Modeling method and device of precipitation prediction model, electronic equipment and storage medium

Abstract

The invention provides a precipitation prediction model modeling method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence and the like. The specific implementation scheme is as follows: based on a return result of a weather forecast center, acquiring circulation variable information of a specified area in each of at least two time periods before a specified moment; based on a pre-collected observation data set of the designated area, obtaining an observation precipitation amount of the designated area in a next time period after a designated moment; and training a precipitation prediction model of the specified area based on the circulation variable information of the specified area in each of at least two time periods before the specified moment and the observed precipitation of the specified area in the next time period after the specified moment. According to the technology disclosed by the invention, the accuracy of the trained rainfall prediction model can be effectively improved.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of computers, in particular to the technical field of artificial intelligence, and more particularly to a modeling method and device of a precipitation prediction model, an electronic device, and a storage medium. BACKGROUND

[0002] For some areas with complex climate conditions and dense population and industry, such as the Yangtze River Basin, drought and flood disasters have a huge impact on economic construction and social development, so the seasonal time scale prediction of summer drought and flood is extremely critical, which is related to the national economy and people's livelihood, and is also the focus of national disaster prevention and mitigation work.

[0003] For seasonal scale precipitation prediction of any specified area, the statistical prediction method can be used to realize precipitation prediction in the prior art. Specifically, a statistical prediction model is established by selecting factors that have an impact on the precipitation of the specified area during the flood season, and then the statistical prediction model is used for precipitation prediction. SUMMARY

[0004] The present disclosure provides a modeling method and device of a precipitation prediction model, an electronic device, and a storage medium.

[0005] According to an aspect of the present disclosure, a modeling method of a precipitation prediction model is provided, comprising:

[0006] Based on the return results of the weather forecast center, circulation variable information of a specified area in each of at least two time periods before a specified time is obtained;

[0007] Based on a pre-collected observation data set of the specified area, an observed precipitation of the specified area in a next time period after the specified time is obtained;

[0008] Based on the circulation variable information of the specified area in each of at least two time periods before the specified time and the observed precipitation of the specified area in the next time period after the specified time, a precipitation prediction model of the specified area is trained.

[0009] According to another aspect of the present disclosure, a modeling device of a precipitation prediction model is provided, comprising:

[0010] The variable acquisition module is configured to obtain circulation variable information of a specified area in each of at least two time periods before a specified time based on return results of a weather forecast center;

[0011] The precipitation acquisition module is configured to obtain an observed precipitation of the specified area in a next time period after the specified time based on a pre-collected observation data set of the specified area;

[0012] The training module is configured to train a precipitation prediction model of the specified area based on circulation variable information of the specified area in each of at least two time periods before a specified time and observed precipitation of the specified area in a next time period after the specified time.

[0013] According to still another aspect of the present disclosure, an electronic device is provided, comprising:

[0014] at least one processor; and

[0015] a memory communicatively connected to the at least one processor; wherein

[0016] The memory stores instructions executable 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 of the aspects and any possible implementation manner described above.

[0017] According to still another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, the computer instructions being used to cause the computer to perform the method of the aspects and any possible implementation manner described above.

[0018] According to still another aspect of the present disclosure, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the method of the aspects and any possible implementation manner described above.

[0019] According to the technology of the present disclosure, the accuracy of the trained precipitation prediction model can be effectively improved.

[0020] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present disclosure, nor to limit the scope of the present disclosure. Other features of the present disclosure will become apparent through the following description. BRIEF DESCRIPTION OF DRAWINGS

[0021] The accompanying drawings are used to better understand the present scheme and do not limit the present disclosure. Among them:

[0022] Figure 1 is a schematic diagram according to a first embodiment of the present disclosure;

[0023] Figure 2 is a schematic diagram according to a second embodiment of the present disclosure;

[0024] Figure 3 is a schematic diagram according to a third embodiment of the present disclosure;

[0025] Figure 4 is a schematic diagram according to a fourth embodiment of the present disclosure;

[0026] Figure 5 is a block diagram of an electronic device for implementing the method of the embodiments of the present disclosure. DETAILED DESCRIPTION

[0027] Exemplary embodiments of the present disclosure are described herein with reference to the accompanying drawings, in which various details of the embodiments of the present disclosure are set forth to assist in the understanding of the present disclosure. It should be appreciated that the embodiments described herein are merely exemplary and are not intended to limit the scope of the present disclosure. Therefore, it will be apparent to those of ordinary skill in the art that various changes and modifications can be made thereto without departing from the scope and spirit of the present disclosure. Also, for the sake of clarity and conciseness, the description below omits the description of well-known functions and structures.

[0028] It is apparent that the described embodiments are part of the embodiments of the present disclosure, but not all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present disclosure.

[0029] It should be noted that the terminal device involved in the embodiments of the present disclosure can include, but is not limited to, a mobile phone, a personal digital assistant (PDA), a wireless handheld device, a tablet computer, and the like. The display device can include, but is not limited to, a personal computer, a television, and the like.

[0030] In addition, the term "and / or" in this paper is only a description of the association relationship between the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent the existence of A alone, the existence of A and B together, and the existence of B alone. In addition, the character " / " in this paper generally represents an "or" relationship between the associated objects before and after it.

[0031] Figure 1 is a schematic diagram according to the first embodiment of the present disclosure; as Figure 1 shown, the embodiment provides a modeling method of a precipitation prediction model, which can specifically include the following steps:

[0032] S101, based on the return result of the weather forecast center, obtaining the circulation variable information of the specified region in each time period of at least two time periods before the specified time;

[0033] The execution subject of the modeling method of the precipitation prediction model of the embodiment can be a modeling device of the precipitation prediction model, which can be an electronic entity, or can also be an application integrated by software.

[0034] The weather forecast center of the embodiment can be the European Centre for Medium-Range Weather Forecasts (ECMWF), which can provide global meteorological data for scientific research, climate modeling, and weather forecasting.

[0035] In the embodiment, the data can be pre-ordered from the ECMWF, and the results returned by the weather forecast center can be obtained. In actual application, data can be ordered from other weather forecast centers to obtain corresponding return results.

[0036] In actual application, real-time weather data can be obtained from the return results obtained from the ECMWF, and weather data of any specified region in a historical period can also be obtained from the ECMWF. Of course, the historical period here refers to a period that has been collected and stored by the ECMWF.

[0037] In actual application, the data in the return results of the weather forecast center is identified by longitude and latitude, and the longitude and latitude can be accurate to 1 degree. As for the selection of data of the specified region, the longitude and latitude of the specified region can be referred to to obtain the data of the specified region.

[0038] The circulation variable information of the embodiment can include the name of the circulation variable, that is, the field name, and can also include the value corresponding to the field. The circulation variable used in the embodiment can refer to a circulation variable related to precipitation.

[0039] In the embodiment, the circulation variable information obtained based on the return results of the weather forecast center can be original data of the weather forecast center, or data obtained by simple reasoning and calculation based on the original data of the weather forecast center. In summary, the accuracy and reliability of the obtained circulation variable information can be effectively ensured.

[0040] S102, based on the pre-collected observation data set of the specified region, obtaining the observed precipitation of the specified region in the next time period after the specified time;

[0041] The observed precipitation of the embodiment refers to the real precipitation data observed at each historical time in the specified region.

[0042] In the embodiment, a professional monitoring tool can be used to monitor the precipitation of each time period in the specified region and generate an observation data set of the specified region. When used, the observed precipitation of the specified region in the next time period after the specified time can be obtained from the observation data set of the specified region, which can ensure the accuracy and effectiveness of the obtained key precipitation.

[0043] S103, training the precipitation prediction model of the specified area based on the circulation variable information of the specified area in each of the at least two time periods before the specified time and the observed precipitation of the specified area in the next time period after the specified time.

[0044] In the embodiment, the accuracy of the circulation variable information obtained from the ECMWF is very high, and the data of the observed precipitation set observed locally is also real and effective. Therefore, training the precipitation prediction model of the specified area based on the circulation variable parameters of the specified area in each of the at least two time periods before the specified time obtained from the ECMWF and the observed precipitation of the specified area in the next time period after the specified time obtained based on the observed data set can effectively ensure the accuracy of the training data source, and further effectively improve the accuracy of the trained precipitation prediction model.

[0045] The length of the time period of the embodiment is the time scale. It can be set according to experience or demand, for example, it can be one month, one week, one day or several hours, that is, according to the implementation principle of the embodiment, a precipitation prediction model of a prediction period of multiple time scales can be modeled.

[0046] The precipitation prediction model of the embodiment can obtain the circulation variable information of the specified area in each of the at least two time periods before the specified time from the weather forecasting center, which can sufficiently ensure the accuracy and reliability of the circulation variable information, and obtain the observed precipitation of the specified area in the next time period after the specified time from the observed data set, which can sufficiently ensure the authenticity of the first observed precipitation data. Further, when the obtained circulation variable information of the specified area in each of the at least two time periods before the specified time and the observed precipitation of the specified area in the next time period after the specified time are used to train the precipitation prediction model of the specified area, the accuracy of the trained precipitation prediction model can be effectively improved.

[0047] Figure 2 is a schematic view according to the second embodiment of the disclosure; the modeling method of the precipitation prediction model of the embodiment can be used to train the precipitation prediction model of the specified area based on the circulation variable information of the specified area in each of the at least two time periods before the specified time obtained from the weather forecasting center and the observed precipitation of the specified area in the next time period after the specified time obtained based on the observed data set. Figure 1 The technical solution of the embodiment is described in more detail as follows. Figure 2 The modeling method of the precipitation prediction model of the embodiment can specifically include the following steps:

[0048] S201, obtaining the circulation variable information of the specified area in each of the at least two time periods before the specified time based on the return result of the weather forecasting center;

[0049] The specific implementation of this step is the same as the above Figure 1The step S101 of the illustrated embodiment is the same.

[0050] For example, the following Table 1 is some circulation variables provided by the present embodiment:

[0051] Table 1

[0052]

[0053]

[0054] As shown in Table 1, the circulation variables of the present embodiment can include pressure-level and single-level. For example, the circulation variables of pressure-level can include at least one of the variables corresponding to each preset pressure height in at least two preset pressure heights (such as 200 hPa, 850 hPa, etc.), such as zonal wind field, meridional wind field, specific humidity field, temperature field, and potential height. The circulation variables of single-level can include at least one of the variables of a specified height, such as temperature of 2m, mean sea level pressure, etc.

[0055] The specified region of the present embodiment can be demarcated based on the longitude and latitude where the region is located. For example, for the Yangtze River Basin region, the region with longitude 70°E-140°E and latitude 15°N-45°N can be focused on, which can cover the Yangtze River Basin and its associated circulation influence region.

[0056] In the present embodiment, the above-mentioned circulation variable information is very rich, which can provide effective data support for the training of the precipitation prediction model, and thus can effectively improve the accuracy of the trained precipitation prediction model.

[0057] S202, based on the return result of the weather forecasting center, obtaining the theoretical precipitation of the specified region in the next time period after the specified time;

[0058] Similarly, the weather forecasting center of the present embodiment can also be ECMWF. For the return result of ECMWF, the theoretical precipitation of any historical time period of the specified region can also be accurately analyzed. Since this precipitation is obtained based on the theoretical analysis of the weather forecasting center, rather than observed in the specified region. Therefore, it is called theoretical precipitation.

[0059] In addition, the return result of the weather forecasting center does not directly identify any region, but identifies data such as circulation variable information and precipitation in grid points. Each grid point corresponds to a longitude range and a dimension range, for example, each grid point can identify a 1°x1° grid.

[0060] For a specified region, such as the Yangtze River Basin, a boundary polygon can be constructed based on the shapefile of its boundary. Then, all the grid points near the obtained polygon are traversed, and whether the center of each grid point is within the polygon is judged in turn. This judgment process can use the ray method, that is, an arbitrary ray is made from the center of the current grid point, and the number of intersection points of the ray with the boundary of the polygon is counted. If the number of intersection points is odd, it is determined that the grid point is within the polygon, and it is saved to an array; otherwise, it is determined that the grid point is outside the polygon. Finally, since the center of some grid points in coastal areas may be located on the sea surface and thus have data default problems, even if these grid points are within the Yangtze River Basin, they also need to be filtered out to ensure that the grid point data within the specified region obtained finally is complete and meets the requirements.

[0061] Then, based on the data of all the grid points included in the specified region, the data of the specified region is obtained. For example, based on the circulation variable information of each grid point in the specified region, a certain mathematical operation such as taking the average value can be performed to obtain the circulation variable information of the specified region. Similarly, based on the theoretical precipitation of each grid point in the specified region, a certain mathematical operation such as accumulation can also be performed to obtain the theoretical precipitation of the specified region.

[0062] Correspondingly, for the acquisition of the value of each circulation variable in each time period of the specified region, the value of the circulation variable of each grid point included in the specified region in the time period can be obtained based on the output result of the weather forecasting center, and then the average value of the values of the circulation variables of the grid points included in the specified region in the time period is taken as the value of the circulation variable of the specified region in the time period. In this way, the value of each circulation variable of the specified region in each time period can be obtained.

[0063] Correspondingly, for the acquisition of the theoretical precipitation of the specified region in each time period, the theoretical precipitation of each grid point included in the specified region in the time period can be obtained based on the output result of the weather forecasting center, and then the theoretical precipitations of the grid points included in the specified region in the time period are accumulated as the theoretical precipitation of the specified region in the time period.

[0064] In the above manner, the circulation variable information of the specified region in each time period before the specified time can be accurately obtained, and the theoretical precipitation of the specified region in the next time period after the specified time can also be accurately obtained

[0065] The data obtained in this embodiment is used for model training, and the obtained data is all historical data, including the next time period after the specified time. Therefore, by using this step, the theoretical precipitation of the specified region in the next time period after the specified time can be accurately obtained.

[0066] Optionally, before step S201, the following steps can also be included: obtaining the return results from ECMWF, then selecting the data of the circulation variables related to precipitation from the data, and performing preprocessing, such as noise removal of some data, standardization processing, unit conversion, etc., which are not limited here.

[0067] S203, training the pre-trained model based on the circulation variable information of the specified region in each time period of at least two time periods before the specified time and the theoretical precipitation of the specified region in the next time period after the specified time;

[0068] The circulation variable information of the specified region in each time period of at least two time periods before the specified time and the theoretical precipitation of the specified region in the next time period after the specified time are used as training data, a plurality of similar training data are collected to construct a training set, and the pre-trained model is trained. The circulation variable information of the specified region in each time period of at least two time periods before the specified time is used as input, and the theoretical precipitation of the specified region in the next time period after the specified time is used as a supervision label. For example, during training, the circulation variable information of the specified region in each time period of at least two time periods before the specified time can be input into the pre-trained model, and the pre-trained model can predict the predicted precipitation of the specified region in the next time period after the specified time based on the input information; then a loss function is constructed for the pre-trained model based on the predicted precipitation and the theoretical precipitation, and the parameters of the pre-trained model are adjusted based on the loss function to make the loss function converge. The pre-trained model is trained using a plurality of similar training data.

[0069] The pre-trained model is trained in the above manner, and after the training is completed, the pre-trained model can also be verified using a verification set and tested using a test set.

[0070] In this embodiment, the training set collected can select the data of ECMWF from 1991 to 2011, correspondingly, the verification set can select the data from 2012 to 2016, and the test set can select the data from 2017 to 2021. In terms of data organization and processing, a three-month sliding window strategy can be used to select seasonal samples, and the circulation variable input of each season is the average value of three months, and the precipitation is the cumulative total of three-month precipitation. Two loss functions can be used during training of the pre-trained model: mean squared error function (Mean Squared Error; MSE) and cosine distance function, and the two loss functions are combined according to a certain proportion, i.e. Loss = L MSE + αL cos , where is a weight parameter, and the prediction effect is better when a is 1. The MSE loss is a necessary loss function in regression prediction, but only the MSE loss will cause the optimization of the model to fall into a local optimal solution, that is, the model is easy to predict an optimal solution suitable for each year, for example, different inputs for each year will predict a similar result, and relatively good results are obtained in the numerical prediction of each year. However, the spatial distribution cannot be correctly predicted. The addition of the cosine distance function enhances the model's ability to fit the spatial type.

[0071] In this embodiment, the circulation variable information of the specified area in each time period before the specified time point and the theoretical precipitation of the specified area in the next time period after the specified time point are all obtained based on the return results of the weather forecast center, which can fully guarantee the authenticity and reliability of the training data, and further effectively improve the accuracy of the pre-training model.

[0072] The pre-training model of this embodiment adopts an innovative dual-module architecture, which combines an adapter module and an improved Residual Network (ResNet) 50, aiming to effectively process meteorological circulation data and improve the accuracy of precipitation prediction.

[0073] The adapter module includes a 3x3 convolution layer, batch normalization (BN), and a convolution block attention module (CBAM), and specifically can include three repeated structures. The 3x3 convolution layer can keep the feature map size unchanged; the BN accelerates the training convergence; and the CBAM can use the dual mechanisms of channel attention and spatial attention to respectively obtain channel statistical information from global average / max pooling and learn spatial importance after channel dimension pooling and convolution, realizing joint modeling of spatial and channel feature maps.

[0074] ResNet50: Load ImageNet pre-training weights and replace the fully connected layer to adapt to the precipitation prediction task.

[0075] In the training, the parameters of ResNet50 except the fully connected layer are frozen to retain the bottom feature extraction capability and prevent small data overfitting, while keeping the adapter trainable to learn to adapt the meteorological data to the ResNet feature space.

[0076] The overall workflow of the pre-training model is as follows: the input data is first converted and enhanced by the adapter, then sent to the ResNet50 network with pre-training weights for deep feature extraction, and finally output the prediction result through the fully connected layer.

[0077] During training, the data is sequentially processed through the feature enhancement of the adapter, the deep feature extraction of ResNet50, and the fully connected layer precipitation prediction after standardization. The entire system is trained end-to-end, and has the advantages of feature alignment enhancement of the CBAM attention mechanism, the transfer learning ability of the pre-trained model, and the anti-overfitting effect of the parameter freezing strategy, and is particularly suitable for processing meteorological circulation data with spatial correlation.

[0078] The pre-trained model trained in this embodiment is trained based on the data in the return results of the weather forecast center, has very strong universality, and can be applied to the prediction of precipitation in any specified region in the world.

[0079] S204, based on the pre-trained model trained in advance for the specified region, a precipitation prediction model for the specified region is constructed;

[0080] For example, when this step is implemented, it can include the following steps:

[0081] (1) Migrate the main structure of the pre-trained model for the specified region to the precipitation prediction model for the specified region as the main structure of the precipitation prediction model for the specified region;

[0082] (2) Construct the fully connected layer of the precipitation prediction model for the specified region.

[0083] The main structure of the precipitation prediction model in this embodiment is migrated from the pre-trained model, that is, the precipitation prediction model also includes an adapter module and ResNet50. The difference is that the fully connected layer is different from that of the pre-trained model, and before use, the fully connected layer of the precipitation prediction model needs to be adaptively trained.

[0084] In this embodiment, by migrating the main structure of the pre-trained model as the main structure of the precipitation prediction model, the construction cost of the precipitation prediction model can be effectively reduced, the accuracy of the constructed precipitation model can be improved, and the training cost of the subsequent precipitation prediction model can be reduced, and the training efficiency can be improved.

[0085] S205, based on the pre-collected observation data set of the specified region, the observation precipitation of the specified region in the next time period after the specified time is obtained;

[0086] The observation data set of this embodiment is based on real data collected locally in the specified region, and compared with the theoretical precipitation of the specified region in the next time period after the specified time obtained in the above step S202, the data is more accurate.

[0087] S206, a first spatial resolution corresponding to the theoretical precipitation is obtained;

[0088] For example, based on the ECMWF return result, when the theoretical precipitation of a specified area is obtained, all grid data can be uniformly interpolated to a 1°x1° grid. That is, the first spatial resolution is 1°x1°, that is, the longitude occupied by each grid is 1, and the latitude is also 1.

[0089] S207, obtaining a second spatial resolution corresponding to the observation data set;

[0090] The observation data set of the embodiment can use the grid precipitation data of CN05.1, and the spatial resolution is 0.25°x0.25°, that is, the longitude occupied by each grid is 0.25, and the latitude is also 0.25.

[0091] S208, detecting whether the first spatial resolution and the second spatial resolution are equal; if not, performing step S209; if equal, performing step S210;

[0092] S209, based on the mapping relationship between the first spatial resolution and the second spatial resolution, mapping the observation precipitation of the specified area in the next time period after the specified time to the space corresponding to the first spatial resolution; performing step S210;

[0093] As in the above embodiment, the first spatial resolution corresponding to the theoretical precipitation is 1°x1°, and the second spatial resolution corresponding to the observation data set is 0.25°x0.25°. At this time, the observation precipitation can be mapped to the space corresponding to the first spatial resolution corresponding to the theoretical precipitation. For example, the four 0.25°x0.25° grid data adjacent to the upper and lower left are interpolated in a 1°x1° grid data, and then the unit uniform precipitation data is obtained, so that the main structure of the migrated precipitation prediction model can more effectively process the observation precipitation, and the predicted precipitation prediction model is more accurate.

[0094] S210, based on the circulation variable information of the specified area in each of the at least two time periods before the specified time and the observation precipitation of the specified area in the next time period after the specified time, adjusting the parameters of the full connection layer of the precipitation prediction model of the specified area, and keeping the main structure parameters of the precipitation prediction model migrated from the pre-training model frozen.

[0095] That is, in the training process of the precipitation prediction model, the main structure of the precipitation prediction model also uses the main structure of the pre-training model, and the parameters are frozen. The whole training process only adjusts the parameters of the full connection layer. The training principle is the same as the training principle of the above pre-training model, and will not be described here.

[0096] The migration learning of the embodiment only updates and optimizes the parameters of the full connection layer of the precipitation prediction model, so as to make full use of the limited observed precipitation, avoid the risk of overfitting, effectively reduce the training difficulty, and improve the training effect.

[0097] Optionally, in the embodiment, the precipitation prediction model can be used for the precipitation of the specified region in the rainy season (which can be in summer), but during training, the observed precipitation of the whole year can be used for training, thereby improving the generalization and stability of the model prediction, and avoiding learning to extreme samples.

[0098] Moreover, the supervision label of the precipitation prediction model in the embodiment is the observed precipitation of the specified region, which is real and reliable, and can effectively improve the accuracy of the trained precipitation prediction model.

[0099] In the field of weather prediction, the precipitation prediction model of the present disclosure can be used to analyze meteorological data and predict precipitation, helping the meteorological department to issue weather warnings in advance and reduce the loss of natural disasters. Specifically, first, historical meteorological data is collected, including atmospheric circulation data, temperature, humidity, etc. Then, these data are cleaned and standardized to meet the requirements of model input. Then, the preprocessed data is used to train the model, and the parameters are adjusted to improve the prediction accuracy. After that, real-time meteorological data is input into the trained model to generate meteorological prediction results for a certain period of time in the future. Finally, the meteorological department issues weather warnings in advance according to these prediction results, guiding the public and relevant departments to take appropriate preventive measures.

[0100] Figure 3 is a schematic diagram according to the third embodiment of the present disclosure; as Figure 3 shown, the embodiment provides a modeling device 300 of a precipitation prediction model, comprising:

[0101] The variable acquisition module 301 is configured to acquire circulation variable information of a specified region in each of at least two time periods before a specified time based on a return result of a weather forecasting center.

[0102] The precipitation acquisition module 302 is configured to acquire observed precipitation of the specified region in a next time period after the specified time based on a pre-acquired observation data set of the specified region.

[0103] The training module 303 is configured to train a precipitation prediction model of the specified region based on the circulation variable information of the specified region in each of the at least two time periods before the specified time and the observed precipitation of the specified region in the next time period after the specified time.

[0104] The modeling device 300 of the precipitation amount prediction model of the embodiment, by adopting the above-mentioned modules, realizes the modeling of the precipitation amount prediction model, and the implementation principle and technical effects are the same as those of the above-mentioned related method embodiments. For details, refer to the description of the above-mentioned related method embodiments, which will not be repeated here.

[0105] Figure 4 is a schematic diagram according to the fourth embodiment of the present disclosure; the modeling device 400 of the precipitation amount prediction model of the embodiment, on the basis of the technical solutions of the above-mentioned Figure 3 further describes the technical solutions of the present disclosure in more detail. As shown in Figure 4 The modeling device 400 of the precipitation amount prediction model of the embodiment includes the same name and function modules as shown above: the variable acquisition module 401, the precipitation amount acquisition module 402, and the training module 403. Figure 3

[0106] In the embodiment, the circulation variable information includes at least one of a zonal wind field, a meridional wind field, a specific humidity field, a temperature field, and a potential height corresponding to the pressure level of each preset air pressure height in at least two preset air pressure heights;

[0107] The circulation variable information further includes at least one of the temperature and the average sea level pressure of the specified height at the ground level.

[0108] Further optionally, as shown in Figure 4 The modeling device 400 of the precipitation amount prediction model of the embodiment further includes:

[0109] The construction module 404 is configured to construct the precipitation amount prediction model of the specified region based on the pre-trained model of the specified region.

[0110] Further optionally, in an embodiment of the present disclosure, the construction module 404 is configured to:

[0111] migrate the main structure of the pre-trained model of the specified region to the precipitation amount prediction model of the specified region as the main structure of the precipitation amount prediction model of the specified region;

[0112] construct the fully connected layer of the precipitation amount prediction model of the specified region.

[0113] Further optionally, in an embodiment of the present disclosure, the training module 403 is configured to:

[0114] ​The parameters of the full connection layer of the precipitation amount prediction model are adjusted based on the circulation variable information of the specified area in each of at least two time periods before the specified time and the observed precipitation amount of the specified area in a next time period after the specified time, while keeping the main structure parameters migrated from the pre-training model in the precipitation amount prediction model frozen.

[0115] Further optionally, as shown in the one embodiment of the present disclosure, Figure 4 The modeling device 400 of the precipitation amount prediction model further includes, in one embodiment of the present disclosure,

[0116] The precipitation amount obtaining module 405 is further configured to obtain the theoretical precipitation amount of the specified area in a next time period after the specified time based on the return result of the weather forecast center.

[0117] The training module 404 is further configured to train the pre-training model based on the circulation variable information of the specified area in each of at least two time periods before the specified time and the theoretical precipitation amount of the specified area in a next time period after the specified time.

[0118] Further optionally, as shown in the one embodiment of the present disclosure, Figure 4 The modeling device 400 of the precipitation amount prediction model further includes, in one embodiment of the present disclosure,

[0119] The resolution obtaining module 406 is configured to obtain a first spatial resolution corresponding to the theoretical precipitation amount.

[0120] The resolution obtaining module 406 is further configured to obtain a second spatial resolution corresponding to the observation data set.

[0121] The detection module 407 is configured to detect whether the first spatial resolution and the second spatial resolution are equal.

[0122] The mapping module 408 is configured to, in response to the first spatial resolution and the second spatial resolution being not equal, map the observed precipitation amount of the specified area in a next time period after the specified time to a space corresponding to the first spatial resolution based on a mapping relationship between the first spatial resolution and the second spatial resolution.

[0123] The modeling device 400 of the precipitation amount prediction model of the present embodiment realizes the modeling of the precipitation amount prediction model by using the above modules, and has the same implementation principle and technical effects as the above related method embodiments. For details, refer to the description of the above related method embodiments, which will not be repeated here.

[0124] In the technical solutions of the present disclosure, the acquisition, storage and application of user personal information comply with relevant laws and regulations and do not violate public order and good customs.

[0125] According to embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0126] Figure 5 A schematic block diagram of an example electronic device 500 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the present disclosure described and / or claimed in this document.

[0127] As shown in Figure 5 The device 500 includes a computing unit 501 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 502 or a computer program loaded into a random access memory (RAM) 503 from a storage unit 508. Various programs and data required for the operation of the device 500 can also be stored in the RAM 503. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0128] Various components in the device 500 are connected to the I / O interface 505, including an input unit 506, such as a keyboard, a mouse, etc., an output unit 507, such as various types of displays, speakers, etc., a storage unit 508, such as a magnetic disk, an optical disk, etc., and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the device 500 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0129] The computing unit 501 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 501 performs various methods and processes described above, such as the above-described methods of the present disclosure. For example, in some embodiments, the above-described methods of the present disclosure can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 508. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded onto the RAM 503 and executed by the computing unit 501, one or more steps of the above-described methods of the present disclosure described above can be performed. Alternatively, in other embodiments, the computing unit 501 can be configured to perform the above-described methods of the present disclosure by any other appropriate means, such as by means of firmware.

[0130] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a complex programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0131] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces a means for implementing the functions / acts specified in the flowcharts and / or block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine and partially on a remote machine or entirely on a remote machine or server.

[0132] In the context of this disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, 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 disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0133] To provide for interaction with a user, the systems and techniques described here can 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 a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0134] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0135] The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0136] It should be understood that the various forms of flow shown above can be used to reorder, add, or remove steps. For example, the steps described in the present disclosure can be performed in parallel, in series, or in a different order, as long as the desired results of the technology disclosed in the present disclosure are achieved, which is not limited herein.

[0137] The above detailed description does not constitute a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A modeling method for a precipitation prediction model, comprising: Based on the reports from the weather forecast center, obtain circulation variable information for a specified area within each of the specified time periods, at least two time periods prior to a specified time. Based on the pre-collected observation dataset of the specified area, obtain the observed precipitation of the specified area in the next time period after the specified time. The precipitation prediction model for the specified area is trained based on the circulation variable information of the specified area in each of the at least two time periods before the specified time and the observed precipitation of the specified area in the next time period after the specified time.

2. The method according to claim 1, wherein, The circulation variable information includes at least one of the following: zonal wind field, meridional wind field, specific humidity field, temperature field, and geopotential height corresponding to each of the at least two preset pressure heights; The circulation variable information also includes at least one of the following: temperature at a specified altitude at ground level and mean sea level pressure.

3. The method according to claim 1, wherein, Before obtaining circulation variable information for a specified area within at least two time periods prior to a specified time, based on reports from weather forecast centers, the method further includes: Based on a pre-trained model pre-trained for the specified area, a precipitation prediction model for the specified area is constructed.

4. The method according to claim 3, wherein, Based on a pre-trained model pre-trained for the specified area, a precipitation prediction model for the specified area is constructed, including: The main structure of the pre-trained model in the specified region is transferred to the precipitation prediction model in the specified region, and used as the main structure of the precipitation prediction model in the specified region; Construct a fully connected layer for the precipitation prediction model of the specified region.

5. The method according to claim 4, wherein, Based on circulation variable information of the specified region in at least two time periods before the specified time, and observed precipitation in the specified region in the next time period after the specified time, a precipitation prediction model for the specified region is trained, including: Based on the circulation variable information of the specified region in each of the at least two time periods before the specified time, and the observed precipitation in the specified region in the next time period after the specified time, the parameters of the fully connected layer of the precipitation prediction model for the specified region are adjusted, while keeping the main structural parameters transferred from the pre-trained model in the precipitation prediction model frozen.

6. The method according to claim 3, wherein, Before constructing the precipitation prediction model for the specified region based on a pre-trained model pre-trained for the specified region, the method further includes: Based on the report from the weather forecast center, obtain the theoretical precipitation for the specified area in the next time period after the specified time. The pre-trained model is trained based on the circulation variable information of the specified region in each of the at least two time periods before the specified time, and the theoretical precipitation of the specified region in the next time period after the specified time.

7. The method according to claim 6, wherein, Before training the precipitation prediction model for the specified region based on circulation variable information of the specified region in each of the at least two time periods before the specified time and observed precipitation in the specified region in the next time period after the specified time, the method further includes: Obtain the first spatial resolution corresponding to the theoretical precipitation; Obtain the second spatial resolution corresponding to the observation dataset; Detect whether the first spatial resolution and the second spatial resolution are equal; When the first spatial resolution and the second spatial resolution are not equal, based on the mapping relationship between the first spatial resolution and the second spatial resolution, the observed precipitation in the specified area within the next time period after the specified time is mapped to the space corresponding to the first spatial resolution.

8. A modeling apparatus for a precipitation prediction model, comprising: The variable acquisition module is used to acquire circulation variable information of a specified area within each of the time periods at least two time periods before a specified time, based on the report results from the weather forecast center. The precipitation acquisition module is used to acquire the observed precipitation in the specified area within the next time period after a specified time, based on the pre-collected observation dataset of the specified area. The training module is used to train a precipitation prediction model for the specified region based on circulation variable information of the specified region in each of the time periods at least two time periods before the specified time, and the observed precipitation of the specified region in the next time period after the specified time.

9. An electronic device, comprising: At least one processor; as well as 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 to enable the at least one processor to perform the method of any one of claims 1-7.

10. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-7.

11. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-7.

Images