Mountainous area field lightning monitoring and early warning method and system

Abstract

The invention relates to the field of lightning early warning, and discloses a mountain area field lightning monitoring and early warning method and system. The mountain area field thunder and lightning monitoring and early warning method comprises the following steps: acquiring multi-source data of a mountain area to form an initial data set; performing topographic correlation correction on the lightning activity monitoring data and the meteorological environment data, constructing a fusion weight model dynamically changing along with topographic features, and performing weighted fusion on multiple types of data in the preprocessed data set to generate a topographic lightning coupling feature data set; inputting the terrain thunder and lightning coupling characteristic data set into an early warning model to obtain a thunder and lightning occurrence probability and intensity prediction result of a grid unit level; in combination with the terrain risk avoiding attributes of the corresponding grid units, generating graded early warning information containing terrain adaptability avoidance guidance; according to the method, the refined topographic features are fully fused, multi-source data dynamic fusion is realized, and the prediction precision of the occurrence probability and intensity of field thunder and lightning in the mountainous area is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of lightning early warning, more particularly, it relates to a mountainous area outdoor lightning monitoring and early warning method and system. BACKGROUND

[0002] As a common sudden natural disaster in mountainous area outdoor environment, lightning has a strong coupling relationship with terrain characteristics. The complex terrain conditions in mountainous areas (such as high-altitude peaks, steep slopes, deep valleys, and differential soil structure) can significantly affect charge accumulation, lightning leader development path, and lightning stroke intensity:

[0003] The high-altitude area has a significantly higher lightning activity frequency than the low-altitude area due to the thin air and intense convective activity, which can form a local strong electric field. The steep slope and valley terrain can change the atmospheric circulation direction, causing the charge to accumulate at the terrain abrupt change, increasing the spatial heterogeneity of lightning strike points.

[0004] The spatial difference in soil resistivity (such as the resistivity difference between rock exposed area and wet grassland area can reach 1-2 orders of magnitude) can directly affect the propagation efficiency of lightning stroke, and further change the lightning risk level of ground facilities and personnel.

[0005] However, the existing lightning monitoring and early warning technology has obvious limitations:

[0006] Insufficient terrain factors: Traditional methods are mostly based on data from plain areas to build models, simplifying terrain characteristics into macro regional attributes (such as "mountainous area" and "plain area"), without quantitative analysis of fine terrain parameters such as slope, altitude, and soil resistivity, resulting in a large error in the prediction results (especially in complex terrain areas, the error can be more than 30%).

[0007] Rough multi-source data fusion: The fusion of lightning monitoring data (such as lightning location), meteorological environment data (such as atmospheric electric field), and terrain data mostly uses fixed weights, without considering the influence of dynamic changes in terrain characteristics on data correlation (such as the sensitivity of atmospheric electric field data in steep slope areas is higher than that in flat areas), making it difficult to reflect the spatio-temporal heterogeneity of lightning activity in mountainous areas.

[0008] Limited practicality of early warning information: Existing early warning mostly only outputs "risk level", without considering the terrain risk avoidance attributes of specific grid elements (such as shelter height, conductive environment, etc.), making it difficult for outdoor workers (such as geological exploration and forest fire prevention personnel) to take targeted avoidance measures based on real-time terrain conditions, increasing the blindness of risk avoidance operations.

[0009] Therefore, there is an urgent need for a lightning monitoring and early warning technology that can deeply integrate fine terrain characteristics, achieve high-precision prediction, and generate terrain-adaptive guidance to meet the safety needs of mountainous area outdoor environments. SUMMARY

[0010] The present application provides a mountainous area outdoor lightning monitoring and early warning method and system, which solves the technical problems of low prediction accuracy and insufficient practicality of risk avoidance guidance in related art mountainous area outdoor lightning early warning due to insufficient integration of fine terrain influence, rough multi-source data fusion, lack of terrain adaptability of early warning information and poor adaptability to mountainous area scenes.

[0011] The present application provides a mountainous area outdoor lightning monitoring and early warning method and system, which solves the technical problems of low prediction accuracy and insufficient practicality of risk avoidance guidance in related art mountainous area outdoor lightning early warning due to insufficient integration of fine terrain influence, rough multi-source data fusion, lack of terrain adaptability of early warning information and poor adaptability to mountainous area scenes.

[0012] Obtain multi-source data of the mountainous area, including lightning activity monitoring data, meteorological environment data and fine terrain data, to form an initial data set;

[0013] Based on the fine terrain data in the initial data set, correct the terrain correlation of the lightning activity monitoring data and the meteorological environment data to obtain a preprocessing data set containing terrain influence factors;

[0014] Based on the terrain influence factors in the preprocessing data set, construct a fusion weight model that dynamically changes with terrain characteristics, and perform weighted fusion on the multi-type data in the preprocessing data set to generate a terrain lightning coupling feature data set;

[0015] Input the terrain lightning coupling feature data set into the early warning model, and through the spatial and temporal feature extraction module in the model with terrain constraint parameters, obtain the lightning occurrence probability and intensity prediction results at the grid cell level;

[0016] Based on the lightning occurrence probability and intensity prediction results at the grid cell level, combine the terrain risk avoidance attributes of the corresponding grid cell to generate hierarchical early warning information containing terrain adaptability avoidance guidance.

[0017] Further, the specific steps of obtaining multi-source data of the mountainous area, including lightning activity monitoring data, meteorological environment data and fine terrain data, to form an initial data set are as follows:

[0018] Obtain fine terrain data, including mountainous area underlying surface terrain data, slope data, elevation data and soil resistivity data, and establish a mountainous area terrain basic database;

[0019] Based on the spatial range of the fine terrain data, collect lightning activity monitoring data: through the ADTD lightning location system composed of multiple lightning location detection sub-stations, real-time lightning data is obtained within the above spatial range;

[0020] In the same terrain space range and time period as the lightning activity monitoring data, radar echo data including reflectivity, moving speed and direction are collected by a meteorological radar; meanwhile, atmospheric electric field monitoring is performed by an atmospheric electric field instrument, and atmospheric electric field intensity value, polarity and electric field change characteristic data caused by lightning discharge are collected once per second;

[0021] The obtained refined terrain data, lightning activity monitoring data and meteorological environment data are spatio-temporally matched to form an initial data set.

[0022] Further, based on the refined terrain data in the initial data set, the lightning activity monitoring data and the meteorological environment data are corrected in terms of terrain correlation to obtain a pre-processing data set containing terrain influence factors, specifically including the following steps:

[0023] From the refined terrain data in the initial data set, key parameters for terrain correlation analysis are extracted;

[0024] Based on the key parameters, spatial correlation processing is performed on the lightning activity monitoring data to obtain the distribution characteristics of the ground flash density and the ground flash intensity in different terrain units, and a mapping relationship between the ground flash parameters and the terrain parameters is established; and for the meteorological environment data, an influence coefficient of the terrain factor on the monitoring value thereof is quantified;

[0025] Based on the mapping relationship between the ground flash and the terrain and the terrain influence coefficient of the meteorological data, a terrain influence factor is generated, and the lightning activity monitoring data and the meteorological environment data are corrected using the terrain influence factor to form a pre-processing data set containing terrain influence factors.

[0026] Further, based on the terrain influence factor in the pre-processing data set, a fusion weight model dynamically changing with the terrain features is constructed, and multiple types of data in the pre-processing data set are weighted and fused to generate a terrain lightning coupling feature data set, specifically including the following steps:

[0027] From the pre-processing data set, terrain influence factors and corresponding terrain grid features are extracted, and terrain complexity of each grid unit is calculated;

[0028] Based on the terrain grid features, terrain feature differences of adjacent grid units are calculated: for each grid unit, several adjacent grids in the periphery are selected, and absolute values of altitude difference, slope difference and soil resistivity difference are calculated, and a weighted sum of the absolute values is defined as a terrain feature difference value to form a terrain feature difference data set;

[0029] Based on the terrain complexity and the terrain feature difference data set, initial weights are comprehensively distributed: the atmospheric electric field data weight is increased in a region with high terrain complexity and large difference, and the lightning activity monitoring data weight is increased in a region with low terrain complexity and small difference;

[0030] Based on the initial weight allocation result, combined with the terrain feature difference data set, a weight dynamic adjustment rule is established: for every 10% increase in terrain feature difference value, the weight correction amplitude increases by 5%; a dynamic fusion weight model is generated;

[0031] Based on the dynamic fusion weight model, the multi-type data in the preprocessed data set after terrain correction is weighted and fused to generate a terrain lightning coupling feature data set.

[0032] Further, the terrain lightning coupling feature data set is input into the early warning model, and through the spatial and temporal feature extraction module in the model, the lightning occurrence probability and intensity prediction results at the grid cell level are obtained, which specifically include the following steps:

[0033] The terrain lightning coupling feature data set is divided into a training set and a test set, wherein the training set is used for model training, and the test set is used for model verification;

[0034] Multi-dimensional input features are extracted from the training set, including lightning activity spatio-temporal sequence features, meteorological environment features, and terrain constraint parameters, to form a structured training feature matrix;

[0035] The structured training feature matrix is input into the early warning model, and through the spatial and temporal feature extraction module built-in the model, the input features are fused and processed: through the spatial feature extraction submodule, the spatial correlation between terrain constraint parameters and lightning activity in different grid cells is captured to generate a spatial feature vector; then through the time series feature extraction submodule, the evolution law of lightning activity in the same grid cell in different time dimensions is mined to generate a time series feature vector; subsequently, through the fusion layer built-in the module, the spatial feature vector and the time series feature vector are associated and integrated, and finally a high-dimensional feature vector of fused terrain spatio-temporal information is output;

[0036] Based on the high-dimensional feature vector, the prediction layer of the early warning model learns the mapping relationship between the features and the lightning occurrence probability and intensity, the model parameters are verified and optimized using the test set, and a trained early warning model is obtained;

[0037] The terrain lightning coupling feature data generated by the fusion processing of real-time monitoring data is input into the spatial and temporal feature extraction module to generate a current high-dimensional feature vector in real time, and after calculation by the prediction layer, the lightning occurrence probability and intensity prediction results at the grid cell level are output.

[0038] Further, based on the lightning occurrence probability and intensity prediction results at the grid cell level, combined with the terrain risk avoidance attributes of the corresponding grid cell, hierarchical early warning information containing terrain adaptability avoidance guidance is generated, which specifically includes the following steps:

[0039] The four-dimensional associated data set is formed by the grid unit coordinates, the lightning occurrence probability, the intensity prediction value, and the terrain avoidance attribute.

[0040] Based on the four-dimensional associated data set, the lightning occurrence probability and the intensity prediction value of each grid unit are extracted, and a risk level matrix is constructed and calibrated according to the following rules: the probability level is divided into three levels, i.e., the first level for the probability ≥80%, the second level for 50%-80%, and the third level for <50%; the intensity level is divided into three levels, i.e., the high intensity level for the peak current ≥100kA, the medium intensity level for 50-100kA, and the low intensity level for <50kA; and the comprehensive risk level of each grid unit is determined by the combination of the probability level and the intensity level.

[0041] Based on the comprehensive risk level of each grid unit, the terrain avoidance attribute parameters in the four-dimensional associated data set are associated, and a mapping rule of the risk level and the terrain adaptation avoidance strategy is established.

[0042] The highest risk level grid unit matches the attribute parameters of “terrain shielding degree ≥0.8 and conductive environment coefficient ≤0.3”, and generates the avoidance guidance of “immediately transfer to a closed terrain shielding area, and stay away from open highlands and conductive bodies”.

[0043] The medium-high risk level grid unit matches the attribute parameters of “terrain closure ≥0.6 and safety area accessibility ≤500m”, and generates the avoidance guidance of “transfer to the designated safety area within 10 minutes, and avoid staying on the hilltop or near the isolated object”.

[0044] The medium-low risk level grid unit matches the attribute parameters of “conductive environment coefficient ≤0.5”, and generates the avoidance guidance of “reduce outdoor activities, stay away from metal facilities, and pay attention to the changes of terrain”.

[0045] Based on the mapping rule, the comprehensive risk level, the corresponding terrain avoidance attribute parameters, and the avoidance guidance of each grid unit are associated through the spatial coordinates, and are integrated into the hierarchical warning information containing the visual grid risk distribution, the adaptive terrain feature description, and the specific operation guidance.

[0046] Further, the attention weighting mechanism is adopted for the association and integration of the spatial feature vector and the time sequence feature vector in the fusion layer of the spatial time sequence feature extraction module, and specifically includes:

[0047] The terrain attention layer calculates the terrain weight of each grid cell based on the slope and altitude parameters of the grid cell, and the weight coefficient of the grid cell with a slope greater than 30° is increased by 20%, and the weight coefficient of the grid cell with an altitude higher than 1000m is increased by 20%;

[0048] The time attention layer calculates the time weight of different time points based on the historical lightning activity law, and the weight coefficient of the feature vector in the peak period of lightning activity is increased by 30%;

[0049] The terrain weight and the time weight are multiplied by matrix to generate the spatio-temporal joint attention weight;

[0050] Based on the spatio-temporal joint attention weight, the spatial feature vector and the time sequence feature vector are weighted and summed to generate a high-dimensional feature vector fused with terrain and spatio-temporal information.

[0051] The second aspect of the present application provides a mountainous area outdoor lightning monitoring and early warning system for executing the mountainous area outdoor lightning monitoring and early warning method, comprising:

[0052] The multi-source data acquisition module is used for acquiring multi-source data of the mountainous area;

[0053] The terrain correlation correction module is used for correcting the terrain correlation of the lightning activity monitoring data and the meteorological environment data based on the refined terrain data in the initial data set, to obtain a preprocessing data set containing terrain influence factors;

[0054] The data fusion module is used for constructing a fusion weight model that dynamically changes with terrain features based on the terrain influence factors in the preprocessing data set, weighting and fusing the multi-type data in the preprocessing data set to generate a terrain lightning coupling feature data set;

[0055] The early warning and prediction module is used for inputting the terrain lightning coupling feature data set into the early warning model, and obtaining the lightning occurrence probability and intensity prediction result of the grid cell level through the spatial and time sequence feature extraction module with the terrain constraint parameters in the model;

[0056] The hierarchical early warning module is used for generating hierarchical early warning information containing terrain adaptability avoidance guidance based on the lightning occurrence probability and intensity prediction result and the terrain risk avoidance attribute of the corresponding grid cell.

[0057] The present application has the beneficial effects that: the present application realizes dynamic fusion of multi-source data by fully integrating fine terrain features, improves the prediction accuracy of lightning occurrence probability and intensity in mountainous areas, and at the same time generates adaptive risk avoidance guidance combined with terrain risk avoidance attributes, enhances the practicality of early warning information, effectively solves the problems of inaccurate prediction and no risk avoidance in complex mountainous area scenes in traditional technologies, and provides reliable technical support for lightning safety protection in mountainous areas. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 is a flowchart of a mountainous area outdoor lightning monitoring and early warning method of the present application. DETAILED DESCRIPTION

[0059] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is merely meant to provide a better understanding of the subject matter described herein and can be changed in function and arrangement without departing from the scope of the present specification. Various processes or components can be omitted, substituted, or added according to various examples. In addition, features described with respect to some examples can be combined in other examples.

[0060] As shown in Figure 1 , a mountainous area outdoor lightning monitoring and early warning method includes:

[0061] Step 1: Obtain multi-source data of mountainous areas, including lightning activity monitoring data, meteorological environment data, and fine terrain data, to form an initial data set;

[0062] Collect lightning activity monitoring data, meteorological environment data, and fine terrain data of mountainous areas, and form an initial data set after spatio-temporal matching.

[0063] Fine terrain data: including mountainous area underlying surface terrain (such as contour lines, slope direction), slope (unit: °), elevation (unit: m), and soil resistivity (unit: Ω·m), obtained by unmanned aerial vehicle aerial survey, GIS map vectorization, or geological exploration, to establish a terrain basic database with 10m x 10m grid precision. Lightning activity monitoring data: based on ADTD (Advanced Digital Thunderstorm Detector) lightning positioning system, composed of at least 3 detection sub-stations, using multi-station time difference positioning method (positioning error ≤ 500m), to collect real-time ground flash data (including return stroke time, location, peak current, steepness).

[0064] Meteorological environment data: meteorological radar data: reflectivity (unit: dBZ), echo top height (unit: km), moving speed, and direction are collected by S-band Doppler radar;

[0065] Atmospheric electric field data: L21 type field mill atmospheric electric field instrument is adopted, and atmospheric electric field intensity (unit: kV / m), polarity and electric field mutation characteristics caused by lightning discharge are collected once per second.

[0066] Space-time matching: based on the spatial grid of terrain data, the lightning data and meteorological data are associated according to the time stamp (accurate to seconds) and spatial coordinates (latitude and longitude), so as to ensure that the data are aligned in the space-time dimension.

[0067] Step 2: Based on the refined terrain data in the initial data set, the terrain correlation correction is performed on the lightning activity monitoring data and meteorological environment data to obtain a preprocessed data set containing terrain influence factors;

[0068] Based on the refined terrain data, the lightning activity monitoring data and meteorological environment data are corrected, the terrain influence factors are introduced, the terrain interference is eliminated and the correlation is strengthened.

[0069] Extracting key terrain parameters: slope (S), altitude (H) and soil resistivity (p) are selected as core terrain parameters from the initial data set.

[0070] Lightning data correction:

[0071] Kriging interpolation algorithm is adopted to perform spatial interpolation on sparse ground flash data to obtain ground flash density (unit: times / (km 2 ·year) and ground flash intensity (peak current average) of each grid unit;

[0072] Establishing the mapping relationship between ground flash parameters and terrain parameters: for example, the positive correlation coefficient r1 between ground flash density and altitude is 0.72 (ground flash is more frequent in high altitude area), and the negative correlation coefficient r2 between ground flash intensity and soil resistivity is -0.65 (low resistivity soil is easy to produce strong return stroke).

[0073] Meteorological data correction:

[0074] Quantifying the influence coefficient of terrain on meteorological parameters: for example, the radar echo reflectivity correction coefficient increases by 5% for every 10° increase in slope (steep slope easily leads to air uplift, enhancing echo intensity);

[0075] Atmospheric electric field intensity correction: for every 1000m increase in altitude, the electric field intensity measurement value is multiplied by the correction coefficient 1.12 (high altitude air is thin, and electric field attenuation is slower).

[0076] Generating terrain influence factors: based on the above mapping relationship and correction coefficient, the terrain influence factor TF is defined:

[0077]

[0078] wherein, a, b, g are weight coefficients (a+b+g=1), H max , S max are the maximum elevation and maximum slope of the region, respectively, p max , p min are the maximum and minimum soil resistivity of the region, respectively, S is the slope, H is the elevation, and p is the soil resistivity.

[0079] The TF is used to weight and correct the lightning data and meteorological data to form a preprocessed data set containing the TF.

[0080] Step 3: Based on the terrain influence factors in the preprocessed data set, a fusion weight model that dynamically changes with terrain features is constructed to weight and fuse the multiple types of data in the preprocessed data set, generating a terrain-lightning coupling feature data set;

[0081] Calculate the terrain complexity and feature difference: Terrain complexity TC: for each grid cell, the slope, elevation standard deviation, and soil resistivity coefficient of variation are weighted to calculate:

[0082]

[0083] (σH is the elevation standard deviation, and σρ is the soil resistivity standard deviation) Terrain feature difference value TD: for each grid cell, select the surrounding 8 adjacent grids, calculate the absolute values of the elevation difference ΔH, slope difference ΔS, and soil resistivity difference Δρ, and the weighted sum is:

[0084] TD = 0.3·|ΔH| + 0.4·|ΔS| + 0.3|Δρ|;

[0085] Construct a dynamic fusion weight model: initial weight distribution: in areas with high terrain complexity (TC>0.6) and large differences (TD>threshold T), the atmospheric electric field data weight is increased to 60% (as it is more sensitive to short-time lightning); otherwise, the lightning activity monitoring data weight is increased to 70%.

[0086] Dynamic adjustment rule: for every 10% increase in terrain feature difference value TD, the weight correction amplitude is increased by 5% (for example, TD increases from 0.2 to 0.22, and the atmospheric electric field weight is corrected from 50% to 52.5%).

[0087] Data fusion: based on the dynamic weight model, the cloud-to-ground lightning density, atmospheric electric field intensity, radar reflectivity, and other data in the preprocessed data set are weighted and fused to generate a terrain-lightning coupling feature data set (containing terrain-lightning correlation features in space-time dimensions).

[0088] Step 4: Input the terrain-lightning coupling feature data set into the warning model, and through the spatial and temporal feature extraction module with terrain constraint parameters integrated in the model, obtain the lightning occurrence probability and intensity prediction results at the grid cell level.

[0089] The terrain lightning coupling feature dataset is input into the early warning model, the terrain constraint parameters are fused through the spatial and temporal feature extraction module, and the lightning occurrence probability and intensity prediction results of the grid unit level are output.

[0090] Dataset division: divided into training set (used for model training) and test set (used for verification) in the ratio of 7:3, and the time span covers different seasons (to ensure that the model adapts to seasonal terrain influence).

[0091] Feature extraction: three types of features are extracted from the training set: spatiotemporal sequence features of lightning activity (such as the time series of lightning frequency in the past 1 hour, spatial distribution entropy);

[0092] Meteorological environment features (such as atmospheric electric field intensity change rate, radar echo movement vector);

[0093] Terrain constraint parameters (slope, elevation, grid distribution of soil resistivity).

[0094] Spatial and temporal feature extraction module: spatial feature extraction submodule: 3-layer CNN (convolutional neural network) is adopted:

[0095] The first layer: 3x3 convolution kernel (step 1), extracting the micro association of terrain parameters and lightning activity within a single grid; The second layer: 5x5 convolution kernel (step 2), capturing the terrain-lightning spatial interaction of 3x3 neighborhood grids; The third layer: combined with terrain attention mechanism (assigning higher weights to grids with slope > 30° or elevation > 1000m), outputting spatial feature vectors. Temporal feature extraction submodule: 2-layer Bi-LSTM (bidirectional long short-term memory network) is adopted, with 64 neurons in each layer, to mine the temporal evolution law of lightning activity in the past 30 minutes, outputting temporal feature vectors. Fusion layer: attention weighting mechanism is used to integrate spatial and temporal features: terrain weight calculation: the weight of grids with slope > 30° is increased by 20%, and the weight of grids with elevation > 1000m is increased by 20%; Time weight calculation: the feature vector weight of lightning activity peak period (such as local 14:00-18:00) is increased by 30%; Spatio-temporal joint weight: the terrain weight and time weight matrices are multiplied to obtain the joint weight; Weighted fusion: based on the joint weight, the spatial feature vector and the temporal feature vector are summed to generate a high-dimensional feature vector. Model training and prediction: the prediction layer adopts a fully connected neural network to learn the mapping relationship between the high-dimensional feature vector and the lightning occurrence probability (a value between 0 and 1) and the intensity (peak current, unit: kA);

[0096] The model parameters are optimized using cross-entropy loss function (for probability prediction) and mean square error loss function (for intensity prediction), and the verification accuracy of the test set is ≥85%; input real-time terrain lightning coupling feature data, and output the lightning occurrence probability and intensity prediction value of each 10m x 10m grid in the next 0-2 hours after model calculation.

[0097] Step 5: Based on the lightning occurrence probability and intensity prediction results at the grid cell level, combined with the terrain risk avoidance attributes of the corresponding grid cell, generate hierarchical warning information containing terrain adaptability avoidance guidelines;

[0098] Based on the prediction results and terrain risk avoidance attributes, generate hierarchical warning information and provide targeted risk avoidance guidelines.

[0099] Data preparation: obtain the prediction results output in step 4, and extract the terrain risk avoidance attributes of each grid: terrain shielding degree: shielding height / grid elevation (the higher the ratio, the better the shielding effect); terrain enclosure: grid perimeter terrain enclosure (0-1, 1 for complete enclosure such as a canyon); conductive environment coefficient: (soil resistivity x 0.6 + surface water probability x 0.4) (the lower the value, the weaker the conductivity, the safer);

[0100] Safety area accessibility: straight-line distance to the nearest low-risk terrain (unit: m).

[0101] Form a four-dimensional associated data set of "grid coordinates-occurrence probability-intensity-terrain risk avoidance attributes".

[0102] Risk level calibration: probability level: P≥80% (first level), 50%≤P<80% (second level), P<50% (third level);

[0103] Intensity level: peak current I≥100kA (high intensity), 50kA≤I<100kA (medium intensity), I<50kA (low intensity);

[0104] Comprehensive risk level: determined by the combination of probability level and intensity level (such as "first level probability + high intensity" corresponding to the highest risk).

[0105] Mapping risk avoidance strategy: highest risk: matching "terrain shielding degree≥0.8, conductive environment coefficient≤0.3" grid, guiding "immediately transfer to closed terrain shielding area (such as behind low rock wall), away from open high ground, isolated trees and metal facilities";

[0106] Medium-high risk: matching "terrain enclosure≥0.6, safety area accessibility≤500m" grid, guiding "transfer to designated safety area (such as leeward of gentle slope) within 10 minutes, avoid staying on slope top or near water";

[0107] Low risk: match the grid of "conductivity environment coefficient ≤ 0.5", guide "reduce outdoor activities, away from wet ground and metal pipelines, keep communication smooth".

[0108] Generate early warning information: integrate grid risk level, terrain features and risk avoidance guidelines to output in the form of visual map (superimposed risk heat map) and text description.

[0109] Embodiment 2

[0110] A mountainous area outdoor lightning monitoring and early warning system for implementing a mountainous area outdoor lightning monitoring and early warning method, comprising:

[0111] Multi-source data acquisition module: for acquiring multi-source data of mountainous areas;

[0112] Terrain correlation correction module: for terrain correlation correction of lightning activity monitoring data and meteorological environment data based on refined terrain data in the initial data set, to obtain a preprocessed data set containing terrain impact factors;

[0113] Data fusion module: for constructing a fusion weight model that dynamically changes with terrain features based on terrain impact factors in the preprocessed data set, weighting and fusing multiple types of data in the preprocessed data set to generate a terrain lightning coupling feature data set;

[0114] Early warning and prediction module: for inputting the terrain lightning coupling feature data set into the early warning model, and obtaining lightning occurrence probability and intensity prediction results at the grid cell level through the spatial and temporal feature extraction module in the model with terrain constraint parameters;

[0115] Hierarchical early warning module: for generating hierarchical early warning information containing terrain adaptability avoidance guidelines based on lightning occurrence probability and intensity prediction results and the terrain risk avoidance attributes of the corresponding grid cell.

[0116] The embodiments of the present application are described above, but the embodiments are not limited to the specific implementation described above, which is only illustrative and not limiting, and those skilled in the art can make many forms under the inspiration of the embodiments, which are all within the protection of the embodiments.

Claims

1. A mountainous field lightning monitoring and early warning method, characterized in that, The method comprises the following steps: Obtain multi-source data of mountainous areas, including lightning activity monitoring data, meteorological environment data, and refined topographic data, to form an initial data set; Based on the refined topographic data in the initial data set, correct the lightning activity monitoring data and meteorological environment data for topographic correlation to obtain a preprocessed data set containing topographic influence factors; Based on the topographic influence factors in the preprocessed data set, construct a fusion weight model that dynamically changes with the topographic features to perform weighted fusion on the multiple types of data in the preprocessed data set, and generate a topographic lightning coupling feature data set; Input the topographic lightning coupling feature data set into the early warning model, and obtain the lightning occurrence probability and intensity prediction results of the grid cell level through the spatial and temporal feature extraction module in the model which incorporates topographic constraint parameters; Based on the lightning occurrence probability and intensity prediction results of the grid cell level, and in combination with the topographic risk avoidance attributes of the corresponding grid cell, generate hierarchical early warning information containing topographic adaptability avoidance guidelines.

2. The mountainous field lightning monitoring and warning method according to claim 1, characterized in that, The specific steps for obtaining the multi-source data of mountainous areas, including lightning activity monitoring data, meteorological environment data, and refined topographic data, to form an initial data set are as follows: Obtain refined topographic data, including mountainous area underlying surface topographic data, slope data, elevation data, and soil resistivity data, and establish a mountainous area topographic database; Based on the spatial range of the refined topographic data, collect lightning activity monitoring data: through the ADTD lightning location system composed of multiple lightning location detection sub-stations, obtain real-time cloud-to-ground (CG) data within the above spatial range; Within the same topographic spatial range and time period as the lightning activity monitoring data, collect radar echo data, including reflectivity, moving speed, and direction, through a weather radar; at the same time, monitor the atmospheric electric field through an atmospheric electric field instrument, and collect atmospheric electric field intensity values, polarity, and electric field change characteristic data caused by lightning discharge every second; Perform spatio-temporal matching on the obtained refined topographic data, lightning activity monitoring data, and meteorological environment data to form an initial data set.

3. The mountainous field lightning monitoring and warning method according to claim 1, characterized in that, Based on the refined topographic data in the initial data set, correct the lightning activity monitoring data and meteorological environment data for topographic correlation to obtain a preprocessed data set containing topographic influence factors, which specifically includes the following steps: From the refined topographic data in the initial data set, extract key parameters for topographic correlation analysis; Based on the key parameters, perform spatial correlation processing on the lightning activity monitoring data to obtain the distribution characteristics of CG density and CG intensity in different topographic elements, and establish a mapping relationship between CG parameters and topographic parameters; for meteorological environment data, quantify the influence coefficient of topographic factors on its monitoring values; Based on the CG and topographic mapping relationship and the topographic influence coefficient of meteorological data, generate topographic influence factors, and use the topographic influence factors to correct the lightning activity monitoring data and meteorological environment data to form a preprocessed data set containing topographic influence factors.

4. The mountainous field lightning monitoring and warning method according to claim 1, characterized in that, Based on the topographic influence factors in the preprocessed data set, construct a fusion weight model that dynamically changes with the topographic features to perform weighted fusion on the multiple types of data in the preprocessed data set, and generate a topographic lightning coupling feature data set, which specifically includes the following steps: Extracting terrain impact factors and corresponding terrain grid features from the preprocessed data set, calculating the terrain complexity of each grid cell; Based on the terrain grid features, the terrain feature difference of adjacent grid cells is calculated: for each grid cell, select several adjacent grids, calculate the absolute value of the altitude difference, slope difference, and soil resistivity difference, and define the weighted sum as the terrain feature difference value to form the terrain feature difference data set; Based on the terrain complexity and terrain feature difference data set, the initial weight is allocated comprehensively: the area with high terrain complexity and large difference increases the weight of the atmospheric electric field data, and vice versa, the weight of the lightning activity monitoring data is increased; Based on the initial weight allocation result, combined with the terrain feature difference data set, the weight dynamic adjustment rule is established: the terrain feature difference value increases by 10%, and the weight correction amplitude increases by 5%; generate a dynamic fusion weight model; Based on the dynamic fusion weight model, the multiple types of data in the preprocessed data set are weighted and fused after terrain correction to generate a terrain lightning coupling feature data set.

5. The mountainous field lightning monitoring and warning method according to claim 1, characterized in that, The terrain lightning coupling feature data set is input into the warning model, and the spatial and temporal feature extraction module with terrain constraint parameters is obtained through the model. The grid cell level lightning occurrence probability and intensity prediction result includes the following steps: Divide the terrain lightning coupling feature data set into training set and test set, where the training set is used for model training and the test set is used for model verification; Extract multi-dimensional input features from the training set, including lightning activity spatio-temporal sequence features, meteorological environment features and terrain constraint parameters, and form a structured training feature matrix; The structured training feature matrix is input into the warning model, and the spatial and temporal feature extraction module built in the model is used to fuse the input features: through the spatial feature extraction submodule, the spatial correlation between terrain constraint parameters and lightning activity in different grid cells is captured to generate a spatial feature vector; then through the time series feature extraction submodule, the evolution law of lightning activity in the same grid cell in different time dimensions is mined to generate a time series feature vector; then through the fusion layer built in the module, the spatial feature vector and the time series feature vector are associated and integrated, and finally a high-dimensional feature vector of fused terrain spatio-temporal information is output; Based on the high-dimensional feature vector, the prediction layer of the warning model learns the mapping relationship between the features and the lightning occurrence probability and intensity, verifies and optimizes the model parameters using the test set, and obtains the trained warning model; The terrain lightning coupling feature data generated by the fusion processing of real-time monitoring data is processed through the spatial and temporal feature extraction module to generate the current high-dimensional feature vector, which is calculated by the prediction layer to output the grid cell level lightning occurrence probability and intensity prediction result.

6. The mountainous field lightning monitoring and warning method according to claim 1, characterized in that, Based on the grid cell level lightning occurrence probability and intensity prediction result, combined with the terrain risk avoidance attribute of the corresponding grid cell, the hierarchical warning information containing terrain adaptability avoidance guidance is generated, which includes the following steps: The four-dimensional associated data set is formed by the grid cell coordinates, the lightning occurrence probability, the intensity prediction value, and the terrain avoidance attribute. Based on the four-dimensional associated data set, the lightning occurrence probability and the intensity prediction value of each grid cell are extracted, and a risk level matrix is constructed and calibrated according to the following rules: the probability level is divided into three levels, i.e., the first level for the probability ≥ 80%, the second level for 50%-80%, and the third level for < 50%; the intensity level is divided into three levels, i.e., the high intensity level for the peak current ≥ 100 kA, the medium intensity level for 50-100 kA, and the low intensity level for < 50 kA; and the comprehensive risk level of each grid cell is determined by the combination of the probability level and the intensity level. Based on the comprehensive risk level of each grid cell, the corresponding terrain avoidance attribute parameters in the four-dimensional associated data set are associated, and a mapping rule of the risk level and the terrain adaptation avoidance strategy is established. The highest risk level grid cell matches the attribute parameters of "terrain shielding degree ≥ 0.8 and conductive environment coefficient ≤ 0.3", and generates the avoidance instruction of "immediately transfer to the closed terrain shielding area, and stay away from the open high ground and the conductive body"; The medium-high risk level grid cell matches the attribute parameters of "terrain closure ≥ 0.6 and safety area accessibility ≤ 500 m", and generates the avoidance instruction of "transfer to the designated safety area within 10 minutes, and avoid staying near the slope top or isolated object"; The medium-low risk level grid cell matches the attribute parameters of "conductive environment coefficient ≤ 0.5", and generates the avoidance instruction of "reduce outdoor activities, stay away from metal facilities, and pay attention to the terrain changes"; Based on the mapping rule, the comprehensive risk level of each grid cell, the corresponding terrain avoidance attribute parameters, and the avoidance instruction are associated through the spatial coordinates, and are integrated into the hierarchical warning information containing the visual grid risk distribution, the adaptive terrain feature description, and the specific operation instruction.

7. The mountainous field lightning monitoring and warning method according to claim 5, characterized in that, The fusion layer of the spatial and temporal feature extraction module uses an attention weighting mechanism for the association and integration of the spatial feature vector and the temporal feature vector, specifically including: The terrain attention layer is used to calculate the terrain weight of each grid cell based on the slope and altitude parameters of the grid cell, and the weight coefficient is increased by 20% for the grid cell with a slope greater than 30° and the grid cell with an altitude higher than 1000 m; The time attention layer is used to calculate the time weight of different time points based on the historical lightning activity law, and the feature vector weight coefficient is increased by 30% during the peak period of lightning activity; The terrain weight and the time weight are multiplied to generate the spatio-temporal joint attention weight. Based on the spatio-temporal joint attention weight, the spatial feature vector and the time sequence feature vector are weighted and summed to generate a high-dimensional feature vector fusing the terrain and spatio-temporal information.

8. A mountainous field lightning monitoring and warning system, characterized in that, The mountainous field lightning monitoring and early warning method according to any one of claims 1-7, comprising: a multi-source data acquisition module for acquiring multi-source data of the mountainous area; a terrain correlation correction module for performing terrain correlation correction on the lightning activity monitoring data and the meteorological environment data based on the refined terrain data in the initial data set to obtain a preprocessed data set containing terrain influence factors; a data fusion module for constructing a fusion weight model dynamically changing with terrain features based on the terrain influence factors in the preprocessed data set, weighting and fusing the multi-type data in the preprocessed data set to generate a terrain lightning coupling feature data set; an early warning and prediction module for inputting the terrain lightning coupling feature data set into an early warning model to obtain grid cell level lightning occurrence probability and intensity prediction results through a spatial time sequence feature extraction module in which terrain constraint parameters are integrated; a hierarchical early warning module for generating hierarchical early warning information containing terrain adaptability avoidance guidelines based on the lightning occurrence probability and intensity prediction results and the terrain risk avoidance attributes of the corresponding grid cell.

Images