CN118095836B

CN118095836B - Weather data analysis method and system for agricultural production state monitoring

Info

Publication number: CN118095836B
Application number: CN202410144329.9A
Authority: CN
Inventors: 刘轻扬; 梁乐宁; 张寅伟; 白玉; 郑巍; 鲁礼文; 韩亚东; 张陆陆; 崔海鹏; 刘晨; 周颖; 张娟; 别文祥
Original assignee: Beijing Tianyi Technology Co ltd
Current assignee: Beijing Tianyi Technology Co ltd
Priority date: 2024-02-01
Filing date: 2024-02-01
Publication date: 2024-09-24
Anticipated expiration: 2044-02-01
Also published as: CN118095836A

Abstract

The embodiment of the application provides a meteorological data analysis method and a system for monitoring agricultural production states, which are characterized in that agricultural production configuration data corresponding to a farmland risk area and target climate monitoring data corresponding to a first agricultural production unit are obtained, target candidate disaster influence parameters corresponding to a target candidate cultivation strategy are determined according to a basic agricultural disaster risk prediction network, and target training disaster influence parameters corresponding to the cultivation strategy are obtained. Then, according to the difference between the two parameters, updating the network parameter information of the agricultural disaster risk prediction network to generate a new agricultural disaster risk prediction network. Thus, the target disaster influence parameters corresponding to the target cultivation strategy can be predicted according to the input agricultural production data and the climate monitoring data. Therefore, the possible disaster risk in the agricultural production and the effect of the corresponding cultivation strategy can be effectively predicted, disaster prevention and control preparation can be carried out in advance in the agricultural production process, and the loss caused by disasters is reduced.

Description

Weather data analysis method and system for agricultural production state monitoring

Technical Field

The invention relates to the technical field of meteorological monitoring, in particular to a meteorological data analysis method and system for monitoring agricultural production states.

Background

Agricultural production is a complex system engineering involving the interactive effects of various climates, soil, plants and management factors. Agricultural disasters such as drought, flood, plant diseases and insect pests and the like form serious threats to agricultural production, and can lead to great reduction and even overall loss of crop yield. Therefore, the method is important to the prediction, prevention and control of agricultural disaster risks.

Traditional agricultural disaster risk prediction mainly depends on empirical judgment or statistical analysis based on historical data, and the methods often neglect complex interaction relations among climate, soil, plants and management factors, so that the accuracy and reliability of a prediction result are low. Moreover, due to the lack of an effective real-time monitoring and early warning mechanism, agricultural producers can often take countermeasures after disasters occur, and cannot take precautions in advance, so that the disaster prevention effect is poor.

In recent years, with the development of artificial intelligence technology, especially the application of deep learning technology, a new solution is provided for agricultural disaster risk prediction. However, most of the deep learning-based agricultural disaster risk prediction models at present mainly focus on the design and training methods of the models, and quality and integrity problems of input data of the models are ignored. Resulting in a non-ideal predictive effect of these models in practical applications.

Disclosure of Invention

In order to at least overcome the above-mentioned shortcomings in the prior art, an object of an embodiment of the present application is to provide a weather data analysis method and system for monitoring agricultural production status,

Acquiring agricultural production configuration data corresponding to a farmland risk area and target climate monitoring data corresponding to a first agricultural production unit, wherein the agricultural production configuration data comprises target agricultural production data corresponding to the farmland risk area and risk early warning requirements;

Determining target candidate disaster influence parameters corresponding to target candidate cultivation strategies based on the target agricultural production data and the target climate monitoring data according to a basic agricultural disaster risk prediction network, wherein the cultivation strategies represent the mode of scheduling agricultural production units to carry out cultivation activities in the farmland risk areas, and the disaster influence parameters represent the disaster degree triggered by the damage measurement dimensions corresponding to the risk early warning requirements according to the cultivation strategies;

acquiring target training disaster influence parameters corresponding to the target candidate cultivation strategies, wherein the target training disaster influence parameters are determined according to disaster parameters and disaster prevention investment parameters generated by scheduling the first agricultural production unit for target cultivation activities based on the target candidate cultivation strategies in the farmland dangerous area, the disaster prevention parameter characterization is determined according to disaster characteristics of the target cultivation activities in which damage measurement dimensions corresponding to the risk early warning requirements are matched, and the disaster prevention investment parameter characterization is used for performing investment corresponding to the target cultivation activities in the damage measurement dimensions corresponding to the risk early warning requirements;

Based on the difference between the target training disaster influence parameters and the target candidate disaster influence parameters, updating network parameter information of the basic agricultural disaster risk prediction network to generate an agricultural disaster risk prediction network, and predicting target disaster influence parameters corresponding to a target cultivation strategy according to input agricultural production data and input climate monitoring data based on the agricultural disaster risk prediction network.

In a possible implementation manner of the first aspect, the target agricultural production data and the target climate monitoring data are agricultural production data and climate monitoring data corresponding to an xth cultivation activity of uninterrupted K cultivation activities, the target cultivation activity is the xth cultivation activity, the target candidate cultivation policy characterizes a manner of scheduling the first agricultural production unit to perform the xth cultivation activity in the farmland risk area, the uninterrupted K cultivation activity is used for enabling the first agricultural production unit to match the risk pre-warning requirement once, and the determining, according to a base agricultural disaster risk prediction network, a target candidate disaster impact parameter corresponding to a target candidate cultivation policy based on the target agricultural production data and the target climate monitoring data includes:

According to the basic agricultural disaster risk prediction network, determining estimated disaster influence parameters corresponding to the x-th round of cultivation activities based on the target agricultural production data and the target climate monitoring data, wherein the estimated disaster influence parameters corresponding to the x-th round of cultivation activities represent the sum of candidate disaster influence parameters respectively corresponding to the x-th round of activities to the K-th round of activities;

Scheduling the first agricultural production unit in the farmland risk area for the x-th round of farming activities based on the target candidate farming strategy;

after the x-th round of cultivation activity is completed, outputting agricultural production data corresponding to the farmland risk area and climate monitoring data corresponding to the first agricultural production unit into agricultural production data and climate monitoring data corresponding to the (x+1) -th round of cultivation activity in the uninterrupted K-round of cultivation activity;

according to the basic agricultural disaster risk prediction network, determining estimated disaster influence parameters corresponding to the (x+1) -th round of cultivation activities based on agricultural production data and climate monitoring data corresponding to the (x+1) -th round of cultivation activities;

And outputting the difference between the estimated disaster impact parameters corresponding to the x-th round of cultivation activities and the estimated disaster impact parameters corresponding to the x+1-th round of cultivation activities as the target candidate disaster impact parameters.

In a possible implementation manner of the first aspect, the target candidate cultivation strategy is determined based on the target agricultural production data and the target climate monitoring data according to the underlying agricultural hazard risk prediction network, and the method further comprises:

Acquiring a cultivation activity sequence, wherein the cultivation activity sequence comprises a plurality of cultivation activities, and the cultivation activities respectively comprise corresponding agricultural production data, climate monitoring data, candidate cultivation strategies and estimated disaster influence parameters;

outputting a candidate cultivation strategy corresponding to the cultivation activity with the maximum corresponding estimated disaster influence parameter from a plurality of cultivation activities simultaneously corresponding to the target agricultural production data and the target climate monitoring data as a template cultivation strategy;

The updating the network parameter information of the basic agricultural disaster risk prediction network based on the difference between the target training disaster influence parameter and the target candidate disaster influence parameter, and generating an agricultural disaster risk prediction network, including:

Updating network parameter information of the basic agricultural disaster risk prediction network based on the difference between the target training disaster impact parameters and the target candidate disaster impact parameters and the difference between the template cultivation strategy and the target candidate cultivation strategy, and generating an agricultural disaster risk prediction network, wherein the agricultural disaster risk prediction network is further used for:

And determining a corresponding cultivation strategy of the agricultural production unit corresponding to the agricultural production unit trend data in a farmland risk area corresponding to the agricultural production data to be tested based on the agricultural production data to be tested and the agricultural production unit trend data.

In a possible implementation manner of the first aspect, the step of determining the continuous K-wheel cultivation activity includes:

Acquiring basic agricultural production data corresponding to the farmland risk area and basic climate monitoring data corresponding to the first agricultural production unit, wherein the basic agricultural production data represents the agricultural production data corresponding to the farmland risk area when the first agricultural production unit is not scheduled to execute any cultivation activity, and the basic climate monitoring data represents the climate monitoring data corresponding to the first agricultural production unit when the first agricultural production unit is not scheduled to execute any cultivation activity;

Outputting the basic agricultural production data and the basic climate monitoring data as agricultural production data and climate monitoring data corresponding to a first-round cultivation activity, and for a y-th-round cultivation activity, determining a candidate cultivation strategy corresponding to the y-th-round cultivation activity based on the agricultural production data and the climate monitoring data corresponding to the y-th-round cultivation activity according to the basic agricultural disaster risk prediction network;

Based on a candidate cultivation strategy corresponding to the y-th cultivation activity, scheduling a first agricultural production unit corresponding to climate monitoring data corresponding to the y-th cultivation activity to perform the y-th cultivation activity in a farmland risk area corresponding to agricultural production data corresponding to the y-th cultivation activity;

if the first agricultural production unit matches the risk early warning requirement after the y-th cultivation activity is completed, outputting a front y-th cultivation activity as the uninterrupted K-wheel cultivation activity;

if the first agricultural production unit does not match the risk early warning requirement after the x-th round of cultivation activity is completed, outputting the agricultural production data corresponding to the farmland risk area and the climate monitoring data corresponding to the first agricultural production unit into the agricultural production data and the climate monitoring data corresponding to the y+1-th round of cultivation activity after the y-th round of cultivation activity is completed.

In a possible implementation manner of the first aspect, the method further includes:

acquiring basic climate monitoring data corresponding to a second agricultural production unit;

Determining cultivation strategies corresponding to uninterrupted T-wheel cultivation activities corresponding to the second agricultural production unit respectively based on basic climate monitoring data corresponding to the second agricultural production unit and basic agricultural production data corresponding to the farmland risk area according to the agricultural disaster risk prediction network, wherein the uninterrupted T-wheel cultivation activities are used for enabling the second agricultural production unit to be matched with the risk early warning requirements once;

Outputting the sum of training disaster influence parameters corresponding to the cultivation strategies corresponding to the uninterrupted T-wheel cultivation activities as disaster risk assessment indexes corresponding to the second agricultural production units in the farmland risk areas, wherein the disaster risk assessment indexes represent assessment indexes of the second agricultural production units matched with the risk early warning requirements in the farmland risk areas.

In a possible implementation manner of the first aspect, the farmland risk area is a disaster concern area, the risk early warning requirement is that the first agricultural production unit is scheduled to a production area corresponding to a disaster-affected production unit in the farmland risk area, and a disaster-affected degree parameter corresponding to the disaster-affected production unit is increased by a set parameter;

the target agricultural production data represents corresponding unit distribution of production units included in the farmland risk area, weather monitoring data corresponding to production units except the first agricultural production unit and farmland label data corresponding to the farmland risk area, the target weather monitoring data represents weather monitoring data corresponding to the first agricultural production unit, and the weather monitoring data comprises crop growth farmland label data and/or crop growth space information;

The disaster recovery parameters include disaster recovery area increasing parameters and disaster loss increasing parameters, and the obtaining the target training disaster influence parameters corresponding to the target candidate cultivation strategy includes:

Determining disaster area increasing parameters, disaster loss increasing parameters and disaster prevention investment parameters corresponding to target farming activities, wherein the target farming activities are the farming activities executed by the first agricultural production units in the farmland risk area based on the target candidate farming strategies, the disaster area increasing parameters represent disaster areas between the first agricultural production units and the disaster recovery production units increased according to the target farming activities, and the disaster loss increasing parameters represent disaster degree parameters of the disaster recovery production units increased according to the target farming activities;

And determining a target training disaster influence parameter corresponding to the target candidate cultivation strategy based on the disaster area increase parameter, the disaster loss increase parameter and the disaster prevention investment parameter.

In a possible implementation manner of the first aspect, the target agricultural production data includes graph data representing a corresponding unit distribution of production units included in the agricultural risk area, a first feature vector representing climate monitoring data corresponding to production units other than the first agricultural production unit, and a sequence feature representing agricultural label data corresponding to the agricultural risk area, the target climate monitoring data includes a second feature vector representing climate monitoring data corresponding to the first agricultural production unit; the map data comprises a plurality of reference farmland identifications and production unit identifications corresponding to production units included in the farmland risk area, and map mapping information of the reference farmland identifications and the production unit identifications in the map data corresponds to position information of the production units included in the farmland risk area;

The determining the target candidate disaster impact parameters corresponding to the target candidate cultivation strategy based on the target agricultural production data and the target climate monitoring data comprises the following steps:

And carrying out extremum aggregation operation on the graph data, fusing graph features corresponding to the graph data after the extremum aggregation operation with the first feature vector and the second feature vector to generate a fused feature vector, and determining target candidate disaster influence parameters corresponding to a target candidate cultivation strategy based on the fused feature vector.

In a possible implementation manner of the first aspect, the determining step of the target candidate cultivation strategy includes:

Determining the target candidate cultivation strategy corresponding to the x-th round of cultivation activities based on the target agricultural production data, the target climate monitoring data and cultivation limiting information, wherein the cultivation limiting information represents negative cultivation activities corresponding to the first agricultural production unit in the farmland risk area, the negative cultivation activities are used for representing cultivation activities which cannot be performed by the first agricultural production unit in the farmland risk area corresponding to the target agricultural production data under the climate monitoring data identified by the target climate monitoring data, and the negative cultivation activities are not included in the x-th round of cultivation activities.

In a possible implementation manner of the first aspect, the target candidate cultivation policy includes candidate sub-cultivation policies corresponding to a plurality of cultivation categories, respectively, the base agricultural disaster risk prediction network includes a base cultivation sub-network corresponding to the plurality of cultivation categories, respectively, and a base disaster influence sub-network for determining an estimated disaster influence parameter, the updating network parameter information of the base agricultural disaster risk prediction network based on a difference between the target training disaster influence parameter and the target candidate disaster influence parameter and a difference between the template cultivation policy and the target candidate cultivation policy, generating an agricultural disaster risk prediction network, includes:

Determining template farming strategies corresponding to the template farming strategies in the plurality of farming categories respectively;

Respectively taking the multiple cultivation categories as target cultivation categories, for the target cultivation categories, updating network parameter information of a basic cultivation sub-network corresponding to the target cultivation categories based on the distinction between candidate sub-cultivation strategies and template sub-cultivation strategies corresponding to the target cultivation categories, and generating cultivation sub-networks corresponding to the target cultivation categories, wherein the cultivation sub-networks are used for determining sub-cultivation strategies of cultivation activities corresponding to the target cultivation categories;

based on the difference between the target training disaster influence parameters and the target candidate disaster influence parameters, updating network parameter information corresponding to the basic disaster influence sub-network to generate a disaster influence sub-network, wherein the disaster influence sub-network is used for determining estimated disaster influence parameters.

According to one aspect of an embodiment of the present application, there is provided a weather monitoring system including a processor and a machine-readable storage medium having stored therein machine-executable instructions loaded and executed by the processor to implement a weather data analysis method for agricultural production status monitoring in any one of the foregoing possible implementations.

According to an aspect of embodiments of the present application, there is provided a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The computer instructions are read from a computer-readable storage medium by a processor of a computer device, which executes the computer instructions, causing the computer device to perform the methods provided in the various alternative implementations of the three aspects described above.

According to the technical scheme provided by the embodiments of the application, the target candidate disaster influence parameters corresponding to the target candidate cultivation strategies are determined based on the basic agricultural disaster risk prediction network, the target agricultural production data and the target climate monitoring data, the target training disaster influence parameters corresponding to the target candidate cultivation strategies are further obtained and the parameters are updated, so that a more accurate agricultural disaster risk prediction network is generated, effective risk prevention and control of cultivation activities in a farmland risk area are realized, agricultural production loss caused by natural disasters is reduced, and the agricultural resource utilization efficiency is improved. The prediction result can be more fit with the actual situation through the determination of the target candidate cultivation strategy and the acquisition of the target training disaster influence parameters, the accuracy of agricultural disaster risk prediction is improved, various factors such as climate factors, farmland environmental characteristics and the like can be fully considered, and the reliability of the prediction result is improved.

That is, by acquiring the agricultural production configuration data corresponding to the farmland risk area and the target climate monitoring data corresponding to the first agricultural production unit, determining a target candidate disaster influence parameter corresponding to a target candidate cultivation strategy according to the basic agricultural disaster risk prediction network, and acquiring a target training disaster influence parameter corresponding to the cultivation strategy. Then, according to the difference between the two parameters, updating the network parameter information of the agricultural disaster risk prediction network to generate a new agricultural disaster risk prediction network. Thus, the target disaster influence parameters corresponding to the target cultivation strategy can be predicted according to the input agricultural production data and the climate monitoring data. Therefore, the possible disaster risk in the agricultural production and the effect of the corresponding cultivation strategy can be effectively predicted, disaster prevention and control preparation can be carried out in advance in the agricultural production process, and the agricultural production loss caused by disasters is reduced. Meanwhile, by adjusting parameters of the agricultural disaster risk prediction network, a prediction model is more accurate, the prediction accuracy of the agricultural disaster risk and the agricultural production efficiency are improved, and the method has important significance in optimizing the agricultural production strategy and improving the agricultural production benefit.

Drawings

For a clearer description of the technical solutions of the embodiments of the present application, reference will be made to the accompanying drawings, which are needed to be activated in the embodiments, and it should be understood that the following drawings only illustrate some embodiments of the present application, and therefore should not be considered as limiting the scope, and other related drawings can be extracted by those skilled in the art without the inventive effort.

FIG. 1 is a schematic flow chart of a meteorological data analysis method for monitoring agricultural production status according to an embodiment of the present application;

FIG. 2 is a schematic block diagram of a weather monitoring system for implementing the weather data analysis method for monitoring agricultural production status according to an embodiment of the present application.

Detailed Description

The following description is presented to enable one of ordinary skill in the art to make and use the application and is provided in the context of a particular application and its requirements. It will be apparent to those having ordinary skill in the art that various changes can be made to the disclosed embodiments and that the general principles defined herein may be applied to other embodiments and applications without departing from the principles and scope of the application. Therefore, the present application is not limited to the described embodiments, but is to be accorded the widest scope consistent with the claims.

Fig. 1 is a flow chart of a weather data analysis method for monitoring an agricultural production status according to an embodiment of the present application, and the weather data analysis method for monitoring an agricultural production status is described in detail below.

Step S110, agricultural production configuration data corresponding to a farmland risk area and target climate monitoring data corresponding to a first agricultural production unit are obtained, wherein the agricultural production configuration data comprises target agricultural production data corresponding to the farmland risk area and risk early warning requirements.

For example, in a field risk area located in a mountain area, agricultural production configuration data of the field risk area needs to be acquired, including target agricultural production data (such as planted crop types, planting areas, etc.) and risk early warning requirements (such as flood control, drought control, pest control, etc.). Meanwhile, target climate monitoring data corresponding to the first agricultural production unit (such as a farm) including temperature, precipitation, humidity and the like are required to be acquired, so that scientific basis is provided for agricultural production.

By way of example, the field risk area may refer to a geographic area with a high risk of agricultural hazards, possibly including risks caused by terrain, climate, soil, etc. For example, in a mountain area, the precipitation amount is large due to the high topography, and the soil erosion is serious, which may be regarded as a dangerous farmland area.

The agricultural production configuration data may refer to data and materials for guiding agricultural production activities, including target agricultural production data and risk early warning requirements, etc. For example, agricultural production configuration data may include crop planting types, planting areas, irrigation facility conditions, pest control measures, and the like. The target agricultural production data may be agricultural production targets, such as the kinds of crops planted, the yields, etc., set for a specific farmland risk area. For example, for a field risk area prone to drought, the target agricultural production data may be to plant crops with greater drought tolerance to reduce the impact of drought on agricultural production. The risk early warning requirement can be an early warning standard and requirement set by a pointer to a specific farmland risk area, and is used for guiding agricultural disaster risk prevention and coping. For example, for a farmland risk area where flood is likely to occur, the risk early warning requirement may be early warning of the rise of water level in advance, so as to ensure the normal operation of farmland drainage facilities, so as to reduce the influence of flood on agricultural production.

Step S120, determining target candidate disaster impact parameters corresponding to target candidate cultivation strategies based on the target agricultural production data and the target climate monitoring data according to a basic agricultural disaster risk prediction network, wherein the cultivation strategies represent modes of scheduling agricultural production units to carry out cultivation activities in the farmland dangerous areas, and the disaster impact parameters represent disaster degrees triggered by damage measurement dimensions corresponding to the risk early warning requirements according to the cultivation strategies.

For example, after the target agricultural production data and the target climate monitoring data are acquired, the weather monitoring system needs to analyze and process the target agricultural production data and the target climate monitoring data according to the basic agricultural disaster risk prediction network to determine target candidate disaster impact parameters corresponding to the target candidate cultivation strategies. For example, in a dangerous area of a farmland where drought is likely to occur, the target candidate cultivation strategy may be to plant crops with stronger drought resistance, and the target candidate disaster influence parameter may be the influence degree of the drought on the crop yield.

Wherein the underlying agricultural hazard risk prediction network may refer to an artificial intelligence model for predicting agricultural hazard risk, typically constructed based on historical data and algorithms. For example, a basic agricultural disaster risk prediction network can be constructed through algorithms such as a neural network and a support vector machine, and is used for predicting disaster influence degrees under different cultivation strategies.

The target candidate cultivation strategy can be a plurality of possible cultivation strategies set by a pointer to a specific farmland risk area and used for evaluating the disaster influence degree. For example, for a field risk area prone to drought, target candidate cultivation strategies may include planting crops with greater drought tolerance, changing planting time, employing water retention irrigation techniques, and so forth.

The target candidate disaster influence parameter can be the disaster degree triggered by the pointer on a specific cultivation strategy under the requirement of matching risk early warning. For example, for a dangerous area of a farm where drought is likely to occur, the target candidate disaster impact parameter for planting crops with greater drought tolerance may be the extent to which drought affects the crop yield.

The cultivation strategy may refer to a manner of scheduling cultivation activities of the agricultural production units in the field risk area, such as planting crop types, planting time, irrigation modes, etc. For example, cultivation strategies may include planting crops that are more drought tolerant, changing planting times, employing water retention irrigation techniques, and the like.

The disaster influence parameters can refer to disaster degrees triggered under the requirement of matching risk early warning according to a cultivation strategy, such as influence degrees of drought on crop yield, damage degrees of flood on farmlands and the like. For example, disaster impact parameters may include the extent to which drought affects crop yields, the extent to which floods damage farms, etc.

The damage measurement dimension may refer to a standard or index for evaluating the extent of disaster impact, such as crop yield, farmland soil quality, agricultural production facilities, etc. For example, damage measurement dimensions may include crop yield, farmland soil quality, agricultural production facilities, and the like.

Step S130, obtaining target training disaster influence parameters corresponding to the target candidate cultivation strategies, wherein the target training disaster influence parameters are determined according to disaster parameters and disaster prevention investment parameters generated by scheduling the first agricultural production unit for target cultivation activities based on the target candidate cultivation strategies in the farmland risk area, the disaster parameters represent disaster characteristics increased according to damage measurement dimensions corresponding to the risk early warning requirements when the target cultivation activities are performed, and the disaster prevention investment parameters represent investment corresponding to the damage measurement dimensions corresponding to the risk early warning requirements when the target cultivation activities are performed.

For example, after determining the target candidate disaster impact parameters corresponding to the target candidate cultivation strategy, the weather monitoring system also needs to acquire the target training disaster impact parameters. Specifically, it is necessary to schedule the first agricultural production unit to perform a target farming activity based on a target candidate farming strategy in the field risk area, and record the generated disaster recovery parameters and disaster prevention input parameters. For example, in a dangerous area of a farmland where drought is likely to occur, crops with stronger drought resistance are planted as target candidate cultivation strategies, and after the strategies are implemented, disaster conditions of the crops in the drought and input disaster prevention resources are required to be recorded.

The target training disaster influence parameters may be disaster influence parameters generated after actual cultivation activities are performed in the farmland risk area based on a target candidate cultivation strategy, and are used for updating an agricultural disaster risk prediction network. For example, based on a target candidate cultivation strategy for planting crops with stronger drought tolerance, after actual cultivation activities are performed in a dangerous area of a farmland, disaster influence parameters, such as the influence degree of drought on crop yield, are collected.

The disaster-affected parameters can be disaster-affected characteristics increased under the requirement of matching risk early warning according to target farming activities, such as the influence degree of drought on crop yield, the damage degree of flood on farmlands and the like. For example, the disaster parameters may include the extent of the effect of drought on crop yield, the extent of damage of flood to farmland, etc.

The disaster prevention input parameters can be disaster prevention input required under the requirement of matching risk early warning according to target cultivation activities, such as irrigation facility input, disaster prevention seed input, disaster prevention training input and the like.

Step S140, updating network parameter information of the basic agricultural disaster risk prediction network based on the difference between the target training disaster influence parameter and the target candidate disaster influence parameter, generating an agricultural disaster risk prediction network, and predicting target disaster influence parameters corresponding to a target cultivation strategy based on the agricultural disaster risk prediction network for input agricultural production data and input climate monitoring data.

For example, after the target training disaster impact parameters are obtained, the meteorological monitoring system needs to compare the target training disaster impact parameters with the target candidate disaster impact parameters, and update the network parameter information of the basic agricultural disaster risk prediction network according to the difference. The updated agricultural disaster risk prediction network can be used for predicting disaster influence parameters triggered by different farming activities, and powerful support is provided for agricultural production. For example, in a dangerous farmland area prone to drought, the updated agricultural disaster risk prediction network can more accurately predict the disaster influence degree of crops with stronger drought resistance, so that a more scientific decision basis is provided.

Illustratively, by comparing the difference between the target training disaster impact parameters and the target candidate disaster impact parameters, for example, whether the predicted drought impact degree on the crop yield deviates from the actual drought impact degree on the crop yield. Then, network parameter information of the basic agricultural disaster risk prediction network is adjusted according to the difference, such as weight and bias items in the neural network are adjusted. The updated agricultural disaster risk prediction network can more accurately predict the disaster influence degree of crops with stronger drought resistance, thereby providing more scientific decision basis.

On the basis, the agricultural insurance decision weight can be generated according to the predicted target disaster influence parameters. Specifically, a regression model may be used, with the target disaster impact parameters as input, to output agricultural insurance decision weights. The agricultural insurance decision weights may be used to formulate agricultural insurance policies. For example, different insurance costs and insurance compensation criteria may be determined based on the generated agricultural insurance decision weights for different agricultural production areas and different cultivation strategies.

Based on the above steps, the technical scheme provided by the embodiment of the application determines the target candidate disaster influence parameters corresponding to the target candidate cultivation strategy based on the basic agricultural disaster risk prediction network, the target agricultural production data and the target climate monitoring data, further acquires the target training disaster influence parameters corresponding to the target candidate cultivation strategy and updates the parameters, thereby generating a more accurate agricultural disaster risk prediction network, realizing effective risk prevention and control of cultivation activities in a farmland risk area, reducing agricultural production loss caused by natural disasters, and improving the utilization efficiency of agricultural resources. The prediction result can be more fit with the actual situation through the determination of the target candidate cultivation strategy and the acquisition of the target training disaster influence parameters, the accuracy of agricultural disaster risk prediction is improved, various factors such as climate factors, farmland environmental characteristics and the like can be fully considered, and the reliability of the prediction result is improved.

In a possible implementation manner, the target agricultural production data and the target climate monitoring data are agricultural production data and climate monitoring data corresponding to an xth round of cultivation activities among uninterrupted K rounds of cultivation activities, the target cultivation activities are the xth round of cultivation activities, and the target candidate cultivation strategy characterizes a manner of scheduling the first agricultural production unit to perform the xth round of cultivation activities in the farmland risk area, wherein the uninterrupted K rounds of cultivation activities are used for enabling the first agricultural production unit to match the risk pre-warning requirement once.

For example, suppose there is a field risk area where there is a need for uninterrupted K rounds of farming activities, each round of activity having different agricultural production data and climate monitoring data. Taking the example of the x-th round of cultivation activity, the target agricultural production data may include the type of crop planted, the planting area, the planting density, etc., and the target climate monitoring data may include temperature, precipitation, humidity, etc.

In the above-mentioned farm risk area, an x-th round of cultivation activity is selected as a target cultivation activity, and a target candidate cultivation strategy is determined to schedule the agricultural production units. For example, the target candidate cultivation strategy may be to plant crops with stronger drought resistance, or to take irrigation measures, etc.

Step S120 may include:

Step S121, according to the basic agricultural disaster risk prediction network, based on the target agricultural production data and the target climate monitoring data, determining estimated disaster impact parameters corresponding to the x-th round of cultivation activities, where the estimated disaster impact parameters corresponding to the x-th round of cultivation activities represent the sum of candidate disaster impact parameters corresponding to the x-th round of activities to the K-th round of activities, respectively.

For example, after determining a target candidate cultivation strategy, it is necessary to predict disaster impact parameters that may be caused by the target candidate cultivation strategy based on the target agricultural production data and the target climate monitoring data using the underlying agricultural disaster risk prediction network. For example, if the target candidate cultivation strategy is to plant crops with a strong drought resistance, disaster influence parameters to which such crops may be subjected under current climate conditions can be predicted by the underlying agricultural disaster risk prediction network.

Step S122, scheduling the first agricultural production unit to perform the x-th round of farming activities in the farmland risk area based on the target candidate farming strategy.

For example, after a target candidate cultivation strategy is obtained, the target candidate cultivation strategy needs to be implemented in a farmland risk area to schedule agricultural production units. For example, if a crop with a high drought resistance is selected as a target candidate cultivation strategy, it is necessary to prompt the simulation of planting such a crop in the farmland.

And step S123, outputting the agricultural production data corresponding to the farmland risk area and the climate monitoring data corresponding to the first agricultural production unit into the agricultural production data and the climate monitoring data corresponding to the (x+1) th cultivation activity in the uninterrupted K-round cultivation activity after the x-th cultivation activity is completed.

For example, after completion of the x-th round of farming activities, it is necessary to collect relevant agricultural production data and climate monitoring data in order to provide a reference for the next round of farming activities. For example, data support may be provided for the x+1 th round of farming campaign by monitoring crop growth, soil humidity, climate conditions, etc.

Step S124, according to the basic agricultural disaster risk prediction network, determining estimated disaster influence parameters corresponding to the (x+1) -th round of cultivation activities based on the agricultural production data and the climate monitoring data corresponding to the (x+1) -th round of cultivation activities.

For example, after agricultural production data and climate monitoring data for an (x+1) th round of farming activities are collected, the disaster impact that may be caused by the round of activities may be predicted by the underlying agricultural disaster risk prediction network. For example, a base agricultural hazard risk prediction network may be utilized to predict the extent of the hazard impact of planting different crops under current climate conditions.

And step S125, outputting the difference between the estimated disaster impact parameters corresponding to the x-th round of cultivation activities and the estimated disaster impact parameters corresponding to the x+1-th round of cultivation activities as the target candidate disaster impact parameters.

For example, the target candidate disaster impact parameters may be obtained by comparing the estimated disaster impact parameters of the x-th and x+1-th cultivation activities. For example, the best crop planting strategy can be selected by comparing the extent of disaster impact of different crops.

In one possible embodiment, the target candidate cultivation strategy is determined from the underlying agricultural hazard risk prediction network based on the target agricultural production data and the target climate monitoring data, the method further comprising:

Step A110, a cultivation activity sequence is obtained, wherein the cultivation activity sequence comprises a plurality of cultivation activities, and the cultivation activities respectively comprise corresponding agricultural production data, climate monitoring data, candidate cultivation strategies and estimated disaster impact parameters.

And step A120, outputting a candidate cultivation strategy corresponding to the cultivation activity with the largest corresponding estimated disaster influence parameter from a plurality of cultivation activities simultaneously corresponding to the target agricultural production data and the target climate monitoring data as a template cultivation strategy.

For example, assuming a field risk area, multiple rounds of farming activities are required, each round having different agricultural production data, climate monitoring data, candidate farming strategies, and estimated disaster impact parameters. The cultivation activities may be arranged in a time sequence to form a cultivation activity sequence.

In the above-mentioned farm risk area, it is necessary to acquire target agricultural production data and target climate monitoring data. For example, the target agricultural production data may include the type of crop planted, the planting area, the planting density, etc., and the target climate monitoring data may include temperature, precipitation, humidity, etc.

After the target agricultural production data and the target climate monitoring data are obtained, a plurality of candidate cultivation strategies need to be determined. For example, a candidate cultivation strategy may be to plant crops that are more drought tolerant, or to take irrigation measures, etc.

For each candidate cultivation strategy, it is necessary to determine its corresponding estimated disaster impact parameters. For example, a base agricultural hazard risk prediction network may be utilized to predict the possible disaster impact each candidate cultivation strategy may have on the basis of target agricultural production data and target climate monitoring data.

After determining the estimated disaster impact parameters of the multiple rounds of cultivation activities, it is necessary to select a candidate cultivation strategy corresponding to the cultivation activity having the largest estimated disaster impact parameter as a template cultivation strategy. For example, if the estimated disaster impact parameter for planting a crop with a strong drought resistance is the largest, the candidate cultivation strategy is used as the template cultivation strategy.

The step S140 may include:

step S141, updating network parameter information of the basic agricultural disaster risk prediction network based on the difference between the target training disaster influence parameter and the target candidate disaster influence parameter and the difference between the template cultivation strategy and the target candidate cultivation strategy, and generating an agricultural disaster risk prediction network. The agricultural disaster risk prediction network is further configured to:

Step S142, based on the agricultural production data to be tested and the agricultural production unit trend data, determining a cultivation strategy corresponding to the agricultural production unit trend data in the farmland risk area corresponding to the agricultural production data to be tested.

For example, after acquiring the target training disaster influence parameters and the target candidate disaster influence parameters, it is necessary to consider the distinction between them. For example, their differences may be calculated in order to update network parameter information of the underlying agricultural hazard risk prediction network. In addition, the distinction between the template farming strategy and the target candidate farming strategy needs to be considered. For example, their differences may be calculated and considered together with differences between the target training disaster impact parameters in order to update the network parameter information of the underlying agricultural disaster risk prediction network. After considering the differences between the target training disaster impact parameters and the target candidate disaster impact parameters and the differences between the template cultivation strategies and the target candidate cultivation strategies, the network parameter information of the basic agricultural disaster risk prediction network needs to be updated. For example, the network parameter information may be updated using a gradient descent algorithm or other optimization algorithm.

After updating the network parameter information of the underlying agricultural hazard risk prediction network, the underlying agricultural hazard risk prediction network may be used to predict a corresponding cultivation policy of the agricultural production unit in the farmland risk area. For example, based on the agricultural production data to be tested and the trend data of the agricultural production units, the agricultural disaster risk prediction network can be input to obtain a corresponding cultivation strategy, such as selecting an optimal crop planting mode, irrigation measures and the like.

In a possible implementation manner, the step of determining the continuous K-wheel farming activities includes:

And step B110, acquiring basic agricultural production data corresponding to the farmland risk area and basic climate monitoring data corresponding to the first agricultural production unit, wherein the basic agricultural production data represents the agricultural production data corresponding to the farmland risk area when the first agricultural production unit is not scheduled to execute any cultivation activity, and the basic climate monitoring data represents the climate monitoring data corresponding to the first agricultural production unit when the first agricultural production unit is not scheduled to execute any cultivation activity.

For example, assuming a field risk area, it is necessary to acquire base agricultural production data and base climate monitoring data. The base agricultural production data may include data of crop yield, soil fertility, etc. of the farmland risk area without any cultivation activity, and the base climate monitoring data may include climate data of the farmland risk area without any cultivation activity, such as temperature, precipitation, etc.

And step B120, outputting the basic agricultural production data and the basic climate monitoring data as agricultural production data and climate monitoring data corresponding to a first-round cultivation activity, and determining a candidate cultivation strategy corresponding to a y-th-round cultivation activity according to the basic agricultural disaster risk prediction network and based on the agricultural production data and the climate monitoring data corresponding to the y-th-round cultivation activity for the y-th-round cultivation activity.

For example, the obtained basic agricultural production data and basic climate monitoring data may be used as agricultural production data and climate monitoring data corresponding to the first-round farming activities.

And, for the y-th round of cultivation activities, it is necessary to determine candidate cultivation strategies corresponding to the y-th round of cultivation activities based on the agricultural production data and the climate monitoring data corresponding to the y-th round of cultivation activities according to the underlying agricultural disaster risk prediction network. For example, crops with stronger drought resistance can be planted or irrigation measures can be taken.

And step B130, based on the candidate cultivation strategy corresponding to the y-th round of cultivation activity, scheduling a first agricultural production unit corresponding to the climate monitoring data corresponding to the y-th round of cultivation activity in a farmland risk area corresponding to the agricultural production data corresponding to the y-th round of cultivation activity to perform the y-th round of cultivation activity.

For example, a first agricultural production unit is scheduled in the agricultural risk area for a y-th round of farming activity based on a determined candidate farming strategy for the y-th round of farming activity. For example, agricultural production units may be arranged to plant crops that are more drought tolerant or to turn on irrigation facilities.

And step B140, outputting the previous y-round cultivation activity as the uninterrupted K-round cultivation activity if the first agricultural production unit matches the risk early warning requirement after the y-round cultivation activity is completed.

After the y-th cultivation activity is completed, whether the first agricultural production unit matches the risk early warning requirement or not needs to be judged. For example, the risk level faced by the agricultural production unit can be estimated based on data such as crop yield, soil fertility, etc., and climate data, and if the risk level is within an acceptable range, the risk early warning requirement is deemed to be matched.

And step B150, if the first agricultural production unit does not match the risk early warning requirement after the x-th round of cultivation activity is completed, outputting the agricultural production data corresponding to the farmland risk area and the climate monitoring data corresponding to the first agricultural production unit into the agricultural production data and the climate monitoring data corresponding to the y+1th round of cultivation activity after the y-th round of cultivation activity is completed.

If the first agricultural production unit matches the risk pre-warning requirement after the y-th round of farming activities is completed, the front y-round of farming activities are output as uninterrupted K-round farming activities. For example, the first 5 cultivation activities may be determined as an uninterrupted sequence of cultivation activities.

If the first agricultural production unit does not match the risk early warning requirement after the y-th cultivation activity is completed, the agricultural production data and the climate monitoring data after the y-th cultivation activity are required to be used as the agricultural production data and the climate monitoring data corresponding to the y+1-th cultivation activity, and the steps 3-6 are continuously executed. For example, if 5 rounds of farming activities are completed, but the risk early warning requirements are not yet matched, then the 6 th round of farming activities need to be continued until the risk early warning requirements are matched.

In one possible embodiment, the method further comprises:

and step C110, acquiring basic climate monitoring data corresponding to the second agricultural production unit.

For example, suppose there is also a field risk area where basic climate monitoring data corresponding to a second agricultural production unit is to be obtained. The base climate monitoring data may include climate data, such as temperature, precipitation, etc., of the agricultural risk area without any farming activities.

And step C120, determining cultivation strategies corresponding to uninterrupted T-wheel cultivation activities corresponding to the second agricultural production units respectively based on basic climate monitoring data corresponding to the second agricultural production units and basic agricultural production data corresponding to the farmland risk areas according to the agricultural disaster risk prediction network, wherein the uninterrupted T-wheel cultivation activities are used for enabling the second agricultural production units to be matched with the risk early warning requirements once.

For the second agricultural production unit, according to the agricultural disaster risk prediction network, determining cultivation strategies corresponding to uninterrupted T-turn cultivation activities corresponding to the second agricultural production unit based on basic climate monitoring data corresponding to the second agricultural production unit and basic agricultural production data corresponding to the farmland dangerous area. For example, crops with stronger drought resistance can be planted or irrigation measures can be taken.

And step C130, outputting the sum of training disaster influence parameters corresponding to cultivation strategies corresponding to the uninterrupted T-turn cultivation activities as disaster risk assessment indexes corresponding to the second agricultural production units in the farmland risk areas, wherein the disaster risk assessment indexes represent assessment indexes of the second agricultural production units matched with the risk early warning requirements in the farmland risk areas.

After the uninterrupted T-turn cultivation activities are completed, disaster risk assessment indexes corresponding to the second agricultural production units in the farmland risk areas need to be calculated. The disaster risk assessment index may represent an assessment index of the second agricultural production unit matching the risk early warning requirement in the farmland risk area. Specifically, the sum of training disaster influence parameters corresponding to cultivation strategies corresponding to uninterrupted T-turn cultivation activities can be used as a disaster risk assessment index.

For example, assuming a field risk area, it is necessary to determine disaster risk assessment indicators of the first agricultural production unit and the second agricultural production unit in the field risk area. For the first agricultural production unit, basic agricultural production data and current climate monitoring data corresponding to the first agricultural production unit can be obtained, and a cultivation strategy corresponding to uninterrupted K-wheel cultivation activities of the first agricultural production unit and candidate disaster influence parameters corresponding to each cultivation strategy are determined through an agricultural disaster risk prediction network. Similarly, for the second agricultural production unit, basic climate monitoring data corresponding to the second agricultural production unit and basic agricultural production data corresponding to the farmland risk area can be obtained, and cultivation strategies corresponding to uninterrupted T-wheel cultivation activities of the second agricultural production unit and training disaster influence parameters corresponding to each cultivation strategy are determined through an agricultural disaster risk prediction network. And finally, taking the sum of candidate disaster influence parameters or the sum of training disaster influence parameters corresponding to the cultivation strategies corresponding to the uninterrupted K-round cultivation activities and the uninterrupted T-round cultivation activities respectively corresponding to the first agricultural production unit and the second agricultural production unit as disaster risk assessment indexes corresponding to the first agricultural production unit and the second agricultural production unit in a farmland risk area.

In a possible implementation manner, the farmland risk area is a disaster concern area, the risk early warning requirement is that the first agricultural production unit is scheduled to a production area corresponding to a disaster-affected production unit in the farmland risk area, and the disaster-affected degree parameter corresponding to the disaster-affected production unit is increased by a set parameter.

Still assume that there is a farm risk area, which is a disaster area of interest, where the first agricultural production unit needs to be scheduled to meet the risk early warning requirement. Specifically, the first agricultural production unit is required to be scheduled to a production area corresponding to the disaster-affected production unit in the farmland risk area, and the disaster-affected degree parameter corresponding to the disaster-affected production unit is increased by a set parameter. For example, the disaster damage degree parameter may include disaster damage area and disaster damage, and the setting parameter may be a fixed value or may be dynamically adjusted according to actual situations.

The target agricultural production data characterizes corresponding unit distribution of production units included in the farmland risk area, climate monitoring data corresponding to production units except the first agricultural production unit and farmland label data corresponding to the farmland risk area, the target climate monitoring data characterizes the climate monitoring data corresponding to the first agricultural production unit, and the climate monitoring data comprises crop growth farmland label data and/or crop growth space information.

For example, the target agricultural production data may characterize a corresponding cell distribution of production cells included in the agricultural risk area, climate monitoring data corresponding to production cells other than the first agricultural production cell, and agricultural label data corresponding to the agricultural risk area. The target climate monitoring data may be indicative of climate monitoring data corresponding to the first agricultural production unit, and the climate monitoring data may include crop growth farmland label data and/or crop growth space information.

The disaster recovery parameters comprise disaster recovery area increasing parameters and disaster damage increasing parameters, the disaster recovery area increasing parameters can represent disaster recovery areas between the first agricultural production units and the disaster recovery production units which are increased according to the target farming activities, and the disaster damage increasing parameters can represent disaster recovery degree parameters of the disaster recovery production units which are increased according to the target farming activities.

In step S130, obtaining a target training disaster impact parameter corresponding to the target candidate cultivation strategy includes:

Step S131, determining disaster area increasing parameters, disaster loss increasing parameters and disaster prevention investment parameters corresponding to target farming activities, wherein the target farming activities are the farming activities executed by the first agricultural production units in the farmland risk area based on the target candidate farming strategies, the disaster area increasing parameters represent disaster areas between the first agricultural production units and the disaster recovery production units increased according to the target farming activities, and the disaster loss increasing parameters represent disaster degree parameters of the disaster recovery production units increased according to the target farming activities.

Step S132, determining a target training disaster influence parameter corresponding to the target candidate cultivation strategy based on the disaster area increase parameter, the disaster damage increase parameter and the disaster prevention investment parameter.

For example, for a target farming activity, it is necessary to determine its corresponding disaster area increase parameter, disaster loss increase parameter, and disaster prevention investment parameter. For example, disaster prevention input parameters may include input cost of irrigation facilities, input cost of pesticides and fertilizers, and the like.

Based on the determined disaster area increasing parameter, disaster damage increasing parameter and disaster prevention investment parameter, a target training disaster influence parameter corresponding to the target candidate cultivation strategy can be determined. For example, the three types of parameters may be used to calculate a target training disaster impact parameter corresponding to the target candidate cultivation strategy by a loss function.

Through the steps, the target training disaster influence parameters corresponding to the target candidate cultivation strategies can be obtained.

In one possible embodiment, the target agricultural production data includes graph data characterizing a corresponding unit distribution of production units included in the agricultural risk area, a first feature vector characterizing climate monitoring data corresponding to production units other than the first agricultural production unit, and a sequence feature characterizing farmland label data corresponding to the agricultural risk area, the target climate monitoring data including a second feature vector characterizing climate monitoring data corresponding to the first agricultural production unit. The map data comprises a plurality of reference farmland identifications and production unit identifications corresponding to production units included in the farmland risk area, and map mapping information of the reference farmland identifications and the production unit identifications in the map data corresponds to position information of the production units included in the farmland risk area.

Wherein the graph data is a commonly used data structure that can be used to represent complex relational networks. In agricultural production, the production units included in the agricultural risk area may be represented as a graph in which nodes represent production units and edges represent relationships between production units. The reference farmland identification and the production unit identification are used for uniquely identifying the production units, and the map mapping information of the production units in the map data can represent the relative positions of the production units in the farmland risk area.

The feature vector is a data structure used to represent a feature of the data. In agricultural production, the feature vector may be used to represent climate monitoring data of production units other than the first agricultural production unit in the agricultural risk area, agricultural label data corresponding to the agricultural risk area, and the like. The feature vector may be a multi-dimensional vector in which each dimension may represent a particular feature.

The unit distribution refers to the distribution condition of the production units in the farmland risk area, and can comprise the number, the types, the planting areas and the like of the production units. In the graph data, the distribution of the production units can be described by the distribution of the nodes and edges.

The farmland tag data refers to data for describing farmland characteristics, and can comprise soil types, terrains, climate types and the like. A sequence feature is a data structure representing a sequence data feature, for example, the sequence feature may be formed from a sequence of values of field tag data.

In summary, the above data can comprehensively describe the condition of the farmland risk area and the characteristics of the production units, thereby helping to formulate a cultivation strategy more suitable for the actual condition of the farmland risk area and reducing the disaster risk.

In step S120, determining a target candidate disaster impact parameter corresponding to the target candidate cultivation policy based on the target agricultural production data and the target climate monitoring data includes: and carrying out extremum aggregation operation on the graph data, fusing graph features corresponding to the graph data after the extremum aggregation operation with the first feature vector and the second feature vector to generate a fused feature vector, and determining target candidate disaster influence parameters corresponding to a target candidate cultivation strategy based on the fused feature vector.

For example, extremum aggregation operations may be performed on the graph data, which may be used to extract meaningful node and side information in the graph data. In particular, some aggregation functions, such as max-pooling, average pooling, etc., may be used to aggregate the node and side information in the graph data to obtain a more compact feature representation.

After the extremum aggregation operation is performed, the graph features corresponding to the obtained graph data can be fused with the first feature vector and the second feature vector to generate a fused feature vector. The fusion characteristic vector can comprehensively consider information such as production unit distribution, climate monitoring data, farmland label data and the like in the farmland risk area, so that the fusion characteristic vector can be used for determining target candidate disaster influence parameters corresponding to a target candidate cultivation strategy. Based on the generated fusion feature vector, some machine learning models, such as a neural network, can be used to determine target candidate disaster impact parameters corresponding to the target candidate cultivation strategy. Specifically, the fusion feature vector can be used as input, and the predicted value of the target candidate disaster influence parameter can be obtained through model training.

In one possible embodiment, the determining step of the target candidate cultivation strategy includes:

For example, target agricultural production data and target climate monitoring data have been previously explained, which may comprehensively describe the condition of the agricultural risk area and the characteristics of the production units.

The cultivation limiting information is information of cultivation activities which cannot be performed by the first agricultural production unit in the farmland risk area. Such information may be obtained by mining and analyzing historical data of agricultural production, e.g., in the first few years of climatic conditions, some farming activities are not possible, etc. The negative farming activities refer to farming activities that the first agricultural production unit cannot do in the risk area of the farmland corresponding to the target agricultural production data under the climate monitoring data identified by the target climate monitoring data.

The target candidate cultivation strategy refers to a cultivation strategy selected for the x-th round of cultivation activities based on the target agricultural production data, the target climate monitoring data and the cultivation definition information. In determining the target candidate cultivation strategy, it is necessary to avoid selecting negative cultivation activities to ensure the feasibility of agricultural production activities.

Thus, the determining of the target candidate cultivation strategy includes determining the target candidate cultivation strategy corresponding to the xth round of cultivation activity based on the target agricultural production data, the target climate monitoring data, and cultivation definition information. The target candidate cultivation strategy determined in this way can adapt to the actual condition of the farmland risk area, and the efficiency and quality of agricultural production are improved.

In a possible implementation manner, the target candidate cultivation strategy includes candidate sub-cultivation strategies corresponding to a plurality of cultivation categories respectively, and the basic agricultural disaster risk prediction network includes a basic cultivation sub-network corresponding to the plurality of cultivation categories respectively, and a basic disaster influence sub-network for determining estimated disaster influence parameters.

The step S141 may include:

Step S1411, determining a template sub-cultivation strategy corresponding to each of the plurality of cultivation categories according to the template cultivation strategy.

Step S1412, regarding the multiple cultivation categories as target cultivation categories, updating network parameter information of a basic cultivation sub-network corresponding to the target cultivation category based on a distinction between a candidate sub-cultivation policy and a template sub-cultivation policy corresponding to the target cultivation category, and generating a cultivation sub-network corresponding to the target cultivation category, wherein the cultivation sub-network is used for determining a sub-cultivation policy of cultivation activities corresponding to the target cultivation category.

Step 1413, updating network parameter information corresponding to the basic disaster influence sub-network based on the difference between the target training disaster influence parameter and the target candidate disaster influence parameter, and generating a disaster influence sub-network, wherein the disaster influence sub-network is used for determining the estimated disaster influence parameter.

For example, the target candidate cultivation strategy includes candidate sub-cultivation strategies corresponding to a plurality of cultivation categories, respectively, and the base agricultural disaster risk prediction network includes a base cultivation sub-network corresponding to a plurality of cultivation categories, respectively, and a base disaster influence sub-network for determining the estimated disaster influence parameters. These networks may be obtained through training and may be used to determine an optimal cultivation strategy.

The template cultivation strategy refers to a cultivation strategy obtained based on historical data and expert experience, and can be used as a reference for training. The template farming strategy may include template farming strategies that correspond to a plurality of farming categories, respectively.

In the above embodiment, for the target cultivation category c, based on the difference between the candidate sub-cultivation policy and the template sub-cultivation policy corresponding to the target cultivation category, the network parameter information of the base cultivation sub-network corresponding to the target cultivation category is updated, and the cultivation sub-network corresponding to the target cultivation category is generated, which may be expressed as:

Δw(c)＝-α*(y(c)-v(c))*(y(c)-v(c))^T

Wherein Δw (c) represents a parameter update vector of the basic tiller network corresponding to the target cultivation category c, y (c) represents a template cultivation policy vector corresponding to the target cultivation category c, v (c) represents a candidate cultivation policy vector corresponding to the target cultivation category c, and α is a learning rate.

The calculation formula for updating the network parameter information of the basic agricultural disaster risk prediction network is as follows:

Δw＝-α*(T-v)*(T-v)^T

wherein Deltaw represents a parameter update vector of the basic agricultural disaster risk prediction network, T represents a target training disaster influence parameter vector, v represents a target candidate disaster influence parameter vector, and alpha is a learning rate.

The calculation formula of the estimated disaster influence parameters is as follows:

Δz＝(I+β*H(v))^-1*Δw

Wherein Δz represents an update vector of the estimated disaster impact parameter, I represents an identity matrix, β is a regularization coefficient, H (v) represents a kernel matrix of the target candidate disaster impact parameter vector, and Δw represents a parameter update vector of the underlying agricultural disaster risk prediction network.

FIG. 2 illustrates a hardware architecture of the weather monitoring system 100 for implementing the weather data analysis method for agricultural production status monitoring according to the embodiment of the present application, as shown in FIG. 2, the weather monitoring system 100 may include a processor 110, a machine-readable storage medium 120, a bus 130, and a communication unit 140.

In an alternative embodiment, the weather monitoring system 100 may be a single server or a group of servers. The server farm may be centralized or distributed (e.g., the weather monitoring system 100 may be a distributed system). In an alternative embodiment, the weather monitoring system 100 may be local or remote. For example, the weather monitoring system 100 may access information and/or data stored in the machine-readable storage medium 120 via a network. As another example, the weather monitoring system 100 may be directly connected to the machine-readable storage medium 120 to access stored information and/or data. In an alternative embodiment, the weather monitoring system 100 may be implemented on a cloud platform. For example only, the cloud platform may include a private cloud, a public cloud, an aggregated cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, or the like, or any combination thereof.

The machine-readable storage medium 120 may store data and/or instructions. In an alternative embodiment, the machine-readable storage medium 120 may store data acquired from an external terminal. In an alternative embodiment, the machine-readable storage medium 120 may store data and/or instructions that are used by the weather monitoring system 100 to perform or use the exemplary methods described herein. In alternative embodiments, machine-readable storage medium 120 may include mass storage, removable storage, volatile read-write memory, read-only memory, and the like, or any combination thereof. Exemplary mass storage devices may include magnetic disks, optical disks, solid state disks, and the like. Exemplary removable memory may include flash drives, floppy disks, optical disks, memory cards, compact disks, tape, and the like.

In a specific implementation, the plurality of processors 110 execute computer executable instructions stored by the machine-readable storage medium 120, so that the processors 110 may execute the weather data analysis method for monitoring agricultural production status according to the above method embodiment, the processors 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processors 110 may be used to control the transceiving actions of the communication unit 140.

The specific implementation process of the processor 110 may refer to the above-mentioned method embodiments executed by the weather monitoring system 100, and its implementation principle and technical effects are similar, which are not described herein again.

In addition, the embodiment of the application also provides a readable storage medium, wherein computer executable instructions are preset in the readable storage medium, and when a processor executes the computer executable instructions, the weather data analysis method for monitoring the agricultural production state is realized.

Similarly, it should be noted that in order to simplify the description of the present disclosure and thereby aid in understanding one or more embodiments of the application, various features are sometimes grouped together in a single embodiment, figure, or description thereof. Similarly, it should be noted that in order to simplify the description of the present disclosure and thereby aid in understanding one or more embodiments of the application, various features are sometimes grouped together in a single embodiment, figure, or description thereof.

Claims

1. The meteorological data analysis method for agricultural production state monitoring is characterized by comprising the following steps of:

Updating network parameter information of the basic agricultural disaster risk prediction network based on the difference between the target training disaster influence parameters and the target candidate disaster influence parameters to generate an agricultural disaster risk prediction network, and predicting target disaster influence parameters corresponding to a target cultivation strategy according to input agricultural production data and input climate monitoring data based on the agricultural disaster risk prediction network;

The target agricultural production data and the target climate monitoring data are agricultural production data and climate monitoring data corresponding to an xth round of cultivation activities in uninterrupted K rounds of cultivation activities, the target cultivation activities are the xth round of cultivation activities, the target candidate cultivation strategies characterize a mode of scheduling the first agricultural production unit to perform the xth round of cultivation activities in the farmland risk area, the uninterrupted K rounds of cultivation activities are used for enabling the first agricultural production unit to match the risk early warning requirement once, and the target candidate disaster influence parameters corresponding to the target candidate cultivation strategies are determined based on the target agricultural production data and the target climate monitoring data according to a basic agricultural disaster risk prediction network, and the method comprises the following steps:

Outputting a difference between the estimated disaster impact parameter corresponding to the x-th round of cultivation activity and the estimated disaster impact parameter corresponding to the x+1-th round of cultivation activity as the target candidate disaster impact parameter;

the farmland risk area is a disaster concerned area, the risk early warning requirement is that the first agricultural production unit is scheduled to a production area corresponding to a disaster-affected production unit in the farmland risk area, and disaster-affected degree parameters corresponding to the disaster-affected production unit are increased by setting parameters;

2. The agricultural production status monitoring oriented meteorological data analysis method of claim 1, wherein the target candidate farming strategy is determined based on the target agricultural production data and the target climate monitoring data in accordance with the underlying agricultural hazard risk prediction network, the method further comprising:

3. The method for analyzing meteorological data for monitoring agricultural production status according to claim 2, wherein the step of determining the continuous K-wheel farming activities comprises:

4. The method for analyzing weather data for monitoring agricultural production status according to claim 3, wherein the method further comprises:

5. The method of claim 1, wherein the target agricultural production data comprises graph data representing a corresponding cell distribution of production cells included in the agricultural risk area, a first feature vector representing climate monitoring data corresponding to production cells other than the first agricultural production cell, and a sequence feature representing farmland label data corresponding to the agricultural risk area, the target climate monitoring data comprising a second feature vector representing climate monitoring data corresponding to the first agricultural production cell; the map data comprises a plurality of reference farmland identifications and production unit identifications corresponding to production units included in the farmland risk area, and map mapping information of the reference farmland identifications and the production unit identifications in the map data corresponds to position information of the production units included in the farmland risk area;

6. The method for analyzing meteorological data for monitoring agricultural production status according to claim 2, wherein the determining step of the target candidate cultivation strategy comprises:

7. The weather data analysis method for agricultural production status monitoring according to claim 2, wherein the target candidate cultivation policy includes candidate sub-cultivation policies corresponding to a plurality of cultivation categories, respectively, the base agricultural disaster risk prediction network includes a base cultivation sub-network corresponding to the plurality of cultivation categories, respectively, and a base disaster influence sub-network for determining an estimated disaster influence parameter, the updating network parameter information of the base agricultural disaster risk prediction network based on a difference between the target training disaster influence parameter and the target candidate disaster influence parameter, and a difference between the template cultivation policy and the target candidate cultivation policy, generating an agricultural disaster risk prediction network, comprising:

8. A weather monitoring system comprising a processor and a machine-readable storage medium having stored therein machine-executable instructions loaded and executed by the processor to implement the agricultural production status monitoring-oriented weather data analysis method of any one of claims 1-7.