Power distribution network satellite meteorological data analysis method, system, equipment and medium
By fusing multi-source satellite data and using deep learning models, combined with the power grid topology, a dynamic prediction and intelligent linkage mechanism was established. This solved the problem of multi-source data fusion and intelligent decision-making in the meteorological disaster early warning system of the distribution network, achieving high-precision disaster prediction and accurate risk assessment, and improving the defense capability and emergency response efficiency of the distribution network.
Patent Information
- Application Number
- CN202510879564.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-27
- Publication Date
- 2025-11-28
AI Technical Summary
Existing meteorological disaster early warning systems for power distribution networks suffer from bottlenecks in multi-source data fusion and intelligent decision-making, making it difficult to balance high-frequency monitoring with high-precision coverage. The accuracy of disaster feature extraction is limited, the prediction model lacks the ability to dynamically model the spatiotemporal evolution, the risk assessment ignores the power grid topology and equipment mechanical characteristics, the emergency response plan relies on human experience, and there is a lack of intelligent linkage mechanisms.
By efficiently fusing and accurately modeling multi-source satellite meteorological data, and using a combination of convolutional neural networks and recurrent neural networks to dynamically model disaster-related characteristics, a hierarchical early warning and emergency response mechanism linked to the power grid topology is established, enabling tower overturning risk assessment and the generation of three-dimensional risk maps.
It has improved the accuracy and reliability of meteorological disaster forecasting, reduced the misjudgment rate of tower collapse and line breakage risks, and realized the active defense capability and emergency response efficiency of the power distribution network under extreme weather conditions.
Smart Images

Figure CN121034018A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of power distribution network, and particularly relates to a power distribution network satellite meteorological data analysis method, system, device and medium. BACKGROUND
[0002] The current power distribution network meteorological disaster early warning system generally faces the bottleneck of multi-source data fusion and intelligent decision-making: the traditional method relies on single satellite data or ground observation stations, and it is difficult to balance high-frequency monitoring and high-precision coverage, resulting in time and space blind area in typhoon path tracking and ice growth prediction; the existing data fusion technology cannot effectively solve the difference in time and space resolution between stationary satellites and polar orbit satellites and cloud noise interference, and the disaster feature extraction accuracy is limited; the disaster prediction model mostly uses static threshold or single network structure, lacks dynamic modeling ability of time and space evolution law, and it is difficult to quantify the confidence interval of the prediction result; the risk assessment link ignores the dynamic coupling of power grid topology and equipment mechanical properties, resulting in high risk misjudgment rate of tower collapse and wire breakage; the emergency disposal scheme relies on artificial experience, and the early warning signal and resource scheduling lack intelligent linkage mechanism, and it is difficult to realize graded and accurate prevention and control. The existing technology has not formed a whole-chain solution of "data fusion-dynamic prediction-topology assessment-emergency linkage", which restricts the active defense capability of the power distribution network under extreme weather. SUMMARY
[0003] In view of the above existing problems, the present application is proposed.
[0004] Therefore, the present application provides a power distribution network satellite meteorological data analysis method, system, device and medium to solve the problems of efficient fusion and accurate modeling of multi-source satellite meteorological data, improve the dynamic prediction capability of the power distribution network to typhoon, icing, rainstorm disasters, and establish a graded early warning and emergency response mechanism linked with the power grid topology.
[0005] To solve the above technical problems, the present application provides the following technical solutions:
[0006] In a first aspect, the present application provides a power distribution network satellite meteorological data analysis method, comprising the following steps:
[0007] Obtain multi-source satellite meteorological data and preprocess to generate a time and space continuous meteorological element distribution map;
[0008] Extract disaster-related features from the meteorological element distribution map;
[0009] Construct a first calculation model to train the disaster-related features and output a disaster distribution confidence interval;
[0010] Map the disaster distribution confidence interval and the power grid topology to generate a tower overturning risk assessment result;
[0011] A multi-level early warning mechanism is established by the tower overturning risk assessment result, a three-dimensional risk map is generated, and a mobile terminal application is deployed.
[0012] As a preferred scheme of the power distribution network satellite meteorological data analysis method, the step of obtaining and preprocessing multi-source satellite meteorological data comprises:
[0013] Visible light band, infrared band and water vapor band time series images are obtained from geostationary meteorological satellites;
[0014] Microwave imager data of polar orbit meteorological satellites are called to invert surface soil moisture and precipitation intensity;
[0015] The surface deformation is monitored by radar data;
[0016] All obtained satellite meteorological data are geometrically corrected and radiometrically calibrated, and the output is multi-source satellite meteorological data;
[0017] The corrected multi-source satellite meteorological data are spatio-temporally registered and information-completed by a data fusion algorithm to generate the meteorological element distribution map.
[0018] The beneficial effects of the preferred technical scheme are that through multi-source satellite data fusion and spatio-temporal registration, the accuracy and continuity of the meteorological element distribution map are improved, and a reliable data foundation is provided for subsequent disaster feature extraction.
[0019] As a preferred scheme of the power distribution network satellite meteorological data analysis method, the step of extracting disaster-related features from the meteorological element distribution map comprises:
[0020] Vegetation coverage anomaly detection data are extracted from visible light band and infrared band data;
[0021] Surface icing thickness inversion data are extracted from microwave imager data;
[0022] Strong convective weather recognition data are extracted from radar data;
[0023] Atmospheric water vapor content analysis data are extracted from water vapor band data;
[0024] The vegetation coverage anomaly detection data, surface icing thickness inversion data, strong convective weather recognition data and atmospheric water vapor content analysis data are feature-fused to form the disaster-related features.
[0025] The beneficial effects of the preferred technical scheme are that through multi-dimensional disaster-related feature extraction and fusion, various meteorological disaster factors affecting the safety of the power distribution network can be comprehensively identified, and the accuracy and coverage of disaster identification are improved.
[0026] As a preferred scheme of the power distribution network satellite meteorological data analysis method, wherein: the step of constructing a first calculation model to train the disaster correlation characteristics comprises:
[0027] The convolutional neural network is used to extract spatial features of the disaster correlation characteristics, and the recurrent neural network is used to model the time sequence of the disaster correlation characteristics, and the first calculation model is constructed;
[0028] The first calculation model is trained using a historical disaster data set;
[0029] The disaster distribution confidence interval is output by the trained first calculation model.
[0030] The beneficial effects of the preferred technical scheme are: through the spatio-temporal feature modeling of the first calculation model, compared with the traditional statistical method, the disaster occurrence probability can be more accurately predicted, and the quantitative confidence interval is provided, and the credibility of the prediction result is enhanced.
[0031] As a preferred scheme of the power distribution network satellite meteorological data analysis method, wherein: the step of constructing a first calculation model to train the disaster correlation characteristics comprises:
[0032] The disaster distribution confidence interval is superimposed on the power distribution network geographic information system, and the risk exposure of the power distribution network equipment is calculated;
[0033] The power distribution network topology relationship is constructed and the key node is identified;
[0034] The risk exposure and the key node analysis are combined to generate the tower overturning risk assessment result.
[0035] The beneficial effects of the preferred technical scheme are: the disaster probability is combined with the power distribution network topology structure, the effective conversion from meteorological disaster to power grid risk is realized, and the high-risk equipment and key node can be accurately positioned.
[0036] As a preferred scheme of the power distribution network satellite meteorological data analysis method, wherein: the step of establishing a multi-level early warning mechanism comprises:
[0037] The multi-level early warning threshold is set according to the tower overturning risk assessment result;
[0038] When the first early warning threshold is reached, the unmanned aerial vehicle inspection system is triggered to automatically inspect the high-risk area;
[0039] When the second early warning threshold is reached, the ice melting device is started to pretreat the line whose ground icing thickness inversion data exceeds the standard;
[0040] When the third early warning threshold is reached, the load transfer scheduling is executed, and the load of the affected line is transferred to the standby line.
[0041] superimpose the disaster-related features on the power distribution network equipment three-dimensional model in a layered transparency manner through a three-dimensional modeling engine to generate the three-dimensional risk map;
[0042] display the tower risk level and disposal scheme through equipment identification technology.
[0043] As a preferred scheme of the power distribution network satellite meteorological data analysis method, the step of generating a three-dimensional risk map through the three-dimensional modeling engine comprises:
[0044] display the vegetation coverage anomaly detection data, ground icing thickness inversion data, ground icing thickness inversion data and atmospheric water vapor content analysis data on the power distribution network equipment three-dimensional model in different color transparency gradient layers;
[0045] color code and identify the tower equipment according to the tower overturning risk assessment result;
[0046] integrate the layered display results to generate the three-dimensional risk map with multi-dimensional risk information superimposed.
[0047] In a second aspect, the present application provides a power distribution network satellite meteorological data analysis system, comprising:
[0048] In a third aspect, the present application provides an electronic device, comprising:
[0049] a data acquisition and preprocessing module for acquiring and preprocessing multi-source satellite meteorological data to generate a meteorological element distribution map;
[0050] a feature extraction module for extracting disaster-related features from the meteorological element distribution map;
[0051] a model training module for training the disaster-related features by constructing a first calculation model to output a disaster distribution confidence interval;
[0052] a risk mapping module for risk space mapping of the disaster distribution confidence interval and the power distribution network topology to generate a tower overturning risk assessment result;
[0053] a warning application module for establishing a multi-level warning mechanism through the tower overturning risk assessment result to generate a three-dimensional risk map
[0054] a memory and a processor;
[0055] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions, which realizes the steps of the power distribution network satellite meteorological data analysis method when the computer executable instructions are executed by the processor.
[0056] In a fourth aspect, the present application provides a computer readable storage medium storing computer executable instructions, which, when executed by a processor, implement the steps of the power distribution network satellite meteorological data analysis method.
[0057] Compared with the prior art, the present application has the following beneficial effects:
[0058] The present application effectively solves the problem of time and space blind area caused by traditional single data source, realizes the cooperative acquisition and space-time registration of static meteorological satellite, polar orbit meteorological satellite and radar data, and overcomes the technical difficulties of resolution difference and cloud noise interference of different satellite systems. Through deep learning model construction, the combination of convolutional neural network and recurrent neural network is realized, which realizes the dynamic space-time modeling of disaster correlation characteristics, not only can accurately predict the typhoon path and icing growth trend, but also can output the quantitative confidence interval, significantly improving the prediction accuracy and reliability.
[0059] The risk space mapping technology established by the present application deeply couples the meteorological disaster probability and the power distribution network topology structure, realizes the accurate conversion from meteorological prediction to power grid risk through key node identification and risk exposure calculation, and effectively reduces the misjudgment rate of tower collapse and wire breakage risk. The intelligent integration of multi-level early warning mechanism and emergency resource linkage response realizes the automatic triggering of unmanned aerial vehicle inspection, ice melting device starting, load transfer scheduling and other emergency measures, forms a complete "data fusion-dynamic prediction-topology evaluation-emergency linkage" whole-chain solution, and significantly enhances the active defense capability and emergency response efficiency of the power distribution network under extreme weather conditions. BRIEF DESCRIPTION OF DRAWINGS
[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.
[0061] Figure 1 The overall flowchart of the power distribution network satellite meteorological data analysis method of an embodiment of the present application.
[0062] Figure 2 The mobile terminal AR device identification and information superposition schematic diagram of the power distribution network satellite meteorological data analysis method of an embodiment of the present application.
[0063] Figure 3 The multi-level early warning control center and emergency linkage schematic diagram of the power distribution network satellite meteorological data analysis method of an embodiment of the present application. DETAILED DESCRIPTION
[0064] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the present application.
[0065] Embodiment 1, reference Figure 1 For an embodiment of the present application, a power distribution network satellite meteorological data analysis method is provided, comprising the following steps S1-S5:
[0066] S1, acquiring multi-source satellite meteorological data and preprocessing to generate a spatiotemporally continuous meteorological element distribution map;
[0067] S2, extracting disaster-related features from the meteorological element distribution map;
[0068] S3, constructing a first calculation model to train the disaster-related features and output a disaster distribution confidence interval;
[0069] S4, mapping the disaster distribution confidence interval and the power distribution network topology in risk space to generate a tower overturning risk assessment result;
[0070] S5, establishing a multi-level early warning mechanism through the tower overturning risk assessment result, generating a three-dimensional risk map and deploying a mobile terminal application.
[0071] It should be noted that the current power distribution network frequently experiences tower collapse and broken line accidents under extreme weather conditions, which seriously affects the safety of power supply. Traditional meteorological monitoring relies on a single data source, which has problems such as incomplete spatiotemporal coverage, insufficient prediction accuracy, and lack of correlation analysis with power grid equipment characteristics, resulting in low risk assessment accuracy. At the same time, the existing early warning system mostly uses static threshold judgment, which cannot adapt to the dynamic evolution characteristics of meteorological disasters, and the emergency response often lags behind the development of disasters, making it difficult to achieve precise prevention and control.
[0072] Therefore, in view of the above existing meteorological monitoring and risk warning problems, through the steps of S1-S5, a multi-source satellite data fusion processing system is constructed to accurately extract multi-dimensional disaster characteristics such as vegetation anomalies, surface icing, severe convective weather, and atmospheric water vapor; a spatiotemporal dynamic prediction mechanism of meteorological disasters is established through a deep learning model to quantitatively output the disaster occurrence probability and confidence interval; the meteorological prediction results are deeply fused with the power distribution network topology to realize accurate mapping from meteorological risk to power grid equipment risk; at the same time, a hierarchical early warning and intelligent linkage mechanism is established based on the risk assessment results to realize active defense and precise emergency response to meteorological disasters of the power distribution network.
[0073] Embodiment 2, reference Figures 1 to 3 For an embodiment of the present application, based on the above embodiment, a power distribution network satellite meteorological data analysis method is provided.
[0074] In the embodiments of the present application, in step S1, the step of acquiring multi-source satellite meteorological data and preprocessing includes A1-A5:
[0075] A1, acquiring time series images of visible light band, infrared band and water vapor band from geostationary meteorological satellite;
[0076] A2, calling the microwave imager data of polar orbit meteorological satellite to invert the surface soil moisture and precipitation intensity;
[0077] A3, monitoring the surface deformation through radar data;
[0078] A4, geometric correction and radiation calibration are performed on all acquired satellite meteorological data, and the output is multi-source satellite meteorological data;
[0079] A5, the corrected multi-source satellite meteorological data is spatio-temporally registered and information is completed through a data fusion algorithm to generate the meteorological element distribution map.
[0080] Specifically, in step A1, the acquisition method of geostationary meteorological satellite time series images can be represented as: acquiring 500-meter spatial resolution visible light band images from GOES-16 satellite to monitor vegetation coverage changes; acquiring 1-kilometer spatial resolution infrared band images to extract surface temperature distribution; synchronously acquiring 2-kilometer spatial resolution water vapor band images to analyze atmospheric water vapor content, and the data acquisition frequency is 10 minutes interval to ensure time series continuity.
[0081] Specifically, in step A2, the processing method of polar orbit meteorological satellite microwave imager data can be represented as:
[0082]
[0083] In the formula, θ is the surface soil moisture; TB 37H , TB 37V are the 37GHz horizontal polarization and vertical polarization brightness temperature values respectively; α, β are calibration coefficients;
[0084] R=γ×(TB 85H -TB 19V )+δ;
[0085] In the formula, R is the precipitation intensity; TB 85H , TB 19V are the 85GHz horizontal polarization and 19GHz vertical polarization brightness temperature values respectively; γ, δ are empirical coefficients.
[0086] It should be noted that by the combination operation of multi-band microwave brightness temperature data, atmospheric interference can be effectively excluded, the inversion accuracy is improved, and reliable ground environment parameters are provided for subsequent disaster feature extraction.
[0087] Specifically, in step A3, the processing method of radar data for monitoring ground deformation is as follows: using Sentinel-1 SARC band radar data, calculating the ground deformation by differential interferometric synthetic aperture radar (DInSAR) technology, and the deformation detection accuracy reaches millimeter level; combined with Doppler radar echo intensity, extracting atmospheric vertical wind shear information, and identifying the development trend of strong convective weather system.
[0088] Specifically, in step A4, the specific implementation of geometric correction and radiometric calibration is as follows: the geometric correction adopts a second-order polynomial model, the number of control points is not less than 8, and the correction accuracy is controlled within ±0.5 pixels; the radiometric calibration adopts MODTRAN atmospheric radiation transfer model, eliminates the influence of atmospheric scattering and absorption, and ensures the radiometric consistency of different sensor data.
[0089] The correction formula can be expressed as:
[0090] X'=a0+a1X+a2Y+a3XY+a4X 2 +a5Y 2 ;
[0091] Y'=b0+b1X+b2Y+b3XY+b4X 2 +b5Y 2 ;
[0092] In the formula, (X', Y') is the corrected coordinate; (X, Y) is the original coordinate; a i , b i are polynomial coefficients.
[0093] Specifically, in step A5, the implementation of the data fusion algorithm is as follows:
[0094] The Kalman filter algorithm is used to align the time series of stationary satellites and polar orbit satellites, and the state transition matrix is:
[0095]
[0096] Where, Δt is the time interval;
[0097] The wavelet transform technology is used for spatial fusion, and the spatial detail information of high-resolution visible light image is fused into low-resolution thermal infrared data:
[0098]
[0099] In the formula, Ifused is a fused image; w i is a weight coefficient; W i is the i-th layer wavelet decomposition; N is the number of decomposition layers.
[0100] It should be noted that through the space-time registration and wavelet fusion technology, a continuous weather element distribution map with high temporal resolution and high spatial resolution can be generated, effectively filling the data missing caused by cloud cover, and providing a complete and reliable data basis for subsequent disaster-related feature extraction.
[0101] In an optional embodiment, in step S1, the multi-source satellite meteorological data is acquired and preprocessed, and the vegetation index product of the MODIS satellite can also be integrated to identify abnormal vegetation coverage trends through NDVI time series analysis, thereby enhancing the timeliness of vegetation disaster monitoring.
[0102] In another optional embodiment, the data fusion algorithm in step S1 can also use deep learning super-resolution reconstruction technology to improve the spatial detail expression capability by using a convolutional neural network to improve low-resolution satellite images to a higher resolution.
[0103] In the embodiments of the present application, in step S2, the step of extracting disaster-related features from the weather element distribution map includes B1-B5:
[0104] B1, vegetation coverage anomaly detection data is extracted from visible light and infrared band data;
[0105] B2, ground ice thickness inversion data is extracted from microwave imager data;
[0106] B3, strong convective weather identification data is extracted from radar data;
[0107] B4, atmospheric water vapor content analysis data is extracted from water vapor band data;
[0108] B5, the vegetation coverage anomaly detection data, the ground ice thickness inversion data, the strong convective weather identification data, and the atmospheric water vapor content analysis data are fused to form the disaster-related features.
[0109] Specifically, in step B1, the extraction method of the vegetation coverage anomaly detection data can be specifically represented as: calculating the normalized vegetation index NDVI, the formula is:
[0110]
[0111] In the formula, NIR is the near-infrared band reflectivity; RED is the red light band reflectivity;
[0112] Abnormal vegetation is identified through time series comparison and analysis:
[0113] ΔNDVI = NDVI current - NDVI historical ;
[0114] Anomaly flag = {1, if |ΔNDVI| > 0.3 \ 0, otherwise;
[0115] In the formula, NDVI current is the vegetation index at the current time; NDVI historical is the historical mean value at the same period; and when the deviation exceeds 30%, it is marked as vegetation coverage anomaly.
[0116] It should be noted that through the NDVI time series change analysis, the vegetation degradation area caused by drought, disease, pest or extreme weather can be effectively identified, thereby providing early warning basis for vegetation lodging risk around the power distribution line.
[0117] Specifically, in step B2, the extraction method of the ground icing thickness inversion data is as follows: an icing thickness inversion model is established based on the 37GHz microwave radiometer brightness temperature data:
[0118]
[0119] In the formula, δ ice is the icing thickness growth; k is a material coefficient, which is 0.12 for the conductor and 0.15 for the tower steel; TB 37H is the 37GHz horizontal polarization brightness temperature; TB baseline is the reference brightness temperature in the ice-free state; T air is the environmental temperature; T freeze is the freezing point temperature; and ΔT is the time interval.
[0120] Icing growth rate The calculation formula is as follows:
[0121]
[0122] In the formula, α ice is the icing coefficient; TB0 is the reference brightness temperature; RH is the relative humidity; and η is the humidity influence index.
[0123] Specifically, in step B3, the extraction method of the severe convective weather identification data is as follows:
[0124] The vertical wind shear is calculated through the Doppler radar data:
[0125]
[0126] In the formula, VWS is the vertical wind shear intensity; V upper , V lowerrespectively are upper and lower layer wind speed; Z upper , respectively are corresponding height layers. lower 6.7
[0127] The strong convection identification criterion is: when the vertical wind shear is greater than 25 m / s and the radar reflectivity is greater than 45 dBZ, the strong convection index is level 3, indicating extremely strong convection weather; when the vertical wind shear is between 15-25 m / s and the radar reflectivity is greater than 35 dBZ, the strong convection index is level 2, indicating strong convection weather; when the vertical wind shear is between 10-15 m / s, the strong convection index is level 1, indicating moderate convection weather; otherwise, the strong convection index is level 0, indicating no obvious convection activity.
[0128] Specifically, in step B4, the extraction mode of the atmospheric water vapor content analysis data is:
[0129] The atmospheric precipitable water is retrieved by using the water vapor band brightness temperature data:
[0130]
[0131] In the formula, PWV is the atmospheric precipitable water; TB 6.7 is the 6.7 μm water vapor absorption band brightness temperature; TB clear is the clear sky reference brightness temperature; TB window is the atmospheric window zone brightness temperature; a, b, c are regression coefficients.
[0132] Specifically, in step B5, the implementation mode of feature fusion is:
[0133] A multi-scale feature fusion algorithm is adopted to combine the four types of disaster features by weighting:
[0134] F disaster = w1 x F vegetation + w2 x F ice + w3 x F convective + w4 x F vapor ;
[0135] In the formula, F disaster is the comprehensive disaster correlation feature; F vegetation , F ice , F convective , F vapor are respectively the vegetation, icing, strong convection, and water vapor feature vectors; w1, w2, w3, w4 are weight coefficients.
[0136] The weight coefficient w i is dynamically calculated by an entropy weight method:
[0137]
[0138] Among them, is the information entropy of the i-th feature; p ik is the standardized value of the k-th sample of the i-th feature.
[0139] Spatial clustering after feature fusion:
[0140] Cluster result = DBSCAN(F disaster , ∈ = 0.5, MinPts = 10);
[0141] In the formula, ∈ is the neighborhood radius; MinPts is the minimum number of samples; and the disaster high-risk area is identified through the density clustering algorithm.
[0142] It should be noted that through the weighted fusion of multi-dimensional features and spatial clustering analysis, the spatial distribution and interaction relationship of various meteorological disaster factors can be comprehensively considered to form complete disaster correlation features, thereby providing high-quality input data for subsequent deep learning models.
[0143] In an optional implementation, the vegetation coverage anomaly detection in step S2 can also combine multi-element vegetation parameters such as leaf area index LAI and vegetation moisture content VMC, and establish a vegetation health degree comprehensive evaluation model through a machine learning algorithm, thereby improving the accuracy of anomaly identification.
[0144] In another optional implementation, the feature fusion in step S2 can also use an attention mechanism to dynamically adjust the weights of various disaster features according to different seasons and geographical locations, thereby enhancing the adaptability of the model to regional disaster characteristics.
[0145] In the embodiments of the present application, in step S3, the step of constructing a first calculation model to train the disaster correlation features includes C1-C3:
[0146] C1, using a convolutional neural network to perform spatial feature extraction on the disaster correlation features, and combining a recurrent neural network to perform time series modeling on the disaster correlation features, thereby constructing the first calculation model;
[0147] C2, using a historical disaster data set to train the first calculation model;
[0148] C3, outputting the disaster distribution confidence interval through the trained first calculation model.
[0149] Specifically, in step C1, the construction of the first calculation model can be specifically represented as follows: a convolutional neural network adopts a ResNet-50 architecture to perform spatial feature extraction on a disaster-related feature map of 512*512 pixels, the input layer contains four channels corresponding to vegetation cover anomaly detection data, ground icing thickness inversion data, severe convective weather identification data and atmospheric water vapor content analysis data, spatial features are extracted layer by layer through five residual modules, and finally a 256-dimensional feature vector is output. A recurrent neural network adopts a bidirectional LSTM architecture to perform time series modeling on 24 consecutive hours of meteorological parameters, the number of hidden layer nodes is set to 128, and the long-term dependence relationship of the time series is captured through forward and backward propagation.
[0150] The implementation of spatial feature extraction is as follows: the input four-channel feature map is first subjected to preliminary feature extraction through a 3*3 convolutional layer, then subjected to batch normalization and ReLU activation function processing, then subjected to deep feature learning in a residual block, each residual block contains two 3*3 convolutional layers and a skip connection, and finally the feature map is compressed into a 256-dimensional vector through global average pooling.
[0151] The implementation of time series modeling is as follows: the spatial feature vector of 24 consecutive hours is taken as the input sequence of the LSTM, the forward LSTM processes the feature vector from time t=0 to t=23 in turn, the backward LSTM processes it in reverse from time t=23 to t=0, the bidirectional hidden state is spliced at each time, and a complete time series feature representation is formed.
[0152] It should be noted that the fusion architecture of the convolutional neural network and the recurrent neural network can simultaneously capture the spatial distribution law and the time evolution trend of the disaster features, thereby providing strong feature representation capability for accurately predicting the disaster probability distribution.
[0153] Specifically, in step C2, the implementation of the historical disaster data set training is as follows: the training data set contains satellite remote sensing data and actual loss records of global important power distribution network disaster events from 2010 to 2024, a total of 150,000 sample points, of which the positive sample accounts for 30%. In the pre-training stage, the global historical disaster data set is used for model initialization, the learning rate is set to 0.001, the batch size is 32, and the training rounds are 100. In the fine-tuning stage, the localized data of the target region is loaded, the tower tilt sensor data, line trip record and the like are introduced as supervision signals, the learning rate is adjusted to 0.0001, and fine-tuning is performed.
[0154] The loss function adopts a weighted combination of cross-entropy loss and mean square error loss: the total loss is equal to the classification loss multiplied by the weight coefficient 0.6 plus the regression loss multiplied by the weight coefficient 0.4, wherein the classification loss is used to judge whether a disaster occurs, and the regression loss is used to predict the disaster intensity.
[0155] The training process adopts an Adam optimizer to update parameters, sets a momentum parameter to 0.9, a weight decay coefficient to 0.0005, and adopts a learning rate decay strategy to multiply the learning rate by 0.1 every 30 rounds. To prevent overfitting, a Dropout technique is used to set a dropout rate of 0.5 in the fully connected layer, and an early stopping strategy is adopted to stop training when the loss of the validation set does not decrease for 10 consecutive rounds.
[0156] Specifically, in step C3, the output mode of the disaster distribution confidence interval is as follows: the trained first calculation model quantifies uncertainty by the Monte Carlo dropout method, keeps the dropout layer activated in the inference stage, and performs forward propagation on the same input for 100 times to obtain 100 prediction results. The mean of the prediction results is calculated as the point estimate of the disaster probability, and the standard deviation is used as the uncertainty measure.
[0157] The confidence interval calculation method is as follows: taking the prediction mean as the center, using 1.96 times the standard deviation as the boundary of the 95% confidence interval, the upper boundary is equal to the prediction mean plus 1.96 times the standard deviation, and the lower boundary is equal to the prediction mean minus 1.96 times the standard deviation. For the region where the disaster probability exceeds the upper boundary, it is marked as a high-risk area, for the region where the disaster probability is between the upper and lower boundaries, it is marked as a medium-risk area, and for the region below the lower boundary, it is marked as a low-risk area.
[0158] The output result contains three dimensions: the spatial distribution of the disaster occurrence probability, the uncertainty distribution, and the risk level distribution. The spatial distribution is displayed in the form of a heat map, the uncertainty distribution reflects the credibility of the model prediction, and the risk level distribution provides an intuitive risk area division.
[0159] It should be noted that the Monte Carlo dropout method can quantify the prediction uncertainty, provide credibility evaluation for decision makers, avoid false judgments caused by excessive confidence of the model, and provide a statistically reliable basis for risk management with the 95% confidence interval.
[0160] In an optional implementation, the first calculation model in step S3 can also use an attention mechanism to enhance the spatial feature extraction capability, identify the relevance between different spatial positions through a self-attention module, and improve the modeling accuracy of complex spatial patterns.
[0161] In another optional implementation, the confidence interval calculation in step S3 can also use a Bayesian neural network method to learn the posterior distribution of parameters through variational inference technology, so as to obtain more accurate uncertainty estimation and confidence interval.
[0162] In the embodiments of the present application, in step S4, the risk space mapping step includes D1-D3:
[0163] D1, superimpose the disaster distribution confidence interval to the power distribution network geographic information system, calculate the risk exposure of the power distribution network equipment;
[0164] D2, build the topology relationship of the power distribution network and identify the key nodes;
[0165] D3, combine the risk exposure and the key node analysis to generate the tower overturning risk assessment result.
[0166] In the embodiments of the present application, the step of risk space mapping in step S4 includes D1-D3:
[0167] D1, superimpose the disaster distribution confidence interval to the power distribution network geographic information system, calculate the risk exposure of the power distribution network equipment;
[0168] D2, build the topology relationship of the power distribution network and identify the key nodes;
[0169] D3, combine the risk exposure and the key node analysis to generate the tower overturning risk assessment result.
[0170] Specifically, in step D1, the calculation method of risk exposure can be specifically represented as: the disaster distribution confidence interval data is accurately superimposed on the power distribution network GIS platform according to the geographic coordinates, and each tower equipment position corresponds to a disaster probability value and a confidence interval range. The risk exposure is calculated by weighted summation: the device risk exposure is equal to the vegetation anomaly risk coefficient multiplied by 0.25 plus the icing risk coefficient multiplied by 0.35 plus the strong convection risk coefficient multiplied by 0.3 plus the water vapor risk coefficient multiplied by 0.1, wherein each risk coefficient is determined according to the disaster probability value at the device position.
[0171] Different types of power distribution equipment adopt different exposure calculation methods: tower equipment mainly considers overturning and broken pole risk, and the weight distribution is icing 35%, strong convection 30%, vegetation 25%, and water vapor 10%; overhead line mainly considers broken line and dancing risk, and the weight distribution is strong convection 40%, icing 30%, vegetation 20%, and water vapor 10%; ground equipment such as transformers mainly considers flood and lightning risk, and the weight distribution is water vapor 40%, strong convection 35%, vegetation 15%, and icing 10%.
[0172] Device vulnerability correction is adjusted by device attribute parameters: the exposure of devices with a tower height exceeding 35 meters is multiplied by a factor of 1.3, the exposure of devices with a service life exceeding 20 years is multiplied by a factor of 1.2, and the exposure of devices located in mountainous or coastal areas is multiplied by a factor of 1.15. The final risk exposure value ranges from 0 to 1, where 0.8 and above is extremely high exposure, 0.6 to 0.8 is high exposure, 0.4 to 0.6 is moderate exposure, and 0.4 and below is low exposure.
[0173] It should be noted that by accurate geographical superposition and differentiated weight setting, the comprehensive meteorological risk faced by different positions and different types of equipment can be accurately quantified, thereby providing reliable basic data for subsequent key node identification and risk assessment.
[0174] Specifically, in step D2, the implementation manner of power distribution network topology relationship construction and key node identification is: a power distribution network topology relationship model is constructed based on a Neo4j graph database, node types include substation, switch station, tower, transformer and the like, edge relationships include power supply line, protection relationship, backup channel and the like connection. Each node in the graph model contains attribute information such as device ID, geographic coordinates, rated capacity, number of power supply users, and each edge contains attributes such as line length, transmission capacity, protection configuration.
[0175] The PageRank algorithm is used for importance sorting in key node identification: the node importance score is equal to 0.15 multiplied by the uniform distribution probability plus 0.85 multiplied by the weighted sum of the importance scores of adjacent nodes, wherein the weight is determined by the transmission capacity and power supply load of the edge. Iterative calculation is performed until the importance score converges, and the convergence condition is that the score difference between two adjacent iterations is less than 0.001.
[0176] The key nodes are screened in combination with power grid operation characteristics: nodes with more than 5,000 power supply users are marked as first-level key nodes, nodes with 2,000 to 5,000 power supply users are marked as second-level key nodes, and hub nodes that undertake the function of converging multiple lines are marked as third-level key nodes. The replaceability of the nodes is also considered, and the importance weight of nodes with multiple backup power supply paths is reduced by 20%, and the importance weight of nodes that operate in isolation or have only a single power supply path is increased by 30%.
[0177] The network centrality analysis is performed by comprehensively evaluating three indexes of degree centrality, closeness centrality and betweenness centrality: the degree centrality reflects the number of direct connections of the node, the closeness centrality reflects the average distance from the node to all other nodes, and the betweenness centrality reflects the frequency of the node appearing in the shortest path. The three centrality indexes are weighted and averaged according to the weights of 0.4, 0.3 and 0.3 to obtain the comprehensive centrality score of the node.
[0178] Specifically, in step D3, the generation manner of the tower overturning risk assessment result is: the risk exposure degree and the key node analysis result are coupled and calculated, and the comprehensive risk assessment score is equal to the risk exposure degree multiplied by 0.6 plus the key node importance score multiplied by 0.4. For first-level key nodes, even if the risk exposure degree is at a medium level, the comprehensive risk assessment score is also increased by 15%; for third-level key nodes, when the risk exposure degree exceeds 0.7, the comprehensive risk assessment score is increased by 25%.
[0179] The overturning risk of the tower is calculated based on the principle of material mechanics: the overturning moment is equal to the wind pressure multiplied by the windward area of the tower multiplied by half of the height of the tower, wherein the wind pressure is equal to 0.5 multiplied by the air density multiplied by the wind resistance coefficient multiplied by the square of the wind speed, the air density is 1.225 kg / m3, and the wind resistance coefficient is 1.2. The anti-overturning moment is determined by the self-weight of the tower and the buried depth of the foundation, and when the overturning moment exceeds 80% of the anti-overturning moment, it is determined as high overturning risk.
[0180] The influence of the ice load on the carrying capacity of the tower is combined: the ice load is equal to the ice thickness multiplied by the length of the conductor multiplied by the unit length ice weight, and the unit length ice weight is calculated according to the 1.2 power of the ice thickness. When the total load obtained by superimposing the ice load and the wind load exceeds 90% of the design carrying capacity of the tower, the overturning risk level of the tower is increased by one level.
[0181] The risk assessment result is divided into four levels: extremely high risk corresponds to a comprehensive score of 0.8 or more and there is a structural weak link, high risk corresponds to a comprehensive score of 0.6 to 0.8, medium risk corresponds to a comprehensive score of 0.4 to 0.6, and low risk corresponds to a comprehensive score of 0.4 or less. Each risk level is equipped with a corresponding warning threshold and disposal suggestion, and the extremely high risk area needs to take preventive power-off measures immediately, and the high risk area needs to strengthen the patrol and emergency preparation.
[0182] It should be noted that through the coupling analysis of the risk exposure degree and the importance of the network topology, combined with the quantitative calculation of the mechanical properties of the equipment, the most vulnerable and most critical risk points in the distribution network can be accurately identified, and scientific basis is provided for precise protection and emergency resource allocation.
[0183] In an optional implementation, the key node identification in step S4 can also use the community discovery algorithm in the complex network theory to identify the functional communities in the distribution network through modularity optimization, and focus on the importance of the connection nodes between the communities.
[0184] In the embodiments of the present application, in step S5, the step of establishing a multi-level warning mechanism includes E1-:
[0185] E1, setting a multi-level warning threshold according to the tower overturning risk assessment result;
[0186] The setting mode of the multi-level early warning threshold can be specifically represented as: according to the comprehensive score of the tower overturning risk assessment result, three-level early warning is divided: the first early warning threshold is set to be between 0.3 and 0.5 of the risk score, corresponding to a mild risk state; the second early warning threshold is set to be between 0.5 and 0.7 of the risk score, corresponding to a moderate risk state; the third early warning threshold is set to be above 0.7 of the risk score, corresponding to a high risk state. The early warning threshold also needs to be dynamically adjusted in combination with the weather forecast information, when strong wind, heavy rain or snow is forecasted within the next 12 hours, the early warning threshold of each level is adjusted downward by 0.1, and when the weather condition is good, the threshold returns to the standard value.
[0187] Different thresholds are set for different equipment types: the early warning threshold of important power transmission towers is reduced by 0.05 than the standard value, the early warning threshold of old equipment is reduced by 0.1 than the standard value, and the early warning threshold of new equipment is increased by 0.05 than the standard value. Considering seasonal factors, the thresholds of each level are lowered by 0.1 in the winter icing high incidence period, and the early warning threshold of strong convective weather is lowered by 0.15 in the summer typhoon season.
[0188] The early warning triggering mechanism adopts a time window smoothing strategy: when the risk score exceeds the early warning threshold for 30 consecutive minutes, the early warning is formally triggered, avoiding false positives caused by transient fluctuations. The early warning release condition is that the risk score is below the early warning threshold for 2 consecutive hours and there is no severe weather in the next 6 hours.
[0189] E2, when the first early warning threshold is reached, triggering the unmanned aerial vehicle inspection system to automatically inspect the high-risk area;
[0190] Specifically, the triggering mode of the unmanned aerial vehicle inspection system is: when the first early warning threshold is reached, the system automatically dispatches the nearest unmanned aerial vehicle base station, and plans the optimal inspection path to cover all the towers and equipment in the first early warning area. The inspection path optimization adopts a genetic algorithm, with the shortest inspection time and the optimal battery endurance as the objective function, and the constraint conditions include unmanned aerial vehicle load limit, flight restricted area avoidance, weather condition suitability, etc.
[0191] The unmanned aerial vehicle is equipped with a high-definition camera, an infrared thermal imager and a laser radar sensor to realize multi-dimensional equipment state detection. The high-definition camera is used to identify tower structure damage, conductor breakage, insulator damage and other visible defects, with a 4K level image resolution and a 10 times optical zoom capability. The infrared thermal imager is used to detect equipment temperature anomalies and identify hot spots and cold joints, with a temperature measurement accuracy of plus or minus 2 degrees Celsius. The laser radar is used to accurately measure the tower perpendicularity and conductor sag, with a ranging accuracy of centimeters.
[0192] Inspection data is transmitted back to the control center in real time, and abnormal areas are automatically marked using image recognition algorithms, with an accuracy rate exceeding 95%. The system establishes a defect classification database, categorizing discovered problems into three levels based on urgency: immediate handling, time-limited handling, and planned handling, and automatically generates inspection reports and handling suggestions.
[0193] E3. When the level II warning threshold is reached, the ice melting device is activated to preprocess the lines where the surface ice thickness inversion data exceeds the standard.
[0194] Specifically, the de-icing device is activated as follows: when the secondary warning threshold is reached and the surface ice thickness inversion data shows that the line ice layer exceeds 5 mm, the system automatically activates the corresponding DC de-icing device. The de-icing device uses the short-circuit current method, artificially creating a short circuit at the end of the line to allow a large current to flow through the conductor, generating heat to melt the ice layer. The de-icing current is dynamically adjusted according to the conductor type and ice thickness. For LGJ-240 conductors, the de-icing current is set to 800 amps when the ice thickness is 5 to 10 mm, and 1200 amps when the ice thickness is 10 to 20 mm.
[0195] The de-icing process employs a pulsed heating strategy, activating power for 5 minutes followed by a 2-minute pause to prevent overheating damage to the conductors. During de-icing, the conductor temperature is monitored in real-time using an infrared thermometer, and de-icing is immediately stopped when the temperature exceeds 80 degrees Celsius. Simultaneously, the process monitors ice shedding, automatically ending the de-icing procedure when the ice thickness drops below 3 millimeters.
[0196] Before starting the de-icing device, it is necessary to confirm that the power line is de-energized, disconnect the relevant circuit breakers via remote control, and install grounding protection at both ends of the line. After de-icing is completed, insulation testing is performed, and power can only be restored after confirming that the line insulation performance is good. The entire de-icing process is automated and unattended, significantly improving emergency response efficiency.
[0197] E4. When the Level 3 warning threshold is reached, load transfer scheduling will be executed to transfer the load of the affected line to the backup line.
[0198] Specifically, the load transfer scheduling is executed as follows: When the level three warning threshold is reached, the system initiates an emergency load transfer procedure. First, it identifies the power supply range and load capacity of the affected lines, and then searches for available backup power supply paths. The load transfer adopts a hierarchical strategy, prioritizing the power supply to important users such as hospitals, schools, and government agencies, followed by residential electricity, and finally considering general industrial and commercial electricity.
[0199] The backup line selection is based on a multi-objective optimization algorithm, which comprehensively considers factors such as line residual capacity, transmission distance, line health status, etc. When the backup line residual capacity is insufficient, a load shedding strategy is adopted, and part of the load is sequentially cut off according to the importance level of the user. The load shedding sequence is: general industrial users, commercial users, non-essential public facilities, and residential electricity, to ensure continuous power supply to critical loads.
[0200] The load transfer operation is remotely executed through the SCADA system, including the sequence of breaker opening and closing, load switch switching, capacitor bank switching, etc. Each operation is provided with a confirmation link, which automatically checks the device state and safety conditions before operation, and monitors the system operation parameters after operation, to ensure the safety and reliability of the transfer process. The entire load transfer process is completed within 15 minutes, minimizing the power outage time and impact.
[0201] E5. superimposing the disaster-related features on the three-dimensional model of the power distribution network equipment in a layered transparency manner through a three-dimensional modeling engine to generate the three-dimensional risk map;
[0202] Specifically, the generation method of the three-dimensional risk map is: a three-dimensional visualization scene of the power distribution network equipment is constructed through the Unity engine, including complete spatial information such as terrain, tower structure, line direction, and power transformation facility. The three-dimensional model uses high-precision modeling technology, the tower model is accurate to the bolt level, the line model considers sag and wind deflection effects, and the terrain model is generated based on digital elevation data with a vertical accuracy of 1 meter.
[0203] The step of generating a three-dimensional risk map through the three-dimensional modeling engine includes E5.1-E5.3:
[0204] E5.1, display the vegetation coverage anomaly detection data, ground icing thickness inversion data, ground icing thickness inversion data, and atmospheric water vapor content analysis data in different color transparency gradient layers on the three-dimensional model of the power distribution network equipment;
[0205] The vegetation coverage anomaly detection data is displayed in green transparency gradient layers, and the transparency varies from 10% to 80% according to the anomaly degree; the ground icing thickness inversion data is displayed in blue transparency gradient layers, and the transparency varies from 20% to 90% according to the icing thickness; the strong convective weather recognition data is displayed in red transparency gradient layers, and the transparency varies from 15% to 85% according to the convective intensity; the atmospheric water vapor content analysis data is displayed in purple transparency gradient layers, and the transparency varies from 5% to 70% according to the water vapor content.
[0206] E5.2, color coding and identifying the tower equipment according to the tower overturning risk assessment results;
[0207] The tower overturning risk assessment results are color-coded: low-risk towers are displayed in green, medium-risk towers are displayed in yellow, high-risk towers are displayed in orange, and extremely high-risk towers are displayed in red with a flashing effect. The tower model also adjusts the display size according to the risk level, with high-risk tower models enlarged by 1.2 times and extremely high-risk tower models enlarged by 1.5 times, enhancing visual recognition.
[0208] E5.3, integrate the hierarchical display results to generate the three-dimensional risk map with multi-dimensional risk information superimposed.
[0209] Various risk information is displayed by layer superimposition technology, and users can display different types of risk information separately or in combination through the interface. The system provides multiple viewing modes, including top view, perspective view, first-person roaming, etc., and users can freely switch viewing angles for risk analysis.
[0210] E6, display tower risk level and disposal scheme through device recognition technology.
[0211] Device recognition and information display are achieved through mobile AR applications. Users use smartphones or tablets to align the tower equipment, and the system automatically identifies the equipment type and number. Device recognition uses a target detection algorithm based on deep learning, and the training data includes multi-angle images of various types of towers, with an accuracy rate of over 98%.
[0212] After successful identification, the risk level identifier is displayed above the device, including risk color, risk value, trend arrow, and other information. The disposal scheme is displayed in the form of text and icons, including specific operation steps, precautions, contact information, etc. For high-risk equipment, the system also provides a voice broadcast function to automatically read the risk status and disposal points.
[0213] It should be noted that through the establishment of a multi-level early warning mechanism, a fully automated process from risk identification to emergency response is achieved, significantly improving the timeliness and accuracy of power distribution network disaster prevention. Three-dimensional visualization technology and mobile applications provide intuitive and convenient means for field operators to obtain risk information.
[0214] In an alternative embodiment, the UAV inspection in step S5 can also be equipped with a meteorological sensor to measure environmental parameters such as wind speed, humidity, and temperature in real time, providing on-site verification data for the risk assessment model and improving the accuracy of the warning.
[0215] In another alternative embodiment, the three-dimensional risk map in step S5 can also integrate virtual reality technology to build an immersive risk scenario simulation environment, providing an emergency drill and training platform for operators.
[0216] In summary, the present application effectively solves the problem of time and space blind area caused by traditional single data source through multi-source satellite data fusion preprocessing technology, realizes the cooperative acquisition and space-time registration of static meteorological satellite, polar orbit meteorological satellite and radar data, and overcomes the technical difficulties of resolution difference of different satellite systems and cloud noise interference. Through deep learning model construction, the combination of convolutional neural network and recurrent neural network is realized, dynamic space-time modeling of disaster correlation characteristics is realized, not only the typhoon path and icing growth trend can be accurately predicted, but also the quantitative confidence interval can be output, which significantly improves the prediction accuracy and reliability.
[0217] The risk space mapping technology established by the present application deeply couples the meteorological disaster probability and the power distribution network topology structure, realizes the accurate conversion from meteorological prediction to power grid risk through key node identification and risk exposure calculation, and effectively reduces the misjudgment rate of tower overturning and wire breaking risk. The intelligent integration of multi-level early warning mechanism and emergency resource linkage response realizes the automatic triggering of unmanned aerial vehicle inspection, ice melting device starting, load transfer scheduling and other emergency measures, forms a complete "data fusion-dynamic prediction-topology evaluation-emergency linkage" whole-chain solution, and significantly enhances the active defense ability and emergency response efficiency of the power distribution network under extreme weather conditions.
[0218] Embodiment 3, the above is a schematic scheme of a power distribution network satellite meteorological data analysis method. It should be noted that the technical scheme of the power distribution network satellite meteorological data analysis system belongs to the same concept as the technical scheme of the power distribution network satellite meteorological data analysis method described above. The technical scheme of the power distribution network satellite meteorological data analysis system in this embodiment is not described in detail, and the description of the technical scheme of the power distribution network satellite meteorological data analysis method can be referred to.
[0219] The present embodiment also provides a power distribution network satellite meteorological data analysis system, comprising:
[0220] The data acquisition and preprocessing module is configured to acquire multi-source satellite meteorological data and perform preprocessing to generate a meteorological element distribution map.
[0221] The feature extraction module is configured to extract disaster correlation features from the meteorological element distribution map.
[0222] The model training module is configured to train the first calculation model on the disaster correlation features and output a disaster distribution confidence interval.
[0223] The risk mapping module is configured to perform risk space mapping on the disaster distribution confidence interval and the power distribution network topology structure to generate a tower overturning risk assessment result.
[0224] The early warning application module is configured to establish a multi-level early warning mechanism based on the tower overturning risk assessment result and generate a three-dimensional risk map.
[0225] The embodiment also provides an electronic device suitable for the case of power distribution network satellite meteorological data analysis, comprising a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the method for implementing power distribution network satellite meteorological data analysis proposed in the above embodiment.
[0226] The embodiment also provides a storage medium having a computer program stored thereon, and the program is executed by a processor to realize the method for implementing power distribution network satellite meteorological data analysis proposed in the above embodiment.
[0227] The storage medium proposed in the embodiment and the method for implementing power distribution network satellite meteorological data analysis proposed in the above embodiment belong to the same inventive concept, and the technical details not described in the embodiment can be referred to the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.
[0228] Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary universal hardware, and of course can also be realized by hardware. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product can be stored in a computer readable storage medium, such as a floppy disk, a ROM, a RAM, a FLASH, a hard disk or an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods of various embodiments of the present application.
[0229] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.
Claims
1. A method for analyzing satellite meteorological data of power distribution networks, characterized in that, Includes the following steps: Acquire multi-source satellite meteorological data and preprocess it to generate a spatiotemporally continuous distribution map of meteorological elements; Disaster-related features were extracted from the meteorological element distribution map. A first computational model is constructed to train the disaster association features and output the disaster distribution confidence interval; The disaster distribution information intervals are mapped to the distribution network topology to generate tower overturning risk assessment results. A multi-level early warning mechanism is established based on the tower overturning risk assessment results, a three-dimensional risk map is generated, and a mobile application is deployed.
2. The method for analyzing satellite meteorological data of power distribution networks as described in claim 1, characterized in that, The steps for acquiring and preprocessing multi-source satellite meteorological data include: Time-series images in the visible light, infrared, and water vapor bands are obtained from geostationary meteorological satellites; Data from the microwave imager of a polar-orbiting meteorological satellite was used to retrieve surface soil moisture and precipitation intensity. Surface deformation is monitored using radar data; Geometric and radiometric corrections are performed on all acquired satellite meteorological data, and the output is multi-source satellite meteorological data; The data fusion algorithm is used to perform spatiotemporal registration and information completion on the corrected multi-source satellite meteorological data to generate the meteorological element distribution map.
3. The satellite meteorological data analysis method for power distribution networks as described in claim 2, characterized in that, The steps for extracting disaster-related features from the meteorological element distribution map include: Extract vegetation cover anomaly detection data from visible light and infrared band data; Surface ice thickness inversion data was extracted from microwave imager data; Data for identifying severe convective weather is extracted from radar data; Atmospheric water vapor content analysis data were extracted from water vapor band data; The disaster-related features are formed by fusing the vegetation cover anomaly detection data, surface ice thickness inversion data, severe convective weather identification data, and atmospheric water vapor content analysis data.
4. The satellite meteorological data analysis method for power distribution networks as described in claim 3, characterized in that, The steps for constructing the first computational model to train the disaster-related features include: A convolutional neural network is used to extract spatial features from the disaster-related features, and a recurrent neural network is used to perform temporal modeling of the disaster-related features to construct the first computational model; The first computational model was trained using a historical disaster dataset; The first computational model, after training, outputs the disaster distribution information interval.
5. The satellite meteorological data analysis method for power distribution networks as described in claim 4, characterized in that, The steps of risk space mapping include: The disaster distribution information intervals are overlaid onto the distribution network geographic information system to calculate the risk exposure of distribution network equipment. Construct the power distribution network topology and identify key nodes; The risk exposure and key node analysis are combined to generate the tower overturning risk assessment result.
6. The satellite meteorological data analysis method for power distribution networks as described in claim 5, characterized in that, The steps to establish a multi-level early warning mechanism include: Multiple early warning thresholds are set based on the tower overturning risk assessment results; When the first-level warning threshold is reached, the drone inspection system is triggered to automatically inspect the high-risk area. When the level II warning threshold is reached, the ice melting device is activated to preprocess the lines where the surface ice thickness inversion data exceeds the standard. When the level 3 warning threshold is reached, load transfer scheduling is executed to transfer the load of the affected line to the backup line; The disaster-related features are overlaid onto the 3D model of the power distribution network equipment using a 3D modeling engine with layered transparency to generate the 3D risk map. The equipment identification technology displays the risk level of the tower and the corresponding handling plan.
7. The satellite meteorological data analysis method for power distribution networks as described in claim 6, characterized in that, The steps for generating a 3D risk map using the 3D modeling engine include: The vegetation cover anomaly detection data, surface ice thickness inversion data, surface ice thickness inversion data, and atmospheric water vapor content analysis data are displayed on the three-dimensional model of the power distribution network equipment using different color transparency gradient layers. Based on the tower overturning risk assessment results, the tower equipment is color-coded and identified. The layered display results are integrated to generate the three-dimensional risk map with multi-dimensional risk information overlay.
8. A satellite meteorological data analysis system for power distribution networks, employing the method described in any one of claims 1-7, characterized in that, include: The data acquisition and preprocessing module is used to acquire and preprocess multi-source satellite meteorological data to generate meteorological element distribution maps; The feature extraction module is used to extract disaster-related features from the meteorological element distribution map; The model training module is used to construct a first computational model to train the disaster association features and output the disaster distribution confidence interval; The risk mapping module is used to perform risk spatial mapping between the disaster distribution information intervals and the distribution network topology to generate tower overturning risk assessment results; The early warning application module is used to establish a multi-level early warning mechanism based on the tower overturning risk assessment results and generate a three-dimensional risk map.
9. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the power distribution network satellite meteorological data analysis method according to any one of claims 1 to 7.
10. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the power distribution network satellite meteorological data analysis method according to any one of claims 1 to 7.
Citation Information
Cited By
Intelligent drought disaster risk estimation system
CN121836400A
An intelligent drought disaster risk estimation system
CN121836400B
