A weather disaster prediction method and device for crops and a storage medium

Abstract

The application discloses a meteorological disaster prediction method and device for crops and a storage medium, comprising: acquiring first meteorological conditions corresponding to each time period in a plurality of sample periods and disaster data corresponding to each sample period; acquiring second meteorological conditions of each time period in a target period, and determining a first correlation relationship based on each first meteorological condition and second meteorological condition; constructing a graph structure based on the first correlation relationship, the disaster data corresponding to each sample period, and a virtual disaster condition corresponding to the target period; determining meteorological features corresponding to the second meteorological conditions according to the second meteorological conditions and a pre-trained feature extraction network; iteratively updating the graph structure using a pre-trained graph neural network model, and predicting based on the pre-trained meteorological disaster prediction model according to the iteratively updated graph structure and meteorological features to determine disaster data corresponding to the target period.

Description

Technical Field

[0001] This application relates to the field of crop growth technology, and in particular to a method, device and storage medium for predicting meteorological disasters for crops. Background Technology

[0002] For crops with long growing seasons and almost complete exposure to the natural environment, forecasting meteorological disasters is extremely important for their normal growth. Crops with long growing seasons include winter wheat and rice.

[0003] Taking winter wheat as an example. Firstly, the key growth periods for winter wheat include the sowing and emergence period, the overwintering period, the greening-jointing period, the heading-flowering period, and the grain-filling and ripening period. The heading-flowering period can be further divided into the booting stage, the heading-flowering stage, and the pre-grain-filling stage. During the booting stage, the crop is tender and sensitive to the environment. For example, low-temperature damage, late frost damage, or drought can all damage seedlings during this stage. Furthermore, hot and dry winds, drought, continuous rain, or excessively moist waterlogged soil can affect the heading and flowering period of winter wheat, thus affecting the number of grains per ear. Finally, hot and dry winds, strong convective weather, or drought can affect the pre-grain-filling stage of winter wheat, thus affecting the grain weight.

[0004] Therefore, based on the above, it can be seen that different weather conditions have different impacts on winter wheat. In other words, in order to further reduce the impact of meteorological disasters on winter wheat, achieve early warning of meteorological disasters, and avoid the risks of weather disasters, meteorological disaster forecasting for crops is an important issue that urgently needs to be addressed.

[0005] However, current technologies can only monitor meteorological disasters affecting crops, and cannot predict them. In other words, current technologies can only provide preventative measures based on disaster data after a disaster has already occurred; they cannot predict meteorological disasters in advance, provide early warnings, or mitigate the risks of meteorological disasters.

[0006] There is currently no effective solution to the technical problems of existing technologies that can only prevent and control meteorological disasters based on monitored meteorological disaster data, but cannot predict meteorological disasters in advance, achieve early warning of meteorological disasters, and avoid meteorological disaster risks. Summary of the Invention

[0007] The embodiments of this disclosure provide a method, apparatus, and storage medium for predicting meteorological disasters for crops, so as to at least solve the technical problem in the prior art that prevention and control can only be carried out based on the meteorological disaster data obtained from monitoring, and that meteorological disasters cannot be predicted in advance, thereby realizing early warning of meteorological disasters and avoiding meteorological disaster risks.

[0008] According to one aspect of the present disclosure, a method for predicting meteorological disasters for crops is provided, comprising: acquiring first meteorological conditions corresponding to each time period within multiple sample periods, and disaster data corresponding to each sample period, wherein the disaster data includes multiple probability values ​​of disasters affecting crops; acquiring second meteorological conditions for each time period within a target period, and determining a first correlation relationship based on each first meteorological condition and the second meteorological condition, wherein the first correlation relationship includes the correlation relationship between the second meteorological condition and each first meteorological condition, and the correlation relationship between any two first meteorological conditions; constructing a graph structure based on the first correlation relationship, the disaster data corresponding to each sample period, and the virtual disaster situation corresponding to the target period; determining meteorological features corresponding to the second meteorological conditions based on the second meteorological conditions and a pre-trained feature extraction network; iteratively updating the graph structure using a pre-trained graph neural network model, and predicting based on the iteratively updated graph structure and meteorological features using a pre-trained meteorological disaster prediction model to determine the disaster data corresponding to the target period.

[0009] According to another aspect of the present disclosure, a storage medium is also provided, the storage medium including a stored program, wherein, when the program is executed, a processor performs any of the methods described above.

[0010] According to another aspect of the present disclosure, a meteorological disaster prediction device for crops is also provided, comprising: a data acquisition module, configured to acquire first meteorological conditions corresponding to each time period within multiple sample periods, and disaster data corresponding to each sample period, wherein the disaster data includes multiple disaster probability values ​​affecting crops; a first correlation determination module, configured to acquire second meteorological conditions for each time period within a target period, and determine a first correlation based on each first meteorological condition and the second meteorological condition, wherein the first correlation includes the correlation between the second meteorological condition and each first meteorological condition, and the correlation between any two first meteorological conditions; a graph structure construction module, configured to construct a graph structure based on the first correlation, the disaster data corresponding to each sample period, and the virtual disaster conditions corresponding to the target period; a meteorological feature determination module, configured to determine meteorological features corresponding to the second meteorological conditions based on the second meteorological conditions and a pre-trained feature extraction network; and a disaster data prediction module, configured to iteratively update the graph structure using a pre-trained graph neural network model, and predict the disaster data corresponding to the target period based on the iteratively updated graph structure and meteorological features using a pre-trained meteorological disaster prediction model.

[0011] According to another aspect of the present disclosure, a meteorological disaster prediction device for crops is also provided, comprising: a processor; and a memory connected to the processor, configured to provide the processor with instructions for processing the following steps: acquiring first meteorological conditions corresponding to each time period within a plurality of sample periods, and disaster data corresponding to each sample period, wherein the disaster data includes multiple probability values ​​of disasters affecting crops; acquiring second meteorological conditions for each time period within a target period, and determining a first correlation relationship based on each first meteorological condition and the second meteorological condition, wherein the first correlation relationship includes the correlation relationship between the second meteorological condition and each first meteorological condition, and the correlation relationship between any two first meteorological conditions; constructing a graph structure based on the first correlation relationship, the disaster data corresponding to each sample period, and the virtual disaster situation corresponding to the target period; determining meteorological features corresponding to the second meteorological condition based on the second meteorological condition and a pre-trained feature extraction network; iteratively updating the graph structure using a pre-trained graph neural network model, and predicting based on the iteratively updated graph structure and meteorological features using a pre-trained meteorological disaster prediction model to determine the disaster data corresponding to the target period.

[0012] This application provides a method for predicting meteorological disasters for crops. First, a processor acquires first meteorological conditions corresponding to each time period within multiple sample periods, as well as disaster data corresponding to each sample period. Then, the processor acquires second meteorological conditions for each time period within a target period and determines a first correlation based on each of the first and second meteorological conditions. Further, the processor constructs a graph structure based on the first correlation, the disaster data corresponding to each sample period, and the virtual disaster situation corresponding to the target period. Then, the processor determines the meteorological features corresponding to the second meteorological conditions based on the second meteorological conditions and a pre-trained feature extraction network. Finally, the processor iteratively updates the graph structure using a pre-trained graph neural network model and predicts the disaster data corresponding to the target period based on the iteratively updated graph structure and meteorological features using a pre-trained meteorological disaster prediction model.

[0013] As described above, this application acquires the first weather conditions corresponding to each time period within multiple sample periods and the second weather conditions for each time period within a target period, and determines the first weather conditions of multiple sample periods and the second weather conditions of the target period, as well as the first weather conditions of any two sample periods. Therefore, in cases where the first weather conditions at each time period within each sample period cause crop damage, a correlation can be constructed based on the first correlation to represent the relationship between the virtual disaster situation of the target period and the disaster data of each sample period.

[0014] Furthermore, since this application also determines the covariance matrix corresponding to the second meteorological condition, and this covariance matrix can represent the correlation between meteorological data of each disaster dimension in the target period, it can deeply explore the spatial correlation between weather data of each disaster dimension. Therefore, by inputting the iteratively updated graph structure and the transformation matrix and eigenvector corresponding to the covariance matrix into the meteorological disaster prediction model, accurate prediction results can be obtained, enabling early warning of weather disasters and mitigation of weather disaster risks.

[0015] This solves the technical problem in existing technologies that can only prevent and control meteorological disasters based on monitored meteorological disaster data, but cannot predict meteorological disasters in advance, thus achieving early warning of meteorological disasters and avoiding meteorological disaster risks. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this disclosure and form part of this application, illustrate exemplary embodiments of this disclosure and are used to explain this disclosure, but do not constitute an undue limitation of this disclosure. In the drawings: Figure 1 This is a hardware structure block diagram of a computing device used to implement the method described in Embodiment 1 of this application; Figure 2 This is a schematic diagram of a meteorological disaster prediction system for crops according to Embodiment 1 of this application; Figure 3 This is a modular schematic diagram of a meteorological disaster prediction system for crops according to Embodiment 1 of this application; Figure 4 This is a flowchart of a meteorological disaster prediction method for crops according to Embodiment 1 of this application; Figure 5 This is a schematic diagram of the diagram structure according to Embodiment 1 of this application; Figure 6 This is a schematic diagram of the feature extraction network according to Embodiment 1 of this application; Figure 7 This is a schematic diagram of the graph neural network model according to Embodiment 1 of this application; Figure 8 This is a schematic diagram of the second MLP model and the meteorological disaster prediction model according to Embodiment 1 of this application; Figure 9 This is a schematic diagram of a meteorological disaster prediction device for crops according to Embodiment 1 of this application; Figure 10 This is a schematic diagram of a meteorological disaster prediction device for crops according to Embodiment 1 of this application. Detailed Implementation

[0017] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this disclosure.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] Example 1 According to this embodiment, a method embodiment for predicting meteorological disasters for crops is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Also, although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than that shown here.

[0020] The method embodiments provided in this example can be executed on mobile terminals, computer terminals, servers, or similar computing devices. Figure 1 A hardware block diagram of a computing device for implementing a method for predicting meteorological disasters affecting crops is shown. Figure 1 As shown, a computing device may include one or more processors (processors may include, but are not limited to, microprocessors such as MCUs or programmable logic devices such as FPGAs), a memory for storing data, a transmission device for communication functions, and an input / output interface. The memory, transmission device, and input / output interface are connected to the processor via a bus. In addition, it may also include a display, keyboard, and cursor control device connected to the input / output interface. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, a computing device may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0021] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element in a computing device. As involved in the embodiments of this disclosure, the data processing circuits serve as processor control (e.g., selection of a variable resistor termination path connected to an interface).

[0022] The memory can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the meteorological disaster prediction method for crops in this embodiment of the present disclosure. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the meteorological disaster prediction method for crops described above. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the computing device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0023] The transmission device is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the computing device's communication provider. In one example, the transmission device includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device may be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0024] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows users to interact with the user interface of the computing device.

[0025] It should be noted here that, in some optional embodiments, the above... Figure 1 The computing device shown may include hardware elements (including circuitry), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware and software elements. It should be noted that... Figure 1This is only one instance of a specific particular instance, and is intended to illustrate the types of components that may exist in the aforementioned computing devices.

[0026] Figure 2 This is a schematic diagram of a meteorological disaster prediction system for crops according to this embodiment. (Refer to...) Figure 2 As shown, the system includes: a terminal device 100, a processor 200 communicatively connected to the terminal device 100, and a memory 300 communicatively connected to the processor 200. Operators can send meteorological disaster prediction requests to the processor 200 through the terminal device 100. In response to the received evaluation request, the processor 200 retrieves from the memory 300 the first weather conditions corresponding to each time period within multiple sample periods, disaster data for each sample period, and the second weather conditions for each time period within the target period, and predicts the disaster data for the target period.

[0027] In addition, the processor 200 is also used to return the predicted disaster data for the target period to the terminal device 100.

[0028] It should be noted that the terminal device 100, processor 200, and memory 300 in the system can all be adapted to the hardware structure described above.

[0029] Figure 3 This is a modular schematic diagram of a meteorological disaster prediction system for crops according to an embodiment of this application. (Reference) Figure 3 As shown, the module includes a data acquisition module, a correlation calculation module, a graph structure construction module, an iterative update module, a feature decomposition module, a fusion and prediction module, and a storage module.

[0030] The data acquisition module is connected to the storage module and is used to obtain the first weather conditions corresponding to each time period within multiple sample periods, disaster data for each sample period, and the second weather conditions for each time period within the target period from the storage module.

[0031] The correlation calculation module is connected to the data acquisition module to calculate the first correlation between the second weather condition and each first weather condition, as well as the first correlation between any two first weather conditions.

[0032] The graph structure construction module communicates with the correlation calculation module to construct a graph structure based on the first correlation relationship, disaster data corresponding to each sample period, and virtual disaster situation corresponding to the target period.

[0033] The iterative update module communicates with the graph structure construction module and is used to iteratively update the graph structure using a pre-trained graph neural network model.

[0034] The feature decomposition module communicates with the data acquisition module and is used to determine the corresponding covariance matrix, transformation matrix and eigenvector based on the second meteorological conditions of each time period within the target period.

[0035] The fusion and prediction module is connected to the iterative update module and the feature decomposition module respectively. It is used to fuse the target feature vector and meteorological features corresponding to the iteratively updated graph structure, and to predict disaster data for the target period based on the fusion results.

[0036] Under the aforementioned operating environment, according to the first aspect of this embodiment, a method for predicting meteorological disasters affecting crops is provided. This method comprises... Figure 3 The system implementation shown. Figure 4 A flowchart illustrating the method is shown below. (Refer to...) Figure 4 As shown, the method includes: S402: Obtain the first meteorological conditions corresponding to each time period within multiple sample periods, as well as the disaster data corresponding to each sample period, wherein the disaster data includes the probability values ​​of various disasters affecting crops; S404: Obtain the second meteorological conditions for each time period within the target period, and determine the first correlation based on each first meteorological condition and the second meteorological conditions, wherein the first correlation includes the correlation between the second meteorological conditions and each first meteorological condition, as well as the correlation between any two first meteorological conditions; S406: Construct a graph structure based on the first correlation, disaster data corresponding to each sample period, and virtual disaster situation corresponding to the target period; S408: Based on the second meteorological condition and the pre-trained feature extraction network, determine the meteorological features corresponding to the second meteorological condition; S410: The graph structure is iteratively updated using a pre-trained graph neural network model, and the disaster prediction model is used to make predictions based on the iteratively updated graph structure and meteorological characteristics to determine the disaster data corresponding to the target period.

[0037] Specifically, firstly, the processor 200 responds to the meteorological disaster prediction request sent by the operator through the terminal device 100, and uses the data acquisition module to obtain the first weather conditions corresponding to each time period within multiple sample periods, as well as the disaster data corresponding to each sample period, from the storage module (S402).

[0038] Furthermore, the sample period is, for example, in years. Each period within the sample period could be, for example, the weather conditions for 90 days from January to March of a year. For example, compared to the first sample period... The corresponding first weather condition .in, This represents the weather conditions on day i of the first sample period, where i = 1 to 90. And within this, . This represents the average temperature corresponding to day i in the first sample period. This represents the maximum temperature corresponding to day i in the first sample period. This represents the rainfall amount corresponding to day i in the first sample period. This represents the maximum wind force corresponding to day i in the first sample period. This represents the solar luminosity corresponding to day i in the first sample period. This represents the illumination duration corresponding to the i-th day of the first sample period. That is, j = 1~6.

[0039] With the second sample period The corresponding first weather condition .in, This represents the weather conditions on day i in the second sample period, where i = 1 to 90. And within this, . This represents the average temperature corresponding to day i in the second sample period. This represents the maximum temperature corresponding to day i in the second sample period. This represents the rainfall amount corresponding to day i in the second sample period. This represents the maximum wind force corresponding to day i in the second sample period. This represents the daylight intensity corresponding to day i in the second sample period. This represents the illumination duration corresponding to the i-th day of the second sample period. That is, j = 1~6.

[0040] And so on.

[0041] With the nth sample period The corresponding first weather condition .in, This represents the weather conditions on day i in the nth sample period, where i = 1 to 90. And within this, . This represents the average temperature corresponding to day i in the nth sample period. This represents the maximum temperature corresponding to day i in the nth sample period. This represents the rainfall amount corresponding to the i-th day of the n-th sample period. This represents the maximum wind force corresponding to day i in the nth sample period. This represents the solar luminosity corresponding to the i-th day of the n-th sample period. This represents the illumination duration corresponding to the i-th day of the n-th sample period. That is, j = 1~6.

[0042] The data acquisition module then collects disaster data corresponding to each sample period from the storage module. This disaster data includes probability values ​​for various disasters affecting crops. For example, the sample period... Disaster situation inside .in, This represents the probability of low-temperature injury occurring within the first sample period; This represents the probability of drought occurring within the first sample period; This indicates the probability of hot, dry winds existing within the first sample period. This represents the probability of continuous rain during the first sample period. This indicates the probability of severe convective weather occurring within the first sample period; This indicates the probability of the presence of a dry, hot wind pattern following rain within the first sample period.

[0043] For example, the sample period Disaster situation inside .in, This indicates the probability of low-temperature injury occurring during the second sample period; This indicates the probability of drought occurring during the second sample period; This indicates the probability of hot, dry winds existing during the second sample period. This indicates the probability of consecutive rainy days occurring during the second sample period. This indicates the probability of severe convective weather occurring within the second sample period; This indicates the probability of a dry, hot wind pattern following rain during the second sample period.

[0044] And so on.

[0045] For example, the sample period Disaster situation inside .in, This represents the probability of low-temperature injury occurring within the nth sample period; This represents the probability of drought occurring within the nth sample period; This represents the probability of hot, dry winds existing within the nth sample period. This represents the probability of consecutive rainy days occurring within the nth sample period; This represents the probability of severe convective weather occurring within the nth sample period; This represents the probability of the presence of a dry, hot wind pattern following rain within the nth sample period.

[0046] Among them, each sample period ~ disaster situation ~ Each element in the array has a value of either 0 or 1. A value of 0 indicates that no meteorological disaster occurred during the corresponding sample period; a value of 1 indicates that a meteorological disaster occurred during the corresponding sample period.

[0047] In addition, within the target period Virtual disaster situation within .in, This indicates the probability of low-temperature damage occurring within the target period; This indicates the probability of drought occurring within the target period; This indicates the probability of hot, dry winds occurring within the target period. This indicates the probability of continuous rain within the target period; This indicates the probability of severe convective weather occurring within the target period; This indicates the probability of a dry, hot wind pattern following rain within the target period.

[0048] Among them, the target period Virtual disaster situation within The values ​​of all elements in the array are equal to 0.

[0049] Based on the above method, the data acquisition module can obtain the first weather conditions corresponding to each time period within multiple sample periods, as well as the disaster data corresponding to each sample period, from the storage module.

[0050] Then, the data acquisition module retrieves the second meteorological conditions for each time period within the target period from the storage module. The target period... The various time periods within this period could be, for example, the weather conditions during the 90 days from January to March of the year. For example, compared to the target period... The corresponding weather conditions on the second day .in, This represents the weather conditions on day i of the target period, where i = 1 to 90. And within this, . This represents the average temperature on day i of the target period. This represents the maximum temperature corresponding to day i of the target period. This represents the rainfall amount corresponding to day i of the target period. This represents the maximum wind force corresponding to day i of the target period. This represents the daylight intensity corresponding to day i of the target period. This represents the duration of illumination corresponding to day i of the target period. That is, j = 1~6.

[0051] Subsequently, the data acquisition module sends the acquired data to the correlation calculation module, which then determines a first correlation based on each of the first and second weather conditions (S404). This first correlation includes the correlation between the second weather condition and each of the first weather conditions, as well as the correlation between any two first weather conditions. Furthermore, the correlation calculation module can, for example, determine each sample period based on Euclidean distance. ~ First weather conditions ~ With target period Second weather conditions The first relationship between them ~ Similarly, the period of each sample can be calculated. ~ First weather conditions ~ With target period Second weather conditions The first relationship between them ~ .

[0052] Similarly, it is also possible to calculate any two sample periods. ~ First weather conditions ~ The primary correlation between them. For example, the sample period. First weather conditions With sample period First weather conditions The first relationship between them Sample period First weather conditions With sample period First weather conditions The first relationship between them ..., sample period First weather conditions With sample period First weather conditions The first relationship between them .

[0053] For example, the sample period First weather conditions With sample period First weather conditions The first relationship between them Sample period First weather conditions With sample period First weather conditions The first relationship between them ..., sample period First weather conditions With sample period First weather conditions The first relationship between them .etc.

[0054] Furthermore, the correlation calculation module can also be based on each primary correlation relationship. ~ Multiple second associations are determined by comparing them with a pre-set association threshold. ~ Similarly, the correlation calculation module can be based on each primary correlation. ~ Multiple second associations are determined by comparing them with a pre-set association threshold. ~ Similarly, the correlation calculation module can determine multiple second correlations based on the first correlation between any two first weather conditions and a pre-set correlation threshold, which will not be elaborated here.

[0055] Specifically, if the first association is greater than the association threshold, the first association is determined as the second association; if the first association is less than or equal to the association threshold, the first association is removed. The above will be described in detail later, and therefore will not be repeated here.

[0056] Furthermore, the graph structure construction module constructs a graph structure based on the first correlation, the disaster data corresponding to each sample period, and the virtual disaster situation corresponding to the target period (S406). Specifically, when the correlation calculation module determines multiple second correlations, the graph structure construction module constructs a graph structure with each second correlation as an edge and the disaster data corresponding to each sample period and the virtual disaster situation corresponding to the target period as nodes.

[0057] Figure 5 This is a schematic diagram of the diagram structure according to an embodiment of this application. (Reference) Figure 5 As shown, the first node With the first node The first side corresponds to First node With the first node The first side corresponds to First node With the first node The first side corresponds to First node With the first node The first side corresponds to First node With the first node The first side corresponds to .etc.

[0058] Then, the feature decomposition module determines the corresponding covariance matrix based on the second weather condition within the target period. This covariance matrix represents the correlation between meteorological data across different disaster dimensions within the target period. The feature decomposition module then performs eigenvalue decomposition on the covariance matrix to obtain the corresponding transformation matrix. Further, the feature decomposition module calculates the corresponding eigenvectors based on the transformation matrix and the covariance matrix. Finally, the feature decomposition module inputs the transformation matrix, eigenvectors, and the second weather condition into a pre-trained feature extraction network to determine the meteorological features corresponding to the second weather condition (S408). The above will be described in detail later, and therefore will not be repeated here.

[0059] Subsequently, the iterative update module uses a pre-trained graph neural network model to iteratively update the graph structure. Specifically, when the graph structure construction module constructs the graph structure, the iterative update module iteratively updates the graph structure. The iterative update module includes a graph neural network model, which is used to iteratively update the graph structure.

[0060] Finally, the fusion and prediction module first uses the second MLP model to fuse the target feature vector and weather features corresponding to the iteratively updated graph structure, generating a fusion result. Then, based on the fusion result and a pre-trained meteorological disaster prediction model (e.g., a neural network model), the fusion and prediction module determines the disaster data corresponding to the target period (S410). For example, the disaster data for the target period... middle, If the value is 1, it indicates that there is low-temperature damage within the target period. If the value is 0, it means that no drought will occur within the target period. If the value is 1, it indicates that there is hot dry air within the target period. If the value is 1, it indicates that there will be continuous rain within the target period. If the value is 1, it indicates that there is severe convective weather within the target period. If the value is 0, it indicates that there is a dry, hot wind pattern similar to that of a ripening crop during the target period.

[0061] This allows operators to take targeted preventative measures and mitigate weather disaster risks based on the predicted probability values ​​of various meteorological disasters within the target period.

[0062] As described in the background section, it can be seen from the above that different meteorological conditions have varying impacts on winter wheat. Therefore, in order to further reduce the impact of meteorological disasters on winter wheat, achieve early warning of meteorological disasters, and mitigate weather disaster risks, meteorological disaster forecasting for crops is a crucial issue that urgently needs to be addressed.

[0063] However, current technologies can only monitor meteorological disasters affecting crops, and cannot predict them. In other words, current technologies can only provide preventative measures based on disaster data after a disaster has already occurred; they cannot predict meteorological disasters in advance, provide early warnings, or mitigate the risks of meteorological disasters.

[0064] In view of this, this application provides a method for predicting meteorological disasters for crops. As described above, this application acquires first meteorological conditions corresponding to each time period within multiple sample periods and second weather conditions for each time period within a target period, and determines a first correlation between the first weather conditions of multiple sample periods and the second weather conditions of the target period, as well as between the first weather conditions of any two sample periods. Therefore, when the first meteorological conditions of each time period within each sample period cause crop damage, a correlation can be constructed based on the first correlation to represent the relationship between the virtual disaster situation of the target period and the disaster data of each sample period.

[0065] Furthermore, since this application also determines the covariance matrix corresponding to the second meteorological condition, and this covariance matrix can represent the correlation between meteorological data of each disaster dimension in the target period, it can deeply explore the spatial correlation between weather data of each disaster dimension. Therefore, by inputting the iteratively updated graph structure and the transformation matrix and eigenvector corresponding to the covariance matrix into the meteorological disaster prediction model, accurate prediction results can be obtained, enabling early warning of weather disasters and mitigation of weather disaster risks.

[0066] This solves the technical problem in existing technologies that can only prevent and control meteorological disasters based on monitored meteorological disaster data, but cannot predict meteorological disasters in advance, thus achieving early warning of meteorological disasters and avoiding meteorological disaster risks.

[0067] Optionally, the operation of constructing a graph structure based on the first association, disaster data corresponding to each sample period, and virtual disaster conditions corresponding to the target period as nodes includes: determining multiple second associations based on each first association and a pre-set association threshold; and constructing a graph structure with each second association as an edge and the disaster data corresponding to each sample period and the virtual disaster conditions corresponding to the target period as nodes. Further optionally, the operation of determining multiple second associations based on each first association and a pre-set association threshold includes: determining the relationship between each first association and the association threshold; determining a first association as a second association if the first association is greater than the association threshold; and removing a first association if the first association is less than or equal to the association threshold.

[0068] Specifically, as can be seen from the above description, the correlation calculation module can, for example, use the Euclidean distance calculation method to calculate the correlation of each sample period. ~ First weather conditions ~ With target period Second weather conditions The first relationship between them ~ .

[0069] For example, the first association Indicates the weather conditions on the second day With sample period First weather conditions The relationships between them, where k = 1 to n. The specific calculation formula is as follows: in, Indicates the weather conditions on the second day With each first weather condition The first relationship between them This represents the second weather condition on day i within the target period. This represents the first weather condition on day i within the k-th sample period.

[0070] Similarly, the correlation calculation module can calculate the first correlation between the first weather conditions of any two sample periods in the same way as described above, which will not be elaborated here.

[0071] The correlation calculation module then maps the determined first correlations to the range of 0 to 1.

[0072] Furthermore, the correlation calculation module determines the relationship between each first correlation and a pre-set correlation threshold. For example, if a first correlation is greater than the correlation threshold, it is identified as a second correlation. Conversely, if a first correlation is less than or equal to the correlation threshold, it is discarded. The correlation threshold can be set by the operator based on historical experience, and is not specifically limited here.

[0073] Therefore, by performing the above-mentioned correlation calculation, the correlation between the current period and each sample period can be determined, and the necessary foundation can be provided for the subsequent construction of a graph structure to represent the meteorological disaster data within the target period and the meteorological disaster data between each sample period.

[0074] Optionally, the operation of determining the meteorological features corresponding to the second meteorological condition based on the second meteorological condition and the pre-trained feature extraction network includes: determining the corresponding covariance matrix based on the second meteorological condition; calculating the corresponding transformation matrix and feature vector based on the covariance matrix; and inputting the transformation matrix, feature vector, and second meteorological condition into the pre-trained feature extraction network to determine the meteorological features corresponding to the second meteorological condition.

[0075] Specifically, the feature decomposition module receives the second meteorological data corresponding to the target period sent by the data acquisition module. And determine the relationship with the second meteorological data. The corresponding covariance matrix P0. Wherein, it corresponds to the second meteorological data. The corresponding matrix form is as follows: Furthermore, the eigenvalue decomposition module determines the second weather condition in matrix form. Determine the corresponding covariance matrix P0∈R 6×6 And the transformation matrix. The specific formula is as follows: in, Indicates the weather conditions of the second day The corresponding covariance matrix, Represents the transformation matrix. This represents the second weather condition on day i within the target period.

[0076] And among them, Represents an orthogonal matrix, a matrix It is a diagonal matrix. Through a diagonal matrix... The feature vector can be obtained. That is, the matrix. The elements on the diagonal form the feature vector. . Represents the trace of a matrix.

[0077] For covariance matrix Feature value extraction is performed, and the specific calculation formula is as follows: Where, diag() represents the determination matrix. Combine the elements on the diagonal into a vector.

[0078] Furthermore, the feature decomposition module inputs the transformation matrix, feature vector, and second meteorological condition into a pre-trained feature extraction network to determine the meteorological features corresponding to the second meteorological condition. Figure 6 This is a schematic diagram of the feature extraction network according to an embodiment of this application. (Reference) Figure 6 As shown, the feature extraction network includes a first feature extraction unit, a second feature extraction unit, and a third MLP model.

[0079] Therefore, the feature decomposition module will determine the second meteorological situation. The data is input to the first feature extraction unit, which uses convolutional layers and fully connected layers to extract features from the second day's weather conditions, thereby generating a first meteorological feature vector corresponding to the second day's weather conditions. Meanwhile, the eigenvalue decomposition module will transform the matrix. The input is fed into the second feature extraction unit, which uses a feedforward neural network model to process the transformation matrix. Feature extraction is performed to generate and transform the matrix. The corresponding second meteorological feature vector Finally, the feature decomposition module determines the first meteorological feature vector. Second meteorological feature vector In this case, the third MLP model is used to analyze the first meteorological feature vector. and, second meteorological feature vector and eigenvectors The data is then fused to generate meteorological feature A corresponding to the second meteorological condition.

[0080] By determining the covariance matrix corresponding to the second meteorological condition, the correlation between meteorological data of each disaster dimension in the target period can be established, thus enabling in-depth exploration of the spatial correlation between weather data of each disaster dimension. This provides a necessary foundation for subsequent prediction of various disaster data within the target period.

[0081] Optionally, the operation of iteratively updating the graph structure using a pre-trained graph neural network model includes: inputting the graph structure into the graph neural network model, and constructing a node attribute vector using the calculation rules corresponding to the nodes, wherein the node attribute vector includes the node attributes of the node, and the product of the edge attributes of two adjacent edges and the node attributes of the corresponding node; and iteratively updating the node attribute vector using multiple first MLP models in the graph neural network model, and generating the iteratively updated graph structure.

[0082] Specifically, after the graph structure building module has constructed the graph structure, the iterative update module inputs the received graph structure into the graph neural network model. Figure 7 This is a schematic diagram of the graph neural network model according to an embodiment of this application. (Reference) Figure 7 As shown, the graph neural network model includes multiple first-order MLP models for iteratively updating the node attributes of nodes in a graph structure.

[0083] Therefore, the iterative update module first constructs node attribute vectors using the computation rules corresponding to the nodes. For example, the iterative update module constructs node attribute vectors corresponding to the nodes in the graph structure according to the computation rules. .in, Representing nodes in a graph structure Node attributes and edges The product of edge attributes, Representing nodes in a graph structure Node attributes and edges The product of the edge properties.

[0084] For example, the iterative update module constructs the node attribute vectors corresponding to the nodes in the graph structure according to the calculation rules. .in, Representing nodes in a graph structure Node attributes and edges The product of edge attributes, Representing nodes in a graph structure Node attributes and edges The product of the edge properties.

[0085] For example, the iterative update module constructs the node attribute vectors corresponding to the nodes in the graph structure according to the calculation rules. .in, Representing nodes in a graph structure Node attributes and edges The product of edge attributes, Representing nodes in a graph structure Node attributes and edges The product of the edge properties.

[0086] And so on. The method for constructing the node attribute vectors corresponding to other nodes is the same as described above, and will not be repeated here.

[0087] Then, multiple third-party MLP models iteratively update the node attribute vectors corresponding to each constructed node, thereby outputting the node features of each node after iterative updates.

[0088] Thus, the iterative update module can determine the updated graph structure based on the node characteristics of each node and the edge attributes of each edge after the iterative update.

[0089] Optionally, the operation of predicting disaster data corresponding to a target period based on a pre-trained meteorological disaster prediction model according to the iteratively updated graph structure and meteorological features includes: determining a target feature vector, wherein the target feature vector includes the node features of the nodes corresponding to the virtual disaster situation after iterative updates, and the product of the edge features of two adjacent edges and the node features of the corresponding nodes; and predicting disaster data corresponding to the target period based on the target feature vector and meteorological features. Further optionally, the operation of predicting disaster data corresponding to the target period based on the target feature vector and the meteorological features includes: fusing the target feature vector and weather features using a second MLP model and determining a fused feature vector; and inputting the fused feature vector into the meteorological disaster prediction model to determine the disaster data corresponding to the target period.

[0090] Specifically, the iterative update module sends the iteratively updated graph structure to the fusion and prediction module, allowing the fusion and prediction module to determine the target features corresponding to the iteratively updated graph structure. The target feature vector includes the node features of the nodes corresponding to the virtual disaster situation after the iterative update, as well as the product of the edge features of two adjacent edges and the node features of their corresponding nodes. For example, the target feature vector... .in, Represents the node features of the node after iterative updates. With edge The product of edge attributes, Represents the node features of the node after iterative updates. With edge The product of the edge properties.

[0091] For example, the target feature vector .in, Represents the node features of the node after iterative updates. With edge The product of edge attributes, Represents the node features of the node after iterative updates. With edge The product of the edge properties.

[0092] Furthermore, the fusion and prediction module uses the second MLP model to fuse the determined multiple target feature vectors with the aforementioned determined weather features, thereby determining the fused feature vector.

[0093] Finally, the fusion and prediction module inputs the fused feature vector into the meteorological disaster prediction model. Figure 8 This is a schematic diagram of the second MLP model and the meteorological disaster prediction model according to the embodiments of this application. (Reference) Figure 8 As shown, after the fusion and prediction module determines the fusion feature vector, the fusion feature vector is input into the meteorological disaster prediction model, which then makes predictions based on the fusion feature vector. This outputs meteorological disaster data for the target period. The meteorological disaster prediction model is a neural network model, which includes an input layer, a fully connected layer, an activation function layer, and an output layer.

[0094] Thus, once operators have determined the meteorological disaster data for the target period, targeted early warnings can be issued based on the meteorological disaster data, making it easier to avoid weather disaster risks.

[0095] Thus, according to the first aspect of this embodiment, the technical effect of being able to predict meteorological disaster data, realize early warning of weather disasters, and facilitate the avoidance of weather disasters is achieved.

[0096] In addition, refer to Figure 1 As shown, according to a second aspect of this embodiment, a storage medium is provided. The storage medium includes a stored program, wherein, when the program is executed, a processor performs any of the methods described above.

[0097] Thus, according to this embodiment, the technical effect of being able to predict meteorological disaster data, realize early warning of weather disasters, and facilitate the avoidance of weather disasters is achieved.

[0098] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0099] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0100] Example 2 Figure 9 A meteorological disaster prediction device 900 for crops according to this embodiment is shown, which corresponds to the method described according to Embodiment 1. (Reference) Figure 9 As shown, the device 900 includes: a data acquisition module 910, used to acquire first meteorological conditions corresponding to each time period within multiple sample periods, and disaster data corresponding to each sample period, wherein the disaster data includes multiple disaster probability values ​​affecting crops; a first correlation determination module 920, used to acquire second meteorological conditions for each time period within a target period, and determine a first correlation based on each first meteorological condition and the second meteorological condition, wherein the first correlation includes the correlation between the second meteorological condition and each first meteorological condition, and the correlation between any two first meteorological conditions; a graph structure construction module 930, used to construct a graph structure based on the first correlation, the disaster data corresponding to each sample period, and the virtual disaster conditions corresponding to the target period; a meteorological feature determination module 940, used to determine the meteorological features corresponding to the second meteorological condition based on the second meteorological condition and a pre-trained feature extraction network; and a disaster data prediction module 950, used to iteratively update the graph structure using a pre-trained graph neural network model, and predict the disaster data corresponding to the target period based on the iteratively updated graph structure and meteorological features using a pre-trained meteorological disaster prediction model.

[0101] Optionally, the graph structure construction module 930 includes: a second association determination module, used to determine multiple second associations based on each first association and a pre-set association threshold; and a graph structure construction submodule, used to construct a graph structure with each second association as an edge and with disaster data corresponding to each sample period and virtual disaster conditions corresponding to the target period as nodes.

[0102] Optionally, the second association determination module includes: a judgment module for judging the relationship between each first association and the association threshold; a third association determination module for determining the first association as the second association when the first association is greater than the association threshold; and a fourth association determination module for removing the first association when the first association is less than or equal to the association threshold.

[0103] Optionally, the meteorological feature determination module 940 includes: a covariance matrix determination module, used to determine the corresponding covariance matrix based on the second meteorological condition; a feature decomposition module, used to calculate the corresponding transformation matrix and feature vector based on the covariance matrix; and a meteorological feature determination submodule, used to input the transformation matrix, feature vector, and second meteorological condition into a pre-trained feature extraction network to determine the meteorological features corresponding to the second meteorological condition.

[0104] Optionally, the disaster data prediction module 950 includes: a node attribute vector construction module, used to input the graph structure into the graph neural network model and construct node attribute vectors using calculation rules corresponding to the nodes, wherein the node attribute vectors include the node attributes of the nodes and the product of the edge attributes of two adjacent edges and the node attributes of the corresponding nodes; and an iterative update module, used to iteratively update the node attribute vectors using multiple first MLP models in the graph neural network model and generate the iteratively updated graph structure.

[0105] Optionally, the disaster data prediction module 950 includes: a target feature vector determination module, used to determine the target feature vector, wherein the target feature vector includes the node features of the node corresponding to the virtual disaster situation after iterative updates, and the product of the edge features of two adjacent edges and the node features of the corresponding node; and a disaster data prediction submodule, used to predict disaster data corresponding to the target period based on the target feature vector and meteorological features.

[0106] Optionally, the disaster data prediction module 950 includes: a fusion feature vector determination module, used to fuse the target feature vector and weather features using a second MLP model, and determine the fusion feature vector; and a disaster data determination module, used to input the fusion feature vector into the meteorological disaster prediction model, and determine the disaster data corresponding to the target period.

[0107] Thus, according to this embodiment, the technical effect of being able to predict meteorological disaster data, realize early warning of weather disasters, and facilitate the avoidance of weather disasters is achieved.

[0108] Example 3 Figure 10A meteorological disaster prediction device 1000 for crops according to this embodiment is shown, which corresponds to the method described in Embodiment 1. Reference Figure 10 As shown, the device 1000 includes: a processor 1010; and a memory 1020 connected to the processor 1010, used to provide the processor 1010 with instructions to process the following steps: acquiring first meteorological conditions corresponding to each time period within multiple sample periods, and disaster data corresponding to each sample period, wherein the disaster data includes multiple disaster probability values ​​affecting crops; acquiring second meteorological conditions for each time period within a target period, and determining a first correlation based on each first meteorological condition and the second meteorological condition, wherein the first correlation includes the correlation between the second meteorological condition and each first meteorological condition, and the correlation between any two first meteorological conditions; constructing a graph structure based on the first correlation, the disaster data corresponding to each sample period, and the virtual disaster conditions corresponding to the target period; determining meteorological features corresponding to the second meteorological condition based on the second meteorological condition and a pre-trained feature extraction network; iteratively updating the graph structure using a pre-trained graph neural network model, and predicting based on the iteratively updated graph structure and meteorological features using a pre-trained meteorological disaster prediction model to determine the disaster data corresponding to the target period.

[0109] Thus, according to this embodiment, the technical effect of being able to predict meteorological disaster data, realize early warning of weather disasters, and facilitate the avoidance of weather disasters is achieved.

[0110] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0111] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0112] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0113] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0114] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0115] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0116] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for predicting meteorological disasters affecting crops, characterized in that, include: Acquire the first meteorological conditions corresponding to each time period within multiple sample periods, and the disaster data corresponding to each sample period, wherein the disaster data includes multiple disaster probability values ​​affecting crops, and wherein the first meteorological conditions include average temperature, maximum temperature, rainfall, maximum wind force, sunlight intensity and / or sunshine duration; The second meteorological conditions for each time period within the target period are obtained, and a first correlation is determined based on each first meteorological condition and the second meteorological conditions, wherein the first correlation includes the correlation between the second meteorological conditions and each first meteorological condition, as well as the correlation between any two first meteorological conditions; Based on the first correlation, the disaster data corresponding to each sample period, and the virtual disaster situation corresponding to the target period, a graph structure is constructed, including: Based on each first association relationship and a pre-set association relationship threshold, multiple second association relationships are determined; and The graph structure is constructed using each second association as an edge and the disaster data corresponding to each sample period and the virtual disaster situation corresponding to the target period as nodes. Based on the second meteorological condition and the pre-trained feature extraction network, the meteorological features corresponding to the second meteorological condition are determined, including: Based on the second meteorological condition, determine the corresponding covariance matrix; Based on the covariance matrix, the corresponding transformation matrix and eigenvectors are calculated; and The transformation matrix, the feature vector, and the second meteorological condition are input into a pre-trained feature extraction network to determine the meteorological features corresponding to the second meteorological condition. The graph structure is iteratively updated using a pre-trained graph neural network model, and a meteorological disaster prediction model is used to predict disaster data corresponding to the target period based on the iteratively updated graph structure and the meteorological characteristics, including: The graph structure is input into the graph neural network model, and a node attribute vector is constructed using the calculation rules corresponding to the nodes. This node attribute vector includes the node attributes of the node, and the product of the edge attributes of two adjacent edges and the corresponding node attributes; and The node attribute vectors are iteratively updated using multiple first MLP models in the graph neural network model, and the iteratively updated graph structure is generated.

2. The method according to claim 1, characterized in that, The operation of determining multiple second relationships based on each first relationship and a pre-set relationship threshold includes: Determine the relationship between each of the first association relationships and the association relationship threshold; If the first association relationship is greater than the association relationship threshold, the first association relationship is determined as the second association relationship; and If the first association is less than or equal to the association threshold, the first association is removed.

3. The method according to claim 1, characterized in that, The operation of determining disaster data corresponding to the target period based on the pre-trained meteorological disaster prediction model, according to the iteratively updated graph structure and meteorological characteristics, includes: Multiple target feature vectors are determined, wherein the target feature vectors include the node features of the node corresponding to the virtual disaster situation after iterative updates, and the product of the edge features of two adjacent edges and the node features of the corresponding node; and Based on the multiple target feature vectors and the meteorological features, disaster data corresponding to the target period is predicted.

4. The method according to claim 3, characterized in that, The operation of predicting disaster data corresponding to the target period based on the multiple target feature vectors and the meteorological features includes: The second MLP model is used to fuse the multiple target feature vectors and weather features, and the fused feature vector is determined; and The fused feature vector is input into the meteorological disaster prediction model, and disaster data corresponding to the target period is determined.

5. A storage medium, characterized in that, The storage medium includes a stored program, wherein, when the program is executed, the method described in any one of claims 1 to 4 is performed by a processor.

6. A meteorological disaster prediction device for crops, characterized in that, include: The data acquisition module is used to acquire the first meteorological conditions corresponding to each time period within multiple sample periods, and the disaster data corresponding to each sample period. The disaster data includes multiple disaster probability values ​​that affect different growth stages of crops. The first meteorological conditions include average temperature, maximum temperature, rainfall, maximum wind force, sunlight intensity and / or sunshine duration. The first correlation determination module is used to obtain the second meteorological conditions for each time period within the target period, and determine the first correlation based on each first meteorological condition and the second meteorological condition, wherein the first correlation includes the correlation between the second meteorological condition and each first meteorological condition, as well as the correlation between any two first meteorological conditions; A graph structure construction module is used to construct a graph structure based on the first association relationship, disaster data corresponding to each sample period, and virtual disaster conditions corresponding to the target period as nodes. The graph structure construction module includes: The second association determination module is used to determine multiple second associations based on each first association and a pre-set association threshold; and The graph structure construction submodule is used to construct the graph structure with each second association as edges and the disaster data corresponding to each sample period and the virtual disaster situation corresponding to the target period as nodes. A meteorological feature determination module is used to determine meteorological features corresponding to the second meteorological condition based on the second meteorological condition and a pre-trained feature extraction network. The meteorological feature determination module includes: The covariance matrix determination module is used to determine the corresponding covariance matrix based on the second meteorological condition. The eigenvalue decomposition module is used to calculate the corresponding transformation matrix and eigenvectors based on the covariance matrix; and The meteorological feature determination submodule is used to input the transformation matrix, the feature vector and the second meteorological condition into a pre-trained feature extraction network to determine the meteorological features corresponding to the second meteorological condition. A disaster data prediction module is used to iteratively update the graph structure using a pre-trained graph neural network model, and to predict disaster data corresponding to the target period based on the iteratively updated graph structure and the meteorological characteristics using a pre-trained meteorological disaster prediction model. The disaster data prediction module includes: A node attribute vector construction module is used to input the graph structure into the graph neural network model and construct node attribute vectors using calculation rules corresponding to the nodes. The node attribute vectors include the node attributes of the nodes and the products of the edge attributes of two adjacent edges and the corresponding node attributes. The iterative update module is used to iteratively update the node attribute vector using multiple first MLP models in the graph neural network model, and generate the iteratively updated graph structure.

7. A meteorological disaster prediction device for crops, characterized in that, include: processor; as well as A memory, connected to the processor, for providing the processor with instructions to perform the following processing steps: Acquire the first meteorological conditions corresponding to each time period within multiple sample periods, and the disaster data corresponding to each sample period, wherein the disaster data includes multiple disaster probability values ​​affecting different growth stages of crops, and wherein the first meteorological conditions include average temperature, maximum temperature, rainfall, maximum wind force, sunlight intensity and / or sunshine duration; The second meteorological conditions for each time period within the target period are obtained, and a first correlation is determined based on each first meteorological condition and the second meteorological conditions, wherein the first correlation includes the correlation between the second meteorological conditions and each first meteorological condition, as well as the correlation between any two first meteorological conditions; Based on the first association relationship, the disaster data corresponding to each sample period, and the virtual disaster situation corresponding to the target period as nodes, a graph structure is constructed, including: Based on each first association relationship and a pre-set association relationship threshold, multiple second association relationships are determined; and The graph structure is constructed using each second association as an edge and the disaster data corresponding to each sample period and the virtual disaster situation corresponding to the target period as nodes. Based on the second meteorological condition and the pre-trained feature extraction network, the meteorological features corresponding to the second meteorological condition are determined, including: Based on the second meteorological condition, determine the corresponding covariance matrix; Based on the covariance matrix, the corresponding transformation matrix and eigenvectors are calculated; and The transformation matrix, the feature vector, and the second meteorological condition are input into a pre-trained feature extraction network to determine the meteorological features corresponding to the second meteorological condition. The graph structure is iteratively updated using a pre-trained graph neural network model, and a meteorological disaster prediction model is used to predict disaster data corresponding to the target period based on the iteratively updated graph structure and the meteorological characteristics, including: The graph structure is input into the graph neural network model, and a node attribute vector is constructed using the calculation rules corresponding to the nodes. The node attribute vector includes the node attributes of the node and the product of the edge attributes of two adjacent edges and the node attributes of the corresponding nodes. The node attribute vector is iteratively updated using multiple first MLP models in the graph neural network model, and the iteratively updated graph structure is generated.

Images