A method and system for predicting meteorological disasters in power areas based on convolution algorithm

Abstract

The invention discloses a method and system for predicting meteorological disasters in a power area based on a convolution algorithm, which relates to the technical field of meteorological prediction, including determining meteorological data in an intersection area of ​​meteorological stations and an unknown area; dividing the power area into a plurality of sub-area maps according to the geographical features of the power area map, calculating the complexity level of the meteorological data in the sub-area map, and classifying the sub-area maps according to the complexity level and assigning a target grid size to each type of sub-area map; determining the impact attribute of each meteorological disaster to establish a meteorological disaster level evaluation standard; configuring model structure parameters according to the meteorological disaster level, and training a convolutional LSTM model; and predicting meteorological disasters using a meteorological disaster prediction model for a power area to achieve visual display of the prediction results. Training the meteorological disaster prediction model for a power area improves the accuracy and adaptability of meteorological disaster prediction for a power area, and ensures the complete and stable operation of the power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological forecasting, and in particular to a method and system for forecasting meteorological disasters in a power area based on a convolution algorithm. Background Art

[0002] With the significant changes in the global climate, extreme weather emergencies are becoming more frequent. Extreme weather conditions such as rainstorms, floods, continuous high temperatures, strong winds and thunderstorms continue to emerge, bringing huge challenges to people's daily lives and work. Against this background, the power system is facing unprecedented pressure. Extreme strong winds, rainstorms, ice and snow, typhoons and other meteorological disasters have caused serious damage to power facilities. High temperature, drought and extremely cold weather have aggravated the tension in power supply. Ice disasters have caused significant losses to companies and people's livelihood. The increase in extreme weather has made the fluctuation of power demand more unpredictable. Heavy rains may cause floods, damage power facilities, and cause power outages in some areas. High temperature weather will cause a large number of electrical equipment such as air conditioners to be activated, and the power load will rise sharply. Strong winds and thunderstorms may damage transmission lines and substations, affecting the stable supply of electricity, making the workload and difficulty of on-site emergency repairs for grassroots power employees continue to rise.

[0003] In view of the above complex and changeable situations, accurate extreme weather forecasting services are urgently needed to improve the monitoring, forecasting and early warning capabilities of meteorological disasters, support the power grid's disaster prevention and mitigation-related pre-disaster differential early warning, disaster tracking and disposal, and post-disaster scientific assessment, and assist grassroots units in making advance arrangements for resource allocation, equipment maintenance, and emergency preparedness, so as to improve the operational reliability of power grid equipment and further ensure the safe, stable and reliable operation of the power system.

[0004] Therefore, how to improve the accuracy and adaptability of meteorological disaster prediction in power areas is a technical problem that needs to be solved. Summary of the invention

[0005] The purpose of the present invention is to solve the problem of low accuracy of meteorological disaster prediction in power area and inability to adapt to changing meteorological conditions in the prior art, and to propose a method for predicting meteorological disasters in power area based on a convolution algorithm.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for predicting meteorological disasters in power areas based on a convolution algorithm, comprising:

[0008] Obtain meteorological data of the power area through the meteorological station, obtain the power area location data, construct a power area map through the power area location data, and mark the meteorological data at the corresponding position of the power area map according to the location of the meteorological data, find the intersection area and the unknown area of ​​the meteorological station in the power area map, determine the meteorological data in the intersection area and the unknown area of ​​the meteorological station, so as to complete the meteorological data of the power area map;

[0009] According to the geographical features of the power area map, the power area is divided into multiple sub-area maps, the complexity level of the meteorological data in the sub-area map is calculated, and the sub-area maps are classified according to the complexity level and the target grid size is assigned to each type of sub-area map. Each type of sub-area map is gridded to different degrees according to the target grid size to obtain a meteorological data set;

[0010] Collect historical meteorological data of each meteorological disaster on the power area map, determine the impact attributes of each meteorological disaster, establish a meteorological disaster level evaluation standard based on the impact attributes of each meteorological disaster, and evaluate the meteorological data set using the meteorological disaster level evaluation standard to obtain the meteorological disaster level;

[0011] The model structure parameters are configured according to the meteorological disaster level, and the convolutional LSTM model is trained according to the meteorological data set, meteorological disaster level, power area map location, and power-related data to obtain the power area meteorological disaster prediction model;

[0012] The meteorological disaster prediction model of the power area is used to predict meteorological disasters, output the prediction results, and map the prediction results to the corresponding positions of the power area map, so as to realize the visualization display of the prediction results.

[0013] In some embodiments of the present application, finding the intersection area of ​​the weather stations and the unknown area in the power area map, and determining the meteorological data in the intersection area of ​​the weather stations and the unknown area include:

[0014] Determine the meteorological collection range of each meteorological station on the power area map, take the intersection of the meteorological collection ranges between the meteorological stations as the intersection range, and determine the center position of the intersection range;

[0015] Calculate the distances from multiple meteorological stations with intersection ranges to the center of the intersection range respectively, assign different weights to the meteorological stations according to the distances, determine the meteorological data from different meteorological stations with different values ​​at the same location within the intersection range, perform weighted average on these meteorological data by the weights of the meteorological stations, and use the weighted average meteorological data as the meteorological data at the location, so as to determine the meteorological data of the intersection area of ​​the meteorological stations;

[0016] An unknown area is determined on the power area map, where the unknown area is an area where meteorological data cannot be obtained, a central position of the unknown area is determined, and a plurality of sample points are randomly selected near the central position;

[0017] The adjacent meteorological stations are determined by determining the critical meteorological collection range of the unknown area, and the covariance between the meteorological station and the center position of the unknown area and multiple sample points is calculated to obtain a comprehensive covariance. The comprehensive covariance is used to screen the adjacent meteorological stations, and the neural network prediction model of the screened meteorological stations is used to predict the meteorological forecast data of the unknown area. The prediction weight of each meteorological station is set according to the comprehensive covariance of the screened meteorological stations, and the meteorological data of the unknown area is determined by weighted averaging the meteorological forecast data and the prediction weight.

[0018] In some embodiments of the present application, the power area is divided into multiple sub-area maps according to the geographical features of the power area map, including:

[0019] The distribution of power facilities in the power area map is obtained, and multiple power unit areas on the power area map are obtained according to the distribution of power facilities. The geographical characteristics of each power unit area are analyzed by GIS software, and the corresponding adjacent distances are defined according to the type of power facilities. The similarity between power unit areas is calculated within the adjacent distance, and similar power unit areas are integrated into a sub-area map through similarity, thereby dividing the complete power area map into multiple sub-area maps.

[0020] In some embodiments of the present application, calculating the complexity level of meteorological data in the sub-region map includes:

[0021] According to the meteorological data in each sub-region map, the coefficient of variation, the correlation coefficient with other meteorological data and the degree of uncertainty of each meteorological data are calculated, and the relevant meteorological data are screened out through the correlation coefficient between each meteorological data and other meteorological data, which are recorded as relevant meteorological data;

[0022] Calculate the complexity level of meteorological data within the sub-region map based on the coefficient of variation, relevant meteorological data and uncertainty level;

[0023] ;in, is the complexity level of the meteorological data in the sub-region map, is the number of meteorological data types in the sub-region map, is the complex weight of the i-th meteorological data, is the coefficient of variation of the i-th meteorological data, is the number of types of meteorological data related to the i-th meteorological data, is the influence weight of the jth related meteorological data of the i-th meteorological data, is the correlation coefficient of the jth related meteorological data of the i-th meteorological data, is the uncertainty degree of the i-th meteorological data, , are the first constant and the second constant respectively, and [] is the rounding symbol.

[0024] In some embodiments of the present application, the sub-region graphs are classified according to the complexity level and a target grid size is assigned to each type of sub-region graph, including:

[0025] The sub-region maps of the same complexity level are planned into the same category, and the target grid size is obtained by mapping the grid size of the complexity level, so as to determine the grid size of each type of sub-region map.

[0026] In some embodiments of the present application, the impact attribute of each meteorological disaster is determined, and the meteorological disaster level evaluation standard is established based on the impact attribute of each meteorological disaster, including:

[0027] Impact attributes include disaster impact range, disaster impact time, and regional impact coefficient;

[0028] Determine the disaster impact scope and disaster impact events through historical meteorological data of each meteorological disaster;

[0029] The historical meteorological data of each meteorological disaster are used to calculate the relationship between the degree of disaster damage in different regions and time on the power area map, so as to obtain the regional impact coefficient;

[0030] The meteorological disaster level evaluation standard is established based on the disaster impact scope, disaster impact time and regional impact coefficient.

[0031] In some embodiments of the present application, the model structure parameters are configured according to the meteorological disaster level, including:

[0032] Determine the highest meteorological disaster level, and configure the model structure parameter group according to the highest meteorological disaster level to determine the convolutional LSTM deep learning network model structure.

[0033] In some embodiments of the present application, a meteorological disaster prediction model for a power area is used to predict meteorological disasters, and the prediction results are output, including:

[0034] The prediction result is the meteorological forecast value. Different meteorological forecast value ranges correspond to different meteorological disaster levels, and corresponding warnings are issued according to the meteorological disaster levels.

[0035] Correspondingly, the present application also provides a power area meteorological disaster prediction system based on a convolution algorithm, comprising:

[0036] A completion module is used to obtain meteorological data of the power area through the meteorological station, obtain the power area location data, construct a power area map through the power area location data, and mark the meteorological data at the corresponding position of the power area map according to the location of the meteorological data, find the intersection area and the unknown area of ​​the meteorological station in the power area map, determine the meteorological data in the intersection area and the unknown area of ​​the meteorological station, so as to complete the meteorological data of the power area map;

[0037] A grid module is used to divide the power area into multiple sub-area maps according to the geographical features of the power area map, calculate the complexity level of the meteorological data in the sub-area map, classify the sub-area maps according to the complexity level, and assign a target grid size to each type of sub-area map. Each type of sub-area map is gridded to different degrees according to the target grid size to obtain a meteorological data set;

[0038] An evaluation module is used to collect historical meteorological data of each meteorological disaster on the power area map, determine the impact attributes of each meteorological disaster, establish a meteorological disaster level evaluation standard based on the impact attributes of each meteorological disaster, and evaluate the meteorological data set using the meteorological disaster level evaluation standard to obtain the meteorological disaster level;

[0039] The prediction module is used to configure the model structure parameters according to the meteorological disaster level, and train the convolutional LSTM model according to the meteorological data set, meteorological disaster level, power area map location, and power-related data to obtain the power area meteorological disaster prediction model;

[0040] The visual module is used to predict meteorological disasters using the power area meteorological disaster prediction model, output the prediction results, and map the prediction results to the corresponding positions of the power area map, thereby realizing the visual display of the prediction results.

[0041] By applying the above technical scheme, meteorological data of the power area is obtained through the meteorological station, the power area location data is obtained, the power area map is constructed through the power area location data, and the meteorological data is marked at the corresponding position of the power area map according to the location of the meteorological data, the intersection area and the unknown area of ​​the meteorological station are found in the power area map, and the meteorological data in the intersection area and the unknown area of ​​the meteorological station are determined to complete the meteorological data of the power area map; the power area is divided into multiple sub-area maps according to the geographical characteristics of the power area map, the complexity level of the meteorological data in the sub-area map is calculated, and the sub-area maps are classified according to the complexity level and a target grid size is assigned to each type of sub-area map, and each type of sub-area map is classified to different degrees according to the target grid size Grid processing to obtain a meteorological data set; collect historical meteorological data of each meteorological disaster on the power area map, determine the impact attributes of each meteorological disaster, establish a meteorological disaster level evaluation standard through the impact attributes of each meteorological disaster, and evaluate the meteorological data set through the meteorological disaster level evaluation standard to obtain the meteorological disaster level; configure the model structure parameters through the meteorological disaster level, and train the convolutional LSTM model according to the meteorological data set, meteorological disaster level, power area map location, and power-related data to obtain the power area meteorological disaster prediction model; use the power area meteorological disaster prediction model to predict meteorological disasters, output the prediction results, and map the prediction results to the corresponding position of the power area map, so as to achieve a visual display of the prediction results. This application completes the meteorological data of the power area map by determining the meteorological data in the intersection area of ​​the meteorological station and the unknown area, so that the meteorological data in the power area map can be improved and the reliability of the meteorological data can be improved. According to the complexity level, the sub-area map is classified and the target grid size of each type of sub-area map is assigned, and different areas are gridded to different degrees to achieve non-uniform gridding, so as to better fit the complex situation of the variable power area. The convolutional LSTM model is trained to obtain the power area meteorological disaster prediction model, which improves the accuracy and adaptability of the power area meteorological disaster prediction and ensures the completely stable operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic diagram of a flow chart of a method for predicting meteorological disasters in power areas based on a convolution algorithm proposed by the present invention;

[0043] Figure 2 This is a structural schematic diagram of a power area meteorological disaster prediction system based on a convolution algorithm proposed by the present invention. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0045] Reference Figure 1 , a method for predicting meteorological disasters in power areas based on a convolution algorithm, comprising the following steps:

[0046] Step S101, obtain the meteorological data of the power area through the meteorological station, obtain the power area location data, construct the power area map through the power area location data, and mark the meteorological data at the corresponding position of the power area map according to the location of the meteorological data, find the intersection area and the unknown area of ​​the meteorological station in the power area map, determine the meteorological data in the intersection area and the unknown area of ​​the meteorological station, so as to complete the meteorological data of the power area map.

[0047] In this embodiment, the meteorological data includes time, precipitation, wind speed, wind direction, temperature, humidity, air pressure visibility, number of lightning and radiation, etc. The power area location data is the GIS coordinates of the power area. The meteorological data here are generally collected by meteorological stations in the power area. The ranges collected by meteorological stations in different locations are different, and there may be intersections, that is, the meteorological data collected by meteorological stations at the same location are different, and there may be certain errors. There may be some places on the power area map where the meteorological station cannot accurately collect meteorological data, which exceeds the collection range. Therefore, it is necessary to determine the meteorological data in the intersection area and unknown area of ​​the meteorological station, complete the meteorological data of the power area map, and improve the meteorological data in the area.

[0048] In some embodiments of the present application, finding the intersection area of ​​the weather stations and the unknown area in the power area map, and determining the meteorological data in the intersection area of ​​the weather stations and the unknown area include:

[0049] Determine the meteorological collection range of each meteorological station on the power area map, take the intersection of the meteorological collection ranges between the meteorological stations as the intersection range, and determine the center position of the intersection range;

[0050] Calculate the distances from multiple meteorological stations with intersection ranges to the center of the intersection range respectively, assign different weights to the meteorological stations according to the distances, determine the meteorological data from different meteorological stations with different values ​​at the same location within the intersection range, perform weighted average on these meteorological data by the weights of the meteorological stations, and use the weighted average meteorological data as the meteorological data at the location, so as to determine the meteorological data of the intersection area of ​​the meteorological stations;

[0051] An unknown area is determined on the power area map, where the unknown area is an area where meteorological data cannot be obtained, a central position of the unknown area is determined, and a plurality of sample points are randomly selected near the central position;

[0052] The adjacent meteorological stations are determined by determining the critical meteorological collection range of the unknown area, and the covariance between the meteorological station and the center position of the unknown area and multiple sample points is calculated to obtain a comprehensive covariance. The comprehensive covariance is used to screen the adjacent meteorological stations, and the neural network prediction model of the screened meteorological stations is used to predict the meteorological forecast data of the unknown area. The prediction weight of each meteorological station is set according to the comprehensive covariance of the screened meteorological stations, and the meteorological data of the unknown area is determined by weighted averaging the meteorological forecast data and the prediction weight.

[0053] In this embodiment, for the intersection area of ​​meteorological stations, the meteorological data is confirmed by the distance from the meteorological station to the center of the intersection range, and the distance weighted average method is used. For unknown areas, the spatial correlation between the meteorological station and the center of the unknown area and the covariance between multiple sample points is used to describe the spatial correlation between the two. Meteorological stations with strong spatial correlation and adjacent to each other are screened out, and prediction weights are assigned according to the spatial correlation. The data of known meteorological stations are learned using a neural network, and a relationship model between meteorological data and spatial position is established, and meteorological data of unknown areas is predicted. Finally, the meteorological forecast data predicted by the model is weighted averaged (multiple sets of meteorological forecast data) to determine the meteorological data of the unknown area.

[0054] Step S102: divide the power area into multiple sub-area maps according to the geographical features of the power area map, calculate the complexity level of the meteorological data in the sub-area map, classify the sub-area maps according to the complexity level and assign a target grid size to each type of sub-area map. Grid processing is performed on each type of sub-area map to different degrees according to the target grid size to obtain a meteorological data set.

[0055] In this embodiment, similar and adjacent geographical features are classified to divide the sub-region map. The grid size is set by judging the complexity of the meteorological data of the sub-region (variation, uncertainty, correlation, etc.). Different grid sizes can be used in different locations in the power region. This grid division method is called "non-uniform grid division", which allows the size and shape of the grid to be adjusted according to different meteorological conditions in the power region. This method can more accurately reflect the characteristics and changes of different regions in the power region, thereby improving the accuracy of meteorological disaster prediction.

[0056] In some embodiments of the present application, the power area is divided into multiple sub-area maps according to the geographical features of the power area map, including:

[0057] The distribution of power facilities in the power area map is obtained, and multiple power unit areas on the power area map are obtained according to the distribution of power facilities. The geographical characteristics of each power unit area are analyzed by GIS software, and the corresponding adjacent distances are defined according to the type of power facilities. The similarity between power unit areas is calculated within the adjacent distance, and similar power unit areas are integrated into a sub-area map through similarity, thereby dividing the complete power area map into multiple sub-area maps.

[0058] In this embodiment, the power facilities include power substations, transmission lines, power plants and other power facilities. Multiple power unit areas are divided according to the distribution. Geographical features include topographical features such as mountains, rivers, and lakes. The corresponding adjacent distance is defined according to the type of power facilities. The distance here refers to the number of power unit areas. Different types of power equipment have different adjacent distances. The similarity between power unit areas is calculated within the adjacent distance, and the similarity of geographical features between different power unit areas is calculated. The power unit areas whose similarity exceeds the threshold are integrated into a sub-area map.

[0059] In some embodiments of the present application, calculating the complexity level of meteorological data in the sub-region map includes:

[0060] According to the meteorological data in each sub-region map, the coefficient of variation, the correlation coefficient with other meteorological data and the degree of uncertainty of each meteorological data are calculated, and the relevant meteorological data are screened out through the correlation coefficient between each meteorological data and other meteorological data, which are recorded as relevant meteorological data;

[0061] Calculate the complexity level of meteorological data within the sub-region map based on the coefficient of variation, relevant meteorological data and uncertainty level;

[0062] ;in, is the complexity level of the meteorological data in the sub-region map, is the number of meteorological data types in the sub-region map, is the complex weight of the i-th meteorological data, is the coefficient of variation of the i-th meteorological data, is the number of types of meteorological data related to the i-th meteorological data, is the influence weight of the jth related meteorological data of the i-th meteorological data, is the correlation coefficient of the jth related meteorological data of the i-th meteorological data, is the uncertainty degree of the i-th meteorological data, , are the first constant and the second constant respectively, and [] is the rounding symbol.

[0063] In this embodiment, the coefficient of variation, the correlation coefficient with other meteorological data and the degree of uncertainty of each meteorological data are calculated based on the meteorological data in each sub-area map. Here, the correlation coefficient with other meteorological data can be expressed by the Pierce correlation coefficient, and the degree of uncertainty (uncertainty of the data, including prediction error, uncertainty range, etc.) can be obtained by the Bayesian method. The Bayesian framework is used to quantify uncertainty, and the uncertainty of each parameter is estimated by combining prior knowledge and observation data.

[0064] It should be noted that the above-mentioned coefficient of variation, correlation coefficient and degree of uncertainty can be calculated by the above-mentioned method or by other methods. The specific method is not limited to this and all belong to the protection scope of this application.

[0065] In some embodiments of the present application, the sub-region graphs are classified according to the complexity level and a target grid size is assigned to each type of sub-region graph, including:

[0066] The sub-region maps of the same complexity level are planned into the same category, and the target grid size is obtained by mapping the grid size of the complexity level, so as to determine the grid size of each type of sub-region map.

[0067] In this embodiment, different complexity levels correspond to different target grid sizes. In areas with high meteorological data complexity, a smaller grid size is used to improve prediction accuracy. In areas with low meteorological data complexity, a larger grid size is used to reduce computing resource consumption.

[0068] Step S103, collect historical meteorological data of each meteorological disaster on the power area map, determine the impact attribute of each meteorological disaster, establish a meteorological disaster level evaluation standard based on the impact attribute of each meteorological disaster, and evaluate the meteorological data set using the meteorological disaster level evaluation standard to obtain the meteorological disaster level.

[0069] In some embodiments of the present application, the impact attribute of each meteorological disaster is determined, and the meteorological disaster level evaluation standard is established based on the impact attribute of each meteorological disaster, including:

[0070] Impact attributes include disaster impact range, disaster impact time, and regional impact coefficient;

[0071] Determine the disaster impact scope and disaster impact events through historical meteorological data of each meteorological disaster;

[0072] The historical meteorological data of each meteorological disaster are used to calculate the relationship between the degree of disaster damage in different regions and time on the power area map, so as to obtain the regional impact coefficient;

[0073] The meteorological disaster level evaluation standard is established based on the disaster impact scope, disaster impact time and regional impact coefficient.

[0074] In this embodiment, the regional impact coefficient refers to an evaluation coefficient obtained by the interaction between the time information of a region and the historical disaster hazard level of the region. The regional coefficient reflects the sensitivity and vulnerability of the region to meteorological disasters.

[0075] In this embodiment, the meteorological disaster level evaluation standard is established according to the disaster impact range, disaster impact time and regional impact coefficient;

[0076] ;in, For the The meteorological disaster level of each meteorological disaster is For the The number of areas affected by each meteorological disaster, For the The impact of meteorological disasters on The regional impact coefficient of a region, For the The impact range of meteorological disasters For the The impact time of various meteorological disasters.

[0077] Step S104, configure the model structure parameters according to the meteorological disaster level, and train the convolutional LSTM model according to the meteorological data set, meteorological disaster level, power area map location, and power-related data to obtain a power area meteorological disaster prediction model.

[0078] In this embodiment, the convolutional LSTM deep learning network model is composed of a CNN convolutional layer, a convolutional LSTM network layer, a pooling layer, a fully connected layer and an output layer connected in sequence, wherein: the CNN convolutional layer uses a VggNet-16 network module, including a Max pooling layer and a ReLu activation layer; the LSTM network layer includes an input gate, a forget gate, a memory cell, an output gate and a hidden state LSTM unit, and learns information data of different lengths in a spatial sequence.

[0079] In some embodiments of the present application, the model structure parameters are configured according to the meteorological disaster level, including:

[0080] Determine the highest meteorological disaster level, and configure the model structure parameter group according to the highest meteorological disaster level to determine the convolutional LSTM deep learning network model structure.

[0081] In this embodiment, generally speaking, the higher the meteorological disaster level, the more complex the model structure may be required, such as increasing the number of convolutional layers and LSTM layers. In order to take into account all levels of meteorological disaster levels, the model structure adapted to the highest meteorological disaster level is adopted. The model structure parameter group includes the number of convolutional layers, the number of LSTM layers, the learning rate, and the hyperparameters. Each meteorological disaster level corresponds to a set of model structure parameter groups, and the parameters between the model structure parameter groups can be adapted to each other.

[0082] Step S105, using the power area meteorological disaster prediction model to predict meteorological disasters, output the prediction results, and map the prediction results to the corresponding positions of the power area map, so as to achieve visual display of the prediction results.

[0083] In some embodiments of the present application, a meteorological disaster prediction model for a power area is used to predict meteorological disasters, and the prediction results are output, including:

[0084] The prediction result is the meteorological forecast value. Different meteorological forecast value ranges correspond to different meteorological disaster levels, and corresponding warnings are issued according to the meteorological disaster levels.

[0085] By applying the above technical scheme, meteorological data of the power area is obtained through the meteorological station, the power area location data is obtained, the power area map is constructed through the power area location data, and the meteorological data is marked at the corresponding position of the power area map according to the location of the meteorological data, the intersection area and the unknown area of ​​the meteorological station are found in the power area map, and the meteorological data in the intersection area and the unknown area of ​​the meteorological station are determined to complete the meteorological data of the power area map; the power area is divided into multiple sub-area maps according to the geographical characteristics of the power area map, the complexity level of the meteorological data in the sub-area map is calculated, and the sub-area maps are classified according to the complexity level and a target grid size is assigned to each type of sub-area map, and each type of sub-area map is classified to different degrees according to the target grid size Grid processing to obtain a meteorological data set; collect historical meteorological data of each meteorological disaster on the power area map, determine the impact attributes of each meteorological disaster, establish a meteorological disaster level evaluation standard through the impact attributes of each meteorological disaster, and evaluate the meteorological data set through the meteorological disaster level evaluation standard to obtain the meteorological disaster level; configure the model structure parameters through the meteorological disaster level, and train the convolutional LSTM model according to the meteorological data set, meteorological disaster level, power area map location, and power-related data to obtain the power area meteorological disaster prediction model; use the power area meteorological disaster prediction model to predict meteorological disasters, output the prediction results, and map the prediction results to the corresponding position of the power area map, so as to achieve a visual display of the prediction results. This application completes the meteorological data of the power area map by determining the meteorological data in the intersection area of ​​the meteorological station and the unknown area, so that the meteorological data in the power area map can be improved and the reliability of the meteorological data can be improved. According to the complexity level, the sub-area map is classified and the target grid size of each type of sub-area map is assigned, and different areas are gridded to different degrees to achieve non-uniform gridding, so as to better fit the complex situation of the variable power area. The convolutional LSTM model is trained to obtain the power area meteorological disaster prediction model, which improves the accuracy and adaptability of the power area meteorological disaster prediction and ensures the completely stable operation of the power grid.

[0086] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present invention can be implemented by hardware, or by software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present invention.

[0087] Correspondingly, the present application also provides a power area meteorological disaster prediction system based on a convolution algorithm, such as Figure 2 As shown, including:

[0088] A completion module is used to obtain meteorological data of the power area through the meteorological station, obtain the power area location data, construct a power area map through the power area location data, and mark the meteorological data at the corresponding position of the power area map according to the location of the meteorological data, find the intersection area and the unknown area of ​​the meteorological station in the power area map, determine the meteorological data in the intersection area and the unknown area of ​​the meteorological station, so as to complete the meteorological data of the power area map;

[0089] A grid module is used to divide the power area into multiple sub-area maps according to the geographical features of the power area map, calculate the complexity level of the meteorological data in the sub-area map, classify the sub-area maps according to the complexity level, and assign a target grid size to each type of sub-area map. Each type of sub-area map is gridded to different degrees according to the target grid size to obtain a meteorological data set;

[0090] An evaluation module is used to collect historical meteorological data of each meteorological disaster on the power area map, determine the impact attributes of each meteorological disaster, establish a meteorological disaster level evaluation standard based on the impact attributes of each meteorological disaster, and evaluate the meteorological data set using the meteorological disaster level evaluation standard to obtain the meteorological disaster level;

[0091] The prediction module is used to configure the model structure parameters according to the meteorological disaster level, and train the convolutional LSTM model according to the meteorological data set, meteorological disaster level, power area map location, and power-related data to obtain the power area meteorological disaster prediction model;

[0092] The visual module is used to predict meteorological disasters using the power area meteorological disaster prediction model, output the prediction results, and map the prediction results to the corresponding positions of the power area map, thereby realizing the visual display of the prediction results.

[0093] Those skilled in the art will appreciate that the modules in the system in the implementation scenario can be distributed in the system of the implementation scenario according to the implementation scenario description, or can be changed accordingly and located in one or more systems different from the implementation scenario. The modules in the above implementation scenario can be combined into one module, or can be further split into multiple submodules.

[0094] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for predicting meteorological disasters in power areas based on a convolution algorithm, characterized in that: include: Obtain meteorological data of the power area through the meteorological station, obtain the power area location data, construct a power area map through the power area location data, and mark the meteorological data at the corresponding position of the power area map according to the location of the meteorological data, find the intersection area and the unknown area of ​​the meteorological station in the power area map, determine the meteorological data in the intersection area and the unknown area of ​​the meteorological station, so as to complete the meteorological data of the power area map; According to the geographical features of the power area map, the power area is divided into multiple sub-area maps, the complexity level of the meteorological data in the sub-area map is calculated, and the sub-area maps are classified according to the complexity level and the target grid size is assigned to each type of sub-area map. Each type of sub-area map is gridded to different degrees according to the target grid size to obtain a meteorological data set; Collect historical meteorological data of each meteorological disaster on the power area map, determine the impact attributes of each meteorological disaster, establish a meteorological disaster level evaluation standard based on the impact attributes of each meteorological disaster, and evaluate the meteorological data set using the meteorological disaster level evaluation standard to obtain the meteorological disaster level; The model structure parameters are configured according to the meteorological disaster level, and the convolutional LSTM model is trained according to the meteorological data set, meteorological disaster level, power area map location, and power-related data to obtain the power area meteorological disaster prediction model; Use the power area meteorological disaster prediction model to predict meteorological disasters, output the prediction results, and map the prediction results to the corresponding positions of the power area map, so as to realize the visualization display of the prediction results; in, Find the intersection area and unknown area of ​​the meteorological stations in the power area map, and determine the meteorological data in the intersection area and unknown area of ​​the meteorological stations, including: Determine the meteorological collection range of each meteorological station on the power area map, take the intersection of the meteorological collection ranges between the meteorological stations as the intersection range, and determine the center position of the intersection range; Calculate the distances from multiple meteorological stations with intersection ranges to the center of the intersection range respectively, assign different weights to the meteorological stations according to the distances, determine the meteorological data from different meteorological stations with different values ​​at the same location within the intersection range, perform weighted average on these meteorological data by the weights of the meteorological stations, and use the weighted average meteorological data as the meteorological data at the location, so as to determine the meteorological data of the intersection area of ​​the meteorological stations; An unknown area is determined on the power area map, where the unknown area is an area where meteorological data cannot be obtained, a central position of the unknown area is determined, and a plurality of sample points are randomly selected near the central position; By determining the critical meteorological collection range of the unknown area, the adjacent meteorological station is determined, and the covariance between the meteorological station and the center position of the unknown area and multiple sample points is calculated to obtain a comprehensive covariance. The comprehensive covariance is used to screen the adjacent meteorological stations, and the meteorological forecast data of the unknown area is predicted according to the neural network prediction model of the screened meteorological stations. The prediction weight of each meteorological station is set according to the comprehensive covariance of the screened meteorological stations, and the meteorological data of the unknown area is determined by weighted averaging the meteorological forecast data and the prediction weight. The complexity level of calculating the meteorological data within the sub-region map includes: According to the meteorological data in each sub-region map, the coefficient of variation, the correlation coefficient with other meteorological data and the degree of uncertainty of each meteorological data are calculated, and the relevant meteorological data are screened out through the correlation coefficient between each meteorological data and other meteorological data, which are recorded as relevant meteorological data; Calculate the complexity level of meteorological data within the sub-region map based on the coefficient of variation, relevant meteorological data and uncertainty level; ;in, is the complexity level of the meteorological data in the sub-region map, is the number of meteorological data types in the sub-region map, is the complex weight of the i-th meteorological data, is the coefficient of variation of the i-th meteorological data, is the number of types of meteorological data related to the i-th meteorological data, is the influence weight of the jth related meteorological data of the i-th meteorological data, is the correlation coefficient of the jth related meteorological data of the i-th meteorological data, is the uncertainty degree of the i-th meteorological data, , are the first constant and the second constant respectively, and [] is the rounding symbol.

2. The method for predicting meteorological disasters in power areas based on a convolution algorithm according to claim 1 is characterized in that: The power region is divided into multiple sub-region maps according to the geographical characteristics of the power region map, including: The distribution of power facilities in the power area map is obtained, and multiple power unit areas on the power area map are obtained according to the distribution of power facilities. The geographical characteristics of each power unit area are analyzed by GIS software, and the corresponding adjacent distances are defined according to the type of power facilities. The similarity between power unit areas is calculated within the adjacent distance, and similar power unit areas are integrated into a sub-area map through similarity, thereby dividing the complete power area map into multiple sub-area maps.

3. The method for predicting meteorological disasters in power areas based on a convolution algorithm according to claim 1 is characterized in that: The sub-region maps are classified according to the complexity level and the target grid size is assigned to each type of sub-region map, including: The sub-region maps of the same complexity level are planned into the same category, and the target grid size is obtained by mapping the grid size of the complexity level, so as to determine the grid size of each type of sub-region map.

4. The method for predicting meteorological disasters in power areas based on a convolution algorithm according to claim 1 is characterized in that: Determine the impact attributes of each meteorological disaster, and establish meteorological disaster level evaluation standards based on the impact attributes of each meteorological disaster, including: Impact attributes include disaster impact range, disaster impact time, and regional impact coefficient; Determine the disaster impact scope and disaster impact events through historical meteorological data of each meteorological disaster; The historical meteorological data of each meteorological disaster are used to calculate the relationship between the degree of disaster damage in different regions and time on the power area map, so as to obtain the regional impact coefficient; The meteorological disaster level evaluation standard is established based on the disaster impact scope, disaster impact time and regional impact coefficient.

5. The method for predicting meteorological disasters in power areas based on a convolution algorithm according to claim 1 is characterized in that: The model structure parameters are configured according to the meteorological disaster level, including: Determine the highest meteorological disaster level, and configure the model structure parameter group according to the highest meteorological disaster level to determine the convolutional LSTM deep learning network model structure.

6. The method for predicting meteorological disasters in power areas based on a convolution algorithm according to claim 1 is characterized in that: Use the power area meteorological disaster prediction model to predict meteorological disasters and output the prediction results, including: The prediction result is the meteorological forecast value. Different meteorological forecast value ranges correspond to different meteorological disaster levels, and corresponding warnings are issued according to the meteorological disaster levels.

7. A system for predicting meteorological disasters in power areas based on a convolution algorithm, used to implement the method for predicting meteorological disasters in power areas based on a convolution algorithm as claimed in any one of claims 1 to 6, characterized in that: include: A completion module is used to obtain meteorological data of the power area through the meteorological station, obtain the power area location data, construct a power area map through the power area location data, and mark the meteorological data at the corresponding position of the power area map according to the location of the meteorological data, find the intersection area and the unknown area of ​​the meteorological station in the power area map, determine the meteorological data in the intersection area and the unknown area of ​​the meteorological station, so as to complete the meteorological data of the power area map; A grid module is used to divide the power area into multiple sub-area maps according to the geographical features of the power area map, calculate the complexity level of the meteorological data in the sub-area map, classify the sub-area maps according to the complexity level, and assign a target grid size to each type of sub-area map. Each type of sub-area map is gridded to different degrees according to the target grid size to obtain a meteorological data set; An evaluation module is used to collect historical meteorological data of each meteorological disaster on the power area map, determine the impact attributes of each meteorological disaster, establish a meteorological disaster level evaluation standard based on the impact attributes of each meteorological disaster, and evaluate the meteorological data set using the meteorological disaster level evaluation standard to obtain the meteorological disaster level; The prediction module is used to configure the model structure parameters according to the meteorological disaster level, and train the convolutional LSTM model according to the meteorological data set, meteorological disaster level, power area map location, and power-related data to obtain the power area meteorological disaster prediction model; The visual module is used to predict meteorological disasters using the power area meteorological disaster prediction model, output the prediction results, and map the prediction results to the corresponding positions of the power area map, thereby realizing the visual display of the prediction results.

Images