Medium-and-long-term rainfall forecasting method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of weather forecast, and discloses a medium-and-long-term rainfall forecast method and device, computer equipment and a storage medium, and the method comprises the steps: firstly obtaining a preset medium-and-long-term numerical forecast data set, and guaranteeing the reliability and stability of a data source; secondly, a rainfall-related meteorological element set is extracted through a preset medium-and-long-term numerical forecasting data set, factors directly related to rainfall can be focused, and some interference variables are eliminated. Furthermore, a precipitation ingredient element set is determined in combination with an ingredient method, and the effects of various meteorological elements in the precipitation forming process are fully considered. And finally, medium-and-long-term rainfall forecasting is carried out through a machine learning method, the advantages of the machine learning algorithm in the aspects of processing a large amount of data and complex relations are fully played, and the rainfall forecasting precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological forecasting, and in particular to a medium- and long-term precipitation forecasting method, device, computer equipment and storage medium. Background Art

[0002] In the past, weather forecasting methods were mainly based on statistical methods of model output of numerical forecast products, empirical rule methods, flow pattern identification methods, etc. For the forecast of high-impact weather (cold waves, blizzards, heat waves, heavy rains, severe convection, etc.), flow pattern identification, i.e., weather situation classification rules, is a generally recognized method. This method compares the weather situation at the time with the typical flow pattern configuration to predict the heavy precipitation area, and it is still widely used today.

[0003] However, the flow pattern identification method has great shortcomings. In the actual process, there are situations where the actual weather situation does not match any typical flow pattern configuration, but heavy precipitation still occurs, resulting in missed reports. Whenever this happens, meteorological experts and forecasters will introduce another flow pattern configuration. However, this method is not only difficult to improve the forecast level of heavy precipitation, but will increase the burden on forecasters. Secondly, when the weather situation is similar to the typical flow pattern in part, the details are not completely consistent, which misleads forecasters and predicts regional heavy precipitation, but it does not actually occur, resulting in false reports. On the other hand, parameterization methods are usually used in numerical forecast models to forecast precipitation, such as cumulus convection, cloud microphysical parameterization, etc. Some forecast factors used in parameterization methods, such as environmental field convective effective potential energy production, vertical velocity, etc., have been proven to be effective. However, the parameterization scheme itself carries many artificial assumptions and introduces a large number of uncertain parameters, which to a certain extent restricts the precipitation forecast performance of the model. Summary of the invention

[0004] In view of this, the present invention provides a medium- and long-term precipitation forecasting method, apparatus, computer equipment and storage medium to solve the problem of low forecasting accuracy of existing precipitation forecasting methods.

[0005] In a first aspect, the present invention provides a medium- to long-term precipitation forecasting method, the method comprising:

[0006] Obtain a preset medium- and long-term numerical forecast data set; extract the preset medium- and long-term numerical forecast data set to obtain a precipitation-related meteorological element set; based on the precipitation-related meteorological element set, use the batching method to determine the initial precipitation batching element set; based on the machine learning method, use the initial precipitation batching element set to perform medium- and long-term precipitation forecasts to obtain a target precipitation forecast data set.

[0007] The medium- and long-term precipitation forecasting method provided by the present invention first obtains a preset medium- and long-term numerical forecast data set to ensure the reliability and stability of the data source. Secondly, by extracting a precipitation-related meteorological element set from the preset medium- and long-term numerical forecast data set, it is possible to focus on factors directly related to precipitation and exclude some interfering variables. Further, the precipitation ingredient element set is determined in combination with the batching method, and the role of various meteorological elements in the precipitation formation process is fully considered. Finally, medium- and long-term precipitation forecasting is performed through a machine learning method, which gives full play to the advantages of machine learning algorithms in processing large amounts of data and complex relationships, and improves the accuracy of precipitation forecasting. Therefore, by implementing the present invention, the batching method is combined with machine learning to achieve real-time precipitation forecasting, greatly improving the prediction accuracy, and further for heavy precipitation areas, meeting the precipitation forecasting needs of strong weather processes such as typhoons and rainstorms, which is of great significance to the development of the meteorological industry.

[0008] In an optional implementation, obtaining a preset medium- and long-term numerical forecast data set includes:

[0009] An initial medium- and long-term numerical forecast data set is obtained; and the initial medium- and long-term numerical forecast data set is processed to obtain a preset medium- and long-term numerical forecast data set.

[0010] The medium- and long-term precipitation forecasting method provided by the present invention can ensure the accuracy and reliability of a preset medium- and long-term numerical forecasting data set by processing an initial medium- and long-term numerical forecasting data set.

[0011] In an optional implementation, based on the precipitation-related meteorological element set, an initial precipitation ingredient element set is determined using an ingredient method, including:

[0012] The preset mathematical model is used to calculate the correlation between the elements in the precipitation-related meteorological element set, where the correlation is one or more of correlation, potential combination pattern and matching relationship; based on the precipitation-related meteorological element set and the correlation, the initial precipitation ingredient element set is determined.

[0013] The medium- and long-term precipitation forecasting method provided by the present invention can provide a deeper understanding of the internal mechanism of precipitation formation by calculating the correlation between various elements. Furthermore, the precipitation-related meteorological element set and the correlation are combined to determine the initial precipitation ingredient element set, which fully considers the role of various meteorological elements in the precipitation formation process and ensures the accuracy and pertinence of the ingredient elements.

[0014] In an optional implementation, based on a machine learning method, a precipitation ingredient element set is used to perform medium- and long-term precipitation forecasting to obtain a target precipitation forecast data set, including:

[0015] Based on the precipitation ingredient element set, the revised initial precipitation forecast data set is obtained after being processed by the precipitation forecast model based on the multi-layer perceptron; the initial precipitation forecast data set is processed by the super-resolution algorithm to obtain the target precipitation forecast data set.

[0016] The medium- and long-term precipitation forecasting method provided by the present invention can fully explore the complex relationships and patterns in the data through the precipitation forecasting model based on the multi-layer perceptron. Furthermore, by using the model to process the precipitation ingredient element set, a revised initial precipitation forecast data set is obtained, which corrects possible errors and deviations to a certain extent. Finally, the spatial resolution and detail information of the data are increased by using the super-resolution algorithm, making the precipitation forecast more precise and accurate.

[0017] In an optional implementation, based on the initial precipitation ingredient element set, after being processed by a precipitation forecast model based on a multi-layer perceptron, a revised initial precipitation forecast data set is obtained, including:

[0018] The historical precipitation ingredient element set is obtained, and the historical precipitation ingredient element set is used for training to obtain a precipitation forecast model based on a multi-layer perceptron; the initial precipitation ingredient element set is processed to obtain a target precipitation ingredient element set; the target precipitation ingredient element set is input into the precipitation forecast model based on a multi-layer perceptron to obtain a revised initial precipitation forecast data set.

[0019] In an optional embodiment, the method further includes:

[0020] The actual precipitation observation dataset is obtained, and the accuracy of the target precipitation forecast dataset is verified based on the actual precipitation observation dataset to obtain the verification result.

[0021] The medium- and long-term precipitation forecasting method provided by the present invention can further verify the accuracy of the target precipitation forecasting dataset through the actual precipitation observation dataset.

[0022] In a second aspect, the present invention provides a medium- and long-term precipitation forecasting device, the device comprising:

[0023] The acquisition module is used to obtain a preset medium- and long-term numerical forecast data set; the extraction module is used to extract the preset medium- and long-term numerical forecast data set to obtain a precipitation-related meteorological element set; the determination module is used to determine the initial precipitation ingredient element set based on the precipitation-related meteorological element set using the ingredient method; the forecast module is used to use the initial precipitation ingredient element set to perform medium- and long-term precipitation forecasting based on a machine learning method to obtain a target precipitation forecast data set.

[0024] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the medium- and long-term precipitation forecasting method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0025] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the medium- and long-term precipitation forecasting method of the first aspect or any corresponding embodiment thereof.

[0026] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the medium- and long-term precipitation forecasting method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0028] Figure 1 is a schematic flow chart of a medium- to long-term precipitation forecasting method according to an embodiment of the present invention;

[0029] Figure 2 is a flow chart of another medium- to long-term precipitation forecasting method according to an embodiment of the present invention;

[0030] Figure 3 is a schematic diagram of a precipitation forecast model based on a multi-layer perceptron according to an embodiment of the present invention;

[0031] Figure 4 is a flow chart of another medium- to long-term precipitation forecasting method according to an embodiment of the present invention;

[0032] Figure 5 is a flow chart of a medium- and long-term precipitation forecasting method combining a batching method and a machine learning method according to an embodiment of the present invention;

[0033] Figure 6 is a structural block diagram of a medium- and long-term precipitation forecasting device according to an embodiment of the present invention;

[0034] Figure 7 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0036] Heavy rain brings a lot of precipitation, which may cause the water level of rivers and lakes to rise sharply, causing floods. Floods may submerge low-lying areas, causing serious damage to houses, farmland and infrastructure, resulting in property losses and casualties. Heavy rainfall can easily lead to mountain torrents and mudslides in mountainous areas. The large amount of water and sediment caused by heavy rain may flow rapidly along the hillside, causing devastating effects on surrounding areas and threatening people's lives and property. Heavy rain can cause waterlogging and mudslides on roads, resulting in flooding or damage to roads and traffic interruptions. This makes transportation difficult and may hinder the arrival of rescue and emergency services. Soil saturation and increased surface water accumulation caused by heavy rain can easily cause land landslides and soil collapse, threatening the safety of houses and people, and also causing damage to land resources. On the other hand, heavy rain may cause sewage systems to overload, sewage overflows, and cause environmental pollution. Pollutants and harmful substances in the water may have a negative impact on people's health and the environment. At the same time, it will also cause soaking and flooding of farmland, which may cause damage and loss of crops, affecting agricultural production and food supply.

[0037] In recent years, the impact of extreme weather on the economy has become increasingly significant. The research on medium- and long-term precipitation forecasts is of great significance to the accuracy of meteorological forecasting decisions. With the development of artificial intelligence and the advancement of technologies such as deep learning, medium- and long-term precipitation research has ushered in more development opportunities. Traditional medium- and long-term precipitation forecasting methods mainly include statistical forecasting, empirical extrapolation methods, and numerical model forecasting. Early forecasts mostly relied on empirical extrapolation methods.

[0038] With the advancement of computer technology, statistical methods and numerical models have been more widely used in the field of precipitation forecasting research. Statistical forecasting is not only used to predict precipitation factors, but also has achieved certain results in the forecast of precipitation situation. However, statistical weather forecasting lacks a physical basis and the stability of statistical relationships is also low, so it still faces many challenges. With the increase of meteorological data and the development of satellite exploration technology, the traditional forecasting methods based on various data have further improved in forecasting effect.

[0039] Traditional medium- and long-term precipitation forecasting methods mainly include statistical precipitation forecasting, precipitation forecasting based on numerical models, and precipitation forecasting based on radar extrapolation. The data used by these methods are not exactly the same. Some use a single type of data, while others combine different types of data to predict precipitation. Parameterization methods are usually used in numerical prediction models to forecast precipitation, such as cumulus convection, cloud microphysical parameterization, etc. Some prediction factors used in parameterization methods, such as environmental field convective effective potential energy production, vertical velocity, etc., have been proven to be effective. However, the parameterization scheme itself carries many artificial assumptions and introduces a large number of uncertain parameters, which to a certain extent restricts the precipitation forecast performance of the model.

[0040] With the advancement of precipitation detection technology, meteorological data for medium- and long-term precipitation forecasts have been further improved. While improving traditional methods, researchers have conducted more in-depth research. However, traditional statistical methods and numerical forecasting methods often provide overly functional and regular results for medium- and long-term precipitation forecasts, which in turn limits their accuracy. The radar-based extrapolation method is affected by geographical factors and the accuracy of radar data, and there are also certain limitations in the research.

[0041] With the deepening of research, domestic weather forecasting mainly relies on model output statistics, empirical rule method and flow type identification method of numerical forecast products.

[0042] According to an embodiment of the present invention, an embodiment of a medium- to long-term precipitation forecasting method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0043] In this embodiment, a medium- to long-term precipitation forecast method is provided, which can be used in electronic devices such as computers, mobile phones, tablet computers, etc. Figure 1 is a flow chart of a medium- to long-term precipitation forecasting method according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0044] Step S101, obtaining a preset medium- and long-term numerical forecast data set.

[0045] Specifically, the preset medium- and long-term numerical forecast data set represents the basic data set used for medium- and long-term precipitation forecast, covering a variety of meteorological elements and precipitation-related data, which may include meteorological element data (temperature, humidity, air pressure, etc.), precipitation-related data (precipitation rate, vertical wind speed, etc.), other related data (geopotential height, meridional wind, zonal wind, etc.) and other data.

[0046] Step S102: extracting a preset medium- and long-term numerical forecast data set to obtain a precipitation-related meteorological element set.

[0047] Among them, the precipitation-related meteorological element set may include precipitation, temperature, humidity, wind field and other related data.

[0048] Specifically, precipitation-related element variables can be extracted from a preset medium- and long-term numerical forecast data set to form a precipitation-related meteorological element set.

[0049] Furthermore, the precipitation-related meteorological element set may be stored in a binary dat format.

[0050] Step S103, based on the precipitation-related meteorological element set, an initial precipitation ingredient element set is determined using an ingredient method.

[0051] Among them, the batching method, also known as the component method, is a method of using numerical forecast products. The batching method believes that precipitation is the result of the combined action of several components. The elements in the batching method are defined as a basic physical quantity or process that directly affects the development and intensity of precipitation events. The correct combination of these basic components can produce expected weather phenomena, thereby providing fixed-point, fixed-time, and quantitative precipitation forecasts. Several basic physical components are ascending forcing, atmospheric instability, and precipitable water.

[0052] Furthermore, the initial precipitation ingredient element set can specifically include convective effective potential energy, environmental field convective effective potential energy production, convective inhibition energy, columnar water vapor convergence and terrain vertical lifting speed:

[0053] 1. Convective Available Potential Energy (CAPE): It represents the potential energy existing in the atmosphere and is a measure of atmospheric instability.

[0054] 2. Production of convective effective potential energy in environmental fields: mainly describes the generation rate or contribution degree of convective effective potential energy under specific environmental field conditions.

[0055] 3. Convective inhibition energy: an indicator that measures the degree of atmospheric resistance to the development of convection.

[0056] 4. Column water vapor convergence: the total amount of water vapor contained in the air column per unit area.

[0057] 5. Vertical uplift speed of terrain: related to wind speed, wind direction and terrain slope.

[0058] Specifically, based on the extracted precipitation-related meteorological element set, the ingredient method can be used to diagnose and calculate the "ingredient" elements that constitute precipitation, and obtain the corresponding initial precipitation ingredient element set.

[0059] Step S104, based on a machine learning method, using the initial precipitation ingredient element set to perform medium- and long-term precipitation forecasts to obtain a target precipitation forecast data set.

[0060] Specifically, by using machine learning methods to make medium- and long-term precipitation forecasts based on the initial precipitation ingredient element set, we can give full play to the advantages of machine learning algorithms in processing large amounts of data and complex relationships, thereby improving the accuracy of precipitation forecasts and making the target precipitation forecast data set more accurate.

[0061] The medium- and long-term precipitation forecasting method provided in this embodiment first obtains a preset medium- and long-term numerical forecast data set to ensure the reliability and stability of the data source. Secondly, by extracting a set of precipitation-related meteorological elements from the preset medium- and long-term numerical forecast data set, it is possible to focus on factors directly related to precipitation and eliminate some interfering variables. Further, the precipitation batching element set is determined in combination with the batching method, and the role of various meteorological elements in the precipitation formation process is fully considered. Finally, medium- and long-term precipitation forecasting is performed through machine learning methods, which gives full play to the advantages of machine learning algorithms in processing large amounts of data and complex relationships, and improves the accuracy of precipitation forecasting. Therefore, by implementing the present invention, the batching method is combined with machine learning to achieve real-time precipitation forecasting, greatly improving the prediction accuracy, and further meeting the precipitation forecasting needs of strong weather processes such as typhoons and rainstorms for heavy precipitation areas, which is of great significance to the development of the meteorological industry.

[0062] In this embodiment, a medium- to long-term precipitation forecast method is provided, which can be used in electronic devices such as computers, mobile phones, tablet computers, etc. Figure 2 is a flow chart of a medium- to long-term precipitation forecasting method according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0063] Step S201, obtain a preset medium- and long-term numerical forecast data set. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0064] Step S202: extract the preset medium- and long-term numerical forecast data set to obtain a set of precipitation-related meteorological elements. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0065] Step S203, based on the precipitation-related meteorological element set, determine the initial precipitation ingredient element set using the ingredient method.

[0066] Specifically, the above step S203 includes:

[0067] Step S2031, using a preset mathematical model to calculate the correlation between the elements in the precipitation related meteorological element set.

[0068] The association relationship may be one or more of a correlation, a potential combination pattern, and a matching relationship.

[0069] Specifically, the calculation of different association relationships may include:

[0070] (1) Relevance

[0071] Correlation analysis can use methods such as Pearson correlation coefficient and Spearman correlation coefficient. Pearson correlation coefficient measures the strength of the linear relationship between two variables, while Spearman correlation coefficient mainly measures the monotonic relationship between variables.

[0072] For example, for the two elements of temperature and precipitation rate, the Pearson correlation coefficient between them is calculated. By collecting temperature and precipitation rate data at multiple time points or spatial locations, the correlation between the two is calculated according to the calculation formula of the correlation coefficient. If the correlation coefficient is close to 1 or -1, it means that there is a strong linear relationship between the two; if it is close to 0, it means that the linear relationship is weak.

[0073] Further, the correlation coefficient calculation method selected can be used to calculate each pair of elements in the precipitation-related meteorological element set. Further, the correlation coefficient value between each pair of elements is recorded, and statistical analysis is performed to determine which elements have significant correlations.

[0074] (2) Potential combination patterns

[0075] Potential combination pattern analysis can use methods such as principal component analysis (PCA) and factor analysis. These methods can reduce the dimensionality of multiple meteorological elements and find potential combination patterns. Taking principal component analysis as an example, the set of precipitation-related meteorological elements can be used as input data, and the original data can be converted into a new coordinate system by calculating the covariance matrix, eigenvalues, and eigenvectors. In the new coordinate system, the first few principal components can usually explain most of the data variation, and these principal components represent the potential combination patterns between meteorological elements.

[0076] Furthermore, the results such as principal components or factor loading matrices can be calculated based on the selected potential combination pattern analysis method, such as principal component analysis or factor analysis. Furthermore, the meaning of the principal components or factors is analyzed to determine the potential combination pattern between meteorological elements. For example, if the first few principal components are mainly composed of temperature, humidity and precipitation rate, then it can be considered that there is a potential combination pattern between these three elements.

[0077] (3) Ratio relationship

[0078] Matching relationship analysis can establish multiple regression models or neural network models, etc., and these models can learn the complex relationship between meteorological elements, so as to determine the matching relationship between them. For example, a multiple linear regression model is established with precipitation as the dependent variable and other meteorological elements as independent variables. By collecting a large amount of data for training, the model can learn the influence of each independent variable on the dependent variable, so as to determine the matching relationship between them.

[0079] In an optional implementation, a statistical analysis method may be used to calculate the correlation between the various elements.

[0080] Step S2032, determining an initial precipitation ingredient element set based on the precipitation-related meteorological element set and the association relationship.

[0081] Specifically, the convective effective potential energy and convective inhibition energy in the initial precipitation ingredient element set can be calculated through the elements of air pressure, temperature and specific humidity. The convective effective potential energy production and columnar water vapor convergence in the environmental field can be calculated through elements such as air pressure, latitudinal / meridional wind, vertical wind speed, temperature, specific humidity and potential height. The terrain vertical uplift velocity can be calculated through terrain and latitudinal / meridional wind.

[0082] Furthermore, the special terrain causes airflow to converge, thus forming forced lift. At the same time, the terrain can also serve as a triggering mechanism for small and medium-scale severe convective systems, causing the release of unstable energy.

[0083] Furthermore, the vertical lift velocity of the terrain is related to the wind speed, the angle between the wind direction and the slope direction, and the terrain slope, and can be calculated using the following equations (1) to (3):

[0084]

[0085] Where: α represents the slope of the slope; β represents the slope direction of the windward slope; and They represent the partial derivatives of the terrain height function h in the x and y directions, respectively, reflecting the changes in the terrain in different directions; ω represents the vertical uplift speed of the terrain; v represents the wind speed; σ represents the angle between the wind direction and the slope direction.

[0086] Furthermore, the wind speed blowing toward the slope surface produces an upward motion (ω<0) that is significant for precipitation due to the influence of slope resistance.

[0087] Furthermore, the calculation of convective effective potential energy needs to take into account the changes in temperature and humidity in the atmosphere. When there is instability in temperature and humidity in the atmosphere, convective motion will occur. The larger the value of, the higher the potential energy in the atmosphere, and the more intense the convective motion will be, as shown in the following relationship (4):

[0088]

[0089] Where: P b and P t Respectively represent the upper and lower limits of air pressure; R d represents the gas constant, which is a constant related to the properties of the atmosphere; T vp Indicates the temperature value after considering the influence of water vapor on temperature; T v Indicates temperature value; T vp -T v Used to reflect temperature instability in the atmosphere.

[0090] Furthermore, the columnar water vapor convergence w corresponds to the amount of precipitation that can be formed when all the water in the air condenses into rain or snow (squeezing the air until there is no water at all), as shown in the following equation (5):

[0091]

[0092] In the formula: g represents the acceleration of gravity; q represents specific humidity, that is, the mass of water vapor contained in unit mass of wet air, which reflects the water vapor content in the air.

[0093] Step S204, based on a machine learning method, using the initial precipitation ingredient element set to perform medium- and long-term precipitation forecasts to obtain a target precipitation forecast data set.

[0094] Specifically, the above step S204 includes:

[0095] Step S2041, based on the precipitation ingredient element set, the precipitation forecast model based on the multi-layer perceptron is processed to obtain a revised initial precipitation forecast data set.

[0096] Specifically, the multi-layer perceptron is a powerful machine learning algorithm that can learn the complex relationship between input data (i.e., precipitation ingredient element set) and output data (i.e., precipitation forecast results). Therefore, through the precipitation forecast model based on the multi-layer perceptron, the complex relationships and patterns in the data can be fully mined.

[0097] Furthermore, by using the model to process the precipitation ingredient element set, the revised initial precipitation forecast data set is obtained, which corrects the possible errors and deviations to a certain extent.

[0098] In some optional implementations, the above step S2041 includes:

[0099] Step a1, obtaining a historical precipitation ingredient element set, and using the historical precipitation ingredient element set for training to obtain a precipitation forecast model based on a multi-layer perceptron.

[0100] Step a2, processing the initial precipitation ingredient element set to obtain the target precipitation ingredient element set.

[0101] Step a3: input the target precipitation ingredient element set into the precipitation forecast model based on the multi-layer perceptron to obtain a revised initial precipitation forecast data set.

[0102] Among them, the precipitation forecast model based on multi-layer perceptron is as follows: Figure 3 shown.

[0103] Specifically, before establishing a precipitation forecast model of a multi-layer perceptron, it is necessary to select features related to precipitation prediction and extract these features, namely the historical precipitation ingredient element set, from the original data as model input.

[0104] Furthermore, based on the extracted features, a precipitation forecast model of a multi-layer perceptron is established. The precipitation forecast model of the multi-layer perceptron is trained using training data to learn the patterns and associations between features. During the training process, an optimization algorithm is used to adjust the weights and parameters of the model to minimize the prediction error.

[0105] The neural network consists of an input layer (the first layer), a hidden layer (the middle layer), and an output layer (the last layer), and the layers are fully connected. The forward propagation algorithm is used to solve the relationship between the outputs of two adjacent layers. The input of a layer is the output of the previous layer. The directional propagation algorithm is used to solve the coefficient relationship matrix and bias of each layer. ReLU is used as the activation function, which introduces nonlinear properties into the neural network so that it can learn complex nonlinear relationships.

[0106] Multilayer perceptron is the basic algorithm of deep neural network, and its loss function is shown in the following equation (6):

[0107]

[0108] Where D = {x i |y i (w T x i +b)<0}.

[0109] Furthermore, the initial precipitation ingredient element set is processed, which may include data normalization, removal of outliers and other operations to ensure the accuracy and reliability of the data.

[0110] Furthermore, by inputting the processed target precipitation ingredient element set into the precipitation forecast model based on the multi-layer perceptron, the revised initial precipitation forecast data set can be obtained.

[0111] In an optional embodiment, by using a convolutional neural network (CNN) deep learning model, the mapping relationship of the image can be learned, thereby achieving high-quality super-resolution reconstruction. CNN is a multi-layer perceptron similar to an artificial neural network, which is often used to analyze visual images and includes a data input layer, a convolution calculation layer, a ReLU excitation layer, a pooling layer, and a fully connected layer. The WRF simulation results with a resolution of 1 km are used as a training set and coarse-grained to 25 km. The convolutional neural network is used to extract the mapping relationship from 25 km to 1 km for the ultra-high resolution output of the corrected precipitation.

[0112] Step S2042: Process the initial precipitation forecast data set using a super-resolution algorithm to obtain a target precipitation forecast data set.

[0113] Among them, super-resolution algorithm refers to super-resolution algorithm, which is used in the meteorological field to improve the spatial resolution and fineness of precipitation forecast data. It usually uses mathematical models and machine learning techniques to infer high-resolution information from low-resolution data.

[0114] Specifically, the super-resolution algorithm can be used to process the corrected precipitation data to increase the spatial resolution and fineness of the data, and obtain refined precipitation data, namely the target precipitation forecast data set.

[0115] Furthermore, the obtained target precipitation forecast data set can be written and stored in the binary text bz2 standard format, and the size of a single file is about 1 MB.

[0116] The medium- and long-term precipitation forecasting method provided in this embodiment can provide a deeper understanding of the internal mechanism of precipitation formation by calculating the correlation between various elements. Furthermore, the precipitation-related meteorological element set and the correlation are combined to determine the initial precipitation ingredient element set, which fully considers the role of various meteorological elements in the precipitation formation process and ensures the accuracy and pertinence of the ingredient elements. Furthermore, through the precipitation forecasting model based on the multi-layer perceptron, the complex relationships and patterns in the data can be fully mined. Furthermore, by using the model to process the precipitation ingredient element set, a revised initial precipitation forecast data set is obtained, which corrects possible errors and deviations to a certain extent. Finally, the spatial resolution and detail information of the data are increased by using the super-resolution algorithm, making the precipitation forecast more refined and accurate.

[0117] In this embodiment, a medium- to long-term precipitation forecast method is provided, which can be used in electronic devices such as computers, mobile phones, tablet computers, etc. Figure 4 is a flow chart of a medium- to long-term precipitation forecasting method according to an embodiment of the present invention. Figure 4 As shown, the process includes the following steps:

[0118] Step S401, obtaining a preset medium- and long-term numerical forecast data set.

[0119] Specifically, the above step S401 includes:

[0120] Step S4011, obtaining an initial medium- and long-term numerical forecast data set.

[0121] The initial medium- and long-term numerical forecast data set refers to the initial medium- and long-term numerical forecast data before processing that is initially acquired during the medium- and long-term precipitation forecast process. For details, please refer to the description of the preset medium- and long-term numerical forecast data set in step S101 above.

[0122] Step S4012, processing the initial medium- and long-term numerical prediction data set to obtain a preset medium- and long-term numerical prediction data set.

[0123] Specifically, operations such as interpolation alignment, standardization, and missing value filling can be performed on the initial medium- and long-term numerical forecast data set to obtain a corresponding preset medium- and long-term numerical forecast data set.

[0124] Step S402: extract the preset medium- and long-term numerical forecast data set to obtain a set of precipitation-related meteorological elements. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0125] Step S403, based on the precipitation related meteorological element set, determine the initial precipitation ingredient element set using the ingredient method. Figure 2 Step S203 of the illustrated embodiment will not be described in detail here.

[0126] Step S404: Based on the machine learning method, the initial precipitation ingredient element set is used to perform medium- and long-term precipitation forecasts to obtain the target precipitation forecast data set. Figure 2 Step S204 of the illustrated embodiment will not be described in detail here.

[0127] Step S405: acquiring an actual precipitation observation dataset, and verifying the accuracy of the target precipitation forecast dataset based on the actual precipitation observation dataset to obtain a verification result.

[0128] Specifically, the precipitation of severe weather processes such as typhoons and rainstorms can be compared based on the obtained target precipitation forecast dataset and the actual precipitation observation dataset to ensure the accuracy and credibility of the target precipitation forecast dataset.

[0129] In an optional implementation, the accuracy of the target precipitation forecast data set can be further verified by back-calculating historical data and comparing the back-calculated forecast data with the measured data.

[0130] The medium- and long-term precipitation forecasting method provided in this embodiment can ensure the accuracy and reliability of the preset medium- and long-term numerical forecasting dataset by processing the initial medium- and long-term numerical forecasting dataset. Furthermore, the accuracy of the target precipitation forecasting dataset can be further verified by the actual precipitation observation dataset.

[0131] In one example, if Figure 5 As shown, a medium- and long-term precipitation forecasting method combining a batching method and a machine learning method is provided, including:

[0132] Step S1: extracting relevant meteorological variables such as precipitation, temperature, humidity, and wind field based on global numerical forecast data;

[0133] Step S2: diagnose and calculate the “ingredients” elements that constitute precipitation;

[0134] Step S3: Based on the precipitation forecast model of the multi-layer perceptron, the corrected precipitation data is obtained;

[0135] Step S4: Based on the corrected precipitation data, a super-resolution algorithm is used to obtain the final refined precipitation data.

[0136] Among them, step S1 specifically includes: obtaining medium- and long-term numerical forecast data; completing the preprocessing of forecast data by screening precipitation-related meteorological elements, interpolation alignment, standardization processing and filling missing values; based on the forecast data, extracting precipitation-related element variables, such as precipitation rate, vertical wind speed, radial / zonal wind, temperature and specific humidity, etc., and storing them in binary dat format.

[0137] Further, step S2 specifically includes: calculating the correlation, potential combination pattern or proportion relationship between different elements through mathematical models or statistical analysis; calculating the ingredients that constitute precipitation: convective effective potential energy, environmental field convective effective potential energy production, convective inhibition energy, columnar water vapor convergence and terrain vertical lifting speed, etc.

[0138] Further, step S3 specifically includes: before inputting the data, preprocessing the data, including data normalization, removing outliers and other operations to ensure the accuracy and reliability of the data; before establishing the precipitation forecast model of the multi-layer perceptron, it is necessary to select features related to precipitation prediction and extract these features from the original data as model input; based on the extracted features, the precipitation forecast model of the multi-layer perceptron is established. The precipitation forecast model of the multi-layer perceptron is trained using training data to learn the patterns and associations between features. During the training process, an optimization algorithm is used to adjust the weights and parameters of the model to minimize the prediction error; after evaluating and verifying the trained precipitation forecast model of the multi-layer perceptron, the corrected precipitation data is obtained.

[0139] It should be noted that a multi-layer perceptron (i.e., a feedforward artificial neural network model) is used, such as Figure 4 A medium- and long-term precipitation forecast model is established as shown.

[0140] The neural network consists of an input layer (the first layer), a hidden layer (the middle layer), and an output layer (the last layer), and the layers are fully connected. The forward propagation algorithm is used to solve the relationship between the outputs of two adjacent layers. The input of a layer is the output of the previous layer. The directional propagation algorithm is used to solve the coefficient relationship matrix and bias of each layer. ReLU is used as the activation function, which introduces nonlinear properties into the neural network so that it can learn complex nonlinear relationships.

[0141] Multilayer perceptron is the basic algorithm of deep neural network, and its loss function is as follows:

[0142]

[0143] Where D = {x i |y i (w T x i +b)<0}.

[0144] The stochastic gradient descent method is used to solve the matrix coefficients.

[0145] The three-dimensional wind field, temperature, and humidity fields are derived from ERA5 reanalysis data with a time resolution of 3 hours and a horizontal resolution of 0.25°0.25°. Observed ground precipitation comes from GPM satellite data with a time resolution of 0.5 hours and a horizontal resolution of 0.1°0.1°. The training data set is data for the entire year of 2021, and the ratio of test to validation data is 7:3. Due to the large sample size, it is impossible to input all the data into the network at one time. The actual training adopts the batch training method, that is, the data set will be randomly divided into multiple batches, each batch containing 256 samples. The number of iterations for the entire training is set to 50.

[0146] Further, step S4 specifically includes: processing the corrected precipitation data based on the super-resolution algorithm to increase the spatial resolution and fineness of the data to obtain refined precipitation data; writing and storing the obtained refined precipitation data in the binary text bz2 standard format, with a single file size of about 1 mb;

[0147] Furthermore, the precipitation of severe weather processes such as typhoons and rainstorms can be compared based on the refined precipitation data and actual observation data to ensure the credibility of this method.

[0148] By using the Convolutional Neural Networks (CNN) deep learning model, the mapping relationship of the image can be learned, thereby achieving high-quality super-resolution reconstruction. CNN is a multi-layer perceptron similar to an artificial neural network, which is often used to analyze visual images. It contains a data input layer, a convolution calculation layer, a ReLU excitation layer, a pooling layer, and a fully connected layer. The WRF simulation results with a resolution of 1 km are used as the training set and coarsened to 25 km. The convolutional neural network is used to extract the mapping relationship from 25 km to 1 km, which is used to correct the ultra-high resolution output of precipitation.

[0149] The medium- and long-term precipitation forecasting method provided in this example combines the batching method and the machine learning method. It has the characteristics of computational flexibility and real-time performance. Combining the batching method with machine learning to realize real-time precipitation forecasting can greatly improve the prediction accuracy, especially for areas with heavy precipitation. It meets the precipitation forecasting needs of strong weather processes such as typhoons and heavy rains, and is of great significance to the development of the meteorological industry.

[0150] In this embodiment, a medium- and long-term precipitation forecast device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0151] This embodiment provides a medium- to long-term precipitation forecasting device. Figure 6 As shown, the device comprises:

[0152] The acquisition module 601 is used to acquire a preset medium- and long-term numerical forecast data set.

[0153] The extraction module 602 is used to extract the preset medium- and long-term numerical forecast data set to obtain a set of precipitation-related meteorological elements.

[0154] The determination module 603 is used to determine the initial precipitation batching element set based on the precipitation related meteorological element set by using the batching method.

[0155] The forecast module 604 is used to perform medium- and long-term precipitation forecasting based on a machine learning method using an initial precipitation ingredient element set to obtain a target precipitation forecast data set.

[0156] In some optional implementations, the acquisition module 601 includes:

[0157] The acquisition submodule is used to obtain the initial medium- and long-term numerical forecast data set.

[0158] The first processing submodule is used to process the initial medium- and long-term numerical prediction data set to obtain a preset medium- and long-term numerical prediction data set.

[0159] In some optional implementations, the determining module 603 includes:

[0160] The calculation submodule is used to calculate the correlation between the elements in the precipitation-related meteorological element set using a preset mathematical model, and the correlation is one or more of correlation, potential combination mode and matching relationship.

[0161] The determination submodule is used to determine the initial precipitation ingredient element set based on the precipitation related meteorological element set and the correlation relationship.

[0162] In some optional implementations, the forecast module 604 includes:

[0163] The second processing submodule is used to obtain a revised initial precipitation forecast data set based on the precipitation ingredient element set after being processed by a precipitation forecast model based on a multi-layer perceptron.

[0164] The third processing submodule is used to process the initial precipitation forecast data set through a super-resolution algorithm to obtain a target precipitation forecast data set.

[0165] In some optional implementations, the second processing submodule includes:

[0166] A training unit is obtained for obtaining a historical precipitation ingredient element set, and the historical precipitation ingredient element set is used for training to obtain a precipitation forecast model based on a multi-layer perceptron.

[0167] The processing unit is used to process the initial precipitation ingredient element set to obtain the target precipitation ingredient element set.

[0168] The input unit is used to input the target precipitation ingredient element set into the precipitation forecast model based on the multi-layer perceptron to obtain a revised initial precipitation forecast data set.

[0169] In some optional embodiments, the device further comprises:

[0170] The acquisition verification module is used to obtain the actual precipitation observation data set, and verify the accuracy of the target precipitation forecast data set based on the actual precipitation observation data set to obtain a verification result.

[0171] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0172] The medium- and long-term precipitation forecast device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0173] The embodiment of the present invention also provides a computer device having the above Figure 6 The medium- and long-term precipitation forecast device shown.

[0174] See also Figure 7 , Figure 7 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 7 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 7 A processor 10 is taken as an example.

[0175] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0176] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0177] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0178] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0179] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0180] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0181] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0182] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A medium- to long-term precipitation forecasting method, characterized in that: The method comprises: Obtain the preset medium- and long-term numerical forecast data set; Extracting the preset medium- and long-term numerical forecast data set to obtain a precipitation-related meteorological element set; Based on the precipitation-related meteorological element set, an initial precipitation ingredient element set is determined using an ingredient method; Based on the machine learning method, the initial precipitation ingredient element set is used to perform medium- and long-term precipitation forecasts to obtain a target precipitation forecast data set.

2. The method according to claim 1, characterized in that Get the preset medium- and long-term numerical forecast data set, including: Obtain the initial medium- and long-term numerical forecast data set; The initial medium- and long-term numerical forecast data set is processed to obtain the preset medium- and long-term numerical forecast data set.

3. The method according to claim 1, characterized in that Based on the precipitation-related meteorological element set, the initial precipitation ingredient element set is determined using the ingredient method, including: Calculate the correlation between the elements in the precipitation-related meteorological element set using a preset mathematical model, where the correlation is one or more of a correlation, a potential combination pattern, and a matching relationship; Based on the precipitation-related meteorological element set and the association relationship, the initial precipitation ingredient element set is determined.

4. The method according to claim 1, characterized in that: Based on the machine learning method, the precipitation ingredient element set is used to perform medium- and long-term precipitation forecasts to obtain a target precipitation forecast data set, including: Based on the precipitation ingredient element set, a revised initial precipitation forecast data set is obtained through processing by a precipitation forecast model based on a multi-layer perceptron; The initial precipitation forecast data set is processed by a super-resolution algorithm to obtain the target precipitation forecast data set.

5. The method according to claim 4, characterized in that Based on the initial precipitation ingredient element set, after being processed by the precipitation forecast model based on the multi-layer perceptron, a revised initial precipitation forecast data set is obtained, including: Acquire a historical precipitation ingredient element set, and use the historical precipitation ingredient element set to train to obtain the precipitation forecast model based on the multi-layer perceptron; Processing the initial precipitation ingredient element set to obtain a target precipitation ingredient element set; The target precipitation ingredient element set is input into the precipitation forecast model based on the multi-layer perceptron to obtain the revised initial precipitation forecast data set.

6. The method according to claim 1, characterized in that The method further comprises: An actual precipitation observation data set is obtained, and the accuracy of the target precipitation forecast data set is verified based on the actual precipitation observation data set to obtain a verification result.

7. A medium- to long-term precipitation forecasting device, characterized in that: The device comprises: An acquisition module is used to acquire a preset medium- and long-term numerical forecast data set; An extraction module, used to extract the preset medium- and long-term numerical forecast data set to obtain a set of precipitation-related meteorological elements; A determination module, used for determining an initial precipitation batching element set by using a batching method based on the precipitation-related meteorological element set; The forecast module is used to perform medium- and long-term precipitation forecasting based on the machine learning method using the initial precipitation ingredient element set to obtain a target precipitation forecast data set.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the medium- and long-term precipitation forecasting method according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the medium- and long-term precipitation forecasting method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises computer instructions, wherein the computer instructions are used to enable a computer to execute the medium- to long-term precipitation forecasting method according to any one of claims 1 to 6.

Images