Agricultural disaster prevention and control method based on digital twinning and related equipment

Abstract

The invention provides an agricultural disaster prevention and control method based on digital twinning and related equipment, and is applied to the technical field of data processing. The method comprises the steps of obtaining target farmland area information, a target event set matched with the target farmland area information and a training sample set; preprocessing the training sample set to generate a training sample set with target feature data; processing a preset disaster distributed ecological hydrological model based on the training sample set with the target feature data to generate a target disaster distributed ecological hydrological model; processing the target event set to generate farmland ecological early warning information of the target farmland area; processing the target farmland area information based on the target disaster distributed ecological hydrological model to generate attribute information of the target farmland area; and processing the attribute information of the target farmland area and the farmland ecological early warning information of the target farmland area based on the target disaster distributed ecological hydrological model, and generating farmland early warning information of the to-be-predicted monitoring point.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and related equipment for agricultural disaster prevention and control based on digital twins. Background Technology

[0002] With the abnormal global climate change, agricultural production faces numerous threats from natural disasters and pests, such as drought, waterlogging, and pests. These disasters not only directly damage crops but also significantly impact the stability of agricultural production, food security, and economic benefits. Drought events are particularly frequent, causing severe losses to society, the economy, and the ecosystem. Regional droughts often have global impacts, making drought one of the most widespread natural disasters globally. The severe impact of drought on society, the economy, and the environment has attracted widespread attention worldwide. Furthermore, the lack of effective plans and guidelines for agricultural disaster prevention and control has stalled efforts. Traditional methods of agricultural disaster prevention and control rely heavily on manual experience and on-site inspections, often resulting in untimely responses, incomplete information, and unscientific measures.

[0003] Although some early warning systems and emergency response plans exist, most systems lack intelligent integration, making it impossible to achieve real-time, comprehensive disaster monitoring and efficient handling. Furthermore, there is currently no complete "engineering hospital" platform for agricultural disaster "diagnosis" and "management," resulting in agricultural disaster prevention and control effectiveness being affected by regional and human factors, lacking universality, and failing to provide real-time, efficient, and objective evaluation of prevention and control results.

[0004] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part by practice of the invention.

[0006] According to one aspect of this application, a method for agricultural disaster prevention and control based on digital twins is provided, comprising: acquiring target farmland area information, a target event set matching the target farmland area information, and a training sample set, wherein the target farmland area information includes real-time crop production factor information within the range of the monitoring point to be predicted within the target farmland area, and the training sample set includes historical production factor information of the monitoring point to be predicted within the target farmland area and real-time crop production factor information of other monitoring points adjacent to the monitoring point to be predicted; preprocessing the training sample set to generate a training sample set with target feature data, wherein the target feature data is used to characterize risk factors affecting crop production; and acquiring a pre-set disaster distributed eco-hydrological model, wherein the pre-set... The disaster-distributed eco-hydrological model is assumed to be generated based on historical crop production information of the target farmland area. The preset disaster-distributed eco-hydrological model is processed using a training sample set containing target feature data to generate the target disaster-distributed eco-hydrological model. The target event set is processed to generate farmland ecological early warning information for the target farmland area. Based on the target disaster-distributed eco-hydrological model, the information of the target farmland area is processed to generate attribute information for the target farmland area, including classification information of the monitoring points to be predicted. Based on the target disaster-distributed eco-hydrological model, the attribute information and the farmland ecological early warning information of the target farmland area are processed to generate farmland early warning information for the monitoring points to be predicted.

[0007] Another aspect of this application is a digital twin-based agricultural disaster prevention and control device, which is used to perform the digital twin-based agricultural disaster prevention and control method according to any one of claims 1 to 7.

[0008] According to another aspect of this application, an electronic device is characterized by comprising: a first processor; and a memory for storing executable instructions of the first processor; wherein the first processor is configured to execute the above-described digital twin-based agricultural disaster prevention method by executing the executable instructions.

[0009] According to another aspect of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a second processor, implements the above-described digital twin-based agricultural disaster prevention and control method.

[0010] According to another aspect of this application, a computer program product is provided, comprising a computer program, characterized in that the computer program, when executed by a third processor, implements the above-described digital twin-based agricultural disaster prevention and control method.

[0011] This application provides a method and related equipment for agricultural disaster prevention based on digital twins. The method involves a server acquiring target farmland area information, a target event set, and a training sample set. The target farmland area information covers real-time crop production factors within the area of ​​the monitoring point to be predicted. The training sample set includes historical production factors of the monitoring point to be predicted and real-time crop production factors of adjacent monitoring points. Next, the training sample set is preprocessed, grouped using a pre-defined mapping table, and features are extracted to determine the original feature library. A training set and a validation set are then defined. A classifier and a pre-defined algorithm are used to obtain prediction results and validation set prediction results, thereby generating target feature data characterizing risk factors affecting crop production. Then, the target event set is processed to generate farmland ecological early warning information. First, the target event type information and risk warning level are determined. Then, historical farmland ecological information is generated to obtain crop production influencing factors and weight information. Based on a pre-defined farmland influence feature set, target feature data is generated, ultimately yielding farmland ecological early warning information.

[0012] Furthermore, a distributed eco-hydrological model for target disasters is used to process information about target farmland areas. On one hand, monitoring point information is generated and grouped, features are extracted to obtain abnormal crop production characteristics and co-occurrence frequencies, and correlation information and monitoring point attribute information are generated. On the other hand, the target farmland area information is processed to generate crop production risk factors, production levels, and classification information for each monitoring point. For monitoring points to be predicted, their attribute information and farmland ecological early warning information are processed based on the distributed eco-hydrological model for target disasters. Grouping information for the monitoring points to be predicted, first and second crop production prediction values ​​are generated, and comprehensive processing yields the crop production prediction value. Then, according to preset crop production prediction rules, farmland early warning information is generated, providing comprehensive early warning and prediction support for agricultural disaster prevention and control.

[0013] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0014] Figure 1 A flowchart illustrating an embodiment of an agricultural disaster prevention and control method based on digital twins provided in this application is shown. Figure 2 A schematic diagram of the structure of an agricultural disaster prevention and control device based on digital twins provided in an embodiment of this application is shown; Figure 3 This illustration shows a schematic diagram of the structure of an electronic device according to an embodiment of this application; Figure 4 A schematic diagram of a storage medium provided in one embodiment of this application is shown. Detailed Implementation

[0015] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0016] In one embodiment, this application also proposes an agricultural disaster prevention and control method and related equipment based on digital twins. Figure 1 A schematic flowchart illustrating an agricultural disaster prevention method based on digital twins according to an embodiment of this application is shown. Figure 1 As shown, this method is applied to a server and includes: S101, obtain target farmland area information, target event set matching the target farmland area information, and training sample set.

[0017] In one implementation, the target farmland area information includes real-time crop production factor information within the area of ​​the monitoring point to be predicted within the target farmland area. The training sample set includes historical production factor information of the monitoring point to be predicted within the target farmland area and real-time crop production factor information of other monitoring points adjacent to the monitoring point to be predicted. The following is an example illustrating the target farmland area information and the training sample set: Example of target farmland area information, for real-time agricultural meteorology: Assuming the target farmland area is located in a plain area, real-time agricultural meteorological data includes current weather conditions (e.g., sunny, cloudy, light rain), wind speed (e.g., 3-5 m / s), wind direction (southwind), etc. This meteorological data directly affects crop growth; for example, sunny days are conducive to photosynthesis, while light rain may provide necessary moisture for crops, but strong winds may damage some vulnerable crops.

[0018] Humidity, for example, the humidity level of the monitoring point within the target farmland area is currently 60%. Different crops have different suitable humidity ranges; most vegetables grow best in environments with 60-80% humidity. If humidity remains too high or too low, it may trigger pests and diseases or affect physiological processes such as crop transpiration.

[0019] Temperature, assuming the current temperature is 25℃. Different crops have very different temperature requirements. 25℃ may be very suitable for the growth of some warm-loving crops (such as corn), but it may be too high for some cold-resistant crops (such as wheat during the winter).

[0020] Soil moisture is expressed as relative soil water content, such as 40% in the current area. Soil moisture directly affects the absorption of water by crop roots. If soil moisture is insufficient (e.g., below 30%), crops may grow slowly due to water shortage; conversely, if it is too high (e.g., above 80%), it may lead to poor soil aeration, affecting root growth.

[0021] Soil fertility is assumed to be measured by the content of major nutrients such as nitrogen, phosphorus, and potassium in the soil. For example, the soil may contain 0.15% nitrogen, 0.1% phosphorus, and 0.2% potassium. Different crops have different fertility requirements at different growth stages. Insufficient fertility can lead to poor crop growth and reduced yield, while excessive fertility may cause environmental pollution or seedling burn.

[0022] Soil pH, assuming the soil pH in this area is 6.5. Most crops thrive in neutral to slightly acidic or slightly alkaline soils; for example, a pH between 6 and 7.5 is suitable for most grain crops and vegetables. If the soil pH is too high or too low, it will affect the availability of nutrients in the soil, thus affecting crop absorption and growth.

[0023] Irrigation rate: Assuming the recent irrigation rate for this area is 10 cubic meters per acre per week. The appropriate irrigation rate depends on factors such as crop type, growth stage, soil moisture, and weather conditions. Insufficient irrigation leads to drought, while excessive irrigation not only wastes water resources but may also cause secondary soil salinization and other problems.

[0024] Remote sensing imagery, such as satellite imagery, can provide information on vegetation cover and crop color in a target farmland area. If remote sensing imagery shows low vegetation cover, it may indicate poor crop growth or problems such as pests and diseases; yellowing crops may suggest a lack of certain nutrients or drought stress.

[0025] Crop height, assuming the average current height of a certain crop (such as rice) is 30 centimeters, is an important indicator for measuring crop growth. Different growth stages have their corresponding normal growth height ranges, and abnormal growth height may reflect adverse factors affecting the crop, such as insufficient sunlight or nutrient deficiency.

[0026] Before the harvest season, there may be some estimated yield data for crops. For example, based on factors such as previous growth conditions and planting density, the estimated wheat yield for a certain field might be 400 kg per acre. Although this data is an estimate, it can be compared with the actual harvest yield to analyze the factors affecting the yield.

[0027] The training sample set is shown below, containing historical production factor information for the monitoring point to be predicted. Taking a specific monitoring point as an example, we review the historical data from the past 5 years: Meteorological data: Average temperature, precipitation, wind speed, etc., for each season over the past 5 years. For example, the average summer temperature in the first year was 28℃, precipitation was 300 mm, and wind speed was 2-4 m / s; the average winter temperature was 5℃, and precipitation was 50 mm, etc. Soil-related data: Changes in soil fertility each year, such as the nitrogen content in the soil being 0.12% in the first year, which changed to 0.1% in the second year due to fertilization and crop absorption; the soil pH value may have fluctuated slightly over the past 5 years, ranging from 6.3 to 6.7. Crop growth data: Average crop height and final yield each year. For example, the average growth height of a certain crop in the first year was 25 cm, with a final yield of 350 kg per mu; with improvements in planting techniques and environmental changes, the average growth height reached 35 cm in the fifth year, and the yield increased to 450 kg per mu.

[0028] Real-time crop production factor information from adjacent monitoring points: Assume there are three monitoring points adjacent to the monitoring point to be predicted: Monitoring point A: Real-time humidity 55%, temperature 24℃, relative soil moisture 35%, fertility (nitrogen 0.14%, phosphorus 0.09%, potassium 0.18%), soil pH 6.4, recent irrigation amount 8 cubic meters / acre per week. Remote sensing imagery shows 80% vegetation cover and crop height 28 cm. Monitoring point B: Humidity 62%, temperature 26℃, relative soil moisture 42%, fertility (nitrogen 0.16%, phosphorus 0.11%, potassium 0.22%), soil pH 6.6, irrigation amount 12 cubic meters / acre per week. Remote sensing imagery shows 85% vegetation cover and crop height 32 cm. Monitoring point C: Humidity 60%, temperature 25℃, soil moisture relative content 38%, fertility nitrogen content 0.15%, phosphorus content 0.1%, potassium content 0.2%, soil pH 6.5, irrigation rate 10 cubic meters / acre per week. Remote sensing imagery shows vegetation cover 82% and crop height 30 cm. These example data illustrate the specific details of the target farmland area information and the content contained in the training sample set, contributing to an understanding of their role in crop production prediction and monitoring.

[0029] S102, preprocess the training sample set to generate a training sample set with target feature data.

[0030] In one implementation, a preset mapping table is obtained, wherein the preset mapping table is used to characterize the correspondence between the historical crop production factor information of the target farmland area and the historical crop production factor information of the monitoring points to be predicted within the target farmland area. Assume that the target farmland area is a mixed area including mountains and plains, and the crop grown is wheat. The historical crop production factor information of the target farmland area includes meteorological aspects: average temperature, precipitation, and sunshine duration in different seasons over the past 10 years. For example, the average spring temperature in mountainous areas is 12-15°C, precipitation is 100-150 mm, and sunshine duration is 8-10 hours; the average spring temperature in plains areas is 15-18°C, precipitation is 80-120 mm, and sunshine duration is 10-12 hours. Soil characteristics: Soil type (mostly loam in mountainous areas, sandy loam in plains), soil fertility (initial nitrogen content in mountainous soils is 0.1%, phosphorus content is 0.08%, and potassium content is 0.15%; nitrogen content in plains soils is 0.12%, phosphorus content is 0.1%, and potassium content is 0.18%), soil pH (pH of mountainous soils is 6.5-7.0, and pH of plains soils is 7.0-7.5). Irrigation: Mountainous areas rely heavily on natural rainfall, with low and irregular irrigation amounts; plains have irrigation systems, with an average irrigation volume of 5-10 cubic meters per acre per week.

[0031] Historical crop production factor information for monitoring points to be predicted. Assume there are three monitoring points: one located at a higher altitude in a mountainous area (Monitoring Point A), one at a lower altitude in a mountainous area (Monitoring Point B), and one located in a plain (Monitoring Point C). Monitoring Point A: Over the past 10 years, the average spring temperature was 12-13°C, precipitation was 120-150 mm, and sunshine duration was 8-9 hours; the soil is loam, with nitrogen content of 0.09%, phosphorus content of 0.07%, potassium content of 0.14%, and a pH value of 6.5; irrigation was low. Monitoring Point B: The average spring temperature was 14-15°C, precipitation was 100-120 mm, and sunshine duration was 9-10 hours; the soil is loam, with nitrogen content of 0.1%, phosphorus content of 0.08%, potassium content of 0.15%, and a pH value of 6.8; irrigation was low. Monitoring point C: Average spring temperature is 16-18°C, precipitation is 80-100 mm, and sunshine duration is 11-12 hours; the soil is sandy loam with a nitrogen content of 0.12%, phosphorus content of 0.1%, potassium content of 0.18%, and a pH of 7.2; irrigation rate is 8 cubic meters per acre per week. A pre-set mapping table will establish the correspondence between these information, for example, mapping the meteorological, soil, and irrigation conditions in mountainous areas to the historical production factor information of the monitoring points to be predicted (A and B) located in mountainous areas, and mapping the conditions in plains areas to the information of monitoring point C. This correspondence can be established using factors such as geographical coordinates, altitude range, and soil type.

[0032] The training sample set is grouped based on a pre-defined mapping table to generate a grouped training sample set. This grouped training sample set includes monitoring points at different altitudes to be predicted. According to the pre-defined mapping table, the training sample set is grouped by altitude: High-altitude group: Includes monitoring point A located at a higher altitude in mountainous areas and its related historical production factor information, such as meteorological, soil, and irrigation data for the region, as well as real-time crop production factor information for adjacent monitoring points in high-altitude areas. Low-altitude group: Includes monitoring point B and its related data, as well as information on adjacent monitoring points in low-altitude mountainous areas. Plain group: Includes monitoring point C and its related data, as well as information on adjacent monitoring points in plain areas. Feature extraction is performed on the grouped training sample sets to determine the original feature library. Features related to crop production are extracted from the grouped training sample sets. Taking the high-altitude group as an example, the extracted features may include: temperature range (12-13°C), precipitation (120-150 mm), sunshine duration (8-9 hours), soil type (loam), soil fertility (nitrogen 0.09%, phosphorus 0.07%, potassium 0.14%), soil pH (6.5), and irrigation amount (low). These features constitute the original feature library for the high-altitude group. Similarly, the low-altitude group and the plain group will each construct their own original feature libraries.

[0033] The original feature library is divided into several feature datasets to generate training and validation sets. Assume the original feature library contains 100 data samples (each containing the aforementioned feature values). For the high-altitude group, the original feature library is divided into training and validation sets in a certain ratio (e.g., 80:20). That is, 80 samples are used as the training set to train the model, and 20 samples are used as the validation set to evaluate the model's performance. A similar division is performed for the low-altitude and plain groups. A classifier is used to predict the results of each validation set based on the original feature library. A classifier, such as a decision tree classifier, is selected. For the high-altitude group's validation set (20 samples), the sample features are input into the decision tree classifier. The classifier classifies and predicts each sample based on previously learned patterns (e.g., based on samples in the training set, features such as low temperature, high precipitation, and low soil fertility may be associated with low yield). For example, predicting whether a sample corresponds to a high, medium, or low crop yield, or predicting the presence of a certain pest or disease risk, these predictions constitute the prediction results for the high-altitude group's validation set. Similarly, predictions were made on the validation sets of the low-altitude group and the plain group, and the corresponding prediction results were obtained.

[0034] The original feature library is divided into training sets using a pre-defined algorithm for training, and the validation set prediction results are obtained. The Support Vector Machine (SVM) algorithm is selected, with the core adaptation objective: to construct a binary classification model (distinguishing between "high-yield" and "low-yield" samples) based on the training sets of different altitude groups, providing a predictive basis for subsequent extraction of crop production risk features. Input training set data for each group (high altitude, low altitude, plains), with 80 samples in each group. Each sample contains target feature dimensions (meteorology: temperature, precipitation; soil: fertility, soil moisture, pH; crop: growth height). Label definition: Sample labels are binary classification variables, where "1" represents high yield (more than 10% above the group average yield) and "0" represents low yield (less than 10% below the group average yield). The radial basis function (RBF) is selected as the kernel function, with initial parameters set as follows: penalty coefficient C = 1.0, kernel function parameter γ = 0.1, and fault tolerance ε = 0.001.

[0035] Training is conducted in groups, with the high-altitude group as an example; the low-altitude and plain groups follow the same process. Z-score standardization is applied to the feature data of the training set to eliminate dimensional differences. The model training iterations are as follows: 80 standardized training samples and their corresponding labels are input. The SVM algorithm finds the optimal hyperplane w·x+b=0 (where w is the feature weight vector and b is the bias term) by maximizing the sample margin. Based on training error feedback, C and γ are dynamically optimized. The iteration stops when the cross-validation accuracy on the validation set improves by ≤0.5% for three consecutive iterations, determining the optimal parameters (e.g., the final parameters are C=2.0 and γ=0.08).

[0036] The trained SVM model is applied to 20 validation set samples. The input is the standardized feature data of the validation set, and the output is the "high-yield / low-yield" predicted label for each sample, i.e., the validation set prediction result for the high-altitude group. The low-altitude and plain groups follow the same "standardization → parameter iteration training → validation set prediction" process, adjusting the optimal parameters only based on their respective training set data (e.g., final parameters for the plain group: C=1.5, γ=0.12). All group models are trained using 5-fold cross-validation to ensure model generalization ability and avoid overfitting. Each group outputs a list of predicted labels for 20 validation set samples (e.g., high-altitude group: [1,0,1,...,0]), which is compared with the classifier's direct prediction results for subsequent extraction of target feature data (risk factors).

[0037] Based on the prediction results and the validation set prediction results, target feature data is generated to characterize the risk factors affecting crop production. The prediction results of the high-altitude group and the validation set prediction results are compared. For example, if a decision tree classifier predicts a sample as high-yield, but the SVM algorithm predicts medium-yield, this may indicate that some influencing factors are causing the difference in results. Features related to these differences are analyzed; for example, it is found that the prediction results of the two algorithms are inconsistent when temperature fluctuations are large and soil fertility is within a certain range. These features related to the differences can be included as part of the risk factors affecting crop production in the target feature data. Similarly, similar analyses are performed on the low-altitude group and the plain group. The results of the three groups are combined to finally generate target feature data to characterize the risk factors affecting crop production. For example, the target feature data may include: temperature fluctuation range (above 5°C), soil fertility below a certain threshold (nitrogen content below 0.1%), and uneven precipitation (coefficient of variation greater than 0.3) as risk factors affecting crop production.

[0038] S103, Obtain the preset disaster distributed eco-hydrological model.

[0039] In one implementation, a pre-defined distributed eco-hydrological model (distributed coupled type) is used, which integrates three core functions: ecological simulation, hydrological cycle, and disaster risk assessment. It is specifically designed for wheat and corn producing areas in river alluvial plains. Focusing on disasters such as floods, droughts, low-temperature freezing damage, and high-temperature heat damage in this region, it enables risk prediction and quantitative assessment of yield impact. It is constructed based on 10 years of historical data of the target farmland area, covering core data such as meteorological (precipitation, temperature), soil (texture, fertility), crops (yield, growth status), and disaster occurrence records.

[0040] The model adopts a three-layer distributed architecture of "region-unit-factor," with hierarchical coupling and independent modeling to ensure accurate adaptation to different regional differences. At the region level (top layer): the target alluvial plain farmland is divided into 100m×100m distributed grid units. Each unit serves as an independent modeling unit, with built-in differentiated attribute labels such as topography (elevation), soil type (loam / sandy soil), and crop type (wheat / corn), achieving refined regional division. At the unit level (middle layer): each grid unit contains three parallel sub-modules, each responsible for calculations and analyses in different dimensions. These modules interact through data interfaces: The meteorological disaster sub-module specifically handles precipitation and temperature-related disasters, including the identification and risk quantification of floods, droughts, low-temperature freezing damage, and high-temperature heat damage. The soil environment sub-module focuses on the dynamic simulation of soil texture, fertility, and water and fertilizer retention capacity, providing basic soil data support for disaster risk assessment. The crop response sub-module quantifies the specific impact of disasters on crop growth and yield based on crop type (wheat / corn) and growth stage. For the factor level (bottom layer): each sub-module contains several core influencing factors and corresponding parameter systems. Data linkage between factors is achieved through correlation algorithms. For example, soil water retention capacity factor and drought risk factor are directly coupled for calculation to ensure the consistency and accuracy of model calculation.

[0041] The information for the meteorological disaster sub-module is as follows: Flooding: The core influencing factors are short-term precipitation and soil drainage capacity. The short-term precipitation threshold is set at 20 mm / h, and the soil drainage capacity coefficient is differentiated according to soil type (0.7 for loam and 0.9 for sandy soil). When the short-term precipitation within a unit exceeds the threshold and the drainage capacity coefficient is below 0.8, it is considered a flood risk, with the risk level increasing with the extent of precipitation exceeding the threshold. Drought: The core influencing factors include the number of consecutive days without effective rainfall, soil water retention capacity, and crop water requirements. The threshold for no effective rainfall is defined as <5 mm / day, and the critical number of days is 15 days. Water requirements are also set according to crop growth stages (4 mm / day for wheat during the jointing stage and 6 mm / day for corn during the tasseling stage). When the number of consecutive days without effective rainfall exceeds the critical number, and the soil water retention capacity cannot meet the average daily water requirements of crops, a drought risk warning is triggered. Low-temperature freezing damage (wheat-specific): The core influencing factors are the lowest winter temperature and its duration. The critical temperature for freezing damage is set at -10℃, and the threshold for severe freezing damage is -15℃ (lasting ≥5 days). Risk assessment is initiated when the temperature falls below the critical temperature. When severe freezing damage conditions are met, the yield loss coefficient is set at 0.5. High-temperature heat damage (maize-specific): The core influencing factors are the highest summer temperature and the number of consecutive days of high temperatures. The critical temperature for heat damage is set at 35℃, and the critical duration is 10 days. When the highest temperature is ≥35℃ for 10 consecutive days, it is considered high-temperature heat damage, and the pollen abortion rate coefficient is set at 0.3, directly related to the decrease in seed setting rate.

[0042] The information for the soil environment submodule is as follows: Soil texture: mainly includes two types: loam (80%) and sandy soil (20%). The water and fertilizer retention capacity coefficient is set at 0.8 for loam and 0.4 for sandy soil. This coefficient directly affects the drainage capacity calculation and drought tolerance assessment. Soil fertility: with nitrogen content (0.1%-0.15%), phosphorus content (0.1%), and potassium content (0.2%) as core indicators, the comprehensive fertility is calculated using the formula "Comprehensive fertility coefficient = 0.5 × nitrogen content + 0.3 × phosphorus content + 0.2 × potassium content". When the coefficient < 0.12, the crop growth rate is judged to have decreased by 20%. Soil water retention capacity: based on the dynamic calculation of soil texture and fertility, the basic water retention capacity of loam is 15 mm / day, and that of sandy soil is 8 mm / day; for each decrease in fertility level, the water retention capacity coefficient is reduced by 10%-20%.

[0043] The information for the crop response submodule is as follows: Wheat: Key growth stages include overwintering (December-February), jointing (March-April), and grain-filling (May). The yield loss coefficients are set at 0.4-0.6 for flooding, 0.6-0.8 for drought, and 0.3-0.5 for frost damage. These loss coefficients are dynamically adjusted according to the severity of the disaster. Maize: Key growth stages include jointing (June-July), tasseling (August), and grain-filling (September). The yield loss coefficients are set at 0.3-0.5 for high-temperature heat damage, 0.4-0.6 for drought, and 0.2-0.3 for insufficient fertility. These loss coefficients are strongly correlated with the disaster sensitivity of each growth stage. Basic yield parameters: The basic yield for wheat is set at 400-500 kg / mu, and for maize at 500-600 kg / mu. Actual yield is calculated using the formula "Actual yield = Basic yield × (1 - Weighted sum of all disaster loss coefficients)", with weights allocated according to the priority of disaster impact.

[0044] Real-time meteorological data (precipitation, temperature), soil data (fertility, humidity), and crop growth stage information are synchronously input into each distributed grid cell to ensure that the data covers all dimensions of the model's computational needs. Three submodules operate simultaneously: the meteorological disaster submodule determines the risk level, the soil environment submodule outputs soil carrying capacity, and the crop response submodule matches the disaster tolerance of the corresponding growth stage. The calculation results from the three submodules are integrated through a coupled algorithm to quantify the probability and magnitude of crop yield loss within the cell, ultimately outputting a regional disaster risk distribution map and yield predictions, providing precise data support for agricultural disaster prevention and control.

[0045] S104. Based on the training sample set with target feature data, the preset disaster distributed eco-hydrological model is processed to generate the target disaster distributed eco-hydrological model.

[0046] In one implementation, the coefficient of variation of rainfall intensity, reflecting the spatiotemporal unevenness of rainfall distribution, is set with a threshold of 0.5. When the value exceeds the threshold, the area is considered to have a high risk of localized flooding or drought, and this needs to be included in the key optimization dimensions of the model. Temperature matching features focus on the compatibility between the number of consecutive high-temperature days and the rice growth stage. Using the heading stage as a key node, a standard of ≥35°C for five consecutive days is set (calibrated based on historical yield impact data). Meeting this condition is considered a high-temperature risk feature. Soil fertility gradient refers to quantifying soil fertility through a comprehensive index of nitrogen, phosphorus, and potassium content. A gradient change threshold of 5% decrease per 100 meters is set. When this threshold is exceeded, the uniformity of rice growth is affected, and this is defined as risk characteristic data.

[0047] Soil drainage capacity was determined by combining soil texture (sandy loam, clay, etc.) and soil structure (porosity). A critical threshold of ≤5 cm drainage in 24 hours was set. Areas below this threshold are prone to waterlogging during periods of heavy rainfall, and this was used as a core feature data point. The difference in altitude was quantified by the rate of change in rice yield corresponding to every 10 meters of altitude change. When the altitude difference led to significant differences in microclimates such as temperature and humidity, it was included as a key feature in the model optimization.

[0048] The initial structure and parameters of the pre-defined disaster-distributed eco-hydrological model are as follows. The model uses 100m×100m grid cells to divide the target farmland area. Each cell independently handles data calculations. The core model includes three parallel modules: a meteorological module, a soil module, and a topographic module. These modules interact in real time through a data interface. For the meteorological module, the initial average rainfall rate is 100 mm per month, and the evaporation rate is 50 mm per month. Parameters related to rainfall unevenness are not yet set. For the soil module, physical property parameters are configured according to soil type: porosity is set to 0.4 for sandy loam and 0.3 for clay. Parameters related to dynamic changes in fertility and drainage capacity thresholds are not included. For the topographic module, the initial elevation is entered based on measured data, and the maximum slope parameter is set to 10°. No linkage calculation relationship between elevation and temperature / humidity has been established.

[0049] The meteorological module has been optimized. For areas with a rainfall intensity variation coefficient > 0.5, a new "rainfall unevenness parameter" (range 0.1-0.9, positively correlated with the variation coefficient) has been added and embedded in the rainfall prediction submodule. This parameter dynamically adjusts the accuracy of rainfall intensity simulation in local areas, precisely adapting to flood and drought risk predictions. Regarding high-temperature risk characteristics, the temperature threshold parameter has been optimized: abandoning the uniform 35°C standard, the threshold standard is subdivided based on regional differences (low / high altitude, orientation) and rice growth stage. For example, the threshold for the heading stage in low-altitude, south-facing areas is adjusted to 33°C, while in high-altitude areas it remains at 35°C. Precise adaptation is achieved through a three-dimensional correlation algorithm of growth stage-region-threshold.

[0050] The soil module has been optimized by adding a "fertility loss factor" (value 0.02-0.08, positively correlated with the gradient descent rate) to the soil fertility dynamic change equation based on soil fertility gradient characteristics. This factor is automatically triggered when a fertility decrease of more than 5% per 100 meters is detected, correcting the fertility decay simulation curve to better reflect actual growth patterns. For soil drainage capacity characteristics, the permeability coefficient in the soil moisture movement equation has been adjusted: the baseline permeability coefficient for sandy loam is set at 0.05 cm / s, and for clay at 0.01 cm / s. When 24-hour drainage is ≤5 cm, the permeability coefficient is corrected by a factor of 0.8, improving the accuracy of waterlogging scenario simulation.

[0051] The terrain module is optimized by establishing a correlation calculation model between altitude, temperature, and humidity based on the altitude difference characteristics. An "altitude temperature lapse rate correction parameter" (base value 0.6°C / 100 meters, dynamically adjusted according to the yield change rate) is introduced. For example, when the yield change rate corresponding to every 10 meters of altitude exceeds 3%, the lapse rate parameter is increased by 0.1°C / 100 meters to accurately reflect the impact of microclimate on rice growth.

[0052] The optimized target model, within a 100m×100m grid cell, can simultaneously integrate real-time meteorological, soil, and topographic data (including rainfall unevenness, fertility dynamics, and altitude-related parameters) to achieve collaborative calculation of multi-dimensional risk factors. The model input data includes real-time meteorological data (rainfall, temperature), soil data (fertility, drainage capacity), topographic data (altitude, slope), and rice growth stage information. Through the optimized parameter system and correlation algorithms, it outputs the probability of rice disaster occurrence, disaster severity, and yield impact prediction within the cell, directly addressing the operational needs of precision farmland management and disaster early warning, and providing data support for the formulation of prevention and control measures.

[0053] S105, process the target event set to generate farmland ecological early warning information for the target farmland area.

[0054] In one implementation, the target event set is processed to generate type information for the target events. This type information characterizes the risk warning level of the target events for the target farmland area. Assuming the target farmland area primarily grows wheat, the target event set includes various events that may affect wheat production. Examples include: Heavy rain events: Rainfall exceeding 50 mm / day within a certain period may lead to waterlogging in the farmland, affecting wheat root respiration and nutrient absorption. Drought events: No effective rainfall for more than 15 consecutive days (daily rainfall less than 5 mm), resulting in a severe drop in soil moisture and causing water shortage for wheat growth. Low-temperature freezing damage events: A sudden drop in winter temperature below -10°C for an extended period may damage wheat seedlings. Wheat rust outbreak: Numerous rust-colored lesions appear on wheat leaves, spreading rapidly and affecting wheat photosynthesis and nutrient synthesis. Aphid infestation: Large numbers of aphids congregate on wheat plants, sucking sap, causing stunted growth and yellowing leaves.

[0055] For rainstorm events, if the rainfall is 50-100 mm / day and the farmland drainage system is good, it is defined as low risk (risk warning level 1); if the rainfall exceeds 100 mm / day, or the farmland drainage is poor, it is defined as high risk (risk warning level 3); and between these two is medium risk (risk warning level 2). For drought events, 15-25 consecutive days without effective rainfall is medium risk (level 2), and more than 25 days is high risk (level 3). For low-temperature freezing damage events, temperatures dropping to -10°C to -15°C for 3-5 days is medium risk (level 2), and temperatures below -15°C for more than 5 days is high risk (level 3). For wheat rust outbreaks, a disease incidence rate of 10%-30% is medium risk (level 2), and more than 30% is high risk (level 3). For aphid infestations, an average of 10-20 aphids per wheat plant is medium risk (level 2), and more than 20 aphids per plant is high risk (level 3).

[0056] The type information of the target event is processed to generate historical farmland ecological information for the target farmland area. Taking rainstorm events as an example, if in the past 10 years, the farmland area experienced 3 high-risk rainstorm events (Level 3), 5 medium-risk (Level 2), and 2 low-risk (Level 1) rainstorm events, and each high-risk rainstorm event resulted in an average decrease of 20% in wheat yield, a 10% decrease after a medium-risk event, and a 5% decrease after a low-risk event, this historical data constitutes part of the historical farmland ecological information regarding the impact of rainstorm events on the target farmland area. For drought events, assuming that in the past 5 years, there were 2 high-risk drought events (Level 3), each resulting in a 30% decrease in wheat yield; and 3 medium-risk (Level 2) drought events, resulting in a 15% decrease in yield, this also constitutes historical farmland ecological information related to drought events. For pest and disease events, such as wheat rust, in the past 8 years, there were 2 high-risk outbreaks (Level 3), reducing wheat yield by 35%; and 4 medium-risk (Level 2) outbreaks, reducing yield by 20%. Regarding aphid infestations, there were 3 high-risk (level 3) infestations, each resulting in a 25% decrease in yield, and 4 medium-risk (level 2) infestations, each resulting in a 15% decrease in yield. These are all components of the historical farmland ecological information for the target farmland area.

[0057] Historical farmland ecological information of the target farmland area is processed to generate crop production influencing factors and their corresponding weights. These factors can be identified from the historical farmland ecological information. For example, for wheat production, heavy rain, drought, low-temperature freezing damage, wheat rust, and aphid infestation are all influencing factors. When calculating weights, these factors can be determined based on their average impact on yield. Assuming that statistical analysis shows drought has the highest average impact on yield (30%, hypothetical value), the weight of drought as an influencing factor can be set to 0.3. Wheat rust has an average impact on yield of 25% (hypothetical value), and its weight is set to 0.25. Heavy rain has an average impact on yield of 15% (hypothetical value), and its weight is set to 0.15. Low-temperature freezing damage has an average impact on yield of 10% (hypothetical value), and its weight is set to 0.1. Aphid infestation has an average impact on yield of 10% (hypothetical value), and its weight is set to 0.1.

[0058] Based on a pre-defined farmland impact feature set, the influencing factors of crop production and their corresponding weights are processed to generate target feature data. It is assumed that the pre-defined farmland impact feature set includes features related to meteorology, pests and diseases, and soil fertility. Taking meteorology as an example, influencing factors such as heavy rain, drought, and low-temperature freezing damage, along with their weights, are integrated with meteorological features in the pre-defined farmland impact feature set. If the pre-defined farmland impact feature set includes features related to the seasonal distribution of rainfall, the rainfall-related portion of the target feature data is adjusted based on a weight of 0.15 for heavy rain events. For example, if the seasonal distribution of rainfall shows a high probability of heavy rain in a certain season, the risk assessment value for wheat production in that season is increased based on a weight of 0.15. Regarding pests and diseases, target feature data is generated based on the weights of wheat rust and aphid infestation (0.25 and 0.1, respectively), combined with features in the pre-defined farmland impact feature set regarding pest and disease transmission pathways and control difficulty. For example, if pest and disease transmission pathways are easy and control is difficult, the pest and disease risk value in the target feature data is further increased based on the weights.

[0059] The target feature data is processed to generate farmland ecological early warning information for the target farmland area. Assume that after integration, the target feature data yields a comprehensive risk assessment value. If this value is between 0 and 0.3, a low-risk farmland ecological early warning is generated, prompting farmers to conduct routine farmland management. If the comprehensive risk assessment value is between 0.3 and 0.6, it is a medium-risk warning, recommending farmers to strengthen monitoring, such as increasing the frequency of pest and disease inspections and paying attention to soil moisture. When the comprehensive risk assessment value exceeds 0.6, it is a high-risk warning, requiring farmers to take emergency measures, such as timely irrigation (to address drought risk), pesticide spraying (to address pest and disease risk), or preparing for drainage in case of impending heavy rain.

[0060] S106, Based on the target disaster distributed eco-hydrological model, the information of the target farmland area is processed to generate the attribute information of the target farmland area.

[0061] In one implementation, information about the target farmland area is processed based on a distributed eco-hydrological model of the target disaster to generate monitoring point information, where the monitoring points are located within the target farmland area. It is assumed that the target farmland area is a mixed terrain region including mountains and plains, where maize is grown. The target farmland area information includes: meteorological data: temperature at different locations: average temperature in mountainous areas is 15-20°C, and average temperature in plains is 20-25°C. Precipitation: annual precipitation in mountainous areas is 800-1000 mm, and annual precipitation in plains is 600-800 mm. Soil data: mountainous soils are mostly loam, with fertility levels of 0.12% nitrogen, 0.08% phosphorus, and 0.15% potassium; plains soils are sandy loam, with 0.1% nitrogen, 0.09% phosphorus, and 0.13% potassium. Crop growth data: average maize plant height in mountainous areas is 1.5-2 meters, and average maize plant height in plains is 2-2.5 meters. Multiple monitoring points were set up within the target farmland area. For example, monitoring point 1 was located at an altitude of 800 meters in the mountainous area, monitoring point 2 was located at an altitude of 500 meters in the mountainous area, and monitoring point 3 was located at an altitude of 100 meters in the plain. The information collected at each monitoring point included local temperature, precipitation, soil fertility, and corn plant height.

[0062] The monitoring point information is grouped to generate several groups, including information on monitoring points at different altitudes. The monitoring points are divided into different groups based on altitude. For example, monitoring points at altitudes of 500-1000 meters are classified as the high-altitude group, which includes monitoring point 1; those at altitudes of 200-500 meters are classified as the mid-altitude group (assuming such monitoring points exist); and those at altitudes of 0-200 meters are classified as the low-altitude group, which includes monitoring point 3. Feature extraction is performed on the grouped information to generate feature datasets for different groups. High-altitude group (taking monitoring point 1 as an example): Meteorological characteristics: average temperature 15°C, annual precipitation 1000 mm. Soil characteristics: loam, nitrogen content 0.12%, phosphorus content 0.08%, potassium content 0.15%. Crop characteristics: corn plant height 1.5 meters. These characteristics constitute the feature dataset for the high-altitude group. Low-altitude group (taking monitoring point 3 as an example): Meteorological characteristics: average temperature 25°C, annual precipitation 600 mm. Soil characteristics: Sandy loam, nitrogen content 0.1%, phosphorus content 0.09%, potassium content 0.13%. Crop characteristics: Maize plant height 2.5 meters. This forms the characteristic dataset for the low-altitude group.

[0063] The feature datasets of different groups were processed to obtain abnormal crop production characteristics and their co-occurrence frequencies. In the high-altitude group, if the corn plant height was generally lower than the normal growth range for this variety (assuming the normal range is 1.8-2.2 meters), then "corn plant height below the normal range" is an abnormal crop production characteristic. In the low-altitude group, if the corn yield was significantly lower than expected (by comparing with historical yield data), then "low yield" is an abnormal crop production characteristic. Taking the high-altitude group as an example, consider the events of "low temperature" (temperature below 18°C) and "corn plant height below the normal range". Assuming that in the past 10 observations, "low temperature" and "corn plant height below the normal range" occurred simultaneously in 6 instances, then the co-occurrence frequency is 6 / 10 = 0.6.

[0064] The abnormal crop production characteristics and their co-occurrence frequencies were processed to generate correlation information, which characterizes the relationship between crop production characteristics at different altitudes and the actual farmland conditions at the target monitoring points. Based on the calculated co-occurrence frequency of 0.6 and the abnormal crop production characteristic "corn plant height below the normal range," it can be concluded that in high-altitude areas, low temperature and corn plant height below the normal range are strongly correlated. This is the correlation information for the high-altitude group. It may also be found that soil nitrogen content is correlated with corn plant height; for example, when nitrogen content is below 0.11%, corn plant height is more likely to be below the normal range, further enriching the correlation information for the high-altitude group. Assuming that in the low-altitude group, the co-occurrence frequency of "high temperature (above 28°C)" and "low yield" is 0.7, this indicates a strong correlation between high temperature and low yield in low-altitude areas. This is the correlation information for the low-altitude group. Based on the correlation information, the feature datasets of different groups are processed to generate attribute information for the monitoring points corresponding to different groups. For monitoring point 1 (located in the high-altitude group), based on correlation information, its attribute information may include: susceptibility to low temperatures leading to shorter maize plant height; and the nitrogen content in soil fertility having a significant impact on maize plant height. For monitoring point 3 (located in the low-altitude group), the attribute information may include: high temperatures may lead to reduced yield; and the impact of temperature on yield needs to be monitored.

[0065] Furthermore, the method includes a formula for calculating the co-occurrence frequency of correlations, which is as follows: F(A,B) ;in, is the number of times events A and B co-occur in the group with altitude i, n is the total number of altitude groups, and N is the total number of samples.

[0066] Taking the aforementioned farmland area as an example, but now considering the impact of different altitudes, we divide the area into two altitude groups: low altitude (plains) and high altitude (hilly areas). Events A and B remain unchanged: Event A: Crops suffer from flooding; Event B: Crop yield is less than 80% of the average yield. The total sample size remains the statistical result of production data from the past 10 years. We assign a weight to the low altitude area. =0.6 (because the planting area in low-altitude areas is larger, and its impact on the overall results is relatively greater), weight of high-altitude areas =0.4.

[0067] In low-altitude areas, events A and B occurred simultaneously in two out of the past 10 years, therefore =2. In high-altitude areas, events A and B occurred simultaneously in one of the past 10 years, therefore... =1. According to the formula F(A,B) Here, n=2 (two altitude groups), N=10. =0.6, =2, =0.4, =1. F(A,B) This indicates that, after considering the weights of different altitudes, the frequency of both crop flooding and yields falling below 80% of the average yield occurring simultaneously is 0.16, or 16%.

[0068] In another implementation, the target farmland area information is processed to generate several monitoring points. These monitoring points include the predicted monitoring points located in the target farmland area information and other monitoring points. Assume the target farmland area is a large, integrated farm cultivating multiple crops. The terrain includes plains and gentle slopes, with loam and sandy loam soil types. Irrigation water comes from nearby rivers and groundwater wells. Meteorological data shows that the average annual precipitation in the area is 800 mm, unevenly distributed seasonally, with concentrated rainfall in summer and less in winter. Based on the area and layout of the farmland, multiple monitoring points are set up. For example, monitoring point A is set up in the plain area near the river, monitoring point B is set up in the middle of the gentle slope, and monitoring point C is set up in the sandy loam area close to the groundwater well, etc. Monitoring point A can be designated as the predicted monitoring point, while monitoring points B and C are designated as other monitoring points. The crop usage information corresponding to each monitoring point is obtained. Monitoring point A: wheat is grown, and the wheat is mainly used to make flour, supplying local bread processing plants and flour mills. Monitoring Point B: Corn is grown, with some used as animal feed for nearby farms and the rest used for corn starch production. Monitoring Point C: Vegetables (e.g., tomatoes) are grown, primarily supplying local farmers' markets and supermarkets.

[0069] Based on the distributed eco-hydrological model of the target disaster, several monitoring points were processed to generate crop production risk factors for each point. Monitoring Point A (Wheat): Meteorological risk factors: Due to concentrated summer rainfall, there is a risk of flooding, affecting the grain-filling stage of wheat; low winter rainfall and drought may affect wheat overwintering and greening. Soil risk factors: Loam soil has good water retention, but improper irrigation may lead to soil waterlogging or salt accumulation. Hydrological risk factors: Located near a river, flooding may inundate farmland and damage wheat crops. Monitoring Point B (Corn): Meteorological risk factors: Heavy summer rainfall may cause corn lodging; drought may affect corn jointing and tasseling. Soil risk factors: Soils on gentle slopes are prone to soil erosion; sandy loam soil has relatively weak fertility retention capacity, which may affect the nutrient supply required for corn growth. Hydrological risk factors: During the dry season, a drop in groundwater levels may affect corn irrigation water sources. Monitoring Point C (Tomato): Meteorological risk factors: Heavy rain may destroy tomato planting ridges; high temperatures may affect fruit quality and yield. Soil risk factors: Sandy loam soil has good aeration but poor fertilizer retention capacity, which may lead to nutrient deficiencies in tomatoes. Hydrological risk factors: Over-reliance on groundwater wells can affect irrigation if wells malfunction or the groundwater level drops.

[0070] Based on the crop use information corresponding to each monitoring point, the crop production risk factors for each monitoring point were processed to generate the crop production level for each monitoring point. Monitoring point A (wheat): For bread processing plants and flour mills, both wheat quality and yield are important. If the risk of flooding or drought causes a decrease in wheat yield of more than 30% or the quality does not meet the standards (e.g., low protein content), the production level is low; if the yield decreases by 10-30% and the quality is slightly affected, the production level is medium; if the yield and quality are only slightly affected (yield decrease of less than 10%), the production level is high. Monitoring Point B (Corn): As a raw material for feed and corn starch, corn yield and starch content are key. If drought and soil fertility issues cause a yield decrease of more than 25% or a reduction in starch content, the production grade is low; if the yield decreases by 10-25% and starch content fluctuates to some extent, the production grade is medium; if the yield and starch content are stable and the yield decrease is less than 10%, the production grade is high. Monitoring Point C (Tomatoes): Supplying farmers' markets and supermarkets, the appearance, taste, and yield of tomatoes all affect their value. If high temperatures and soil fertility issues lead to poor fruit quality (e.g., poor appearance, sour taste) and a yield decrease of more than 20%, the production grade is low; if fruit quality is somewhat affected and the yield decreases by 10-20%, the production grade is medium; if the fruit quality is good and the yield decrease is less than 10%, the production grade is high.

[0071] The crop production levels at each monitoring point are processed to generate classification information for each point. This classification information characterizes the expected production progress of crops at each monitoring point. For monitoring point A (wheat), if the production level is high, the classification information might be "Expected production progress is normal, and high-quality wheat is expected to be harvested on time"; if it is medium, the classification information might be "Production progress may be slightly delayed, requiring monitoring of wheat growth to ensure yield and quality"; if it is low, the classification information might be "Production progress may be severely hampered, potentially failing to meet the needs of flour mills and bread processing plants." For monitoring point B (corn), when the production level is high, the classification information is "Corn production progress is good, meeting feed and starch production needs"; when it is medium, the classification information is "Corn production may face some challenges, requiring measures to ensure yield and starch content"; when it is low, the classification information is "Corn production is encountering serious problems, potentially affecting feed and starch supply." For monitoring point C (tomatoes), when the production level is high, the classification information is "Tomatoes are growing normally and are expected to be supplied to the market as high-quality products on time"; when it is medium, the classification information is "Tomato production may be affected to some extent, and management needs to be strengthened to ensure market supply"; when it is low, the classification information is "Tomato production faces significant risks and may not be able to meet market demand".

[0072] S107, Based on the target disaster distributed eco-hydrological model, the attribute information of the target farmland area and the farmland ecological early warning information of the target farmland area are processed to generate farmland early warning information for the monitoring points to be predicted.

[0073] In one implementation, the attribute information of the target farmland area is processed based on a distributed eco-hydrological model of the target disaster to generate attribute information for the monitoring points to be predicted. This attribute information includes grouping information for the monitoring points. It is assumed that the target farmland area is a topographically complex agricultural region with different altitude zones and soil types. The region is divided into three main groups: a high-altitude mountain group, a mid-altitude hill group, and a low-altitude plain group. Each group has different characteristics. For example, the high-altitude mountain group has lower temperatures and relatively poor soil but abundant water resources; the mid-altitude hill group has more undulating terrain, moderate soil fertility, and is susceptible to soil erosion; the low-altitude plain group has fertile soil and convenient irrigation but may face flood risks. It is assumed that the monitoring point to be predicted is located within the mid-altitude hill group. This grouping information indicates that the monitoring point possesses the general characteristics of the mid-altitude hill group, such as undulating terrain, moderate soil fertility, and susceptibility to soil erosion.

[0074] Based on the distributed eco-hydrological model of the target disaster, the attribute information of the monitoring points to be predicted is processed to generate the first crop production prediction value for the monitoring points. According to the formula... ,in, Let m be the weight coefficient of the j-th factor in the i-th group in the target disaster distribution eco-hydrological model, and m be the total number of influencing factors and n be the total number of groups. This represents the value (such as soil moisture, temperature, etc.) related to the j-th factor affecting crop production in the i-th group (e.g., different altitude groups). Assume there are three groups (n=3) and five factors affecting crop production (m=5): soil moisture, temperature, sunshine duration, soil fertility, and irrigation amount. This refers to the weighting coefficient of the j-th factor in the i-th group of the target disaster distribution eco-hydrological model. For example, in the mid-altitude hilly group, this is the weighting coefficient for soil moisture. It could be 0.2, the weighting factor for temperature. The weighting factor for illumination duration is 0.3. The weighting coefficient for soil fertility is 0.2. The weighting factor for irrigation amount is 0.2. It is 0.1.

[0075] This is the value associated with the i-th factor affecting crop production in the i-th group. For example, in the mid-altitude hilly group where the monitoring point to be predicted is located, the current soil moisture measurement is 60% (corresponding to...). ), temperature is 20°C (corresponding to The duration of light exposure is 8 hours per day (corresponding to) The soil fertility test result was at a moderate level (assuming a corresponding value of 0.6). The irrigation volume is 5 cubic meters per week (corresponding to) The predicted value of the first crop production for the monitoring points in the mid-altitude hilly area was calculated using the formula. : (It is assumed here that factors such as temperature and duration of sunlight require certain numerical processing during the calculation, for example, converting temperature into an impact coefficient on crop production.)

[0076] Based on the distributed eco-hydrological model of the target disaster, the farmland ecological early warning information of the target farmland area is processed to generate the second crop production prediction value for the monitoring points to be predicted. According to the formula... ,in, This represents the value of the kth factor related to farmland ecological early warning (e.g., the risk coefficient determined based on the target event type, the comprehensive impact value calculated based on the machine weights of crop production impact factors, etc.). For the k-th factor in the distributed hydrological model of the target disaster, , This represents the total number of factors related to farmland ecological early warning. Assume there are three factors (l=3) related to farmland ecological early warning: pest and disease risk coefficient, soil erosion risk coefficient, and extreme weather risk coefficient. This refers to the weighting coefficient for the k-th factor in a distributed hydrological model of a target disaster. For example, the weighting coefficient for the risk coefficient of pests and diseases. The weight of the soil erosion risk coefficient may be 0.4. The weight of the extreme weather risk coefficient is 0.3. It is 0.3. This is the k-th factor value related to farmland ecological early warning. For example, the current pest and disease risk coefficient assessment value for the monitoring point to be predicted is 0.7 (meaning there is a certain risk of pests and diseases, which may reduce yield by 30%). The soil erosion risk coefficient is 0.8 (corresponding to...). The extreme weather risk coefficient is 0.9 (assuming there is no extreme weather risk in the near future, but there is still a certain possibility, corresponding to...). ).

[0077] Calculate the predicted value of the second crop production at the monitoring point to be predicted using the formula. : The predicted crop production values ​​for the first and second monitoring points are processed to generate the predicted crop production values ​​for the monitoring points; according to the formula... ,in and Weights are assigned to the accuracy of crop production forecasts. Assumptions =0.6, This indicates that the first crop production forecast has a relatively high importance in the overall forecast. The previously calculated first crop production forecast is known. Second crop production forecast Substitute the values ​​into the formula to calculate the predicted crop production values ​​for the monitoring points to be predicted: .

[0078] Based on preset crop production forecasting rules, the predicted crop production values ​​for the monitoring points are processed to generate farmland early warning information for these points. The preset crop production forecasting rules are as follows: if the predicted crop production value is greater than 0.8, it is a low-risk warning, indicating that crop production is basically normal and only routine management is needed; if the predicted value is between 0.6 and 0.8, it is a medium-risk warning, requiring strengthened monitoring and preventative measures; if the predicted value is less than 0.6, it is a high-risk warning, requiring emergency measures to ensure crop production. Based on the calculated predicted crop production values ​​for the monitoring points, the farmland early warning information is determined according to the preset rules. For example, if the predicted value is less than 0.8, then the farmland early warning information for the monitoring point is a medium-risk warning, requiring strengthened monitoring of crop growth at that point, and measures such as increasing the frequency of pest and disease inspections, checking soil erosion, and paying attention to weather forecasts.

[0079] This application acquires target farmland area information, a target event set, and a training sample set. The target farmland area information covers real-time crop production factors within the area of ​​the monitoring point to be predicted. The training sample set includes historical production factors of the monitoring point to be predicted and real-time crop production factors of adjacent monitoring points. Next, the training sample set is preprocessed, grouped using a pre-defined mapping table, features are extracted to determine the original feature library, and training and validation sets are divided. A classifier and a pre-defined algorithm are used to obtain prediction results and validation set prediction results, thereby generating target feature data characterizing risk factors affecting crop production. Then, the target event set is processed to generate farmland ecological early warning information. First, the target event type information and risk warning level are determined, then historical farmland ecological information is generated to obtain crop production influencing factors and weight information. Based on the pre-defined farmland influence feature set, target feature data is generated, thus obtaining farmland ecological early warning information.

[0080] Furthermore, a distributed eco-hydrological model for target disasters is used to process information about target farmland areas. On one hand, monitoring point information is generated and grouped, features are extracted to obtain abnormal crop production characteristics and co-occurrence frequencies, and correlation information and monitoring point attribute information are generated. On the other hand, the target farmland area information is processed to generate crop production risk factors, production levels, and classification information for each monitoring point. For monitoring points to be predicted, their attribute information and farmland ecological early warning information are processed based on the distributed eco-hydrological model for target disasters. Grouping information for the monitoring points to be predicted, first and second crop production prediction values ​​are generated, and comprehensive processing yields the crop production prediction value. Then, according to preset crop production prediction rules, farmland early warning information is generated, providing comprehensive early warning and prediction support for agricultural disaster prevention and control.

[0081] In one implementation, such as Figure 2 As shown, this application also provides an agricultural disaster prevention and control device based on digital twins, comprising: The acquisition module 201 is used to acquire target farmland area information, a set of target events matching the target farmland area information, and a training sample set. The target farmland area information includes real-time crop production factor information within the area of ​​the monitoring point to be predicted within the target farmland area. The training sample set includes historical production factor information of the monitoring point to be predicted within the target farmland area and real-time crop production factor information of other monitoring points adjacent to the monitoring point to be predicted. The module also acquires a preset disaster distributed eco-hydrological model, which is generated based on historical crop production information of the target farmland area. Processing module 202 is used to preprocess the training sample set to generate a training sample set with target feature data, wherein the target feature data is used to characterize risk factors affecting crop production; process the preset disaster distributed eco-hydrological model based on the training sample set with target feature data to generate a target disaster distributed eco-hydrological model; process the target event set to generate farmland ecological early warning information for the target farmland area; process the target farmland area information based on the target disaster distributed eco-hydrological model to generate attribute information for the target farmland area, wherein the attribute information of the target farmland area includes classification information of the monitoring points to be predicted; and process the attribute information and farmland ecological early warning information of the target farmland area based on the target disaster distributed eco-hydrological model to generate farmland early warning information for the monitoring points to be predicted.

[0082] This application provides an electronic device, such as... Figure 3 As shown, the electronic device 3 includes a first processor 300, a memory 301, a bus 302, and a communication interface 303. The first processor 300, the communication interface 303, and the memory 301 are connected through the bus 302. The memory 301 stores a computer program that can run on the first processor 300. When the first processor 300 runs the computer program, it executes the agricultural disaster prevention and control method based on digital twin provided in any of the foregoing embodiments of this application.

[0083] The memory 301 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 303 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network.

[0084] Bus 302 can be an ISA bus, PCI bus, or EISA bus, etc. Buses can be divided into address buses, data buses, control buses, etc. Memory 301 is used to store programs. After receiving execution instructions, the first processor 300 executes the program. The agricultural disaster prevention method based on digital twins disclosed in any of the foregoing embodiments of this application can be applied to the first processor 300, or implemented by the first processor 300.

[0085] The first processor 300 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the first processor 300 or by instructions in software form. The first processor 300 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be implemented by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 301. The first processor 300 reads the information in memory 301 and, in conjunction with its hardware, completes the steps of the above method.

[0086] The electronic devices provided in the above embodiments of this application and the agricultural disaster prevention and control method based on digital twins provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0087] This application provides a computer-readable storage medium, such as... Figure 4 As shown, the computer-readable storage medium 401 stores a computer program, which is read and executed by the second processor 402 to implement the aforementioned agricultural disaster prevention and control method based on digital twins.

[0088] The technical solutions of this application embodiment, in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, 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 an electronic device (which may be an air conditioner, refrigeration unit, personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of this application embodiment. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0089] The computer-readable storage medium provided in the above embodiments of this application and the agricultural disaster prevention and control method based on digital twins provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application stored therein.

[0090] The computer program products provided in the above embodiments of this application and the agricultural disaster prevention and control method based on digital twins provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0091] The various embodiments in this application are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for evaluating the agricultural disaster prevention and control method, electronic device, electronic device, and readable storage medium based on digital twins are basically similar to the above-described embodiments of the agricultural disaster prevention and control method based on digital twins, so the description is relatively simple. Relevant parts can be referred to the descriptions of the above-described embodiments of the agricultural disaster prevention and control method based on digital twins.

[0092] While this application discloses the above information, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of this application; therefore, the scope of protection of this application shall be determined by the scope defined in the claims.

Claims

1. A method for agricultural disaster prevention and control based on digital twins, characterized in that, include: Acquire target farmland area information, target event set matching the target farmland area information, and training sample set. The target farmland area information includes real-time crop production factor information within the range of the monitoring point to be predicted within the target farmland area. The training sample set includes historical production factor information of the monitoring point to be predicted within the target farmland area and real-time crop production factor information of other monitoring points adjacent to the monitoring point to be predicted. The training sample set is preprocessed to generate a training sample set with target feature data, where the target feature data is used to characterize the risk factors affecting crop production. Obtain a preset disaster distributed eco-hydrological model, which is generated based on historical crop production information of the target farmland area; The preset disaster distributed eco-hydrological model is processed based on the training sample set with target feature data to generate the target disaster distributed eco-hydrological model. Process the target event set to generate farmland ecological early warning information for the target farmland area; Based on the distributed eco-hydrological model of the target disaster, the information of the target farmland area is processed to generate the attribute information of the target farmland area, which includes the classification information of the monitoring points to be predicted. Based on the distributed eco-hydrological model of the target disaster, the attribute information and farmland ecological early warning information of the target farmland area are processed to generate farmland early warning information for the monitoring points to be predicted.

2. The method as claimed in claim 1, characterized in that, The training sample set is preprocessed to generate a training sample set with target feature data, including: Obtain a preset mapping table, which is used to represent the correspondence between the historical crop production factor information of the target farmland area and the historical crop production factor information of the monitoring points to be predicted located in the target farmland area; The training sample set is grouped based on a preset mapping table to generate a grouped training sample set, which includes monitoring points to be predicted at different altitudes. Feature extraction is performed on the grouped training sample set to determine the original feature library; Based on the original feature library, divide each feature dataset to generate training and validation sets; The original feature library is divided into various validation sets using a classifier for prediction, and the prediction results are determined. The original feature library is divided into training sets using a preset algorithm to obtain the validation set class prediction results. Based on the prediction results and the validation set prediction results, target feature data is generated. The target feature data is used to characterize the risk factors affecting crop production.

3. The method as described in claim 1, characterized in that, The target event set is processed to generate farmland ecological early warning information for the target farmland area, including: The target event set is processed to generate target event type information, which is used to characterize the risk warning level of the target event for the target farmland area; The type information of the target event is processed to generate historical farmland ecological information of the target farmland area; The historical farmland ecological information of the target farmland area is processed to generate crop production influencing factors and corresponding weight information. Based on a preset set of farmland impact features, the impact factors on crop production and the corresponding weight information of the impact factors on crop production are processed to generate target feature data. The target feature data is processed to generate farmland ecological early warning information for the target farmland area.

4. The method as claimed in claim 1, characterized in that, Based on the distributed eco-hydrological model of the target disaster, the information of the target farmland area is processed to generate the attribute information of the target farmland area, including: Information on the target farmland area is processed based on the target disaster distributed eco-hydrological model to generate monitoring point information, which consists of monitoring points located within the target farmland area. The monitoring point information is grouped to generate several groups, including monitoring point information at different altitudes. The grouping information is processed by feature extraction to generate feature datasets for different groups; The feature datasets of different groups are processed to obtain abnormal crop production characteristics and co-occurrence frequencies of correlations; The abnormal crop production characteristics and co-occurrence frequency of correlations are processed to generate correlation information, which is used to characterize the correlation between crop production characteristics at different altitudes and the actual farmland conditions at the target monitoring points. Based on the correlation information, the feature datasets of different groups are processed to generate attribute information of the monitoring points corresponding to different groups; The method includes a formula for calculating the co-occurrence frequency of correlations, which is as follows: F(A,B) ; in, is the number of times events A and B co-occur in the group with altitude i, n is the total number of altitude groups, and N is the total number of samples.

5. The method as described in claim 4, characterized in that, Based on the distributed eco-hydrological model of the target disaster, information about the target farmland area is processed to generate attribute information of the target farmland area, which also includes: The target farmland area information is processed to generate several monitoring points, including the monitoring points to be predicted and other monitoring points located in the target farmland area information. Obtain information on the crop usage corresponding to each monitoring point; Based on the target disaster distributed eco-hydrological model, several monitoring points were processed to generate crop production risk factors for each monitoring point; Based on the crop use information corresponding to each monitoring point, the crop production risk factors of each monitoring point are processed to generate the crop production level of each monitoring point; The crop production level of each monitoring point is processed to generate classification information for each monitoring point. The classification information of each monitoring point is used to characterize the expected crop production progress information of each monitoring point.

6. The method as described in claim 5, characterized in that, Based on the distributed eco-hydrological model of the target disaster, the attribute information and farmland ecological early warning information of the target farmland area are processed to generate farmland early warning information for the monitoring points to be predicted, including: The attribute information of the target farmland area is processed based on the target disaster distributed eco-hydrological model to generate the attribute information of the monitoring points to be predicted. The attribute information of the monitoring points to be predicted includes the grouping information of the monitoring points to be predicted. Based on the distributed eco-hydrological model of the target disaster, the attribute information of the monitoring point to be predicted is processed to generate the first crop production prediction value of the monitoring point to be predicted. Based on the distributed eco-hydrological model of the target disaster, the farmland ecological early warning information of the target farmland area is processed to generate the second crop production prediction value of the monitoring point to be predicted; The predicted values ​​of the first crop production at the monitoring point to be predicted and the predicted values ​​of the second crop production at the monitoring point to be predicted are processed to generate the predicted values ​​of the crop production at the monitoring point to be predicted. Based on preset crop production forecasting rules, the predicted crop production values ​​of the monitoring points to be predicted are processed to generate early warning information for farmland at the monitoring points to be predicted.

7. The method of claim 6, characterized in that, Based on the distributed eco-hydrological model for target disasters, the attribute information and early warning information of the target farmland area are processed to generate early warning information for the monitoring points to be predicted. This also includes: The target disaster distributed eco-hydrological model includes a calculation formula for obtaining the first crop production forecast value. The calculation formula is as follows: ; in, Let m be the weight coefficient of the j-th factor in the i-th group in the target disaster distribution eco-hydrological model, and m be the total number of influencing factors and n be the total number of groups. This represents the value in the i-th group that is related to the j-th factor affecting crop production; The target disaster distributed eco-hydrological model includes a calculation formula for obtaining the second crop production forecast value. The calculation formula is as follows: ; in, Let k be the value of the factor related to farmland ecological early warning. For the k-th factor in the distributed hydrological model of the target disaster, , This represents the total number of factors related to farmland ecological early warning. The target disaster distributed eco-hydrological model includes a calculation formula for obtaining crop production forecasts. The calculation formula is as follows: ; in, and Weighting of crop production forecast accuracy.

8. An agricultural disaster prevention and control device based on digital twins, characterized in that, The device is used to perform the agricultural disaster prevention and control method based on digital twins as described in any one of claims 1 to 7.

9. An electronic device, characterized in that, include: First processor; And memory for storing executable instructions of the first processor; The first processor is configured to execute the agricultural disaster prevention and control method based on digital twins as described in any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the second processor, it implements the agricultural disaster prevention and control method based on digital twins as described in any one of claims 1 to 7.

Images