal scheduling scheme for emergency resources is calculated based on the heuristic algorithm. According to the resource demand of the disposal process and the current distribution of resources, the improved ant colony algorithm or genetic algorithm is used to calculate the optimal resource scheduling scheme. This scheme needs to consider multiple constraints, such as resource quantity restrictions, transportation time requirements, cost control, etc. The algorithm finds a balance between multiple goals by designing a reasonable fitness function and constraint processing mechanism to obtain a practical and feasible scheduling scheme.

[0131] Finally, the emergency response process and the optimal scheduling plan are combined to generate the final emergency response plan. The plan is presented in the form of pictures and texts, including a disposal flow chart, a resource allocation table, key time nodes, and the division of responsibilities. At the same time, the plan also includes different levels of emergency response plans, which can be dynamically adjusted according to actual conditions. For example, when encountering heavy rainstorms, a complete disposal plan including equipment protection, personnel evacuation, emergency repairs, etc. will be generated, and the execution time, required resources, and responsible personnel of each link will be clarified.

[0132] The present application also provides a GIS-based intelligent monitoring and early warning device for power grid meteorological disasters, including:

[0133] A data processing module is used to obtain historical multi-source data and perform standardization processing to obtain a four-dimensional data model including longitude, latitude, altitude and time dimensions, wherein the historical multi-source data includes power grid equipment data, meteorological data and equipment operation data;

[0134] A model building module, used to build a disaster risk assessment model based on the four-dimensional data model using a deep learning method; the disaster risk assessment model includes a meteorological disaster type identification model, a power grid vulnerability assessment model and a disaster impact prediction model;

[0135] A risk assessment module, used to input real-time multi-source data into the disaster risk assessment model, identify the type of meteorological disaster based on the meteorological disaster type identification model, assess the equipment vulnerability based on the power grid vulnerability assessment model, predict the scope and extent of disaster impact based on the meteorological disaster type, the equipment vulnerability assessment result and the disaster impact prediction model, and obtain a disaster risk assessment result;

[0136] An early warning decision module is used to make early warning decisions based on the disaster risk assessment results and generate early warning information;

[0137] The emergency response module is used to generate an emergency response plan based on the warning information.

[0138] The present application also provides a computer device, the computer device comprising:

[0139] at least one processor; and,

[0140] a memory communicatively connected to the at least one processor; wherein,

[0141] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned GIS-based power grid meteorological disaster intelligent monitoring and early warning method.

[0142] An embodiment of the present application also provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the above-mentioned GIS-based power grid meteorological disaster intelligent monitoring and early warning method.

[0143] An embodiment of the present application also provides a computer program product, including computer instructions, characterized in that when the computer instructions are executed by a processor, the steps of the above-mentioned GIS-based power grid meteorological disaster intelligent monitoring and early warning method are implemented.

[0144] Based on the technical features described in the above specific implementation methods, the GIS-based intelligent monitoring and early warning device for power grid meteorological disasters of the present application is as follows: Figure 5As shown, it includes the following functional modules:

[0145] The data processing module is used to acquire and process historical multi-source data. The module is connected to the power grid SCADA system, meteorological monitoring system and equipment status monitoring system through a standard data interface, and adopts a multi-threaded parallel processing mechanism to ensure real-time data collection and efficient processing. The module integrates functional units such as data cleaning, standardization conversion and time-space alignment, which can convert heterogeneous data from different sources into a four-dimensional data model in a unified format.

[0146] The model building module is responsible for the construction and training of the disaster risk assessment model. This module adopts a modular design concept and decomposes the complex assessment task into three subtasks: meteorological disaster identification, equipment vulnerability assessment, and disaster impact prediction. Each subtask is handled by a dedicated deep learning model, which exchanges data and integrates results through predefined interfaces. The module also contains functional units for model training and verification, supporting incremental learning and model updates.

[0147] The risk assessment module performs real-time risk assessment tasks. The module receives pre-processed real-time data and distributes it to the corresponding assessment model. Parallel computing technology is used to improve processing efficiency and ensure that the assessment results can be output in a timely manner. The module also implements the visualization function of the assessment process, which facilitates operation and maintenance personnel to monitor the assessment status and intermediate results.

[0148] The early warning decision module is responsible for the generation and release of early warning information. The module implements the risk level calculation function based on multi-level fuzzy comprehensive evaluation and integrates dynamic programming algorithm to optimize early warning strategy. The module provides an intuitive early warning information display interface through the WebGIS platform and supports a variety of early warning information push methods, such as SMS, email, system notification, etc.

[0149] The emergency response module is used to intelligently generate emergency response plans. The module maintains a real-time updated knowledge graph database that stores various emergency plans and response experiences. The module uses a graph computing engine to achieve efficient knowledge retrieval and reasoning, and uses an optimization algorithm to calculate resource scheduling plans. The final generated response plan is presented through a visual interface, and supports dynamic adjustment of the plan and real-time feedback.

[0150] In summary, through the above technical solution, this application not only realizes the intelligence and automation of power grid meteorological disaster monitoring and early warning, but also significantly improves data processing efficiency, early warning accuracy and emergency response effect. The implementation of this solution has important practical significance for ensuring the safe and stable operation of the power grid and reducing the losses caused by meteorological disasters.

[0151] Next, the specific application scenarios of this application are described:

[0152] Scenario 1: Typhoon disaster warning scenario

[0153] During the typhoon season, meteorological monitoring data is obtained in real time through the data interface of the meteorological department. When it is detected that a tropical depression develops into a typhoon, the meteorological disaster type identification model promptly identifies the development trend of the typhoon. At the same time, the status of power grid equipment in the area that may be affected by the typhoon is analyzed, with special attention paid to the vulnerability of key facilities such as transmission lines and substations. For example, for transmission lines of 35kV and above in coastal areas, their wind resistance and structural stability are evaluated. Based on the evaluation results, graded warning information is generated, and disposal plans including line reinforcement, personnel evacuation, and emergency material allocation are formulated.

[0154] Scenario 2: Lightning Disaster Protection Scenario

[0155] In seasons with frequent thunderstorms, lightning activity is monitored in real time through weather radar and other equipment. When a severe convective weather system is identified, the model predicts the probability of occurrence and impact range of lightning disasters, pays special attention to the status of lightning protection facilities, and evaluates the effectiveness of substation lightning protection equipment. For example, when it is predicted that strong lightning activity will occur in a certain substation area, the operating status of the lightning arrester in the area is automatically checked, the reliability of the grounding system is evaluated, and an emergency plan including equipment protection, monitoring encryption and other measures is generated.

[0156] Scenario 3: Snow and ice disaster response scenario

[0157] In cold regions or in winter, focus on monitoring meteorological factors such as temperature, humidity, and precipitation. When an icing disaster is predicted, assess the icing risk of the transmission line. Through the grid vulnerability assessment model, calculate the load-bearing capacity of different types of transmission lines in different regions and predict the possible thickness of ice. Based on the assessment results, generate emergency response plans including deicing plans and load transfer strategies. For example, when it is predicted that a certain transmission line may be severely iced, plan the ice melting path in advance and deploy deicing equipment and personnel.

[0158] Scenario 4: Secondary disaster prevention scenario caused by heavy rain

[0159] During the flood season, we comprehensively evaluate the risks of secondary disasters such as landslides and loose foundations caused by heavy rains through multi-source data analysis. In particular, for power transmission lines and substations in mountainous areas, we combine geological data, historical disaster records and other information to assess the risk of equipment damage. For example, when a landslide is predicted to occur near a substation in a mountainous area, we issue a warning in a timely manner and generate a response plan including equipment transfer and emergency power supply plan.

[0160] The practice of these application scenarios shows that the monitoring and early warning method provided by this application can effectively respond to different types of meteorological disaster threats and provide all-round protection for the safe operation of the power grid. Through the intelligent analysis and early warning of the system, operation and maintenance personnel can take preventive measures in advance and greatly reduce the losses caused by meteorological disasters. At the same time, the self-learning ability enables the accuracy of early warning and the effectiveness of disposal plans to be continuously improved, providing strong support for the intelligent operation and maintenance of the power grid.

[0161] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A GIS-based intelligent monitoring and early warning method for power grid meteorological disasters, characterized in that: include: Acquiring historical multi-source data, wherein the historical multi-source data includes power grid equipment data, meteorological data, and equipment operation data; Standardizing the historical multi-source data to obtain a four-dimensional data model including longitude, latitude, altitude and time dimensions; Based on the four-dimensional data model, a disaster risk assessment model is constructed using a deep learning method, wherein the disaster risk assessment model includes a meteorological disaster type identification model, a power grid vulnerability assessment model, and a disaster impact prediction model; Inputting real-time multi-source data into the disaster risk assessment model, identifying the type of meteorological disaster based on the meteorological disaster type identification model, assessing the vulnerability of equipment based on the power grid vulnerability assessment model, and predicting the scope and extent of disaster impact based on the meteorological disaster type, the equipment vulnerability assessment results and the disaster impact prediction model to obtain a disaster risk assessment result; Make early warning decisions based on the disaster risk assessment results and generate early warning information; An emergency response plan is generated based on the warning information.

2. The method according to claim 1, characterized in that The standardization of historical multi-source data includes: Standardize the formats of data from different sources and establish a unified data description standard; Use statistically based outlier detection methods and time series interpolation algorithms for data cleaning and correction; Based on spatiotemporal indexing technology, the alignment and fusion of multi-source data are achieved, and a four-dimensional data model including longitude, latitude, altitude and time dimensions is constructed.

3. The method according to claim 1, characterized in that The method of constructing a disaster risk assessment model based on the four-dimensional data model using a deep learning method includes: Using a convolutional neural network to build a meteorological disaster type recognition model, the input of the meteorological disaster type recognition model is the meteorological data, and the output is the meteorological disaster type; A power grid vulnerability assessment model is established by using a graph neural network, wherein the input of the power grid vulnerability assessment model is the power grid equipment data and equipment operation data, and the output is the equipment vulnerability assessment result; A Lipschitz neural network is used to construct a disaster impact prediction model, the input of which is the type of meteorological disaster and the equipment vulnerability assessment results, and the output is the disaster risk assessment results including the scope and degree of disaster impact.

4. The method according to claim 3, characterized in that The Lipschitz neural network is used to construct a disaster impact prediction model, including: Design a multi-layer Lipschitz neural network structure and constrain the weight matrix of each layer of the network through spectral normalization technology; Quantify the sensitivity of multi-layer Lipschitz neural network structure to different input variables through a quality assessment mechanism based on the extended Jacobian matrix; Design an adaptive loss function including prediction error term, Lipschitz constraint term and regularization term; The disaster impact prediction model is optimized based on the optimization strategy of adaptive loss function and gradient clipping.

5. The method according to claim 3, characterized in that The method of establishing a power grid vulnerability assessment model using a graph neural network includes: The power grid topology structure in the power grid equipment data is modeled as a graph structure, where nodes represent equipment and edges represent line connection relationships; Capture the interaction relationship between nodes through graph convolutional layers and message passing mechanism; Introducing a temporal attention layer to model the temporal dependencies of device states; The vulnerability score of each node is output through a multi-layer perceptron.

6. The method according to claim 1, characterized in that The step of making a warning decision based on the disaster risk assessment result and generating warning information includes: Based on the disaster risk assessment results, a multi-level fuzzy comprehensive evaluation method is used to calculate the risk level; Based on the risk level, a dynamic programming algorithm is used to optimize the warning triggering and release conditions; The risk level, warning trigger and release conditions are input into a visual warning display system based on WebGIS to generate warning information.

7. The method according to claim 6, characterized in that The use of a dynamic programming algorithm to optimize the warning triggering and release conditions according to the risk level includes: Modeling the early warning process as a multi-stage decision-making problem; Constructing a state transfer equation, calculating the optimal warning time and warning level corresponding to different risk levels, and constructing warning triggering and release conditions for different risk levels according to the optimal warning time and warning level; Through a reinforcement learning-based strategy, the warning triggering and release conditions are continuously optimized by learning historical warning effect feedback.

8. The method according to claim 6, characterized in that The generating of an emergency response plan based on the early warning information includes: Construct a knowledge graph of emergency plans and establish the relationship between event types, disposal measures and resource requirements; Based on the risk level and warning triggering and release conditions in the warning information, a step-by-step emergency response process is generated using graph reasoning and case matching algorithms; According to the emergency response process, an optimal scheduling scheme for emergency resources is calculated based on a heuristic algorithm; The emergency response process and the optimal scheduling plan are combined to generate a final emergency response plan.

9. The method according to claim 8, characterized in that The construction of the emergency plan knowledge graph includes: Use multi-level semantic network structure to build knowledge graph; Extract key knowledge from historical emergency plan documents, incident handling reports and expert experience.

10. A GIS-based intelligent monitoring and early warning device for power grid meteorological disasters, characterized in that: include: A data processing module is used to obtain historical multi-source data and perform standardization processing to obtain a four-dimensional data model including longitude, latitude, altitude and time dimensions, wherein the historical multi-source data includes power grid equipment data, meteorological data and equipment operation data; A model building module, used to build a disaster risk assessment model based on the four-dimensional data model using a deep learning method; the disaster risk assessment model includes a meteorological disaster type identification model, a power grid vulnerability assessment model and a disaster impact prediction model; A risk assessment module, used to input real-time multi-source data into the disaster risk assessment model, identify the type of meteorological disaster based on the meteorological disaster type identification model, assess the equipment vulnerability based on the power grid vulnerability assessment model, predict the scope and extent of disaster impact based on the meteorological disaster type, the equipment vulnerability assessment result and the disaster impact prediction model, and obtain a disaster risk assessment result; An early warning decision module is used to make early warning decisions based on the disaster risk assessment results and generate early warning information; The emergency response module is used to generate an emergency response plan based on the warning information.

Images