Precipitation forecasting method and device, computer equipment, storage medium and program product
Through the neural network model, low-resolution and high-resolution data features are learned within the preset time, and high-resolution forecasts are generated only when heavy precipitation occurs, solving the problem of high-resolution mode delay and achieving efficient and real-time precipitation forecasts .
Patent Information
- Application Number
- CN202411398180.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-09
- Publication Date
- 2025-01-03
AI Technical Summary
High-resolution mode requires a large amount of computing resources and storage capacity in weather forecasts, resulting in delay problems and it is difficult to predict high-impact precipitation events in real time.
The characteristics of low-resolution and high-resolution forecast data are learned through neural network models, and high-resolution forecast data are generated only when heavy precipitation occurs within the preset aging, reducing unnecessary computing resource consumption.
It reduces the calculation time of high-resolution forecast data, reduces the delay in heavy precipitation forecasts, saves computing resources, and improves the real-time and efficiency of forecasts.
Smart Images
Figure CN119247516A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of weather forecasting, and in particular to a precipitation forecasting method, apparatus, computer equipment, storage medium, and program product. Background Art
[0002] With rapid socioeconomic development, economic activities are becoming increasingly diverse, and people are becoming increasingly sensitive to high-impact weather events, such as heavy rain. High-intensity, widespread rainfall often causes traffic jams or even disruptions, impacting agricultural and food production, causing damage to the ecological environment, and threatening people's lives and property.
[0003] Currently, weather forecasts primarily use high-resolution and low-resolution models. High-resolution models use a finer spatial grid to segment the Earth's surface and atmosphere, while low-resolution models use a coarser spatial grid to segment the Earth's surface and atmosphere. Because high-resolution models can more precisely capture and simulate small-scale weather phenomena, they are often used to forecast heavy and extreme precipitation.
[0004] However, high-resolution modes usually consume more computing resources and storage capacity, so there will be a large delay in weather forecasting. Summary of the Invention
[0005] Based on this, it is necessary to provide a precipitation forecasting method, device, computer equipment, storage medium and program product that can reduce forecast delay in response to the above technical problems.
[0006] In a first aspect, the present application provides a precipitation forecasting method. The method comprises:
[0007] Determine whether heavy rainfall will occur within a first preset forecast timeframe based on pending low-resolution forecast data for the target area;
[0008] When it is determined that heavy rainfall will occur within the first preset forecast time limit, the low-resolution forecast data to be reported for the target area is input into the neural network model to obtain high-resolution forecast data to be reported corresponding to the low-resolution forecast data to be reported;
[0009] Precipitation forecast is made based on the pending high-resolution forecast data.
[0010] In one embodiment, before inputting the low-resolution forecast data of the target area to be reported into the neural network model, the method further includes:
[0011] Obtain historical low-resolution forecast data and historical high-resolution forecast data for the target area;
[0012] Determine whether heavy rainfall will occur within the second preset forecast timeframe based on historical low-resolution forecast data for the target area;
[0013] When it is determined that heavy rainfall will occur within the second preset forecast time period, the initial neural network model is trained based on the historical low-resolution forecast data and the historical high-resolution forecast data to obtain the neural network model.
[0014] In one embodiment, determining whether heavy precipitation will occur within a second preset forecast time period based on historical low-resolution forecast data of the target area includes:
[0015] Obtaining the most recent first low-resolution forecast data from the historical low-resolution forecast data;
[0016] If the first low-resolution forecast data indicates that rainfall will occur within the second preset forecast time period, then calculating the precipitation area and precipitation intensity within the second preset forecast time period based on the historical low-resolution forecast data;
[0017] When the precipitation area reaches a preset precipitation area threshold, and / or the precipitation intensity reaches a preset precipitation intensity threshold, it is determined that heavy precipitation will occur within the second preset forecast time limit.
[0018] In one embodiment, the initial neural network model is trained based on the historical low-resolution forecast data and the historical high-resolution forecast data to obtain the neural network model, including:
[0019] Acquire second forecast data of the second preset forecast time within a preset time period from the historical low-resolution forecast data and the historical high-resolution forecast data; the second forecast data includes the second low-resolution forecast data and the second high-resolution forecast data;
[0020] If the data amount of the second forecast data reaches a preset data amount, the initial neural network model is trained based on the second forecast data to obtain the neural network model.
[0021] In one embodiment, the method further comprises:
[0022] If the data volume of the second forecast data does not reach the preset data volume, the preset time duration is extended to obtain a target time duration; the target time duration includes the preset time duration, and the data volume of the third forecast data of the second preset forecast time validity within the target time duration reaches the preset data volume;
[0023] The initial neural network model is trained based on the third forecast data to obtain the neural network model.
[0024] In one embodiment, the method further comprises:
[0025] Acquiring initial forecast data for the target area; the initial forecast data includes low-resolution forecast data and initial high-resolution forecast data;
[0026] The precipitation element data in the initial forecast data is standardized based on a logarithmic function to obtain intermediate forecast data;
[0027] The other element data in the intermediate forecast data are normalized to their maximum and minimum values to obtain historical forecast data; the historical forecast data includes historical low-resolution forecast data and historical high-resolution forecast data.
[0028] In a second aspect, the present application also provides a precipitation forecasting device. The device comprises:
[0029] A first determination module is configured to determine whether heavy rainfall will occur within a first preset forecast time period based on pending low-resolution forecast data for a target area;
[0030] A first acquisition module is configured to input the pending low-resolution forecast data of the target area into a neural network model when it is determined that heavy rainfall will occur within the first preset forecast time limit, so as to obtain the pending high-resolution forecast data corresponding to the pending low-resolution forecast data;
[0031] The forecast module is used to make precipitation forecast based on the high-resolution forecast data to be reported.
[0032] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any of the above methods when executing the computer program.
[0033] In a fourth aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any of the above methods when executed by a processor.
[0034] In a fifth aspect, the present application also provides a computer program product, comprising a computer program, which implements the steps of any of the above methods when executed by a processor.
[0035] The above-mentioned precipitation forecast method, device, computer equipment, storage medium and program product determine whether heavy precipitation will occur within the first preset forecast time limit based on the low-resolution forecast data to be reported in the target area. When it is determined that heavy precipitation will occur within the first preset forecast time limit, the low-resolution forecast data to be reported in the target area is input into the neural network model to obtain the high-resolution forecast data to be reported corresponding to the low-resolution forecast data to be reported, and precipitation forecast is performed based on the high-resolution forecast data to be reported. In the embodiment of the present application, the characteristics of the low-resolution forecast data and the high-resolution forecast data are learned by the neural network model, so that the corresponding high-resolution forecast data can be obtained according to the low-resolution forecast data through the neural network model, which reduces the calculation time of the high-resolution forecast data, thereby reducing the delay in the heavy precipitation forecast. In addition, the generation of high-resolution forecast data is triggered only when heavy precipitation will occur within the preset forecast time limit, further saving computing resources. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] FIG1 is a diagram showing the internal structure of a computer device provided in an embodiment of the present application;
[0037] FIG2 is a flow chart of a precipitation forecasting method provided in an embodiment of the present application;
[0038] FIG3 is a flow chart of a neural network model training method provided in an embodiment of the present application;
[0039] FIG4 is a schematic diagram of a neural network model provided in an embodiment of the present application;
[0040] FIG5 is a schematic diagram of another neural network model provided in an embodiment of the present application;
[0041] FIG6 is a flow chart of a method for determining heavy rainfall according to an embodiment of the present application;
[0042] FIG7 is a flow chart of another neural network model training method provided in an embodiment of the present application;
[0043] FIG8 is a flow chart of another neural network model training method provided in an embodiment of the present application;
[0044] FIG9 is a flow chart of a method for obtaining historical forecast data according to an embodiment of the present application;
[0045] FIG10 is a flow chart of a precipitation forecasting method based on a numerical model and a deep learning model provided in an embodiment of the present application;
[0046] FIG11 is a structural block diagram of a precipitation forecasting device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0047] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.
[0048] With rapid socioeconomic development, economic activities are becoming increasingly diverse, and people are becoming increasingly sensitive to high-impact weather events, such as heavy rain. High-intensity, widespread rainfall often causes traffic jams or even disruptions, impacting agricultural and food production, causing damage to the ecological environment, and threatening people's lives and property.
[0049] Currently, weather forecasts primarily use high-resolution and low-resolution models. High-resolution models use a finer spatial grid to segment the Earth's surface and atmosphere, while low-resolution models use a coarser spatial grid to segment the Earth's surface and atmosphere. Because high-resolution models can more precisely capture and simulate small-scale weather phenomena, they are often used to forecast heavy and extreme precipitation.
[0050] However, high-resolution modes usually consume more computing resources and storage capacity, so there will be a large delay in weather forecasting.
[0051] For example, assume the high-resolution model has a resolution of 3 km, and the low-resolution model has a resolution of 1 km. Both high- and low-resolution models run eight times daily, every three hours, at 02:00, 05:00, 08:00, 11:00, 14:00, 17:00, 20:00, and 23:00. The forecast validity period is 0-72 hours, with a forecast interval of 1 hour. The low-resolution forecast model arrives faster, while the high-resolution forecast model has a relatively longer delay.
[0052] The precipitation forecasting method provided in an embodiment of the present application can be applied in an application environment as shown in Figure 1. Figure 1 is an internal structure diagram of a computer device provided in an embodiment of the present application. The computer device can be a server, and its internal structure diagram can be as shown in Figure 1. The computer device includes a processor, a memory, and a network interface connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a precipitation forecasting method is implemented.
[0053] Those skilled in the art will understand that the structure shown in FIG1 is merely a block diagram of a portion of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.
[0054] In one embodiment, as shown in FIG2 , FIG2 is a flow chart of a precipitation forecasting method provided in an embodiment of the present application. The method can be applied to the computer device in FIG1 , and the method includes the following steps:
[0055] S201, determining whether heavy rainfall will occur within a first preset forecast time period based on pending low-resolution forecast data for a target area.
[0056] For example, assume that the arrival time of the latest low-resolution forecast data is At time , the low-resolution forecast data to be reported can be Timely arrival forecast data.
[0057] Optionally, you can The low-resolution forecast data to be reported at the moment is used to determine whether precipitation will occur within the first preset forecast time limit. If it is determined that precipitation will occur within the first preset forecast time limit, the average precipitation area and precipitation intensity are calculated based on the low-resolution forecast data to be reported. If the calculated average precipitation area reaches a preset area threshold and / or the precipitation intensity reaches a preset precipitation intensity threshold, it is determined whether heavy precipitation will occur within the first preset forecast time limit.
[0058] For example, according to The low-resolution forecast data to be reported at the moment is determined Will there be precipitation within 0-12 hours in the future? If precipitation will occur within 0-12 hours in the future, the average precipitation area and precipitation intensity are calculated based on the low-resolution forecast data to be reported. If the calculated average precipitation area reaches the preset area threshold, and / or the precipitation intensity reaches the preset precipitation intensity threshold, it is determined that There will be heavy rainfall in the next 0-12 hours.
[0059] In one possible implementation, it is possible to obtain Time and The target area is determined based on the historical forecast data before the time, and the meteorological element data in the historical forecast data. The historical forecast data includes historical low-resolution forecast data and historical high-resolution forecast data.
[0060] S202, when it is determined that heavy rainfall will occur within the first preset forecast time period, the low-resolution forecast data to be reported for the target area is input into the neural network model to obtain high-resolution forecast data to be reported corresponding to the low-resolution forecast data to be reported.
[0061] For example, in determining If there will be heavy rainfall within 0-12 hours in the future, you can The low-resolution forecast data to be reported of the target area at each moment is used as the input of the neural network model to obtain the high-resolution forecast data to be reported corresponding to the low-resolution forecast data to be reported output by the neural network model.
[0062] The neural network model is trained based on historical low-resolution forecast data and historical high-resolution forecast data. Optionally, the neural network model may be, for example, a convolutional neural network (CNN) model.
[0063] S203: Precipitation forecasting is performed based on the pending high-resolution forecast data.
[0064] Optionally, the pending high-resolution forecast data can be released through various channels such as official websites, social media, and mobile apps to achieve precipitation forecasting.
[0065] In an embodiment of the present application, by determining whether heavy precipitation will occur within the first preset forecast time limit based on the pending low-resolution forecast data of the target area, when it is determined that heavy precipitation will occur within the first preset forecast time limit, the pending low-resolution forecast data of the target area is input into the neural network model to obtain the pending high-resolution forecast data corresponding to the pending low-resolution forecast data, and precipitation forecast is performed based on the pending high-resolution forecast data. In an embodiment of the present application, the characteristics of the low-resolution forecast data and the high-resolution forecast data are learned by the neural network model, so that the corresponding high-resolution forecast data can be obtained according to the low-resolution forecast data through the neural network model, which reduces the calculation time of the high-resolution forecast data, thereby reducing the delay in the heavy precipitation forecast. In addition, the generation of the high-resolution forecast data is triggered only when heavy precipitation will occur within the preset forecast time limit, further saving computing resources.
[0066] 3, which is a flow chart of a neural network model training method provided by an embodiment of the present application. Before the above S202, the method further includes the following steps:
[0067] S301, obtaining historical low-resolution forecast data and historical high-resolution forecast data of the target area.
[0068] For example, historical low-resolution forecast data and historical high-resolution forecast data for the target area can be obtained through observation equipment such as meteorological satellites, meteorological radars, and ground meteorological stations.
[0069] S302: Determine whether heavy rainfall will occur within a second preset forecast time period based on historical low-resolution forecast data of the target area.
[0070] In one embodiment, the most recent first low-resolution forecast data can be obtained from historical low-resolution forecast data for the target area, and then whether precipitation will occur within a second preset forecast time period can be determined based on the first low-resolution forecast data. If the first low-resolution forecast data indicates that rainfall will occur within the second preset forecast time period, the precipitation area and precipitation intensity within the second preset forecast time period are calculated based on the historical low-resolution forecast data. When the precipitation area reaches a preset precipitation area threshold and / or the precipitation intensity reaches a preset precipitation intensity threshold, it is determined that heavy precipitation will occur within the second preset forecast time period.
[0071] For example, it can be obtained from the historical low-resolution forecast data of the target area The first low-resolution forecast data at the moment, and then according to The first low-resolution forecast data at the moment is determined Whether there will be precipitation within 0-12 hours in the future. If there will be precipitation within 0-12 hours in the future, then according to Time and Historical low-resolution forecast data before time, calculate The precipitation area and precipitation intensity in the next 0-12 hours, When the precipitation area in the next 0-12 hours reaches the preset precipitation area threshold, and / or the precipitation intensity reaches the preset precipitation intensity threshold, determine There will be heavy rainfall in the next 0-12 hours.
[0072] S303: When it is determined that heavy rainfall will occur within the second preset forecast time period, the initial neural network model is trained based on the historical low-resolution forecast data and the historical high-resolution forecast data to obtain a neural network model.
[0073] In an embodiment of the present application, the occurrence of heavy rainfall within the second preset forecast time limit is used as a trigger condition for modeling. That is, when it is determined that heavy rainfall will occur within the second preset forecast time limit, the initial neural network model is trained based on historical low-resolution forecast data and historical high-resolution forecast data to obtain a neural network model.
[0074] Optionally, the forecast time within 3 days can be obtained from all historical low-resolution forecast data and historical high-resolution forecast data. The second forecast data within 0-12 hours in the future, and the forecast time within these 3 days The second forecast data within 0-12 hours in the future is used as a training set. The initial neural network model is then trained based on the training set to obtain a neural network model.
[0075] For example, referring to Figures 4-5, Figure 4 is a schematic diagram of a neural network model provided in an embodiment of the present application, and Figure 5 is a schematic diagram of another neural network model provided in an embodiment of the present application. As shown in Figure 4, the initial neural network model in the embodiment of the present application can be a convolutional neural network (CNN) model, which has three hidden layers and uses ReLU as the activation function. The convolutional neural network (CNN) first performs a convolution operation on the input data through a convolutional layer to extract typical features. The typical features are then resampled through a pooling layer to reduce the data dimension while retaining key features. The data is then sorted and preprocessed into the format required for neural network training. The model parameters can be optimized using a stochastic gradient descent (SGD) optimizer, and the mean squared error (MSE) loss function is used to measure the difference between the predicted value and the true value. The data is then nonlinearly transformed using an activation function, and finally output through a fully connected layer.
[0076] Alternatively, in actual applications, an initial neural network model with 4-7 hidden layers can be adopted based on actual machine performance and arrival time requirements, as shown in Figure 5. Adding hidden layers can improve the model's ability to handle complex problems and help it learn more complex feature representations.
[0077] In an embodiment of the present application, historical low-resolution forecast data and historical high-resolution forecast data of the target area are obtained, and based on the historical low-resolution forecast data of the target area, it is determined whether heavy precipitation will occur within the second preset forecast time limit. When it is determined that heavy precipitation will occur within the second preset forecast time limit, the initial neural network model is trained based on the historical low-resolution forecast data and the historical high-resolution forecast data to obtain a neural network model, so that the characteristics of the low-resolution forecast data and the high-resolution forecast data can be obtained through machine learning, and the corresponding high-resolution forecast data can be obtained according to the low-resolution forecast data through the neural network model, thereby reducing the calculation time of the high-resolution forecast data, thereby reducing the delay in the heavy precipitation forecast. In addition, by training the neural network model based on the historical data within the preset time of the target area, the seasonal variation characteristics of precipitation can also be considered, and it is targeted in time. At the same time, triggering modeling through trigger conditions can also improve the flexibility of the above method and further save computing resources.
[0078] Referring to Figure 6, Figure 6 is a flow chart illustrating a method for determining heavy rainfall according to an embodiment of the present application. This embodiment relates to a possible implementation method for determining whether heavy rainfall will occur within a second preset forecast timeframe based on historical low-resolution forecast data for a target area. Based on the above embodiment, S302 includes the following steps:
[0079] S601: Acquire the most recent first low-resolution forecast data from historical low-resolution forecast data.
[0080] For example, it can be obtained from the historical low-resolution forecast data of the target area The first low-resolution forecast data at the moment, and then according to The first low-resolution forecast data at the moment is determined Whether there will be precipitation in the next 0-12 hours.
[0081] S602: If the first low-resolution forecast data indicates that rainfall will occur within a second preset forecast time period, the precipitation area and precipitation intensity within the second preset forecast time period are calculated based on the historical low-resolution forecast data.
[0082] For example, if If there will be precipitation within 0-12 hours in the future, then according to Time and Historical low-resolution forecast data before time, calculate The precipitation area and precipitation intensity within 0-12 hours in the future.
[0083] S603: When the precipitation area reaches a preset precipitation area threshold, and / or the precipitation intensity reaches a preset precipitation intensity threshold, it is determined that heavy precipitation will occur within a second preset forecast time period.
[0084] For example, in When the precipitation area in the next 0-12 hours reaches the preset precipitation area threshold, and / or the precipitation intensity reaches the preset precipitation intensity threshold, determine There will be heavy rainfall in the next 0-12 hours.
[0085] Optionally, when the precipitation area reaches a preset precipitation area threshold, it can be determined that heavy precipitation will occur within the second preset forecast time period.
[0086] Alternatively, when the precipitation intensity reaches a preset precipitation intensity threshold, it can be determined that heavy precipitation will occur within the second preset forecast time period.
[0087] Alternatively, when the precipitation area reaches a preset precipitation area threshold and the precipitation intensity reaches a preset precipitation intensity threshold, it can be determined that heavy precipitation will occur within the second preset forecast time limit.
[0088] In an embodiment of the present application, the most recent first low-resolution forecast data is obtained from the historical low-resolution forecast data. If the first low-resolution forecast data indicates that rainfall will occur within the second preset forecast time limit, the precipitation area and precipitation intensity within the second preset forecast time limit are calculated based on the historical low-resolution forecast data. When the precipitation area reaches the preset precipitation area threshold, and / or the precipitation intensity reaches the preset precipitation intensity threshold, it is determined that heavy precipitation will occur within the second preset forecast time limit. Since, in the embodiment of the present application, whether heavy precipitation will occur within the second preset forecast time limit is determined based on the most recent first low-resolution forecast data, whether heavy precipitation will occur within the second preset forecast time limit can be used as a trigger condition for modeling, and modeling is triggered when heavy precipitation occurs within the second preset forecast time limit, thereby avoiding wasting computing resources when there is no precipitation or the precipitation is weak, and further reducing the delay in heavy precipitation forecast.
[0089] Referring to FIG. 7 , FIG. 7 is a flow chart illustrating another neural network model training method provided in an embodiment of the present application. This embodiment relates to a possible implementation method of how to train an initial neural network model based on historical low-resolution forecast data and historical high-resolution forecast data to obtain a neural network model. Based on the above embodiment, the above S303 includes the following steps:
[0090] S701 , obtaining second forecast data of a second preset forecast time within a preset time length from historical low-resolution forecast data and historical high-resolution forecast data.
[0091] The second forecast data includes second low-resolution forecast data and second high-resolution forecast data.
[0092] In one embodiment, assuming that the preset time period is 3 days, the forecast time within 3 days can be obtained from all historical low-resolution forecast data and historical high-resolution forecast data. Second forecast data within 0-12 hours in the future.
[0093] For example, if the forecast validity period of historical forecast data is 0-72h and the forecast interval is 1h, each low-resolution model product and high-resolution model product generated will contain 72 sets of forecast data. According to the above rules, the resolution model products generated within 3 days will involve Forecast data for the next 0-12 hours can be obtained from the resolution model products generated within 3 days. The forecast data within 0-12 hours in the future is used as the second forecast data.
[0094] S702: If the amount of the second forecast data reaches a preset amount, the initial neural network model is trained based on the second forecast data to obtain a neural network model.
[0095] For example, assuming that the forecast validity of historical forecast data is 0-72 hours, the forecast interval is 1 hour, and the high and low resolution versions are run 8 times a day, then the number of There are 3×8×12=288 groups of forecast data within 0-12 hours in the future, that is, the amount of data of the second forecast data is 288 groups.
[0096] In one embodiment, a preset data volume can be set according to actual needs. If the data volume of the second forecast data reaches the preset data volume, the initial neural network model is trained based on the second forecast data to obtain a neural network model.
[0097] In an embodiment of the present application, second forecast data of a second preset forecast time within a preset time period is obtained from historical low-resolution forecast data and historical high-resolution forecast data. If the amount of the second forecast data reaches the preset amount, the initial neural network model is trained based on the second forecast data to obtain a neural network model. The neural network model can obtain the characteristics of the low-resolution forecast data and the high-resolution forecast data through machine learning, and then the corresponding high-resolution forecast data can be obtained from the low-resolution forecast data through the neural network model, thereby reducing the calculation time of the high-resolution forecast data and thus reducing the delay in the heavy precipitation forecast. In addition, the training of the neural network model based on historical data within the preset time of the target area can also take into account the seasonal variation characteristics of precipitation and be targeted in time.
[0098] Referring to Figure 8, Figure 8 is a flow chart of another neural network model training method provided in an embodiment of the present application. Based on the above embodiment, the method further includes the following steps:
[0099] S801: If the data volume of the second forecast data does not reach the preset data volume, the preset duration is extended to obtain the target duration.
[0100] The target duration includes a preset duration, and the data volume of the third forecast data of the second preset forecast time limit within the target duration reaches the preset data volume.
[0101] For example, if the preset data volume is 600, it means that 600 sets of samples are needed to train the neural network model, but There are only 3×8×12=288 groups of forecast data within 0-12 hours in the future. At this time, the preset time can be expanded to 7 days as the target time. There may be 7×8×12=672 sets of forecast data within 0-12 hours in the future, reaching the preset data volume of 600.
[0102] S802: Train the initial neural network model based on the third forecast data to obtain a neural network model.
[0103] In one embodiment, the above 672 sets of data can be screened to filter out samples with little or weak precipitation, and then the remaining data can be used as the third forecast data, and the initial neural network model can be trained based on the third forecast data to obtain a neural network model.
[0104] In an embodiment of the present application, if the amount of data of the second forecast data does not reach the preset amount of data, the preset time length is expanded to obtain the target time length, and then the initial neural network model is trained based on the third forecast data of the second preset forecast time within the target time length to obtain a neural network model, which can obtain the characteristics of low-resolution forecast data and high-resolution forecast data through machine learning, and then can obtain corresponding high-resolution forecast data according to the low-resolution forecast data through the neural network model, thereby reducing the calculation time of the high-resolution forecast data, thereby reducing the delay in the heavy precipitation forecast. In addition, training the neural network model based on historical data within the preset time of the target area can also take into account the seasonal variation characteristics of precipitation, and is targeted in time. At the same time, expanding the amount of data in the training set can also improve the performance of the neural network model and improve the accuracy of precipitation forecasts.
[0105] Referring to Figure 9, Figure 9 is a flow chart of a method for obtaining historical forecast data provided by an embodiment of the present application. Based on the above embodiment, the method further includes the following steps:
[0106] S901: Obtain initial forecast data for the target area.
[0107] Among them, the initial forecast data includes low-resolution forecast data and initial high-resolution forecast data.
[0108] For example, initial forecast data for the target area can be obtained through observation equipment such as meteorological satellites, meteorological radars, and ground meteorological stations.
[0109] S902: Standardize the precipitation element data in the initial forecast data based on a logarithmic function to obtain intermediate forecast data.
[0110] For example, the precipitation element data in the initial forecast data can be standardized based on the following formula (1):
[0111] (1)
[0112] In formula (1), is the original precipitation, is the standardized precipitation.
[0113] In one embodiment, 1.5 times of the climate extreme value of the 1-hour precipitation forecast for each region can be taken as the maximum value, and 0 as the minimum value, and the standardized precipitation element data can be converted into grayscale data to obtain intermediate forecast data.
[0114] S903: Perform maximum and minimum value normalization processing on other element data in the intermediate forecast data to obtain historical forecast data.
[0115] Among them, historical forecast data includes historical low-resolution forecast data and historical high-resolution forecast data.
[0116] Exemplarily, a maximum value and a minimum value may be set, and other element data in the intermediate forecast data may be normalized according to the maximum value and the minimum value to obtain historical forecast data.
[0117] In an embodiment of the present application, initial forecast data for the target area is obtained, precipitation element data in the initial forecast data is standardized based on a logarithmic function to obtain intermediate forecast data, and other element data in the intermediate forecast data are normalized to their maximum and minimum values to obtain historical forecast data, thereby enabling preprocessing of the historical forecast data, reducing the impact of irrelevant data and noise on the precipitation forecast, and improving the accuracy of the precipitation forecast.
[0118] Referring to FIG10 , FIG10 is a flow chart of a precipitation forecasting method based on a numerical model and a deep learning model provided in an embodiment of the present application. The method includes the following steps:
[0119] S1001: Standardize precipitation element data in the initial forecast data based on a logarithmic function to obtain intermediate forecast data.
[0120] S1002, performing maximum and minimum value normalization processing on other element data in the intermediate forecast data to obtain historical forecast data.
[0121] S1003: Determine whether heavy rainfall will occur within a second preset forecast time period based on historical low-resolution forecast data of the target area.
[0122] S1004: When it is determined that heavy rainfall will occur within the second preset forecast time period, the initial neural network model is trained based on the historical low-resolution forecast data and the historical high-resolution forecast data to obtain a neural network model.
[0123] S1005: Determine whether heavy rainfall will occur within a first preset forecast time period based on the pending low-resolution forecast data for the target area.
[0124] S1006, when it is determined that heavy rainfall will occur within the first preset forecast time period, the low-resolution forecast data to be reported for the target area is input into the neural network model to obtain the high-resolution forecast data to be reported corresponding to the low-resolution forecast data to be reported.
[0125] S1007: Precipitation forecast based on the pending high-resolution forecast data.
[0126] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.
[0127] Based on the same inventive concept, embodiments of the present application also provide a precipitation forecast device for implementing the aforementioned precipitation forecast method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more precipitation forecast device embodiments provided below can be found in the above-described limitations of the precipitation forecast method and will not be further elaborated here.
[0128] In one embodiment, as shown in FIG11 , FIG11 is a structural block diagram of a precipitation forecasting device provided in an embodiment of the present application. The device 1100 includes:
[0129] The first determining module 1101 is configured to determine whether heavy rainfall will occur within a first preset forecast time period based on pending low-resolution forecast data for a target area.
[0130] The first acquisition module 1102 is used to input the low-resolution forecast data to be reported in the target area into the neural network model when it is determined that heavy rainfall will occur within the first preset forecast time period, so as to obtain the high-resolution forecast data to be reported corresponding to the low-resolution forecast data to be reported.
[0131] The forecast module 1103 is used to make precipitation forecast based on the high-resolution forecast data to be reported.
[0132] In one embodiment, the apparatus 1100 further includes:
[0133] The second acquisition module is used to obtain historical low-resolution forecast data and historical high-resolution forecast data of the target area.
[0134] The second determination module is used to determine whether heavy rainfall will occur within a second preset forecast time period based on historical low-resolution forecast data of the target area.
[0135] The first training module is used to train the initial neural network model based on historical low-resolution forecast data and historical high-resolution forecast data to obtain a neural network model when it is determined that heavy rainfall will occur within the second preset forecast time period.
[0136] In one embodiment, the second determining module includes:
[0137] The first acquisition unit is used to acquire the most recent first low-resolution forecast data from the historical low-resolution forecast data.
[0138] A calculation unit is used to calculate the precipitation area and precipitation intensity within the second preset forecast time period based on the historical low-resolution forecast data if the first low-resolution forecast data indicates that rainfall will occur within the second preset forecast time period.
[0139] The first determining unit is configured to determine that heavy precipitation will occur within a second preset forecast time period when the precipitation area reaches a preset precipitation area threshold and / or the precipitation intensity reaches a preset precipitation intensity threshold.
[0140] In one embodiment, the first training module includes:
[0141] The second acquisition unit is used to acquire second forecast data of a second preset forecast time within a preset time length from the historical low-resolution forecast data and the historical high-resolution forecast data; the second forecast data includes the second low-resolution forecast data and the second high-resolution forecast data.
[0142] A training unit is used to train the initial neural network model based on the second forecast data to obtain a neural network model if the data amount of the second forecast data reaches a preset data amount.
[0143] In one embodiment, the apparatus 1100 further includes:
[0144] The third acquisition module is used to extend the preset time length to obtain the target time length if the data volume of the second forecast data does not reach the preset data volume; the target time length includes the preset time length, and the data volume of the third forecast data of the second preset forecast time length within the target time length reaches the preset data volume.
[0145] The second training module is used to train the initial neural network model based on the third forecast data to obtain a neural network model.
[0146] In one embodiment, the apparatus 1100 further includes:
[0147] The fourth acquisition module is used to acquire initial forecast data of the target area; the initial forecast data includes low-resolution forecast data and initial high-resolution forecast data.
[0148] The first preprocessing module is used to perform standardization processing on the precipitation element data in the initial forecast data based on a logarithmic function to obtain intermediate forecast data.
[0149] The second preprocessing module is used to perform maximum and minimum value normalization processing on other element data in the intermediate forecast data to obtain historical forecast data; the historical forecast data includes historical low-resolution forecast data and historical high-resolution forecast data.
[0150] Each module in the precipitation forecast device described above may be implemented in whole or in part through software, hardware, or a combination thereof. Each module may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.
[0151] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:
[0152] Determine whether heavy rainfall will occur within a first preset forecast timeframe based on pending low-resolution forecast data for the target area;
[0153] When it is determined that heavy rainfall will occur within the first preset forecast time period, the low-resolution forecast data to be reported for the target area is input into the neural network model to obtain the high-resolution forecast data to be reported corresponding to the low-resolution forecast data to be reported;
[0154] Precipitation forecast is made based on pending high-resolution forecast data.
[0155] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0156] Obtain historical low-resolution forecast data and historical high-resolution forecast data for the target area;
[0157] Determine whether heavy rainfall will occur within the second preset forecast timeframe based on historical low-resolution forecast data for the target area;
[0158] When it is determined that heavy rainfall will occur within the second preset forecast time period, the initial neural network model is trained based on historical low-resolution forecast data and historical high-resolution forecast data to obtain a neural network model.
[0159] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0160] Obtaining the most recent first low-resolution forecast data from historical low-resolution forecast data;
[0161] If the first low-resolution forecast data indicates that rainfall will occur within the second preset forecast time period, then calculating the precipitation area and precipitation intensity within the second preset forecast time period based on the historical low-resolution forecast data;
[0162] When the precipitation area reaches a preset precipitation area threshold, and / or the precipitation intensity reaches a preset precipitation intensity threshold, it is determined that heavy precipitation will occur within the second preset forecast time period.
[0163] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0164] Acquire second forecast data of a second preset forecast time within a preset time period from the historical low-resolution forecast data and the historical high-resolution forecast data; the second forecast data includes the second low-resolution forecast data and the second high-resolution forecast data;
[0165] If the data amount of the second forecast data reaches a preset data amount, the initial neural network model is trained based on the second forecast data to obtain a neural network model.
[0166] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0167] If the data volume of the second forecast data does not reach the preset data volume, the preset time period is extended to obtain a target time period; the target time period includes the preset time period, and the data volume of the third forecast data of the second preset forecast time period within the target time period reaches the preset data volume;
[0168] The initial neural network model is trained based on the third forecast data to obtain a neural network model.
[0169] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0170] Obtaining initial forecast data for the target area; the initial forecast data includes low-resolution forecast data and initial high-resolution forecast data;
[0171] The precipitation element data in the initial forecast data are normalized based on the logarithmic function to obtain the intermediate forecast data;
[0172] The maximum and minimum values of other element data in the intermediate forecast data are normalized to obtain historical forecast data; the historical forecast data includes historical low-resolution forecast data and historical high-resolution forecast data.
[0173] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:
[0174] Determine whether heavy rainfall will occur within a first preset forecast timeframe based on pending low-resolution forecast data for the target area;
[0175] When it is determined that heavy rainfall will occur within the first preset forecast time period, the low-resolution forecast data to be reported for the target area is input into the neural network model to obtain the high-resolution forecast data to be reported corresponding to the low-resolution forecast data to be reported;
[0176] Precipitation forecast is made based on pending high-resolution forecast data.
[0177] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0178] Obtain historical low-resolution forecast data and historical high-resolution forecast data for the target area;
[0179] Determine whether heavy rainfall will occur within the second preset forecast timeframe based on historical low-resolution forecast data for the target area;
[0180] When it is determined that heavy rainfall will occur within the second preset forecast time period, the initial neural network model is trained based on historical low-resolution forecast data and historical high-resolution forecast data to obtain a neural network model.
[0181] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0182] Obtaining the most recent first low-resolution forecast data from historical low-resolution forecast data;
[0183] If the first low-resolution forecast data indicates that rainfall will occur within the second preset forecast time period, then calculating the precipitation area and precipitation intensity within the second preset forecast time period based on the historical low-resolution forecast data;
[0184] When the precipitation area reaches a preset precipitation area threshold, and / or the precipitation intensity reaches a preset precipitation intensity threshold, it is determined that heavy precipitation will occur within the second preset forecast time period.
[0185] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0186] Acquire second forecast data of a second preset forecast time within a preset time period from the historical low-resolution forecast data and the historical high-resolution forecast data; the second forecast data includes the second low-resolution forecast data and the second high-resolution forecast data;
[0187] If the data amount of the second forecast data reaches a preset data amount, the initial neural network model is trained based on the second forecast data to obtain a neural network model.
[0188] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0189] If the data volume of the second forecast data does not reach the preset data volume, the preset time period is extended to obtain a target time period; the target time period includes the preset time period, and the data volume of the third forecast data of the second preset forecast time period within the target time period reaches the preset data volume;
[0190] The initial neural network model is trained based on the third forecast data to obtain a neural network model.
[0191] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0192] Obtaining initial forecast data for the target area; the initial forecast data includes low-resolution forecast data and initial high-resolution forecast data;
[0193] The precipitation element data in the initial forecast data are normalized based on the logarithmic function to obtain the intermediate forecast data;
[0194] The maximum and minimum values of other element data in the intermediate forecast data are normalized to obtain historical forecast data; the historical forecast data includes historical low-resolution forecast data and historical high-resolution forecast data.
[0195] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:
[0196] Determine whether heavy rainfall will occur within a first preset forecast timeframe based on pending low-resolution forecast data for the target area;
[0197] When it is determined that heavy rainfall will occur within the first preset forecast time period, the low-resolution forecast data to be reported for the target area is input into the neural network model to obtain the high-resolution forecast data to be reported corresponding to the low-resolution forecast data to be reported;
[0198] Precipitation forecast is made based on pending high-resolution forecast data.
[0199] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0200] Obtain historical low-resolution forecast data and historical high-resolution forecast data for the target area;
[0201] Determine whether heavy rainfall will occur within the second preset forecast timeframe based on historical low-resolution forecast data for the target area;
[0202] When it is determined that heavy rainfall will occur within the second preset forecast time period, the initial neural network model is trained based on historical low-resolution forecast data and historical high-resolution forecast data to obtain a neural network model.
[0203] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0204] Obtaining the most recent first low-resolution forecast data from historical low-resolution forecast data;
[0205] If the first low-resolution forecast data indicates that rainfall will occur within the second preset forecast time period, then calculating the precipitation area and precipitation intensity within the second preset forecast time period based on the historical low-resolution forecast data;
[0206] When the precipitation area reaches a preset precipitation area threshold, and / or the precipitation intensity reaches a preset precipitation intensity threshold, it is determined that heavy precipitation will occur within the second preset forecast time period.
[0207] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0208] Acquire second forecast data of a second preset forecast time within a preset time period from the historical low-resolution forecast data and the historical high-resolution forecast data; the second forecast data includes the second low-resolution forecast data and the second high-resolution forecast data;
[0209] If the data amount of the second forecast data reaches a preset data amount, the initial neural network model is trained based on the second forecast data to obtain a neural network model.
[0210] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0211] If the data volume of the second forecast data does not reach the preset data volume, the preset time period is extended to obtain a target time period; the target time period includes the preset time period, and the data volume of the third forecast data of the second preset forecast time period within the target time period reaches the preset data volume;
[0212] The initial neural network model is trained based on the third forecast data to obtain a neural network model.
[0213] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:
[0214] Obtaining initial forecast data for the target area; the initial forecast data includes low-resolution forecast data and initial high-resolution forecast data;
[0215] The precipitation element data in the initial forecast data are normalized based on the logarithmic function to obtain the intermediate forecast data;
[0216] The maximum and minimum values of other element data in the intermediate forecast data are normalized to obtain historical forecast data; the historical forecast data includes historical low-resolution forecast data and historical high-resolution forecast data.
[0217] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), data processing logic devices based on quantum computing, and the like.
[0218] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0219] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.
Claims
1. A precipitation forecasting method, characterized in that: The method comprises: Determine whether heavy precipitation will occur within the first preset forecast time period based on the pending low-resolution forecast data of the target area; When it is determined that heavy rainfall will occur within the first preset forecast time period, the low-resolution forecast data to be reported for the target area is input into the neural network model to obtain the high-resolution forecast data to be reported corresponding to the low-resolution forecast data to be reported; Precipitation forecast is performed based on the pending high-resolution forecast data.
2. The method according to claim 1, characterized in that Before inputting the low-resolution forecast data to be reported for the target area into the neural network model, the method further comprises: Acquiring historical low-resolution forecast data and historical high-resolution forecast data for the target area; Determining whether heavy precipitation will occur within a second preset forecast time period based on historical low-resolution forecast data of the target area; When it is determined that heavy precipitation will occur within the second preset forecast time period, the initial neural network model is trained based on the historical low-resolution forecast data and the historical high-resolution forecast data to obtain the neural network model.
3. The method according to claim 2, characterized in that The determining, based on the historical low-resolution forecast data of the target area, whether heavy precipitation will occur within the second preset forecast time period includes: Acquire the most recent first low-resolution forecast data from the historical low-resolution forecast data; If the first low-resolution forecast data indicates that rainfall will occur within the second preset forecast time period, then calculating the precipitation area and precipitation intensity within the second preset forecast time period based on the historical low-resolution forecast data; When the precipitation area reaches a preset precipitation area threshold, and / or the precipitation intensity reaches a preset precipitation intensity threshold, it is determined that heavy precipitation will occur within a second preset forecast time period.
4. The method according to claim 2, characterized in that: The training of the initial neural network model based on the historical low-resolution forecast data and the historical high-resolution forecast data to obtain the neural network model includes: Acquire second forecast data of the second preset forecast time within a preset time length from the historical low-resolution forecast data and the historical high-resolution forecast data; the second forecast data includes second low-resolution forecast data and second high-resolution forecast data; If the data volume of the second forecast data reaches a preset data volume, the initial neural network model is trained based on the second forecast data to obtain the neural network model.
5. The method according to claim 4, characterized in that The method further comprises: If the data volume of the second forecast data does not reach the preset data volume, the preset duration is extended to obtain a target duration; the target duration includes the preset duration, and the data volume of the third forecast data of the second preset forecast time within the target duration reaches the preset data volume; An initial neural network model is trained based on the third forecast data to obtain the neural network model.
6. The method according to any one of claims 1 to 5, characterized in that: The method further comprises: Acquiring initial forecast data for the target area; the initial forecast data includes low-resolution forecast data and initial high-resolution forecast data; Standardizing the precipitation element data in the initial forecast data based on a logarithmic function to obtain intermediate forecast data; The other element data in the intermediate forecast data are normalized by maximum and minimum values to obtain historical forecast data; the historical forecast data includes historical low-resolution forecast data and historical high-resolution forecast data.
7. A precipitation forecasting device, characterized in that: The device comprises: A first determination module is used to determine whether heavy precipitation will occur within a first preset forecast time period according to the pending low-resolution forecast data of the target area; A first acquisition module is used for inputting the low-resolution forecast data to be reported in the target area into the neural network model to obtain the high-resolution forecast data to be reported corresponding to the low-resolution forecast data to be reported when it is determined that heavy precipitation will occur within the first preset forecast time limit; A forecast module is used to make precipitation forecast based on the high-resolution forecast data to be reported.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.