Double-layer cascade dynamic constraint optimization agricultural disaster prediction method and system

Abstract

The invention discloses a double-layer cascade dynamic constraint optimization agricultural disaster prediction method and system, and the method comprises the steps: obtaining and carrying out the stage processing of satellite remote sensing data and external reference data, and constructing a prediction center and influence region data set; establishing an early warning model by using the central data set to predict agricultural disasters and trigger alarms; combining the two data sets to construct a space-time diffusion model, predicting a development trend and formulating a strategy; a dynamic optimization model is set and evaluated, and a sampling strategy is dynamically adjusted. According to the method, a double-layer cascade prediction model and a dynamic constraint optimization strategy are adopted, balance of agricultural disaster prediction precision and efficiency is realized, disaster hotspots are rapidly positioned and a global trend is predicted through collaborative prediction of hierarchical positioning early warning and a space-time diffusion model, and dependence on full-dimensional data is reduced; a self-adaptive sampling strategy is adopted to reduce redundancy calculation, dynamic updating is kept in a disaster high-incidence period, and a hierarchical early warning and prevention and control strategy and interpretable output provide clear guidance for decision makers.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural disaster prediction, and in particular to a double-layer cascaded dynamic constraint optimization agricultural disaster prediction method and system. Background Art

[0002] Agricultural disaster prediction technology plays a vital role in the agricultural sector, particularly in monitoring and providing early warning for disasters such as drought, floods, low temperatures, frost damage, typhoons, and hail. Existing agricultural disaster prediction technologies primarily rely on mathematical statistics, such as time series analysis and multivariate linear regression. These technologies analyze historical disaster data to build prediction models, thereby providing early warning information for agricultural production. However, these technical solutions have several limitations in practical application:

[0003] First, the models suffer from poor interpretability. Existing global spatiotemporal models based on neural networks are often viewed as "black boxes," meaning their internal workings are opaque and difficult to provide decision makers with clear actionable insights. This lack of transparency limits their application in real-world decision-making.

[0004] Second, existing models rely heavily on data integrity. Errors in the data fusion process, such as misalignment, can invalidate the entire model's predictions, impacting both accuracy and reliability.

[0005] Third, the complexity of data processing. Existing methods may encounter technical and resource challenges when processing and integrating large amounts of heterogeneous data, increasing the difficulty and cost of implementation.

[0006] Fourth, the real-time and dynamic adaptability are insufficient. Under rapidly changing environmental conditions, the existing models may not be able to update and adapt to new data in a timely manner, thus affecting the timeliness and accuracy of the prediction.

[0007] In summary, while existing agricultural disaster prediction technologies can provide early warning information to a certain extent, they still have significant room for improvement in practical applications due to poor model interpretability, a high reliance on data integrity, complex data processing, and inadequate real-time and dynamic adaptability. Therefore, developing more accurate, reliable, transparent, and adaptable agricultural disaster prediction methods and technologies is particularly important. Summary of the Invention

[0008] The embodiment of the present invention provides a two-layer cascaded dynamic constraint optimization agricultural disaster prediction method and system to solve the technical problem of the existing technology that the agricultural disaster prediction accuracy and real-time performance cannot be balanced.

[0009] A two-layer cascaded dynamic constraint optimization agricultural disaster prediction method includes:

[0010] Obtain satellite remote sensing data and external reference data of the target area and perform hierarchical processing to generate prediction center dataset and impact area dataset;

[0011] Build a hierarchical positioning regression warning model based on the prediction center data set, predict agricultural disasters, and trigger the alarm process;

[0012] Based on the prediction center dataset and the impact area dataset, a spatiotemporal diffusion dynamic constraint prediction model is constructed to predict future agricultural disaster trends and formulate prevention and control strategies. A dynamic constraint optimization model with multiple constraints is also set up, and the optimized model is evaluated in multiple dimensions.

[0013] The sampling strategy for agricultural disaster prediction is dynamically adjusted using a multi-source data fusion dynamic constraint optimization model.

[0014] A two-layer cascaded dynamic constraint optimization agricultural disaster prediction system, comprising:

[0015] A multi-source hierarchical data generation model is used to obtain satellite remote sensing data and external reference data of the target area and perform hierarchical processing to generate a prediction center dataset and an impact area dataset;

[0016] A hierarchical early warning trigger model is used to build a hierarchical positioning regression early warning model based on the prediction center data set, predict agricultural disasters, and trigger the alarm process;

[0017] The spatiotemporal dynamic constraint prediction optimization model is used to build a spatiotemporal diffusion dynamic constraint prediction model based on the prediction center dataset and the impact area dataset to predict future agricultural disaster development trends and formulate prevention and control strategies; a dynamic constraint optimization model with multiple constraints is set up, and the optimized model is evaluated in multiple dimensions;

[0018] The multi-source fusion constrained sampling model is used to dynamically adjust the sampling strategy for agricultural disaster prediction using a multi-source data fusion dynamic constraint optimization model.

[0019] The present invention has the following beneficial effects:

[0020] The present invention achieves a balance between prediction accuracy and efficiency through a two-layer cascaded hierarchical prediction model (a hierarchical positioning regression early warning model and a spatiotemporal diffusion dynamic constraint prediction model) and a dynamic constraint optimization strategy. "Local-global" collaborative prediction is achieved through hierarchical modeling. The underlying regression model can quickly locate disaster hotspots and output an interpretable disaster level; the upper spatiotemporal diffusion model integrates historical diffusion laws and real-time correlation features to predict global trends. In this way, it avoids the interpretation difficulties caused by multi-layer nonlinear mapping of traditional neural networks, and reduces the strong dependence on full-dimensional spatiotemporal data by separating local features and diffusion laws. Even if data in some areas is missing, the prediction continuity can still be maintained through spatiotemporal correlation. The adaptive sampling strategy of the present invention (spatial classification, temporal dynamic adjustment, and multi-source data fusion) reduces redundant calculations, allowing the model to maintain minute-level updates during high-incidence periods of disasters, which is significantly better than the hour-level response delay of traditional methods.

[0021] The present invention effectively copes with data noise, environmental mutations and resource limitations through multi-source data fusion and dynamic parameter optimization (the PPO algorithm corrects model parameters in real time). Its spatial adaptive strategy increases sampling density in high-risk areas and combines it with a significance detection algorithm to mark abnormal areas, avoiding the global failure of traditional methods caused by data bias; the temporal adaptive mechanism increases the sampling frequency during high-incidence periods of disasters, ensuring that the model captures dynamic changes in a timely manner, so that the system still maintains high reliability in complex and changeable agricultural disaster scenarios.

[0022] The present invention provides decision makers with a clear basis for action through a hierarchical early warning and prevention and control strategy (priority response in high-risk areas, dynamic monitoring of low-risk areas) and interpretable output (disaster level, diffusion direction, and key influencing factors). BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0024] Figure 1 This is a flow chart of a two-layer cascaded dynamic constraint optimization agricultural disaster prediction method according to one embodiment of the present invention;

[0025] Figure 2 1 is a schematic diagram of a two-layer cascaded dynamic constraint optimization agricultural disaster prediction system according to an embodiment of the present invention;

[0026] Figure 3 is another schematic diagram of a two-layer cascaded dynamic constraint optimization agricultural disaster prediction system according to one embodiment of the present invention;

[0027] Figure 4 FIG. 1 is a schematic diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort shall fall within the scope of protection of the present invention.

[0029] The agricultural disaster prediction method of the double-layer cascade dynamic constraint optimization provided by the embodiment of the present invention can be applied as follows: Figure 3 Specifically, the double-layer cascaded dynamic constraint optimization agricultural disaster prediction method is applied in a double-layer cascaded dynamic constraint optimization agricultural disaster prediction system, which includes the following: Figure 3 The client and server shown communicate over a network to achieve a balance between accuracy and real-time performance in agricultural disaster prediction. The client, also known as the user end, is the program that corresponds to the server and provides local services to clients. The client can be installed on, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented as a standalone server or a server cluster consisting of multiple servers.

[0030] In one embodiment, if Figure 1 As shown in the figure, a two-layer cascade dynamic constraint optimization method for agricultural disaster prediction is provided. Figure 3 The server in the example is used as an example, and the steps are as follows:

[0031] S1. Obtain satellite remote sensing data and external reference data of the target area and perform hierarchical processing to generate a prediction center dataset and an impact area dataset.

[0032] In one embodiment, step S1 further includes the following steps:

[0033] S101. Using the core geographic coordinates of the target area as the target center, set a hierarchical threshold range. The hierarchical threshold includes a first threshold and a second threshold. The hierarchical threshold range includes a first threshold range with the target center as the center point and the first threshold as the radius, and a second threshold range with the second threshold as the radius. The first threshold range is larger than the second threshold range. It can be understood that by setting the first and second thresholds, the prediction center area is precisely delineated, which helps focus resources and attention on areas most likely to be affected by disasters, thereby improving the accuracy and efficiency of predictions.

[0034] S102: Divide the annular zone outside the second threshold range but within the first threshold range into four impact sub-regions. It can be understood that through hierarchical processing and regional division, monitoring and early warning resources can be allocated more effectively to ensure that key areas receive adequate attention and protection.

[0035] S103: Using the satellite remote sensing data within the second threshold range as a prediction center data set, and using the data of the four impact sub-regions as an impact region data set.

[0036] Understandably, acquired satellite remote sensing data includes acquisition time, longitude and latitude, satellite zenith angle, tropospheric water vapor channel brightness temperature, longwave infrared channel brightness temperature, and tropospheric water vapor channel average specific humidity. External reference data includes weather station data and historical disaster records. By setting a classification threshold range, the acquired satellite remote sensing data and external reference data are graded to generate a prediction center dataset (also called the first dataset) and an impact area dataset (also called the second dataset). This reduces reliance on a single data source. Even if data in some areas is missing or ambiguous, the spread of disasters can still be captured through data from other areas.

[0037] S2. Build a hierarchical positioning regression warning model based on the prediction center dataset, predict agricultural disasters, and trigger the alert process. The hierarchical positioning regression warning model built based on the prediction center dataset (the first dataset) is called the first-level agricultural disaster prediction model. It can quickly locate the disaster site and trigger the alert process, improving the accuracy and timeliness of disaster warnings and addressing the issues of delayed initial disaster response and poor interpretability.

[0038] In one embodiment, step S2 further includes the following steps:

[0039] S201. Based on the prediction center dataset, a regression algorithm is used to train a hierarchical positioning regression warning model, with satellite remote sensing data as input and disaster probability and severity as output. Real-time satellite data is then fed into the hierarchical positioning regression warning model to generate current agricultural disaster forecasts and warning information. The use of a regression algorithm balances prediction accuracy with model interpretability, enabling predictions of the likelihood, severity, severity, and probability of a disaster based on the input remote sensing data, facilitating rapid response by decision makers.

[0040] S202: Push the warning information to the local agricultural department or emergency command system through a standardized interface, triggering the alert process. This shortens the time from prediction to action, improving the efficiency of disaster response and avoiding the slow response caused by relying on the completeness of the spatiotemporal data cube.

[0041] S203: The current agricultural disaster prediction results are compared with the actual disaster data observed on the ground in time and space to obtain an evaluation score. It can be understood that data verification ensures the credibility of the model output and avoids "false positives" or "missing negatives".

[0042] S204. Set the evaluation index and evaluation threshold; if the evaluation score is within the preset evaluation threshold, the model is determined to be valid, and the current model parameters and the current prediction center data set are stored in the cloud database; if the evaluation score is lower than the preset evaluation threshold, the model correction process is triggered. Understandably, the evaluation threshold can be set according to demand. For example, the evaluation threshold is set to 0.8. If the evaluation score meets the standard (such as ≥0.8), the current model parameters are stored in the cloud database; if the score is insufficient (such as <0.8), the cloud backup parameters are automatically used for correction or the manual correction process is started to realize dynamic correction of model parameters. The adaptability of the model is improved through the feedback loop, and there will be no global failure due to data fusion errors.

[0043] S3. A spatiotemporal diffusion dynamic constraint prediction model is constructed based on the prediction center dataset and the impact area dataset to predict future agricultural disaster trends and formulate prevention and control strategies. A dynamic constraint optimization model with multiple constraints is then established, and the optimized model is evaluated from multiple dimensions. This spatiotemporal diffusion dynamic constraint prediction model, also known as the secondary agricultural disaster prediction model, integrates data from the central and surrounding areas to more accurately predict how disasters will spread in space and time, providing more precise guidance for disaster management and resource allocation.

[0044] In one embodiment, step S3 further includes the following steps:

[0045] S301: Fusion of the prediction center dataset and the impact area dataset to construct a spatiotemporal dynamic constraint prediction model to predict current agricultural disaster trends. This model analyzes the patterns of disaster spread by fusing the prediction center and impact area data. By analyzing the weight of the impact of disaster status in adjacent regions on the prediction center, the direction and intensity of future disaster spread are predicted.

[0046] In one embodiment, step S301 further includes the following sub-steps:

[0047] S3011: The real-time updated prediction center dataset is integrated with the impact area dataset to construct an input feature matrix representing the current disaster status. Through feature screening and normalization, key disaster characteristics are extracted to generate a quantitative description of the current disaster. This step ensures that the model can make predictions based on the latest data, improving the timeliness and accuracy of predictions.

[0048] S3012. Analyze the impact of impact region datasets (e.g., rainfall at surrounding weather stations) on prediction center datasets (e.g., satellite remote sensing temperature and humidity). This analysis analyzes the spread and variation of agricultural disasters across different regions. Quantify the potential impact of disaster conditions in adjacent regions on the prediction center to predict future agricultural disaster trends. This quantified impact weight can capture the directionality of disaster spread, clarify its path, and more accurately predict its development, providing a basis for prevention and control strategies.

[0049] S302. Based on the current development trends of agricultural disasters, target areas are divided into high-risk (core area), medium-risk (potential spread area), and low-risk (safe zone) areas, and tiered prevention and control strategies are developed for each area. Based on the prediction results, the system automatically matches emergency plans to a database (e.g., "initiate emergency irrigation in high-risk areas," "evacuate personnel in medium-risk areas") and generates executable prevention and control instructions (e.g., immediately allocate a water pump to a specific village). This tiered management approach allows for more targeted resource allocation and implementation, improving the efficiency and effectiveness of disaster response.

[0050] S303: Dynamically constraining the spatiotemporal diffusion dynamic constraint prediction model with multiple constraints. Using the real-time prediction error as a feedback signal, the PPO algorithm dynamically adjusts model parameters in real time to optimize the model. Dynamic parameter adjustment through real-time error feedback (e.g., the residual between the predicted value and the true value) can improve the model's adaptability to changing environments, thereby enhancing the accuracy and reliability of the prediction.

[0051] S304. A series of evaluation indicators are set to perform a multi-dimensional evaluation of the optimized model. These indicators include prediction accuracy, response speed, and resource consumption, which are used to evaluate the predictive performance, timeliness, and stability of the optimization results. This multi-dimensional evaluation provides a comprehensive understanding of the model's performance, ensuring the predictive performance, timeliness, and stability of the optimization results.

[0052] S4. Dynamically adjust the sampling strategy for agricultural disaster prediction using a multi-source data fusion dynamic constraint optimization model. This intelligently adjusts the sampling frequency and points based on the real-time disaster development and the performance of the prediction model to optimize the data collection process.

[0053] In one embodiment, step S4 further includes the following steps:

[0054] The mathematical expression of the multi-source data fusion dynamic constraint optimization model is:

[0055]

[0056] The constraint condition is: st. Resource consumption ≤ C max , real-time ≤ T max ;

[0057] Where θ represents the model parameters, E represents the expected value, i.e., the statistical mean, L(·) represents the loss function, which is used to measure the difference between the predicted value and the true value, y represents the true value, i.e., the actual data observed, and f θ (x) represents the predicted value, λ1 represents the regularization strength coefficient, which controls the model complexity; λ2 represents the variance penalty coefficient, which controls the variance of the model parameters; Var(θ 1：T ) represents the variable θ 1：T The variance of C represents the uncertainty of the model parameters; max Indicates the maximum allowable value of resource consumption, T max Indicates the maximum allowed value of the time interval.

[0058] In one embodiment, to further improve the adaptability and accuracy of the model, an adaptive sampling strategy is used for dynamic constraints, including three types of constraints that work together to improve the accuracy and reliability of predictions:

[0059] Spatially adaptive sampling mechanism: Based on historical disaster distribution maps, it identifies high-frequency disaster areas, increases sampling density in these areas, and dynamically marks anomalous areas in satellite imagery using a saliency detection algorithm. This improves data representation capabilities in high-risk areas, enhances the model's sensitivity to disaster-stricken areas, and ensures high-resolution and frequent data updates in key areas.

[0060] Temporal dynamic adaptive sampling mechanism: Based on historical disaster cycles (e.g., the peak period of rainy season and drought is June-September), a time series prediction model (e.g., LSTM) is constructed to predict the most active disaster period. The sampling frequency of satellite data is increased during high-risk periods and reduced during low-risk periods. For example, the sampling frequency of satellite data can be increased to once per hour during high-risk periods and reduced to once per day during low-risk periods. This is synchronized with real-time data from meteorological stations. The sampling frequency is increased during high-risk seasons and reduced during non-critical periods. This balances computing resources with the timeliness of disaster response, enabling adaptive adjustment of the sampling frequency.

[0061] Multi-source data fusion: Combined with meteorological station data to trigger satellite data sampling; it can improve the data representation capabilities of high-risk areas and enhance the model's sensitivity to disaster core areas.

[0062] As can be understood, step S4 and all its substeps can, first, reduce data redundancy and unnecessary data collection through intelligent sampling strategies, thereby reducing data processing complexity and computational costs. Second, the optimized sampling strategy can more quickly respond to changes in disaster development, providing decision makers with more timely forecast information, thereby improving response speed. Furthermore, it enhances decision support, providing more accurate disaster forecast information, and helping decision makers develop more effective disaster response and resource allocation strategies.

[0063] By dynamically adjusting the sampling strategy, step S4 improves the model's robustness, ensuring stable prediction performance across diverse environments and conditions. This is crucial for agricultural disaster prediction systems, as it ensures reliable predictions under a wide range of circumstances. Finally, by optimizing resource usage and improving prediction efficiency, step S4 supports the long-term sustainability of agricultural disaster prediction systems.

[0064] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0065] In one embodiment, a two-layer cascaded dynamic constraint optimization agricultural disaster prediction system is provided, and the two-layer cascaded dynamic constraint optimization agricultural disaster prediction system corresponds one-to-one with the two-layer cascaded dynamic constraint optimization agricultural disaster prediction method in the above embodiment. Figure 2 As shown, the two-layer cascade dynamic constraint optimization agricultural disaster prediction system includes:

[0066] A multi-source hierarchical data generation model 100 is used to obtain satellite remote sensing data and external reference data of a target area and perform hierarchical processing to generate a prediction center dataset and an impact area dataset;

[0067] The hierarchical warning trigger model 200 is used to build a hierarchical positioning regression warning model based on the prediction center data set, predict agricultural disasters, and trigger the alarm process;

[0068] The spatiotemporal dynamic constraint prediction optimization model 300 is used to construct a spatiotemporal diffusion dynamic constraint prediction model based on the prediction center dataset and the impact area dataset to predict future agricultural disaster trends and formulate prevention and control strategies. It also sets a dynamic constraint optimization model with multiple constraints and conducts multi-dimensional evaluation of the optimized model.

[0069] Multi-source fusion constrained sampling model 400 is used to dynamically adjust the sampling strategy for agricultural disaster prediction using a multi-source data fusion dynamic constraint optimization model.

[0070] Regarding the specific limitations of the agricultural disaster prediction system with dynamic constraint optimization of the double-layer cascade, please refer to the limitations of the agricultural disaster prediction method with dynamic constraint optimization of the double-layer cascade above, which will not be repeated here. The various modules in the above-mentioned agricultural disaster prediction system with dynamic constraint optimization of the double-layer cascade can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

[0071] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, memory, network interface and database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a two-layer cascaded dynamic constraint optimization agricultural disaster prediction method is implemented.

[0072] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0073] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A two-layer cascaded dynamic constraint optimization agricultural disaster prediction method, characterized in that: include: S1. Obtain satellite remote sensing data and external reference data of the target area and perform hierarchical processing to generate a prediction center dataset and an impact area dataset; S2. Build a hierarchical positioning regression warning model based on the prediction center data set, predict agricultural disasters, and trigger the alarm process; S3. Build a spatiotemporal diffusion dynamic constraint prediction model based on the prediction center dataset and the impact area dataset to predict future agricultural disaster trends and formulate prevention and control strategies; And set up a dynamic constraint optimization model with multiple constraints, and conduct multi-dimensional evaluation of the optimized model; S4. Use the multi-source data fusion dynamic constraint optimization model to dynamically adjust the sampling strategy for agricultural disaster prediction.

2. The agricultural disaster prediction method of the double-layer cascade dynamic constraint optimization according to claim 1 is characterized in that: The step S1 includes the following sub-steps: S101, using the core geographic coordinates of the target area as the target center, setting a grading threshold range, the grading threshold including a first threshold and a second threshold, the grading threshold range including a first threshold range with the target center as the center point and the first threshold as the radius, and a second threshold range with the second threshold as the radius; wherein the first threshold range is larger than the second threshold range; S102, dividing the annular zone outside the second threshold range but within the first threshold range into four impact sub-regions; S103: Using the satellite remote sensing data within the second threshold range as a prediction center data set, and using the data of the four impact sub-regions as an impact region data set.

3. The agricultural disaster prediction method of the double-layer cascade dynamic constraint optimization according to claim 2 is characterized in that: The step S2 includes the following sub-steps: S201. Based on the prediction center data set, a regression algorithm is used to train a hierarchical positioning regression warning model, with the input being satellite remote sensing data and the output being the probability and level of disaster occurrence; Input real-time satellite data into the hierarchical positioning regression early warning model to generate current agricultural disaster prediction results and early warning information; S202: Push the warning information to the local agricultural department or emergency command system through a standardized interface to trigger the alarm process; S203, performing temporal and spatial matching comparison between the current agricultural disaster prediction results and the actual disaster data observed on the ground to obtain an evaluation score; S204. Set evaluation indicators and evaluation thresholds; if the evaluation score is within the preset evaluation threshold, the model is determined to be valid, and the current model parameters and the current prediction center data set are stored in the cloud database; if the evaluation score is lower than the preset evaluation threshold, the model correction process is triggered.

4. The agricultural disaster prediction method of the double-layer cascade dynamic constraint optimization according to claim 3 is characterized in that: The step S3 includes the following sub-steps: S301, integrate the prediction center dataset and the impact area dataset to build a spatiotemporal diffusion dynamic constraint prediction model to predict the current development trend of agricultural disasters; S302. Based on the current development trend of agricultural disasters, the target area is divided into high-risk, medium-risk, and low-risk areas, and graded prevention and control strategies are formulated for each area. Based on the prediction results, the emergency plan library is automatically matched to generate executable prevention and control instructions. S303, setting dynamic constraints of multiple constraints on the spatiotemporal diffusion dynamic constraint prediction model, using the real-time prediction error as a feedback signal, and dynamically adjusting the model parameters in real time through the PPO algorithm to optimize the model; S304. Set a series of evaluation indicators to conduct multi-dimensional evaluation of the optimized model. The series of indicators include prediction accuracy, response speed and resource consumption, which are used to evaluate the prediction performance, timeliness and stability of the optimization results.

5. The agricultural disaster prediction method of the double-layer cascade dynamic constraint optimization according to claim 4 is characterized in that: The step S301 further includes the following sub-steps: S3011. Fuse the real-time updated prediction center dataset with the impact area dataset to construct an input feature matrix representing the current disaster status; extract key disaster features through feature screening and normalization processing to generate a quantitative description of the current disaster; S3012. Analyze the impact of the impact area dataset on the prediction center dataset, analyze the spread and changes of agricultural disasters between different regions, and quantify the potential impact weight of the disaster status of adjacent regions on the prediction center to predict the development trend of future agricultural disasters.

6. The agricultural disaster prediction method of the double-layer cascade dynamic constraint optimization according to claim 1 is characterized in that: The mathematical expression of the multi-source data fusion dynamic constraint optimization model in step S4 is: The constraint condition is: st resource consumption ≤ C max , real-time ≤ T max ; Where θ represents the model parameters, E represents the expected value, i.e., the statistical mean, L(·) represents the loss function, which is used to measure the difference between the predicted value and the true value, y represents the true value, i.e., the actual data observed, and f θ (x) represents the predicted value, λ1 represents the regularization strength coefficient, which controls the model complexity; λ2 represents the variance penalty coefficient, which controls the variance of the model parameters; Var(θ 1：T ) represents the variable θ 1：T The variance of C represents the uncertainty of the model parameters; max Indicates the maximum allowable value of resource consumption, T max Indicates the maximum allowed value of the time interval.

7. The agricultural disaster prediction method of the double-layer cascade dynamic constraint optimization according to claim 6 is characterized in that: The sampling strategy in step S4 adopts an adaptive sampling strategy and performs dynamic constraints, and its constraints include: A spatially adaptive sampling mechanism identifies high-frequency disaster areas based on historical disaster distribution maps, increases sampling density in high-frequency disaster areas, and dynamically marks abnormal areas in satellite images using a saliency detection algorithm. A time-dynamic adaptive sampling mechanism builds a time series prediction model based on historical disaster cycles to predict active disaster periods; the sampling frequency of satellite data is increased during high-risk disaster periods, and reduced during low-risk periods; Multi-source data fusion, combined with weather station data, triggers satellite data sampling.

8. A two-layer cascaded dynamic constraint optimization agricultural disaster prediction system, characterized by: include: A multi-source hierarchical data generation model is used to obtain satellite remote sensing data and external reference data of the target area and perform hierarchical processing to generate a prediction center dataset and an impact area dataset; A hierarchical early warning trigger model is used to build a hierarchical positioning regression early warning model based on the prediction center data set, predict agricultural disasters, and trigger the alarm process; A spatiotemporal dynamic constraint prediction optimization model is used to build a spatiotemporal diffusion dynamic constraint prediction model based on the prediction center dataset and the impact area dataset to predict future agricultural disaster trends and formulate prevention and control strategies; And set up a dynamic constraint optimization model with multiple constraints, and conduct multi-dimensional evaluation of the optimized model; The multi-source fusion constrained sampling model is used to dynamically adjust the sampling strategy for agricultural disaster prediction using a multi-source data fusion dynamic constraint optimization model.

Images