Thunderstorm and gale weather forecasting method, device and equipment and storage medium

Abstract

The embodiment of the invention provides a thunderstorm and gale weather forecasting method and device, equipment and a storage medium. The method comprises the steps of obtaining ground observation data and ERA5 reanalysis data of a target area in a historical time period, constructing a thunderstorm and gale weather forecast data set according to the ground observation data and the ERA5 reanalysis data, constructing an extreme thunderstorm and gale weather forecast data set according to the ground observation data, and constructing a thunderstorm and gale weather forecast data set according to the extreme thunderstorm and gale weather forecast data set. A thunderstorm and gale weather forecast model based on an AI algorithm is constructed, an extreme thunderstorm and gale weather forecast model based on a batching algorithm is constructed according to an extreme thunderstorm and gale weather forecast data set, and then the thunderstorm and gale weather forecast model and the extreme thunderstorm and gale weather forecast model are combined to refine forecast thunderstorm and gale weather.

Description

Technical Field

[0001] The present disclosure relates to the technical field of meteorological services, and in particular to a method, device, equipment and storage medium for forecasting thunderstorm and gale weather. Background Art

[0002] Severe convective weather refers to meteorologically defined disastrous weather that occurs suddenly, moves quickly, is violent, and has strong destructive power. It mainly includes thunderstorms, gale, hail, tornadoes, and local heavy rainfall. Severe convective weather occurs in small and medium-scale weather systems with a small spatial scale. The general horizontal range is about 10 kilometers to 200 to 300 kilometers, and the horizontal scale is generally less than 200 kilometers. Some horizontal ranges are only tens of meters to more than ten kilometers. Its life history is short and has obvious suddenness, about one hour to more than ten hours, and the shorter ones are only a few minutes to one hour. It often occurs in convective cloud systems or single convective cloud blocks. When severe convective weather comes, it is often accompanied by severe weather such as lightning, thunder, strong winds and heavy rain, which can cause houses to collapse, crops and trees to be destroyed, telecommunications and transportation to be damaged, and even cause casualties.

[0003] It should be noted that the current nowadays forecast of thunderstorm and gale weather mainly relies on radar and other observation data, combined with storm identification and tracking technology for radar extrapolation forecasting. However, this scheme is difficult to depict the generation, development, weakening and dissipation of thunderstorm and gale weather systems, and the forecasting effect is limited. Summary of the invention

[0004] Embodiments of the present disclosure provide a thunderstorm and gale weather forecasting method, device, equipment and storage medium.

[0005] In a first aspect, an embodiment of the present disclosure provides a thunderstorm and gale weather forecast model training method, the method comprising:

[0006] Obtain ground observation data and ERA5 reanalysis data for the target area within the historical period;

[0007] Based on ground observation data and ERA5 reanalysis data, a thunderstorm and gale weather forecast dataset is constructed; based on ground observation data, an extreme thunderstorm and gale weather forecast dataset is constructed;

[0008] Based on the thunderstorm and gale weather forecast data set, a thunderstorm and gale weather forecast model based on the AI ​​algorithm is constructed; based on the extreme thunderstorm and gale weather forecast data set, an extreme thunderstorm and gale weather forecast model based on the batching algorithm is constructed.

[0009] In some implementations of the first aspect, a thunderstorm and gale weather forecast dataset is constructed based on ground observation data and ERA5 reanalysis data, including:

[0010] Based on K-Means cluster analysis, manual labeling and time neighborhood expansion budget algorithm, the ground observation data is quality controlled to construct positive and negative samples of thunderstorm and gale weather.

[0011] The ERA5 reanalysis data are processed to generate grid-by-grid and hourly atmospheric vertical profiles. Based on this, the physical quantities corresponding to the positive and negative samples of thunderstorm and gale weather are calculated based on the SHARPpy sounding analysis and diagnostic package. The physical quantities are added to the corresponding samples, and the subsequent samples are quality controlled based on the percentiles of the physical quantities. The physical quantities of the samples after quality control are subjected to time lag processing, and a thunderstorm and gale weather forecast dataset is constructed based on the samples after time lag processing.

[0012] In some implementations of the first aspect, constructing an extreme thunderstorm and gale weather forecast dataset based on ground observation data includes:

[0013] Extreme thunderstorm and gale data are screened from ground observation data, and extreme thunderstorm and gale samples are constructed based on them. Then, an extreme thunderstorm and gale weather forecast dataset is constructed based on the extreme thunderstorm and gale samples.

[0014] In some implementations of the first aspect, a thunderstorm and gale weather forecast model based on an AI algorithm is constructed according to a thunderstorm and gale weather forecast dataset, including:

[0015] According to the thunderstorm and gale weather forecast data set, XGboost, LightGBM, and Random Forest are trained respectively to obtain the corresponding thunderstorm and gale weather forecast sub-model;

[0016] A plurality of thunderstorm and gale weather forecast sub-models are integrated to obtain a thunderstorm and gale weather forecast model.

[0017] In some implementable embodiments of the first aspect, a plurality of thunderstorm and gale weather forecast sub-models are integrated to obtain a thunderstorm and gale weather forecast model, including:

[0018] Conduct comparative experiments on multiple thunderstorm and gale weather forecast sub-models to determine the performance data of multiple thunderstorm and gale weather forecast sub-models;

[0019] Calculate weight coefficients of multiple thunderstorm and gale weather forecast sub-models according to performance data of multiple thunderstorm and gale weather forecast sub-models;

[0020] According to the weight coefficients of multiple thunderstorm and gale weather forecast sub-models, multiple thunderstorm and gale weather forecast sub-models are integrated to obtain a thunderstorm and gale weather forecast model.

[0021] In a second aspect, an embodiment of the present disclosure provides a method for forecasting thunderstorm and gale weather, the method comprising:

[0022] Utilize multi-scale numerical forecast model products to calculate the thunderstorm and gale weather forecast factor data of the target area to be input into the thunderstorm and gale weather forecast model;

[0023] Inputting the thunderstorm and gale weather forecast factor data into the thunderstorm and gale weather forecast model, and the thunderstorm and gale weather forecast model processes the thunderstorm and gale weather forecast factor data to obtain the thunderstorm and gale weather forecast result;

[0024] If the thunderstorm and gale weather forecast result is an extreme thunderstorm and gale weather forecast result, the extreme thunderstorm and gale weather forecast factor data of the target area to be input into the extreme thunderstorm and gale weather forecast model is calculated using the multi-scale numerical forecast model product;

[0025] Inputting the extreme thunderstorm and gale weather forecast factor data into the extreme thunderstorm and gale weather forecast model, and the extreme thunderstorm and gale weather forecast model processes the extreme thunderstorm and gale weather forecast factor data to obtain the final extreme thunderstorm and gale weather forecast result;

[0026] If the thunderstorm and gale weather forecast result is not an extreme thunderstorm and gale weather forecast result, it will be determined as the final thunderstorm and gale weather forecast result;

[0027] Among them, the thunderstorm and gale weather forecast model and the extreme thunderstorm and gale weather forecast model are obtained based on the training method as described above.

[0028] In a third aspect, an embodiment of the present disclosure provides a thunderstorm and gale weather forecast model training device, the device comprising:

[0029] The acquisition module is used to obtain ground observation data and ERA5 reanalysis data of the target area in the historical time period;

[0030] The construction module is used to construct a thunderstorm and gale weather forecast dataset based on ground observation data and ERA5 reanalysis data; and to construct an extreme thunderstorm and gale weather forecast dataset based on ground observation data;

[0031] The construction module is also used to build a thunderstorm and gale weather forecast model based on the AI ​​algorithm according to the thunderstorm and gale weather forecast data set; and to build an extreme thunderstorm and gale weather forecast model based on the batching algorithm according to the extreme thunderstorm and gale weather forecast data set.

[0032] In a fourth aspect, an embodiment of the present disclosure provides a thunderstorm and gale weather forecast device, the device comprising:

[0033] A calculation module is used to calculate the thunderstorm and gale weather forecast factor data of the target area to be input into the thunderstorm and gale weather forecast model by using the multi-scale numerical forecast model product;

[0034] The forecast module is used to input the thunderstorm and gale weather forecast factor data into the thunderstorm and gale weather forecast model, and the thunderstorm and gale weather forecast model processes the thunderstorm and gale weather forecast factor data to obtain the thunderstorm and gale weather forecast result;

[0035] The calculation module is further used to calculate the extreme thunderstorm and gale weather forecast factor data of the target area to be input into the extreme thunderstorm and gale weather forecast model by using the multi-scale numerical forecast model product if the thunderstorm and gale weather forecast result is an extreme thunderstorm and gale weather forecast result;

[0036] The forecast module is also used to input the extreme thunderstorm and gale weather forecast factor data into the extreme thunderstorm and gale weather forecast model, and the extreme thunderstorm and gale weather forecast model processes the extreme thunderstorm and gale weather forecast factor data to obtain the final extreme thunderstorm and gale weather forecast result;

[0037] A determination module, for determining the thunderstorm and gale weather forecast result as the final thunderstorm and gale weather forecast result if the thunderstorm and gale weather forecast result is not an extreme thunderstorm and gale weather forecast result;

[0038] Among them, the thunderstorm and gale weather forecast model and the extreme thunderstorm and gale weather forecast model are obtained based on the training method as described above.

[0039] In a fifth aspect, an embodiment of the present disclosure provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method as described above.

[0040] In a sixth aspect, an embodiment of the present disclosure provides a non-transitory computer-readable storage medium storing computer instructions, where the computer instructions are used to cause a computer to execute the method as described above.

[0041] In the embodiments of the present disclosure, a thunderstorm and gale weather forecast model based on an AI algorithm can be constructed according to a thunderstorm and gale weather forecast data set, and an extreme thunderstorm and gale weather forecast model based on a batching algorithm can be constructed according to an extreme thunderstorm and gale weather forecast data set. The thunderstorm and gale weather forecast model and the extreme thunderstorm and gale weather forecast model can then be combined to refine the forecast of thunderstorm and gale weather.

[0042] It should be understood that the contents described in the summary of the invention are not intended to limit the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, among which:

[0044] Figure 1 A flowchart of a thunderstorm and gale weather forecast model training method provided by an embodiment of the present disclosure;

[0045] Figure 2 A flow chart of a thunderstorm and gale weather forecasting method provided in an embodiment of the present disclosure;

[0046] Figure 3 Flowchart for constructing a thunderstorm and gale weather forecast dataset;

[0047] Figure 4 This is a schematic diagram of sample distribution after quality control outliers;

[0048] Figure 5 It is the box plot of sample pressure and temperature;

[0049] Figure 6 It is a three-dimensional scatter plot of positive samples and cluster analysis-samples after manual adjustment;

[0050] Figure 7 The box plot of maximum wind speed and precipitation of positive samples of stations corresponding to only 0 and 1 samples is selected;

[0051] Figure 8 Schematic diagram of the time variation characteristics of positive samples;

[0052] Fig. 9 is the time difference between the start of strong wind and the start of precipitation;

[0053] Fig.10 Schematic diagram of the conceptual model of thunderstorm and gale weather;

[0054] Fig.11 A structural diagram of a thunderstorm and gale weather forecast model training device provided in an embodiment of the present disclosure;

[0055] Fig.12 A structural diagram of a thunderstorm and gale weather forecasting device provided in an embodiment of the present disclosure;

[0056] Fig.13 A structural diagram of an exemplary electronic device capable of implementing an embodiment of the present disclosure. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0058] In addition, the term "and / or" in this article is only a description of the association relationship between the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0059] In view of the problems in the background technology, the embodiments of the present disclosure provide a thunderstorm and gale weather forecast method, device, equipment and storage medium. Ground observation data and ERA5 reanalysis data of the target area in the historical time period can be obtained, and a thunderstorm and gale weather forecast data set can be constructed based on the ground observation data and the ERA5 reanalysis data. According to the ground observation data, an extreme thunderstorm and gale weather forecast data set can be constructed. According to the thunderstorm and gale weather forecast data set, a thunderstorm and gale weather forecast model based on an AI algorithm can be constructed. According to the extreme thunderstorm and gale weather forecast data set, an extreme thunderstorm and gale weather forecast model based on a batching algorithm can be constructed, and then the thunderstorm and gale weather forecast model and the extreme thunderstorm and gale weather forecast model can be combined to refine the forecast of thunderstorm and gale weather.

[0060] In conjunction with the accompanying drawings, a thunderstorm and gale weather forecast method, device, equipment and storage medium provided by the embodiments of the present disclosure are described in detail through specific embodiments.

[0061] Figure 1 A flow chart of a thunderstorm and gale weather forecast model training method provided by an embodiment of the present disclosure, such as Figure 1 As shown, the thunderstorm and gale weather forecast model training method 100 may include the following steps:

[0062] S110, obtaining ground observation data and ERA5 reanalysis data of the target area within a historical period.

[0063] Among them, the ground observation data are the hourly ground observation data of the national automatic meteorological observation station after quality control, which can include meteorological elements such as temperature, relative humidity, air pressure, precipitation, and maximum wind speed; the ERA5 reanalysis data has 24 vertical layers, with a temporal resolution of 1h and a spatial resolution of 0.25°×0.25°, which can include meteorological elements such as temperature, potential, relative humidity, U wind, and V wind.

[0064] S120: Construct a thunderstorm and gale weather forecast dataset based on ground observation data and ERA5 reanalysis data; construct an extreme thunderstorm and gale weather forecast dataset based on ground observation data.

[0065] In some embodiments, the ground observation data can be quality controlled based on K-Means cluster analysis, manual labeling and time neighborhood expansion budget algorithm to construct positive and negative samples of thunderstorm and gale weather (above level 8), and then the ERA5 reanalysis data can be processed to generate grid-by-grid and hourly atmospheric vertical profiles. The physical quantities corresponding to the positive and negative samples of thunderstorm and gale weather are calculated based on the SHARPpy sounding analysis and diagnostic package, and the physical quantities are added to the corresponding samples. The subsequent samples are quality controlled based on the percentiles of the physical quantities, and the physical quantities of the quality-controlled samples are time-lagged. A thunderstorm and gale weather forecast dataset is constructed based on the samples after the time-lagged processing.

[0066] In some embodiments, extreme thunderstorm and gale data (for example, above level 10 or above level 12) can be screened from ground observation data, and extreme thunderstorm and gale samples can be constructed based on them, thereby constructing an extreme thunderstorm and gale weather forecast data set.

[0067] S130, based on the thunderstorm and gale weather forecast data set, construct a thunderstorm and gale weather forecast model based on the AI ​​algorithm; based on the extreme thunderstorm and gale weather forecast data set, construct an extreme thunderstorm and gale weather forecast model based on the batching algorithm.

[0068] In some embodiments, XGboost, LightGBM, and Random Forest can be trained respectively according to the thunderstorm and gale weather forecast data set to obtain the corresponding thunderstorm and gale weather forecast sub-models, and then multiple thunderstorm and gale weather forecast sub-models are integrated to obtain a thunderstorm and gale weather forecast model. Exemplarily, the integration process can be specifically as follows:

[0069] A comparative experiment is conducted on multiple thunderstorm and high wind weather forecast sub-models to determine the performance data of the multiple thunderstorm and high wind weather forecast sub-models, and then the weight coefficients of the multiple thunderstorm and high wind weather forecast sub-models are calculated based on the performance data of the multiple thunderstorm and high wind weather forecast sub-models. Then, based on the weight coefficients of the multiple thunderstorm and high wind weather forecast sub-models, the multiple thunderstorm and high wind weather forecast sub-models are integrated to obtain a thunderstorm and high wind weather forecast model.

[0070] Among them, the comparative experiment settings can be shown in Table 1:

[0071] Table 1

[0072]

[0073] In some embodiments, according to the extreme thunderstorm and gale weather forecast data set, the physical quantity parameter values ​​in the n samples in the extreme thunderstorm and gale weather forecast data set can be converted into binary variables of 0 and 1 (the occurrence of thunderstorm and gale is represented by "1" and "0" binary variables according to the physical quantity threshold and the idea of ​​random probability. When the physical quantity factor meets the threshold, the physical quantity is converted to 1; when the threshold is not met, the physical quantity is converted to 0), forming an n×k "0,1" matrix. Assuming that the function of the control weight is divided into two cases where f is 1 (meets the threshold) or 0 (does not meet the threshold), the extreme thunderstorm and gale weather forecast model can be expressed as:

[0074] P = f 1 × 1 +f 2 × 2 +f 3 × 3 +…+f k × k (1)

[0075] Where: x 1 , x 2 , …, x k (%) are the contribution rates of k forecast factors, and P is the probability of occurrence of extreme thunderstorms and gales. If the forecast factor meets the corresponding threshold, then f≡1, and the corresponding factor contribution rate is used as the probability of occurrence of extreme thunderstorms and gales. If the more forecast factors meet the threshold, the greater the cumulative value of the contribution rate, the more favorable the conditions for the occurrence of extreme thunderstorms and gales, and the higher the probability of occurrence; otherwise, the lower the probability.

[0076] Optionally, the evaluation indicators used here to evaluate the performance of the extreme thunderstorm and gale weather forecast model may be the hit rate (POD), the false alarm rate (FAR), the bias (Bias), the risk score (TS), and the fairness skill score (ETS).

[0077] In the embodiments of the present disclosure, a thunderstorm and gale weather forecast model based on an AI algorithm can be constructed according to a thunderstorm and gale weather forecast data set, and an extreme thunderstorm and gale weather forecast model based on a batching algorithm can be constructed according to an extreme thunderstorm and gale weather forecast data set, thereby facilitating the subsequent combination of the thunderstorm and gale weather forecast model and the extreme thunderstorm and gale weather forecast model to refine the forecast of thunderstorm and gale weather.

[0078] Based on the thunderstorm and gale weather forecast model training method 100 provided in the embodiment of the present disclosure, the embodiment of the present disclosure also provides a thunderstorm and gale weather forecast method, such as Figure 2 As shown, the thunderstorm and gale weather forecasting method 200 may include the following steps:

[0079] S210, using a multi-scale numerical forecast model product, calculating thunderstorm and gale weather forecast factor data of a target area to be input into a thunderstorm and gale weather forecast model.

[0080] Among them, the multi-scale numerical forecast model products can be EC, CMA-meso and other multi-scale numerical forecast model products; the thunderstorm and gale weather forecast factors are various physical quantities included in the samples input during the training of the thunderstorm and gale weather forecast model.

[0081] S220, inputting the thunderstorm and gale weather forecast factor data into a thunderstorm and gale weather forecast model, and the thunderstorm and gale weather forecast model processes the thunderstorm and gale weather forecast factor data to obtain a thunderstorm and gale weather forecast result.

[0082] S230: If the thunderstorm and gale weather forecast result is an extreme thunderstorm and gale weather forecast result, a multi-scale numerical forecast model product is used to calculate extreme thunderstorm and gale weather forecast factor data of the target area to be input into the extreme thunderstorm and gale weather forecast model.

[0083] Among them, the multi-scale numerical forecast model products can be EC, CMA-meso and other multi-scale numerical forecast model products; the extreme thunderstorm and gale weather forecast factors are various physical quantities included in the samples input during the extreme thunderstorm and gale weather forecast model training.

[0084] S240, inputting the extreme thunderstorm and gale weather forecast factor data into the extreme thunderstorm and gale weather forecast model, and the extreme thunderstorm and gale weather forecast model processes the extreme thunderstorm and gale weather forecast factor data to obtain a final extreme thunderstorm and gale weather forecast result.

[0085] S250: If the thunderstorm and gale weather forecast result is not an extreme thunderstorm and gale weather forecast result, determine it as the final thunderstorm and gale weather forecast result.

[0086] Among them, the thunderstorm and gale weather forecast model and the extreme thunderstorm and gale weather forecast model are obtained based on the training method as described above.

[0087] In the embodiments of the present disclosure, the trained thunderstorm and gale weather forecast model and the extreme thunderstorm and gale weather forecast model may be combined to refine the forecast of thunderstorm and gale weather.

[0088] For further understanding, the thunderstorm and gale weather forecast model training method 100 and the thunderstorm and gale weather forecast method 200 are described in detail below in conjunction with a specific embodiment:

[0089] 1. Data Source

[0090] (1) Ground observation data: Hourly ground observation data from the national automatic meteorological observation stations in the Beijing-Tianjin-Hebei region after quality control, including meteorological elements such as temperature, relative humidity, air pressure, precipitation, and maximum wind speed.

[0091] (2) ERA5 reanalysis data: temperature, potential, relative humidity, U wind, V wind and other meteorological elements, 24 vertical layers, time resolution of 1 hour, spatial resolution of 0.25°×0.25°.

[0092] The above data will be used for modeling from May to September 2017-2021, and for verification from May to September 2022.

[0093] 2. Establishment of the thunderstorm and gale weather forecast dataset in the North China Plain

[0094] 2.1 Sample processing

[0095] The project uses relative humidity, air pressure, precipitation, maximum wind speed and other factors to screen severe convective weather samples. The main steps are as follows: Figure 3 As shown:

[0096] (1) Extreme value quality control of ground observation data: retain samples with temperature between -10 and 45°C, maximum wind speed less than 56 m / s, air pressure between 700 and 1050 hPa, and hourly rainfall less than 200 mm.

[0097] (2) Sample labeling: For the ground observation data after quality control, the samples are labeled according to the thresholds of maximum wind speed 17m / s and hourly precipitation 20mm. The samples are divided into thunderstorm and strong wind samples (HWIND: maximum wind speed ≥17m / s, hourly precipitation <20mm), short-term heavy precipitation samples (HRAIN: maximum wind speed <17m / s, hourly precipitation ≥20mm), thunderstorm and strong wind and short-term heavy precipitation mixed samples (Both: maximum wind speed ≥17m / s, hourly precipitation ≥20mm) and other samples (Negative), among which HWIND, HRIAN and Both are positive samples.

[0098] Figure 4 This is the sample distribution after quality control of outliers. According to statistics, severe convective weather often occurs in a concomitant form. First, we should distinguish between single disasters and mixed disasters. In addition, the number of strong winds in April and May is the highest, accounting for more than 70% of the whole year, but whether it is a systematic strong wind caused by cold air or a thunderstorm strong wind caused by strong convection needs further distinction.

[0099] (3) Positive sample optimization:

[0100] a. Screening modeling stations. Statistics were collected on the ground temperature, air pressure, humidity and other factors of all samples, and it was found that some samples had abnormally low air pressure and temperature ( Figure 5). K-Means was used to cluster the positive samples (17m / s or above 20mm / h) (only using ground temperature, precipitation, wind speed, and air pressure), and they were divided into three categories: 0, 1, and 2. The results showed that the air pressure of the mountain station samples (category 2) was significantly lower than that of other samples. First, stations with large terrain differences were eliminated. The results are as follows: Figure 6 As shown in the figure, before the green scattered points were removed, there were abnormal classes (green scattered points) obtained by cluster analysis, with a total of 3497 samples with low air pressure (29 stations), and after the removal, 8755 samples were obtained from 149 stations.

[0101] b. Positive sample adjustment. Figure 7 After clustering in step (3), all pure thunderstorm and strong wind (HWIND) samples (no short and strong) are classified as 0. Pure HRAIN samples (no strong wind) are mainly classified as 1, but there are also 0. Therefore, the 0-class samples in the classification of heavy precipitation samples (HRAIN) with rainfall intensity slightly greater than 20 mm / h and wind speed of about 17 m / s are adjusted to mixed types (Both), so that the number of mixed samples can be further expanded.

[0102] c. Based on the manual marking results and the relative humidity threshold (25%), typical system gale samples are eliminated.

[0103] d. Eliminate unlabeled samples. The project proposed a temporal neighbor dilation algorithm (TNDA) to clean samples that are missing manual labels and eliminate systematic strong winds and isolated samples. The algorithm draws on the idea of ​​the image dilation algorithm. The specific implementation steps are as follows: 1) According to expert knowledge, all positive samples are pre-labeled, where samples with hourly precipitation of 8 mm and relative humidity above 60% are labeled as 1, and samples with hourly precipitation of more than 20 mm are labeled as 2; 2) Perform a temporal neighborhood dilation operation. For each labeled sample, if there are still unlabeled samples within the time window (±6h), these unlabeled samples are marked as 0; 3) Perform multiple iterations to obtain the final labeled samples. After the iteration, all unlabeled samples are eliminated.

[0104] e. Finally, 5088 positive samples were obtained, including 2418 HWIND samples, 2347 HRAIN samples, and 323 Both samples.

[0105] (4) Negative sample cleaning: Reduce the number of samples by cleaning negative samples. The specific steps are as follows:

[0106] a. Quality control of outliers: According to the upper and lower thresholds of the box plot of the positive sample ground elements, retain the negative samples with temperature > 8°C, air pressure 890-1025hPa, and relative humidity > 25%.

[0107] b. Spatiotemporal probability sampling. Resample the negative samples according to the temporal and spatial probability distribution characteristics of the positive samples to ensure that the positive and negative samples have similar spatiotemporal probability density distribution characteristics and that the samples are representative. When sampling, set the proportion of negative samples to 10 times that of positive samples.

[0108] c. Cluster analysis. The K-Means algorithm was used to cluster the remaining negative samples (divided into 3 categories), and a few station clusters were eliminated (a total of 41,212 samples (27 stations), corresponding to high altitude areas and low pressure samples), and finally 284,625 negative samples were obtained.

[0109] 2.2 Time-Lagged Dataset

[0110] SHARPpy (Sounding and Hodograph Analysis Routines Package in python) is a sounding analysis and diagnostic package developed by NOAA / SPC (Storm Prediction Center).

[0111] Firstly, physical quantities that can describe the occurrence and development characteristics of severe weather are calculated based on SHARPpy. Then, when using ERA5 hourly reanalysis data to calculate diagnostic physical parameters, the grid data are first interpolated to nearby stations based on the temporal and spatial correspondence, and the air temperature (℃), dew point temperature (℃), air pressure (hPa), geopotential height, wind direction (°), and wind speed (kts, 1kts = 0.514m / s) are input to construct vertical profiles. 75 physical quantities that characterize thermal conditions, water vapor conditions, and dynamic conditions 0-6h before the occurrence of severe weather events are calculated using the diagnostic package.

[0112] Based on the 75 physical quantities calculated from the ERA5 reanalysis data, we first eliminated physical quantities (9) with small variance and high correlation through variance analysis and cross-correlation analysis, and then statistically analyzed the percentiles of the remaining physical quantities. The 0.5 and 99.5 percentiles of the positive samples were used as the lower and upper limits, and the samples exceeding the threshold were eliminated. Finally, 5033 positive samples and 105034 negative samples were obtained, of which 2392, 2324, and 317 were HWIND, HRAIN, and Both samples, respectively. The ratio of positive to negative samples was close to 1:20, which was significantly improved compared with the original ratio (1:277). Considering that there may be a lagged correlation between the occurrence of severe convective weather and physical quantities, the physical quantities of each sample were subjected to time lag processing, that is, the physical quantities t hours ago were used to reconstruct the features and establish a time lag data set as the input for model training. The project t was 0, 3, and 6 hours, and the time lag data set had a total of 198 input features.

[0113] 3. Temporal and spatial distribution characteristics of thunderstorm and gale and conceptual model of thunderstorm and gale weather

[0114] 3.1 Spatiotemporal distribution characteristics

[0115] 3.1.1 Seasonal variation characteristics

[0116] Figure 8 The temporal distribution characteristics of thunderstorms and gale from May to September 2017-2021 were obtained based on the dataset. Figure 8 The left side shows the frequency of occurrence by month. It can be seen that July is the month with the most severe weather, and it is also the month with the highest frequency of thunderstorms and gale, accounting for more than 30% of the total; August is second, accounting for more than 20%; the frequency of occurrence in May and June is similar, about 20%; September is the least, only 50%. Affected by the climate background, short-term heavy precipitation is mainly concentrated in July and August, accounting for 85% of the total, and the frequency of occurrence in other months is significantly less.

[0117] 3.1.2 Daily variation characteristics

[0118] Figure 8 The right side shows the frequency of occurrence by hour. It can be seen that thunderstorms and strong winds are concentrated from 15:00 in the afternoon to 23:00 at night, with a peak at 18:00, accounting for 10.4% of the total, and thunderstorms and strong winds occur less frequently from 5:00 to 11:00. The frequency of thunderstorms and strong winds has obvious daily variation characteristics. This may be because under normal circumstances, the near-ground temperature rises significantly during the day, making the stratification state of the atmosphere more unstable, which is conducive to the occurrence of thunderstorms and strong winds. The concentrated time of short-term heavy precipitation can be extended from 15:00 to 05:00 in the second half of the night, while the mixed weather appears later than thunderstorms and strong winds, from 17:00 to 23:00.

[0119] 3.1.3 Characteristics of meteorological elements changes

[0120] When a thunderstorm or gale occurs or one hour before it occurs, negative temperature changes generally occur, and the hourly negative temperature changes are generally above 2°C, with a maximum of 8-9°C. When a thunderstorm or gale occurs or one hour before it occurs, positive pressure changes generally occur, and the hourly pressure changes are generally around 2hPa, with a maximum of more than 5hPa.

[0121] In 90% of cases, thunderstorms and gale were observed together with precipitation, and in 10% of cases, there was no precipitation. In North China, most cases of thunderstorms and gale were accompanied by precipitation. Before the gale, the rainfall intensity exceeded 10mm / h in a few cases, not exceeding 20%, and the maximum rainfall intensity was 95.2mm / h. Fig. 9For the observational facts when thunderstorms and strong winds are accompanied by precipitation, the difference between the initial time when the wind speed at the station reaches the strong wind standard (gusts of level 8) (called the strong wind outbreak time) and the precipitation start time is given. Among them, positive values ​​indicate that the thunderstorm and strong wind outbreak precedes the ground precipitation; negative values, on the contrary; 0 means that strong gusts and precipitation are observed at the same time. The statistical results show that 42.78% of the cases have started precipitation before the thunderstorm and strong wind outbreak, 2.10% of the cases where strong winds and precipitation are observed at the same time, and 44.36% of the strong winds occur before the precipitation starts. Therefore, most of the strong wind cases in Tianjin are thunderstorm and strong wind processes accompanied by precipitation. 42.78% of the cases occurred in the front downdraft area and the rear downdraft area, with the characteristics of downburst. The cases affected by the passage of downburst and gust front are close.

[0122] 3.1.4 Spatial distribution

[0123] The high-frequency area of ​​thunderstorms and gale weather is consistent with the Taihang Mountains and Yinshan Mountains, and decreases from the western part of Beijing-Tianjin-Hebei to the east. Thunderstorms and gale weather are more common in the western mountainous areas of Beijing-Tianjin-Hebei, and less common in the central plains. The southeastern coastal area of ​​Hebei Province is also a high-incidence area of ​​thunderstorms and gale weather.

[0124] 3.1.5 Wind speed intensity and wind direction frequency

[0125] Statistics on wind speeds during thunderstorms and gale events in the past five years show that the maximum wind speed was 41.4m / s; among them, strong thunderstorms and gale events (>25m / s) occurred 167 times, accounting for 4.7%; ordinary thunderstorms and gale events (17-19m / s) occurred 1841 times, accounting for 53.8% of the total; the gusts of thunderstorms and gale events in North China were mostly below 20m / s, with average intensity.

[0126] The statistical frequency of thunderstorms and gale wind directions shows that thunderstorms and gale winds have occurred in all wind directions, among which north wind (including northerly wind, northeasterly wind and northwesterly wind) is the most common, accounting for 71.1% of the total. Due to the influence of topography, coastal areas have more thunderstorms and gale winds with easterly winds.

[0127] 3.2 Thunderstorm and strong wind convection environmental conditions

[0128] The convective environmental conditions were analyzed and diagnosed using sounding and ground automatic stations at 08:00 and 20:00 before the occurrence of thunderstorms and gale.

[0129] 3.2.1 Impact system

[0130] Based on the 500hPa weather influencing system (excluding typhoon systems), thunderstorms and gale in North China are mainly affected by the northeastern cold vortex, westerly trough and the edge of the subtropical high. When affected by the northeastern cold vortex and westerly trough, thunderstorms and gale occur in front of or behind the trough; thunderstorms and gale affected by the southerly airflow or shear line on the western edge of the western Pacific subtropical high (hereinafter referred to as the subtropical high) are classified as subtropical high edge type. According to statistics, from 2017 to 2021, the thunderstorm and gale influence systems are: cold vortex accounts for 41%, westerly trough accounts for 50%, and subtropical high and others account for 9%. Among them, the subtropical high edge type mainly appears in July and August, and the proportions of cold vortex and westerly trough are similar in all months.

[0131] 3.2.2 Convective parameter statistics

[0132] There is no sounding data in Tianjin, so in actual forecasts, forecasters usually use soundings from Beijing and Leting to analyze severe convective weather. But in fact, there are great differences in the boundary layer due to the different underlying surfaces. According to meteorological principles, when North China is controlled by the same weather system and the weather situation is relatively stable and seldom changes, the atmospheric structure of the two places outside the boundary layer is not much different, and sounding data above 850hPa can be used to analyze severe convective weather in Tianjin; while in the boundary layer, the thermal and dynamic structures are quite different, and the sounding data from Beijing or Leting cannot be directly applied for analysis. Therefore, the encrypted and corrected sounding data are formed by the hourly encrypted automatic station data of Tianjin and the sounding data from Beijing and Leting above 850hPa (the proximity principle is used when selecting sounding data), and various convective parameters are calculated. By analyzing the extreme values ​​12h before the occurrence of thunderstorms and gales, the forecast indicators and reference thresholds with indicative significance when thunderstorms and gales occur are refined.

[0133] According to the physical meaning of convective parameters and the forecaster's experience, the project uses K index (KI), Sha'a index (SI), 3-6km vertical wind shear as physical quantities to characterize the environmental background, as well as physical quantities with mesoscale characteristics: lift index (LI), best convective effective potential energy (BCAPE), 0-3km vertical wind shear and strong wind index (WINDEX). The reference threshold is obtained by comparing the distribution and changes of flow parameter samples. At the same time, since thunderstorms and strong winds in Tianjin are mainly affected by high-altitude cold vortices and high-altitude low troughs, the two weather types account for about 90% in total, so the physical quantity parameters of the two weather types are also counted.

[0134] 3.2.3 Constructing a conceptual model of thunderstorm and gale weather

[0135] Fig.10 The conceptual model of thunderstorm and gale weather is constructed. Under the favorable influencing system, the unstable environmental field cooperates with the convergence and lifting of the mesoscale system, which may lead to thunderstorm and gale weather.

[0136] 4. Build a probability forecast model for thunderstorm and gale weather based on AI algorithm.

[0137] Based on the constructed thunderstorm and gale weather forecast dataset for the North China Plain, classification training, 0-24 hour forecast, and algorithm verification and evaluation of thunderstorm and gale are carried out. Sample selection includes 0 station: no thunderstorm and gale. 3-11 stations: thunderstorm and gale occur in a certain area, and more than 12 stations: large-scale thunderstorm and gale.

[0138] Here, we will select XGboost, LightGBM, and Random Forest algorithms commonly used in the literature on severe convection classification forecasting to establish thunderstorm and gale weather forecast sub-models, and carry out comparative experiments. According to the advantages and disadvantages of the performance of the three models, we will provide a final multi-model integration solution, and then obtain a thunderstorm and gale weather forecast model to improve the model classification accuracy and model generalization ability.

[0139] 5. Construct an extreme thunderstorm and gale weather forecast dataset.

[0140] Extreme thunderstorm and gale data (for example, above level 10 or above level 12) are selected from ground observation data, and extreme thunderstorm and gale samples are constructed based on them, and then an extreme thunderstorm and gale weather forecast dataset is constructed.

[0141] 6. Construct a probability forecast model for extreme thunderstorm and gale weather based on the batching method.

[0142] According to the extreme thunderstorm and gale weather forecast data set, the physical quantity parameter values ​​in the n samples in the extreme thunderstorm and gale weather forecast data set can be converted into binary variables of 0 and 1 according to the physical quantity threshold and using the random probability idea (the occurrence of thunderstorm and gale is represented by "1" and "0" binary variables. When the physical quantity factor meets the threshold, the physical quantity is converted to 1; when it does not meet the threshold, the physical quantity is converted to 0), forming an n×k "0,1" matrix. Assuming that the function of the control weight is divided into two cases, f is 1 (meets the threshold) or 0 (does not meet the threshold), the extreme thunderstorm and gale weather forecast model can be expressed as:

[0143] P = f 1 × 1 +f 2 × 2 +f 3 × 3 +…+f k × k (1)

[0144] Where: x 1 , x 2 , …, x k(%) are the contribution rates of k forecast factors, and P is the probability of occurrence of extreme thunderstorms and gales. If the forecast factor meets the corresponding threshold, then f≡1, and the corresponding factor contribution rate is used as the probability of occurrence of extreme thunderstorms and gales. If the more forecast factors meet the threshold, the greater the cumulative value of the contribution rate, the more favorable the conditions for the occurrence of extreme thunderstorms and gales, and the higher the probability of occurrence; otherwise, the lower the probability.

[0145] Optionally, the evaluation indicators used here to evaluate the performance of the extreme thunderstorm and gale weather forecast model may be the hit rate (POD), the false alarm rate (FAR), the bias (Bias), the risk score (TS), and the fairness skill score (ETS).

[0146] 7. Relying on the TIPTOP integrated platform, the constructed thunderstorm and gale weather forecast model and extreme thunderstorm and gale weather forecast model will be commercialized

[0147] The multi-scale numerical forecast model products such as EC and CMA-meso are used to calculate the physical quantities required by the model and use them as input to drive the AI ​​model and the batching method model to make probability forecasts of thunderstorms and gale. At the same time, the thunderstorm and gale weather conceptual model is used to eliminate the forecast results, and finally an objective classification grid forecast product for thunderstorms and gale is generated.

[0148] Relying on the TIPTOP integrated platform, the hourly thunderstorm and gale probability forecast products are integrated into the platform's objective forecast and monitoring and early warning modules, providing objective probability forecast products for thunderstorms and gale with a spatial resolution of 1km and a temporal resolution of 1 hour for 0-24 hours, and realizing the visualization of grid probability forecast products, providing technical support and platform support for the fine monitoring, early warning and forecasting of thunderstorms and gale.

[0149] In summary, according to the embodiments of the present invention, at least the following technical effects are achieved:

[0150] Construct a thunderstorm and gale weather forecast dataset. The cluster analysis algorithm is used to extract thunderstorm and gale samples. In similar studies, the extraction of thunderstorm and gale samples mostly uses lightning data to perform quality control on gale observation data, while this project uses algorithms such as K-Means cluster analysis and time neighborhood expansion budget (TNDA) to classify the sample set in an unsupervised mode. The advancement lies in the more objective extraction method, simple and fast algorithm, and good classification effect.

[0151] For large sample sizes of thunderstorms and gale cases, an objective forecast model is established based on the AI ​​algorithm. For small sample sizes, a batching method forecast model is constructed. The forecast problems of different types of thunderstorms and gale are solved separately. The combined application of artificial intelligence algorithms and batching algorithms has the advantage over current business methods in that it can not only predict and classify multiple factors based on large data sets through intelligent screening and combination to obtain classified forecast results for severe convective weather, but also extract small sample extreme weather characteristics to obtain graded forecast products.

[0152] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present disclosure is not limited by the order of the actions described, because according to the present disclosure, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present disclosure.

[0153] The above is an introduction to the method embodiment. The following is a further explanation of the scheme disclosed in the present invention through an apparatus embodiment.

[0154] Fig.11 A structural diagram of a thunderstorm and gale weather forecast model training device provided by an embodiment of the present disclosure, such as Fig.11 As shown, the thunderstorm and gale weather forecast model training device 1100 may include:

[0155] The acquisition module 1110 is used to obtain ground observation data and ERA5 reanalysis data of the target area within a historical period.

[0156] The construction module 1120 is used to construct a thunderstorm and gale weather forecast dataset based on ground observation data and ERA5 reanalysis data; and to construct an extreme thunderstorm and gale weather forecast dataset based on ground observation data.

[0157] Construction module 1120 is also used to construct a thunderstorm and gale weather forecast model based on the AI ​​algorithm according to the thunderstorm and gale weather forecast data set; and to construct an extreme thunderstorm and gale weather forecast model based on the batching algorithm according to the extreme thunderstorm and gale weather forecast data set.

[0158] Understandably, Fig.11 Each module / unit in the thunderstorm and gale weather forecast model training device 1100 shown has the function of realizing Figure 1 The functions of the various steps in the thunderstorm and gale weather forecast model training method 100 shown can achieve their corresponding technical effects, and for the sake of brevity, they will not be repeated here.

[0159] Fig.12A structural diagram of a thunderstorm and gale weather forecast device provided in an embodiment of the present disclosure, such as Fig.12 As shown, the thunderstorm and gale weather forecast device 1200 may include:

[0160] The calculation module 1210 is used to calculate the thunderstorm and gale weather forecast factor data of the target area to be input into the thunderstorm and gale weather forecast model by using the multi-scale numerical forecast model product.

[0161] The forecast module 1220 is used to input the thunderstorm and gale weather forecast factor data into the thunderstorm and gale weather forecast model, and the thunderstorm and gale weather forecast model processes the thunderstorm and gale weather forecast factor data to obtain the thunderstorm and gale weather forecast result.

[0162] The calculation module 1210 is also used to calculate the extreme thunderstorm and gale weather forecast factor data of the target area to be input into the extreme thunderstorm and gale weather forecast model using the multi-scale numerical forecast model product if the thunderstorm and gale weather forecast result is an extreme thunderstorm and gale weather forecast result.

[0163] The forecast module 1220 is also used to input the extreme thunderstorm and gale weather forecast factor data into the extreme thunderstorm and gale weather forecast model, and the extreme thunderstorm and gale weather forecast model processes the extreme thunderstorm and gale weather forecast factor data to obtain the final extreme thunderstorm and gale weather forecast result.

[0164] The determination module 1230 is used to determine the thunderstorm and gale weather forecast result as the final thunderstorm and gale weather forecast result if the thunderstorm and gale weather forecast result is not an extreme thunderstorm and gale weather forecast result.

[0165] Among them, the thunderstorm and gale weather forecast model and the extreme thunderstorm and gale weather forecast model are obtained based on the training method as described above.

[0166] Understandably, Fig.12 Each module / unit in the thunderstorm and gale weather forecast device 1200 shown has the function of realizing Figure 2 The functions of the various steps in the thunderstorm and gale weather forecasting method 200 shown can achieve their corresponding technical effects, and for the sake of brevity, they will not be repeated here.

[0167] Fig.131 is a block diagram of an exemplary electronic device capable of implementing an embodiment of the present disclosure. Electronic device 1300 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 1300 may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0168] like Fig.13 As shown, the electronic device 1300 may include a computing unit 1301, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1302 or a computer program loaded from a storage unit 1308 into a random access memory (RAM) 1303. In the RAM 1303, various programs and data required for the operation of the electronic device 1300 may also be stored. The computing unit 1301, the ROM 1302, and the RAM 1303 are connected to each other via a bus 1304. An input / output (I / O) interface 1305 is also connected to the bus 1304.

[0169] Multiple components in the electronic device 1300 are connected to the I / O interface 1305, including: an input unit 1306, such as a keyboard, a mouse, etc.; an output unit 1307, such as various types of displays, speakers, etc.; a storage unit 1308, such as a disk, an optical disk, etc.; and a communication unit 1309, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1309 allows the electronic device 1300 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0170] The computing unit 1301 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1301 performs the various methods and processes described above, such as method 100 or method 200. For example, in some embodiments, method 100 or method 200 may be implemented as a computer program product, including a computer program, which is tangibly contained in a computer-readable medium, such as a storage unit 1308. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1300 via ROM 1302 and / or communication unit 1309. When the computer program is loaded into RAM 1303 and executed by the computing unit 1301, one or more steps of the method 100 or method 200 described above may be performed. Alternatively, in other embodiments, the computing unit 1301 may be configured to execute the method 100 or the method 200 in any other appropriate manner (eg, by means of firmware).

[0171] The various embodiments described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs, which may be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general programmable processor, which may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0172] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0173] In the context of the present disclosure, a computer-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a computer-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0174] It should be noted that the present disclosure also provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute method 100 or method 200 and achieve the corresponding technical effects achieved by the embodiments of the present disclosure executing the methods thereof. For the sake of brevity, they will not be repeated here.

[0175] In addition, the present disclosure also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the method 100 or the method 200 is implemented.

[0176] To provide interaction with a user, the above-described embodiments may be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0177] The embodiments described above can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0178] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0179] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0180] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A thunderstorm and gale weather forecast model training method, characterized in that: The method comprises: Obtain ground observation data and ERA5 reanalysis data for the target area within the historical period; Based on the ground observation data and the ERA5 reanalysis data, a thunderstorm and gale weather forecast dataset is constructed; based on the ground observation data, an extreme thunderstorm and gale weather forecast dataset is constructed; According to the thunderstorm and gale weather forecast data set, a thunderstorm and gale weather forecast model based on the AI ​​algorithm is constructed; according to the extreme thunderstorm and gale weather forecast data set, an extreme thunderstorm and gale weather forecast model based on the batching algorithm is constructed.

2. The method according to claim 1, characterized in that The thunderstorm and gale weather forecast data set is constructed based on the ground observation data and the ERA5 reanalysis data, including: Based on K-Means cluster analysis, manual labeling and time neighborhood expansion budget algorithm, the ground observation data is quality controlled to construct positive and negative samples of thunderstorm and gale weather; The ERA5 reanalysis data are processed to generate grid-by-grid and hourly atmospheric vertical profiles. Based on this, the physical quantities corresponding to the positive and negative samples of thunderstorm and gale weather are calculated based on the SHARPpy sounding analysis and diagnostic package, and the physical quantities are added to the corresponding samples. The subsequent samples are quality controlled based on the percentiles of the physical quantities, and the physical quantities of the samples after quality control are subjected to time lag processing. Based on the samples after time lag processing, a thunderstorm and gale weather forecast dataset is constructed.

3. The method according to claim 1, characterized in that The extreme thunderstorm and gale weather forecast data set is constructed based on the ground observation data, including: Extreme thunderstorm and gale data are selected from the ground observation data, and extreme thunderstorm and gale samples are constructed based on them, and then an extreme thunderstorm and gale weather forecast data set is constructed based on the extreme thunderstorm and gale samples.

4. The method according to claim 1, characterized in that The method of constructing a thunderstorm and gale weather forecast model based on an AI algorithm according to the thunderstorm and gale weather forecast data set includes: According to the thunderstorm and gale weather forecast data set, XGboost, LightGBM, and Random Forest are trained respectively to obtain the corresponding thunderstorm and gale weather forecast sub-models; A plurality of thunderstorm and gale weather forecast sub-models are integrated to obtain a thunderstorm and gale weather forecast model.

5. The method according to claim 4, characterized in that The multiple thunderstorm and gale weather forecast sub-models are integrated to obtain a thunderstorm and gale weather forecast model, including: Conduct comparative experiments on multiple thunderstorm and gale weather forecast sub-models to determine the performance data of multiple thunderstorm and gale weather forecast sub-models; Calculate weight coefficients of multiple thunderstorm and gale weather forecast sub-models according to performance data of multiple thunderstorm and gale weather forecast sub-models; According to the weight coefficients of multiple thunderstorm and gale weather forecast sub-models, multiple thunderstorm and gale weather forecast sub-models are integrated to obtain a thunderstorm and gale weather forecast model.

6. A thunderstorm and gale weather forecast method, characterized in that: The method comprises: Utilize multi-scale numerical forecast model products to calculate the thunderstorm and gale weather forecast factor data of the target area to be input into the thunderstorm and gale weather forecast model; Inputting the thunderstorm and gale weather forecast factor data into a thunderstorm and gale weather forecast model, and processing the thunderstorm and gale weather forecast factor data by the thunderstorm and gale weather forecast model to obtain a thunderstorm and gale weather forecast result; If the thunderstorm and gale weather forecast result is an extreme thunderstorm and gale weather forecast result, the extreme thunderstorm and gale weather forecast factor data of the target area to be input into the extreme thunderstorm and gale weather forecast model is calculated by using a multi-scale numerical forecast model product; Inputting the extreme thunderstorm and gale weather forecast factor data into the extreme thunderstorm and gale weather forecast model, and the extreme thunderstorm and gale weather forecast model processes the extreme thunderstorm and gale weather forecast factor data to obtain a final extreme thunderstorm and gale weather forecast result; If the thunderstorm and gale weather forecast result is not an extreme thunderstorm and gale weather forecast result, it is determined as the final thunderstorm and gale weather forecast result; Wherein, the thunderstorm and gale weather forecast model and the extreme thunderstorm and gale weather forecast model are obtained based on the thunderstorm and gale weather forecast model training method described in any one of claims 1-5.

7. A thunderstorm and gale weather forecast model training device, characterized in that: The device comprises: The acquisition module is used to obtain ground observation data and ERA5 reanalysis data of the target area in the historical time period; A construction module, for constructing a thunderstorm and gale weather forecast dataset based on the ground observation data and the ERA5 reanalysis data; and constructing an extreme thunderstorm and gale weather forecast dataset based on the ground observation data; The construction module is also used to construct a thunderstorm and gale weather forecast model based on the AI ​​algorithm according to the thunderstorm and gale weather forecast data set; and to construct an extreme thunderstorm and gale weather forecast model based on the batching algorithm according to the extreme thunderstorm and gale weather forecast data set.

8. A thunderstorm and gale weather forecast device, characterized in that: The device comprises: A calculation module is used to calculate the thunderstorm and gale weather forecast factor data of the target area to be input into the thunderstorm and gale weather forecast model by using the multi-scale numerical forecast model product; A forecast module, used for inputting the thunderstorm and gale weather forecast factor data into a thunderstorm and gale weather forecast model, and the thunderstorm and gale weather forecast model processes the thunderstorm and gale weather forecast factor data to obtain a thunderstorm and gale weather forecast result; The calculation module is further used to calculate the extreme thunderstorm and gale weather forecast factor data of the target area to be input into the extreme thunderstorm and gale weather forecast model by using the multi-scale numerical forecast model product if the thunderstorm and gale weather forecast result is an extreme thunderstorm and gale weather forecast result; The forecast module is further used to input the extreme thunderstorm and gale weather forecast factor data into the extreme thunderstorm and gale weather forecast model, and the extreme thunderstorm and gale weather forecast model processes the extreme thunderstorm and gale weather forecast factor data to obtain a final extreme thunderstorm and gale weather forecast result; A determination module, configured to determine the thunderstorm and gale weather forecast result as the final thunderstorm and gale weather forecast result if the thunderstorm and gale weather forecast result is not an extreme thunderstorm and gale weather forecast result; Wherein, the thunderstorm and gale weather forecast model and the extreme thunderstorm and gale weather forecast model are obtained based on the thunderstorm and gale weather forecast model training method described in any one of claims 1-5.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to make a computer execute the method according to any one of claims 1 to 7.

Images