Rainfall prediction method, system and device and storage medium

Abstract

The invention discloses a rainfall prediction method, system and device, and a storage medium. The method comprises the following steps: obtaining a first historical rainfall data set of a ground observation station and a second historical rainfall data set of a radar observation station; determining first predicted rainfall data of the current day in the background field based on the background field data and the first historical rainfall data set, and determining second predicted rainfall data after the current day in the background field based on the background field data and the second historical rainfall data set; inputting the first predicted rainfall data and the second historical rainfall data set into a convolutional neural network to obtain third predicted rainfall data; adjusting parameters of a convolutional neural network according to the second predicted rainfall data and the third predicted rainfall data to obtain a trained convolutional neural network; and inputting rainfall data before the prediction day into the trained convolutional neural network to obtain target rainfall data of the prediction day. The method can be widely applied to the technical field of artificial intelligence.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a rainfall prediction method, system, device and storage medium. Background Art

[0002] Precipitation prediction is a complex and important research task. Its technical background covers multiple fields such as meteorology, statistics, satellite remote sensing technology, machine learning and deep learning, and a variety of mature technologies have been developed, such as numerical weather forecasting, meteorological radar technology, satellite remote sensing technology, etc. Meteorological radar technology uses echo intensity to monitor rainfall distribution in real time and realizes short-term forecasting through radar puzzle, but it is mainly suitable for short-term forecasting; satellite remote sensing technology provides information on the spatiotemporal distribution of clouds and precipitation through polar orbits and geostationary satellites, which supplements the insufficiency of radar data. In meteorology, rainfall prediction mainly relies on theoretical models based on atmospheric dynamics and thermodynamics. These models simulate physical variables such as atmospheric motion, humidity, and temperature through numerical weather forecasting (NWP), and realize long-term and large-scale predictions through high-performance computing. On the other hand, early statistical methods, such as time series analysis and linear regression, also provided preliminary solutions for rainfall prediction by establishing the relationship between historical data and rainfall. With the development of observation technology, meteorological satellites and radars have gradually become important data sources for rainfall forecasting. Satellites can provide macro information on clouds and humidity, while radars can capture local precipitation intensity. The combination of the two has greatly improved the accuracy of rainfall forecasting.

[0003] For different regions and time ranges, the methods of rainfall prediction are different. For short time (0-6 hours) and areas with relatively stable climate, some relatively simple models can complete the prediction task well. However, for areas with rapid changes in atmospheric state and longer time ranges (4-6 hours or more), the predicted physical model is required to have strong simulation capabilities, such as simulating precipitation processes. Such models belong to the category of numerical weather prediction (NWP). NWP models simulate physical changes through partial differential equations and approximate related equations through numerical methods. As one of the NWP models, the Weather Research and Forecasting Model (WRF) can provide parameter configurations for simulation of different physical processes, making the simulation conditions more similar to the corresponding regions. For different regions, the optimal grid resolution and WRF model parameter configuration may be quite different and need to be tried many times. For example, in order to obtain the best precipitation forecast in British Columbia, Canada, relevant researchers have studied more than 100 possible model configurations; some studies have also found that in some complex terrains, high-resolution (3 km) grids are better for predicting precipitation values, but low-resolution (27 km) grids perform better overall.

[0004] In the related technologies, rainfall prediction technology shows a diversified development trend of combining traditional physical models with modern data-driven models. The traditional numerical weather forecast (NWP) method is based on the principles of atmospheric dynamics and thermodynamics and can simulate rainfall conditions over a large range and over a long time scale. However, due to its high computational complexity and sensitivity to initial conditions, its accuracy in local rainfall and short-term rainfall prediction is limited. Therefore, there are still technical problems that need to be solved in the related technologies. Summary of the invention

[0005] The purpose of this application is to solve one of the technical problems existing in the prior art to at least a certain extent.

[0006] To this end, an object of an embodiment of the present application is to provide a rainfall prediction method, system, device and storage medium, which can improve the accuracy of rainfall prediction.

[0007] In order to achieve the above-mentioned technical objectives, the technical solution adopted by the embodiment of the present application includes: a rainfall prediction method, comprising the following steps: obtaining a first historical rainfall data set of a ground observation station and a second historical rainfall data set of a radar observation station; wherein the first historical rainfall data set is a set of historical rainfall data observed by the ground observation station before the current day; the second historical rainfall data set is a set of the first historical rainfall data set and the rainfall data of the current day; based on the background field data and the first historical rainfall data set, determining the first predicted rainfall data of the current day under the background field, and based on the background field data and the second historical rainfall data set, determining the second predicted rainfall data after the current day under the background field; inputting the first predicted rainfall data and the second historical rainfall data set into a convolutional neural network to obtain third predicted rainfall data after the current day; adjusting the parameters of the convolutional neural network according to the second predicted rainfall data and the third predicted rainfall data to obtain a trained convolutional neural network; inputting the rainfall data before the prediction day into the trained convolutional neural network to obtain the target rainfall data for the prediction day.

[0008] The present application can obtain a first historical rainfall data set of a ground observation station and a second historical rainfall data set of a radar observation station; wherein the first historical rainfall data set is a set of historical rainfall data observed by the ground observation station before the current day; the second historical rainfall data set is a set of the first historical rainfall data set and the rainfall data of the current day; based on the background field data and the first historical rainfall data set, the first predicted rainfall data of the current day under the background field is determined, and based on the background field data and the second historical rainfall data set, the second predicted rainfall data after the current day under the background field is determined; the first predicted rainfall data and the second historical rainfall data set are input into a convolutional neural network to obtain the third predicted rainfall data after the current day; according to the second predicted rainfall data and the third predicted rainfall data, the parameters of the convolutional neural network are adjusted to obtain a trained convolutional neural network; the rainfall data before the predicted day is input into the trained convolutional neural network to obtain the target rainfall data of the predicted day. The present application proposes a new prediction method, which fully considers the historical rainfall data before the current day and the rainfall data of the current day to predict the rainfall data of the predicted day, and the present application can improve the prediction accuracy.

[0009] In addition, a rainfall prediction method according to the above embodiment of the present invention may also have the following additional technical features:

[0010] Further, in the embodiment of the present application, determining the first predicted rainfall data of the current day under the background field based on the background field data and the first historical rainfall data set specifically includes:

[0011] configuring the configuration parameters of the weather research and forecasting model and sending the background field data and the physical boundary conditions to the weather research and forecasting model to obtain a simulation model;

[0012] The simulation model is run and the first historical rainfall data set is input into the simulation model to obtain the first predicted rainfall data.

[0013] Further, in the embodiment of the present application, determining the second predicted rainfall data after the current day under the background field based on the background field data and the second historical rainfall data set specifically includes:

[0014] configuring the configuration parameters of the weather research and forecasting model and sending the background field data and the physical boundary conditions to the weather research and forecasting model to obtain a simulation model;

[0015] The simulation model is run and a second historical rainfall data set is input into the simulation model to obtain the second predicted rainfall data.

[0016] Further, in the embodiment of the present application, the first predicted rainfall data and the second historical rainfall data set are input into a convolutional neural network to obtain third predicted rainfall data after the current day, specifically including:

[0017] Performing data preprocessing on the first predicted rainfall data and the second historical rainfall data set to obtain training data;

[0018] The training data is input into the convolutional neural network to obtain the third predicted rainfall data.

[0019] Further, in the embodiment of the present application, the first predicted rainfall data and the second historical rainfall data set are subjected to data preprocessing to obtain training data, specifically including:

[0020] Performing data alignment processing on the first predicted rainfall data and the second historical rainfall data set,

[0021] Furthermore, the first predicted rainfall data and the second historical rainfall data set are processed for data format adjustment to obtain training data.

[0022] Further, in the embodiment of the present application, adjusting the parameters of the convolutional neural network according to the second predicted rainfall data and the third predicted rainfall data to obtain a trained convolutional neural network specifically includes:

[0023] determining a training error according to the second predicted rainfall data and the third predicted rainfall data;

[0024] When the training error is greater than a preset error, the parameters of the convolutional neural network are adjusted and the convolutional neural network is retrained until the training error is less than or equal to the preset error, thereby obtaining the trained convolutional neural network.

[0025] Further, in the embodiment of the present application, determining the training error according to the second predicted rainfall data and the third predicted rainfall data specifically includes:

[0026] Subtracting the second predicted rainfall data from the third predicted rainfall data to obtain a first difference;

[0027] The first difference is used as the training error.

[0028] On the other hand, the embodiment of the present application further provides a rainfall prediction system, including:

[0029] A first processing unit is used to obtain a first historical rainfall data set of a ground observation station and a second historical rainfall data set of a radar observation station; wherein the first historical rainfall data set is a set of historical rainfall data observed by the ground observation station before the current day; and the second historical rainfall data set is a set of the first historical rainfall data set and rainfall data of the current day;

[0030] A second processing unit is used to determine first predicted rainfall data for the current day under the background field based on the background field data and the first historical rainfall data set, and to determine second predicted rainfall data after the current day under the background field based on the background field data and the second historical rainfall data set;

[0031] A third processing unit, configured to input the first predicted rainfall data and the second historical rainfall data set into a convolutional neural network to obtain third predicted rainfall data after the current day;

[0032] a fourth processing unit, configured to adjust the parameters of the convolutional neural network according to the second predicted rainfall data and the third predicted rainfall data, so as to obtain a trained convolutional neural network;

[0033] The fifth processing unit is used to input the rainfall data before the forecast day into the trained convolutional neural network to obtain the target rainfall data of the forecast day.

[0034] On the other hand, the present application also provides a rainfall prediction device, comprising:

[0035] at least one processor;

[0036] at least one memory for storing at least one program;

[0037] When the at least one program is executed by the at least one processor, the at least one processor implements a rainfall prediction method as described in any one of the invention contents.

[0038] In addition, the present application also provides a computer-readable storage medium, which stores processor-executable instructions. When the processor-executable instructions are executed by the processor, they are used to execute a rainfall prediction method as described in any one of the above items.

[0039] The advantages and benefits of the present application will be partially given in the following description, and partially become apparent from the following description, or be understood through the practice of the present application:

[0040] The present application can obtain a first historical rainfall data set of a ground observation station and a second historical rainfall data set of a radar observation station; wherein the first historical rainfall data set is a set of historical rainfall data observed by the ground observation station before the current day; the second historical rainfall data set is a set of the first historical rainfall data set and the rainfall data of the current day; based on the background field data and the first historical rainfall data set, the first predicted rainfall data of the current day under the background field is determined, and based on the background field data and the second historical rainfall data set, the second predicted rainfall data after the current day under the background field is determined; the first predicted rainfall data and the second historical rainfall data set are input into a convolutional neural network to obtain the third predicted rainfall data after the current day; according to the second predicted rainfall data and the third predicted rainfall data, the parameters of the convolutional neural network are adjusted to obtain a trained convolutional neural network; the rainfall data before the predicted day is input into the trained convolutional neural network to obtain the target rainfall data of the predicted day. The present application proposes a new prediction method, which fully considers the historical rainfall data before the current day and the rainfall data of the current day to predict the rainfall data of the predicted day, and the present application can improve the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A schematic diagram of the steps of a rainfall prediction method in a specific embodiment of the present invention;

[0042] Figure 2 A schematic diagram of a rainfall prediction method in a specific embodiment of the present invention

[0043] Figure 3 It is a structural schematic diagram of a rainfall prediction system in another specific embodiment of the present invention;

[0044] Figure 4 It is a structural schematic diagram of a rainfall prediction device in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0045] The embodiments of the present invention are described in detail below with reference to the accompanying drawings to illustrate the principles and processes of the rainfall prediction method, system, device and storage medium in the embodiments of the present invention.

[0046] First, the terms of this application are explained:

[0047] 1.NWP (Numerical Weather Prediction) model: Numerical weather forecast model, a general term for weather forecasting methods based on physics, mathematics and computer science. It simulates future changes in atmospheric conditions by solving basic physical equations that describe atmospheric motion (such as fluid dynamics equations and thermodynamics equations). In rainfall forecasting, the NWP model predicts the timing, intensity and distribution of precipitation by simulating the evolution of parameters such as humidity, airflow and temperature in the atmosphere.

[0048] 2.WRF (Weather Research and Forecasting Model): Weather Research and Forecasting Model, a specific model in the NWP model, specifically used for weather research and actual weather forecasting tasks.

[0049] 3. RMSE (Root Mean Square Error): Root mean square error, a statistical indicator that measures the deviation between the predicted value and the actual observed value. The calculation method is the square root of the average of the squares of the prediction errors.

[0050] 4. HR (hit rate): Hit rate, used to evaluate the accuracy of the prediction model. The calculation method is the true positive divided by the sum of the true positive and false negative.

[0051] 5. FAR (false alarm rate): False alarm rate, used to evaluate the error rate of the model. In the prediction model, low FAR means good prediction effect. The calculation method is false positive examples divided by the sum of true positive examples and false positive examples.

[0052] 6. CSI (Critical Success Index): Critical success index is an indicator used to evaluate the performance of a prediction model or classifier. It is calculated by dividing the true positives by the sum of the true positives, false positives, and false negatives.

[0053] 7. FSS (Fractional Skill Score): An indicator used to evaluate spatial prediction performance. It is mainly used in fields such as meteorology and environmental modeling that require verification of spatial distribution. The calculation method is relatively complex. The closer the FSS is to 1, the closer the prediction result is to the true value.

[0054] 8. Grid: The basic unit used to discretize continuous space in numerical simulation, dividing the study area into a regular grid structure.

[0055] 9. Resolution: Resolution is the spatial scale of the grid, which refers to the size of each grid cell, usually expressed in distance (such as kilometers or meters).

[0056] 10.ECMWF (European Centre for Medium-Range Weather Forecasts): European Centre for Medium-Range Weather Forecasts. It is an international organization headquartered in Reading, UK, specializing in the research, development and operation of medium-range weather forecasts (i.e. the next 3-10 days). ECMWF is widely regarded as one of the most advanced numerical weather forecasting institutions in the world.

[0057] 11. Machine Learning: A general term for technologies that allow computers to automatically learn patterns from data and make predictions or decisions without being explicitly programmed.

[0058] 12. Deep learning: Deep learning is a machine learning method that automatically learns the features and patterns in data through multi-layer neural networks to perform classification, regression or generation tasks.

[0059] 13. CNN (Convolutional Neural Network): A convolutional neural network is a deep learning model that is particularly good at processing images and spatiotemporal data. In precipitation prediction, CNN learns the difference between real data and predicted data to generate an AI model to calibrate the predicted data.

[0060] Precipitation prediction is a complex and important research task. Its technical background covers multiple fields such as meteorology, statistics, satellite remote sensing technology, machine learning and deep learning, and a variety of mature technologies have been developed, such as numerical weather forecasting, meteorological radar technology, satellite remote sensing technology, etc. Meteorological radar technology uses echo intensity to monitor rainfall distribution in real time and realizes short-term forecasting through radar puzzle, but it is mainly suitable for short-term forecasting; satellite remote sensing technology provides information on the spatiotemporal distribution of clouds and precipitation through polar orbits and geostationary satellites, which supplements the insufficiency of radar data. In meteorology, rainfall prediction mainly relies on theoretical models based on atmospheric dynamics and thermodynamics. These models simulate physical variables such as atmospheric motion, humidity, and temperature through numerical weather forecasting (NWP), and realize long-term and large-scale predictions through high-performance computing. On the other hand, early statistical methods, such as time series analysis and linear regression, also provided preliminary solutions for rainfall prediction by establishing the relationship between historical data and rainfall. With the development of observation technology, meteorological satellites and radars have gradually become important data sources for rainfall forecasting. Satellites can provide macro information on clouds and humidity, while radars can capture local precipitation intensity. The combination of the two has greatly improved the accuracy of rainfall forecasting.

[0061] For different regions and time ranges, the methods of rainfall prediction are different. For short time (0-6 hours) and areas with relatively stable climate, some relatively simple models can complete the prediction task well. However, for areas with rapid changes in atmospheric state and longer time ranges (4-6 hours or more), the predicted physical model is required to have strong simulation capabilities, such as simulating precipitation processes. Such models belong to the category of numerical weather prediction (NWP). NWP models simulate physical changes through partial differential equations and approximate related equations through numerical methods. As one of the NWP models, the Weather Research and Forecasting Model (WRF) can provide parameter configurations for simulation of different physical processes, making the simulation conditions more similar to the corresponding regions. For different regions, the optimal grid resolution and WRF model parameter configuration may be quite different and need to be tried many times. For example, in order to obtain the best precipitation forecast in British Columbia, Canada, relevant researchers have studied more than 100 possible model configurations; some studies have also found that in some complex terrains, high-resolution (3 km) grids are better for predicting precipitation values, but low-resolution (27 km) grids perform better overall.

[0062] In the related technologies, rainfall prediction technology shows a diversified development trend of combining traditional physical models with modern data-driven models. The traditional numerical weather forecast (NWP) method is based on the principles of atmospheric dynamics and thermodynamics and can simulate rainfall conditions over a large range and over a long time scale. However, due to its high computational complexity and sensitivity to initial conditions, its accuracy in local rainfall and short-term rainfall prediction is limited. Therefore, there are still technical problems that need to be solved in the related technologies.

[0063] In view of the above-mentioned defects of the prior art, Figure 1 , this application provides a rainfall prediction method. Figure 1 In the method, the method may include the following steps S101-S105.

[0064] S101. Obtain a first historical rainfall data set of a ground observation station and a second historical rainfall data set of a radar observation station; wherein the first historical rainfall data set is a set of historical rainfall data observed by the ground observation station before the current day; and the second historical rainfall data set is a set of the first historical rainfall data set and the rainfall data of the current day.

[0065] S102. Determine first predicted rainfall data for the current day under the background field based on the background field data and the first historical rainfall data set, and determine second predicted rainfall data after the current day under the background field based on the background field data and the second historical rainfall data set.

[0066] S103: Input the first predicted rainfall data and the second historical rainfall data set into a convolutional neural network to obtain third predicted rainfall data after the current day.

[0067] S104. Adjust the parameters of the convolutional neural network according to the second predicted rainfall data and the third predicted rainfall data to obtain a trained convolutional neural network.

[0068] S105, inputting the rainfall data before the forecast day into the trained convolutional neural network to obtain the target rainfall data for the forecast day.

[0069] It is understandable that the predicted day is a day after the current day, and the predicted day and the current day are two adjacent days. The current day can be any day.

[0070] In some feasible embodiments of the present application, the processor can be connected to the acquisition unit via a wired or wireless connection. After the connection is established, the processor can obtain a first historical rainfall data set of the ground observation station and a second historical rainfall data set of the radar observation station from the acquisition unit; wherein the first historical rainfall data set is a set of historical rainfall data observed by the ground observation station before the current day; the second historical rainfall data set is a set of the first historical rainfall data set and the rainfall data of the current day; based on the background field data and the first historical rainfall data set, the processor can determine the first predicted rainfall data of the current day under the background field, and based on the background field data and the second historical rainfall data set, determine the second predicted rainfall data after the current day under the background field; the first predicted rainfall data and the second historical rainfall data set are input into the convolutional neural network, and the processor can obtain the third predicted rainfall data after the current day; according to the second predicted rainfall data and the third predicted rainfall data, the processor can adjust the parameters of the convolutional neural network to obtain a trained convolutional neural network; the rainfall data before the prediction day is input into the trained convolutional neural network, and the processor can obtain the target rainfall data for the prediction day.

[0071] It should be noted that the rainfall data may be the rainfall amount of the day. The wired connection method may include the connection between the mobile device and the processing module, the connection between the processing module and the hardware device, and the wired connection between other devices known or developed in the future and the processing module; and the wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (Ultra Wide Band) connection, and other wireless connection methods known or developed in the future.

[0072] The present application can obtain a first historical rainfall data set of a ground observation station and a second historical rainfall data set of a radar observation station; wherein the first historical rainfall data set is a set of historical rainfall data observed by the ground observation station before the current day; the second historical rainfall data set is a set of the first historical rainfall data set and the rainfall data of the current day; based on the background field data and the first historical rainfall data set, the first predicted rainfall data of the current day under the background field is determined, and based on the background field data and the second historical rainfall data set, the second predicted rainfall data after the current day under the background field is determined; the first predicted rainfall data and the second historical rainfall data set are input into a convolutional neural network to obtain the third predicted rainfall data after the current day; according to the second predicted rainfall data and the third predicted rainfall data, the parameters of the convolutional neural network are adjusted to obtain a trained convolutional neural network; the rainfall data before the predicted day is input into the trained convolutional neural network to obtain the target rainfall data of the predicted day. The present application proposes a new prediction method, which fully considers the historical rainfall data before the current day and the rainfall data of the current day to predict the rainfall data of the predicted day, and the present application can improve the prediction accuracy.

[0073] Further, in an embodiment of the present application, based on the background field data and the first historical rainfall data set, the process of determining the first predicted rainfall data of the current day under the background field may include steps S201 and S202.

[0074] S201, configure the configuration parameters of the weather research and forecast model and send the background field data and physical boundary conditions to the weather research and forecast model to obtain a simulation model.

[0075] S202, running a simulation model and inputting a first historical rainfall data set into the simulation model to obtain first predicted rainfall data.

[0076] It can be understood that the simulation model can be a weather research and forecasting model, and the first predicted rainfall data can be the rainfall data of the current day obtained by running the weather research and forecasting model.

[0077] Further, in the embodiment of the present application, based on the background field data and the second historical rainfall data set, the process of determining the second predicted rainfall data after the current day under the background field may include steps S301 and S302.

[0078] S301, configure the configuration parameters of the weather research and forecast model and send the background field data and physical boundary conditions to the weather research and forecast model to obtain a simulation model.

[0079] S302, running the simulation model and inputting a second historical rainfall data set into the simulation model to obtain second predicted rainfall data.

[0080] It can be understood that the simulation model can be a weather research and forecasting model, and the second prediction data can be the rainfall data of the predicted day obtained by running the weather research and forecasting model, and the rainfall data is a predicted value.

[0081] Furthermore, in an embodiment of the present application, the process of inputting the first predicted rainfall data and the second historical rainfall data set into a convolutional neural network to obtain the third predicted rainfall data after the current day may include steps S401 and S402.

[0082] S401. Preprocess the first predicted rainfall data and the second historical rainfall data set to obtain training data.

[0083] S402: Input the training data into the convolutional neural network to obtain third predicted rainfall data.

[0084] Furthermore, in the embodiment of the present application, the process of performing data preprocessing on the first predicted rainfall data and the second historical rainfall data set to obtain training data may include step S501.

[0085] S501, performing data alignment processing on the first predicted rainfall data and the second historical rainfall data set,

[0086] Furthermore, the first predicted rainfall data and the second historical rainfall data set are processed for data format adjustment to obtain training data.

[0087] Furthermore, in an embodiment of the present application, the process of adjusting the parameters of the convolutional neural network according to the second predicted rainfall data and the third predicted rainfall data to obtain a trained convolutional neural network may include steps S601 and S602.

[0088] S601. Determine a training error according to the second predicted rainfall data and the third predicted rainfall data.

[0089] S602: When the training error is greater than the preset error, adjust the parameters of the convolutional neural network and retrain the convolutional neural network until the training error is less than or equal to the preset error, thereby obtaining a trained convolutional neural network.

[0090] Further, in the embodiment of the present application, the process of determining the training error according to the second predicted rainfall data and the third predicted rainfall data may include steps S701 and S702.

[0091] S701. Subtract the second predicted rainfall data from the third predicted rainfall data to obtain a first difference.

[0092] S702: Use the first difference as a training error.

[0093] The following is combined with Figure 2 The principle of this application is explained.

[0094] Reference Figure 2 This embodiment proposes a method for selecting the optimal grid resolution through a preliminary experiment and calibrating the prediction results through a convolutional neural network (CNN) training model. The innovation lies in conducting a preliminary experiment to obtain the overall optimal spatial resolution, and on this basis, using CNN to calibrate the prediction results.

[0095] The specific steps are as follows:

[0096] 1. Data preparation and processing: Through ground observation stations, the precipitation in the forecast area is collected periodically (such as every hour), and the corresponding radar observation data is collected through the National Environment Agency (radar observation data is more accurate and generally used as the true value), and the corresponding background field, boundary conditions and other data are obtained from the database provided by the European Center for Medium-Range Weather Forecasts (ECMWF). The data is classified and used for pre-experiments, numerical simulation of rainfall prediction, and verification of model effects.

[0097] 2. Preliminary experiment and spatial resolution selection: Set several possible spatial resolutions and select a small number of days for preliminary experiments. By analyzing the deviations of rainfall intensity, temporal distribution, and spatial coverage from observed data, as well as evaluation indicators such as CSI and FSS, the optimal resolution that takes into account both accuracy and computational efficiency is selected and set as the spatial resolution in the subsequent experiments.

[0098] 3. WRF numerical simulation: In this technical solution, the WRF version used is v4.5. Based on the optimal resolution and physical scheme determined by the preliminary experiment, the WRF mode is run to perform numerical simulation of the rainfall process in the target area and output rainfall prediction results (such as cumulative rainfall and rainfall intensity time series).

[0099] 4. CNN configuration and training: Match the rainfall data output by the WRF simulation with the corresponding radar observation data to generate input-output sample pairs. In this technical solution, the CNN model consists of five convolutional layers, one fully connected layer and an output layer. Each convolutional layer contains at least 32 filters and uses ReLu as the activation function. The input of the first layer consists of various physical parameters extracted from the WRF model and the grid resolution. In the convolution operation, a randomly initialized convolution kernel is used to act on the input features. The feature map is obtained through the output of the first layer, and then these feature maps are used as the input of the second layer. In subsequent layers, this process is repeated. The output of the fifth convolutional layer is passed to the fully connected layer. The last layer converts the data into a precipitation intensity map consistent with the spatial resolution and outputs it. Use the constructed model and the aforementioned input data to train the CNN model, optimize the loss function, and verify the performance of the model on an independent test set. Optimize the model performance (by adjusting the hyperparameters) based on the evaluation results.

[0100] Practical application and performance evaluation: In actual rainfall prediction, the data directly predicted by WRF and the predicted data obtained after CNN calibration and radar observation data are compared respectively. The results are presented in the form of high-resolution spatial rainfall maps, and the improvement effect of the invented method is evaluated by evaluation indicators such as RMSE, CSI, and FSS. The magnitude of statistical improvement for different rainfall times is analyzed, the universality of the method is analyzed, and the calibrated results are converted into standardized meteorological products (such as rainfall intensity forecast maps, rainfall probability forecasts), and provided to meteorological departments or related users in the form of images or data files.

[0101] In some embodiments, taking Singapore as an example, the specific application process of the technical solution of this embodiment is demonstrated in detail. In terms of data preparation, we periodically collected rainfall observation data observed by various observation stations in Singapore from 2020 to 2021, and focused on a day containing heavy rain weather. The data came from the National Environment Agency of Singapore. In addition to the ground observation data, we also obtained the same rainfall data through radar observation data provided by the National Environment Agency, and used it as real data. At the same time, we used the 12UTC initial and boundary conditions of the high-resolution deterministic model from the European Center for Medium-Range Weather Forecasts (ECMWF) for all experiments. The ECMWF model input data is at intervals of 3 hours and has a spatial resolution of 0.1°. In order to test the impact of WRF spatial resolution on Singapore precipitation simulation, multiple domain resolutions are set. In this embodiment, the version of WRF is set to v4.5. During training, the most representative 15 days were selected from the obtained data, including different seasons and rainfall levels, for pre-experiments.

[0102] After obtaining the data, the domain resolutions of 1km, 3km, 9km, and 12km are taken respectively, and the WRF model is used for preliminary experiments. The prediction results are evaluated using evaluation indicators such as CSI and FSS, and the spatial resolution with the best comprehensive effect is selected. In this embodiment, the spatial resolution of 1km has a better effect, and the scores of CSI and FSS are the highest in each type of weather. Therefore, 1km is used as the spatial resolution of WRF, and numerical simulation is performed through observation data to obtain prediction data. Then, the rainfall data output by the WRF simulation is matched with the corresponding radar observation data through the obtained prediction data and the corresponding radar observation data to generate an input-output sample pair. After the data processing is completed, the CNN model is used for training, in which the configuration of CNN is the same as in the technical solution, and a model that can be used to calibrate the prediction results is obtained. Finally, a comparative experiment is carried out in the data set set as the validation set, and the results of numerical prediction directly using the WRF model and the results after calibration using the AI ​​model are compared through indicators such as CSI and FSS.

[0103] The results show that calibration through the CNN model has improved all indicators compared to directly using WRF. Specifically, in this embodiment, the comprehensive improvement of the FSS score of the method in this embodiment is 10.1% compared to the traditional method, the comprehensive improvement of the CSI score is 9.8%, and the overall RMSE is improved by 11.2%.

[0104] In addition, refer to Figure 3 ,and Figure 1Corresponding to the method, a rainfall prediction system is also provided in an embodiment of the present application. The system may include a first processing unit 1001, a second processing unit 1002, a third processing unit 1003, a fourth processing unit 1004 and a fifth processing unit 1005. Among them, the first processing unit 1001 can be used to obtain a first historical rainfall data set of a ground observation station and a second historical rainfall data set of a radar observation station; wherein the first historical rainfall data set is a set of historical rainfall data observed by the ground observation station before the current day; the second historical rainfall data set is a set of the first historical rainfall data set and the rainfall data of the current day. The second processing unit 1002 can be used to determine the first predicted rainfall data of the current day under the background field based on the background field data and the first historical rainfall data set, and determine the second predicted rainfall data after the current day under the background field based on the background field data and the second historical rainfall data set. The third processing unit 1003 can be used to input the first predicted rainfall data and the second historical rainfall data set into the convolutional neural network to obtain the third predicted rainfall data after the current day; the fourth processing unit 1004 can be used to adjust the parameters of the convolutional neural network according to the second predicted rainfall data and the third predicted rainfall data to obtain a trained convolutional neural network. The fifth processing unit 1005 can be used to input the rainfall data before the predicted day into the trained convolutional neural network to obtain the target rainfall data for the predicted day.

[0105] It should be noted that the first processing unit may be any integrated circuit unit or microprocessor unit obtained by integrating a chip having a processing function and its peripheral circuits through existing integration technology. The first processing unit and the second processing unit may also be any integrated circuit module or microprocessor module obtained by integrating a chip having a processing function and its peripheral circuits through existing integration technology. The first processing unit and the second processing unit may also include one or more memories.

[0106] It should be noted that the contents of the above-mentioned rainfall prediction method embodiment are all applicable to the present rainfall prediction system embodiment. The functions specifically implemented by the present rainfall prediction system embodiment are the same as those of the above-mentioned rainfall prediction method embodiment, and the beneficial effects achieved are also the same as those achieved by the above-mentioned rainfall prediction method embodiment.

[0107] and Figure 1 Corresponding to the method, the embodiment of the present application also provides a rainfall prediction device, the specific structure of which can be referred to Figure 4 ,include:

[0108] at least one processor 1011;

[0109] At least one memory 1012, used to store at least one program;

[0110] When the at least one program is executed by the at least one processor, the at least one processor implements the rainfall prediction method.

[0111] The contents of the above method embodiments are all applicable to the present device embodiments. The functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0112] and Figure 1 Corresponding to the method, an embodiment of the present application further provides a computer-readable storage medium, which stores instructions executable by a processor, and the processor-executable instructions are used to execute the rainfall prediction method when executed by the processor.

[0113] The contents of the above-mentioned rainfall prediction method embodiment are all applicable to the present storage medium embodiment. The functions specifically implemented by the present storage medium embodiment are the same as those of the above-mentioned rainfall prediction method embodiment, and the beneficial effects achieved are also the same as those achieved by the above-mentioned rainfall prediction method embodiment.

[0114] In some optional embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the application is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Optional embodiments are expected, wherein the order of various operations is changed and the sub-operation described as a part of a larger operation is performed independently.

[0115] In addition, although the present application is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present application. More specifically, in view of the properties, functions, and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional techniques of the engineer. Therefore, those skilled in the art can implement the present application set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present application, which is determined by the full scope of the attached claims and their equivalents.

[0116] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several programs to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0117] The logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable programs for implementing the logical functions, and may be embodied in any computer-readable medium for use by a program execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch and execute a program from a program execution system, device or apparatus), or in conjunction with such program execution systems, devices or apparatuses. For purposes of this specification, a "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by a program execution system, device or apparatus, or in conjunction with such program execution systems, devices or apparatuses.

[0118] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0119] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable program execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0120] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0121] Although the embodiments of the present application have been shown and described, those skilled in the art will appreciate that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present application, and that the scope of the present application is defined by the claims and their equivalents.

[0122] The above is a specific description of the preferred implementation of the present application, but the present application is not limited to the described embodiments. Technical personnel familiar with the field may make various equivalent modifications or substitutions without violating the spirit of the present application. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present application.

Claims

1. A rainfall prediction method, characterized in that: The following steps are involved: Acquire a first historical rainfall data set of a ground observation station and a second historical rainfall data set of a radar observation station; wherein the first historical rainfall data set is a set of historical rainfall data observed by the ground observation station before the current day; and the second historical rainfall data set is a set of the first historical rainfall data set and rainfall data of the current day; Determine first predicted rainfall data for the current day under the background field based on the background field data and the first historical rainfall data set, and determine second predicted rainfall data for the background field after the current day based on the background field data and the second historical rainfall data set; Inputting the first predicted rainfall data and the second historical rainfall data set into a convolutional neural network to obtain third predicted rainfall data after the current day; According to the second predicted rainfall data and the third predicted rainfall data, adjusting the parameters of the convolutional neural network to obtain a trained convolutional neural network; The rainfall data before the forecast day is input into the trained convolutional neural network to obtain the target rainfall data for the forecast day.

2. A rainfall prediction method according to claim 1, characterized in that: The determining, based on the background field data and the first historical rainfall data set, first predicted rainfall data for the current day under the background field specifically includes: configuring the configuration parameters of the weather research and forecasting model and sending the background field data and the physical boundary conditions to the weather research and forecasting model to obtain a simulation model; The simulation model is run and the first historical rainfall data set is input into the simulation model to obtain the first predicted rainfall data.

3. A rainfall prediction method according to claim 1, characterized in that: The determining, based on the background field data and the second historical rainfall data set, second predicted rainfall data after the current day under the background field specifically includes: configuring the configuration parameters of the weather research and forecasting model and sending the background field data and the physical boundary conditions to the weather research and forecasting model to obtain a simulation model; The simulation model is run and a second historical rainfall data set is input into the simulation model to obtain the second predicted rainfall data.

4. A rainfall prediction method according to claim 1, characterized in that: The first predicted rainfall data and the second historical rainfall data set are input into a convolutional neural network to obtain third predicted rainfall data after the current day, specifically including: Performing data preprocessing on the first predicted rainfall data and the second historical rainfall data set to obtain training data; The training data is input into the convolutional neural network to obtain the third predicted rainfall data.

5. A rainfall prediction method according to claim 4, characterized in that: The step of performing data preprocessing on the first predicted rainfall data and the second historical rainfall data set to obtain training data specifically includes: Performing data alignment processing on the first predicted rainfall data and the second historical rainfall data set, Furthermore, the first predicted rainfall data and the second historical rainfall data set are processed for data format adjustment to obtain training data.

6. A rainfall prediction method according to claim 1, characterized in that: The step of adjusting the parameters of the convolutional neural network according to the second predicted rainfall data and the third predicted rainfall data to obtain a trained convolutional neural network specifically includes: determining a training error according to the second predicted rainfall data and the third predicted rainfall data; When the training error is greater than a preset error, the parameters of the convolutional neural network are adjusted and the convolutional neural network is retrained until the training error is less than or equal to the preset error, thereby obtaining the trained convolutional neural network.

7. A rainfall prediction method according to claim 1, characterized in that: The determining of the training error according to the second predicted rainfall data and the third predicted rainfall data specifically includes: Subtracting the second predicted rainfall data from the third predicted rainfall data to obtain a first difference; The first difference is used as the training error.

8. A rainfall prediction system, characterized in that: include: A first processing unit is used to obtain a first historical rainfall data set of a ground observation station and a second historical rainfall data set of a radar observation station; The first historical rainfall data set is a set of historical rainfall data observed by a ground observation station before the current day; the second historical rainfall data set is a set of the first historical rainfall data set and the rainfall data of the current day; A second processing unit is used to determine first predicted rainfall data for the current day under the background field based on the background field data and the first historical rainfall data set, and to determine second predicted rainfall data after the current day under the background field based on the background field data and the second historical rainfall data set; A third processing unit, configured to input the first predicted rainfall data and the second historical rainfall data set into a convolutional neural network to obtain third predicted rainfall data after the current day; a fourth processing unit, configured to adjust the parameters of the convolutional neural network according to the second predicted rainfall data and the third predicted rainfall data, so as to obtain a trained convolutional neural network; The fifth processing unit is used to input the rainfall data before the forecast day into the trained convolutional neural network to obtain the target rainfall data of the forecast day.

9. A rainfall prediction device, characterized in that include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements a rainfall prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium storing instructions executable by a processor, characterized in that: The processor-executable instructions are used to execute a rainfall prediction method as described in any one of claims 1-7 when executed by the processor.

Images