Search

CN-122022012-A - Crop yield prediction system and method based on automated modeling intelligent agent

CN122022012ACN 122022012 ACN122022012 ACN 122022012ACN-122022012-A

Abstract

The application provides a crop yield prediction system and method based on an automated modeling intelligent agent. The method comprises a data acquisition module, a characteristic construction module, an automatic modeling module, a multi-scale yield prediction module, a supply chain safety evaluation module and a strategy output module, wherein the data acquisition module is used for performing standardized processing and space-time alignment on multi-source data to generate a unified data set, the characteristic construction module is used for constructing a characteristic set for prediction according to a stage rule and a time window rule related to a growth process of target crops, the automatic modeling module is used for generating a region self-adaptive yield prediction model corresponding to different regions, the multi-scale yield prediction module is used for performing scale summarization on yield prediction results of region scales to generate yield prediction results of higher scales, the supply chain safety evaluation module is used for generating supply and demand evaluation results and risk identifications according to preset supply and demand evaluation rules, and the strategy output module is used for generating strategy suggestions or instruction information for grain supply-keeping scheduling. The application can improve modeling and prediction updating efficiency, enhance cross-regional prediction adaptability and improve timeliness of the guaranteed scheduling decision.

Inventors

  • ZENG ZIYIN
  • XIE SHULEI

Assignees

  • 厦门屿智未来科技有限公司

Dates

Publication Date
20260512
Application Date
20251230

Claims (10)

  1. 1. A crop yield prediction system based on automated modeling agents, comprising: The data acquisition module is used for acquiring multi-source data related to the growth and supply chain operation of the target crops, and performing standardized processing and space-time alignment on the multi-source data to generate a unified data set; The characteristic construction module is used for constructing a characteristic set for prediction according to a stage rule and a time window rule related to the growth process of the target crops based on the unified data set; the automatic modeling module is used for executing candidate model selection and parameter optimization based on the feature set by an automatic modeling agent, generating an area self-adaptive yield prediction model corresponding to different areas, and carrying out iterative update on the area self-adaptive yield prediction model when new data arrives; The multi-scale yield prediction module is used for outputting a yield prediction result of a regional scale by utilizing the regional adaptive yield prediction model, and performing scale summarization on the yield prediction result of the regional scale to generate a yield prediction result of a higher scale; the supply chain safety evaluation module is used for correlating the yield prediction result with supply chain operation data and generating a supply and demand evaluation result and a risk identification according to a preset supply and demand evaluation rule; And the strategy output module is used for generating strategy suggestions or instruction information for grain supply and demand scheduling according to the supply and demand evaluation result and the risk identification.
  2. 2. The system of claim 1, wherein the acquiring multi-source data related to target crop growth and supply chain operation and performing normalization processing and space-time alignment on the multi-source data to generate a unified data set comprises: Acquiring remote sensing observation data related to the growth of target crops, and generating remote sensing index data representing the growth vigor of the crops based on the remote sensing observation data; acquiring meteorological observation data related to the growth of a target crop, performing space aggregation on the meteorological observation data according to a preset region granularity, and performing time aggregation according to a preset time slice granularity to generate meteorological element data; Acquiring area data and historical yield data corresponding to a target crop, and acquiring supply chain operation data related to supply chain operation; performing field caliber unification, unit conversion, missing data processing and abnormal data processing on the multi-source data to complete the standardized processing; And performing association and alignment on the multi-source data subjected to standardization processing based on the region identification and the time slice identification to form a data record set with a preset structure as the unified data set.
  3. 3. The system of claim 1, wherein the constructing a feature set for prediction according to a stage rule and a time window rule related to a target crop growing process based on the unified data set comprises: determining a plurality of growth stages corresponding to the growth process of the target crops in the unified data set according to the stage rule, and segmenting the unified data set according to the growth stages; Performing stage aggregation processing on remote sensing index data and meteorological element data in each growth stage to generate stage characteristics representing crop growth vigor and meteorological conditions in the growth stage; performing rolling aggregation processing on the unified data set according to the time window rule to generate window characteristics representing the change of the unified data set in a rolling time window; And combining the stage characteristics with the window characteristics to obtain the characteristic set for prediction.
  4. 4. The system of claim 1, wherein the performing, by the automated modeling agent, candidate model selection and parameter optimization based on the feature set, generating region-adaptive yield prediction models corresponding to different regions comprises: Respectively constructing a history sample set aiming at different areas based on the feature set, and executing rolling division on the history sample set according to time sequence to form a training data set and a verification data set; training candidate models in a preset candidate model set, and determining parameter configuration corresponding to the candidate models by adopting a preset parameter optimizing strategy; And comparing the evaluation results of different candidate models and parameter configurations on the verification data set according to a preset evaluation rule, and selecting the candidate models meeting the preferred conditions and the corresponding parameter configurations as the region self-adaptive yield prediction models associated with the corresponding regions.
  5. 5. The system of claim 4, wherein iteratively updating the region adaptive yield prediction model as new data arrives comprises: accessing new data corresponding to the unified data set, and performing feature update on the new data based on the stage rule and the time window rule to form an incremental sample; Judging whether to start model updating according to a preset updating triggering condition, wherein the updating triggering condition comprises at least one of the time span covered by newly added data meeting a threshold condition and the model prediction error change meeting the threshold condition; and when the model update is judged to be started, executing retraining update or incremental update on the area self-adaptive yield prediction model based on the incremental sample so as to obtain an updated area self-adaptive yield prediction model.
  6. 6. The system of claim 5, wherein the region adaptive yield prediction model comprises: a feature input substructure for receiving the feature set and generating a corresponding feature representation; The time sequence fusion substructure is used for fusing the feature representations corresponding to different time slices to obtain a fused feature representation; and the regression prediction substructure is used for outputting a unit yield prediction result associated with the corresponding region based on the fusion characteristic representation.
  7. 7. The system of claim 1, wherein outputting the regional-scale yield prediction results using the regional adaptive yield prediction model and performing a scale summary of the regional-scale yield prediction results to generate higher-scale yield prediction results comprises: Inputting a feature set corresponding to a target area into the area self-adaptive yield prediction model, and outputting a unit yield prediction result of the target area; Performing fusion operation based on the unit yield prediction result and the area data corresponding to the unified data set to obtain a total yield prediction result of the target area, wherein the total yield prediction result is used as a yield prediction result of the area scale; And carrying out weighted summarization or accumulated summarization on the total yield prediction results of the plurality of areas according to a preset summarization rule, and generating the yield prediction results with higher scales.
  8. 8. The system of claim 1, wherein the correlating the yield forecast with supply chain operational data generates supply and demand assessment results and risk identifications in accordance with preset supply and demand assessment rules, comprising: matching the yield prediction result with the supply chain operation data based on a region identifier and a time slice identifier; According to the preset supply and demand evaluation rule, carrying out supply and demand balance calculation on the available quantity represented by the yield prediction result, the inventory quantity represented by the supply chain operation data, the arrival supply quantity and the processing consumption quantity to obtain supply and demand gap information comprising gap scale information and gap time window information as the supply and demand evaluation result; And determining a risk category or a risk level corresponding to the region identifier and the time slice identifier based on the supply and demand gap information, and generating the risk identifier.
  9. 9. The system of claim 1, wherein the generating policy advice or instruction information for food conservation and dispatch based on the supply and demand assessment result and the risk identification comprises: Determining a target supply strategy type from a preset supply strategy set based on gap scale information and gap time window information in the supply and demand evaluation result and combining the risk identification; performing parameterization configuration on the target reserve strategy type according to a preset constraint rule to determine an execution time window and a resource scheduling parameter corresponding to the target reserve strategy type, wherein the resource scheduling parameter comprises at least one of an import rhythm adjustment parameter, a reserve parameter, a transport capacity allocation parameter and a processing scheduling parameter; and outputting strategy advice or instruction information containing the execution time window and the resource scheduling parameters.
  10. 10. A crop yield prediction method based on an automated modeling agent according to any one of claims 1 to 9, comprising: acquiring multi-source data related to growth and supply chain operation of a target crop, and performing standardized processing and space-time alignment on the multi-source data to generate a unified data set; Constructing a feature set for prediction according to a stage rule and a time window rule related to the growth process of the target crops based on the unified data set; Executing candidate model selection and parameter optimization by an automatic modeling agent based on the feature set, generating an area self-adaptive yield prediction model corresponding to different areas, and carrying out iterative update on the area self-adaptive yield prediction model when new data arrives; Outputting a yield prediction result of the regional scale by using the regional adaptive yield prediction model, and performing scale summarization on the yield prediction result of the regional scale to generate a yield prediction result of a higher scale; correlating the yield prediction result with supply chain operation data, and generating a supply and demand evaluation result and a risk identification according to a preset supply and demand evaluation rule; and generating strategy suggestions or instruction information for grain supply and demand scheduling according to the supply and demand assessment result and the risk identification.

Description

Crop yield prediction system and method based on automated modeling intelligent agent Technical Field The application relates to the technical field of intelligent analysis of agricultural data, in particular to a crop yield prediction system and method based on an automatic modeling intelligent body. Background Crops (e.g., soybeans) are important sources of grain and oil and feed, and the yield variation directly affects supply chain decisions such as importation scheduling, reserve planning, press scheduling, port and inland logistics capacity allocation, and feed cost control. Countries, enterprises and trade bodies often need to grasp the production trends and risks of major production areas several months in advance. The existing crop yield prediction mainly comprises two types, namely a statistical model such as linear regression and stepwise regression, wherein a prediction relation is generally constructed based on a weather mean value, a historical single yield mean value and the like, and a subjective research and judgment mode which depends on research and experience judgment, for example, estimation is formed through observation of sowing area, pod bearing and growth vigor. However, the method still has the defects that the response capability of the statistical model to sudden factors such as extreme weather, drought, plant diseases and insect pests and the like is limited, the hydrothermal structures and characteristic sensitivity differences of different areas are obvious, the model migration and generalization depend on a large number of manual adjustments to participate in model reconstruction, the engineering cost is high, the experience research and judgment mode is slow to update, the cost is high, the subjectivity is strong, and the week-level or month-level high-frequency dynamic update is difficult to realize. Meanwhile, the existing prediction stays at output of the yield value, a closed-loop mechanism for linking a prediction result with supply chain operation information such as inventory, squeezing, arrival rhythm and the like is lacking, and an executable supply-keeping scheduling basis is difficult to directly form. Disclosure of Invention In view of the above, the embodiment of the application provides a crop yield prediction system and method based on an automated modeling intelligent agent, so as to solve the problems of low efficiency of manual modeling, poor regional generalization capability and difficult linkage of prediction results for supply-keeping scheduling in the prior art. The first aspect of the embodiment of the application provides a crop yield prediction system based on an automated modeling agent, which comprises a data acquisition module, a characteristic construction module, an automated modeling module, a supply chain safety evaluation module and a strategy evaluation module, wherein the data acquisition module is used for acquiring multisource data related to target crop growth and supply chain operation, performing standardized processing and time-space alignment on the multisource data to generate a unified data set, the characteristic construction module is used for constructing a characteristic set for prediction according to a stage rule and a time window rule related to a target crop growth process based on the unified data set, the automated modeling agent is used for performing candidate model selection and parameter optimization based on the characteristic set to generate a region adaptive yield prediction model corresponding to different regions, and performing iterative updating on the region adaptive yield prediction model when new added data arrives, the multiscale yield prediction module is used for outputting a region-scale yield prediction result by utilizing the region adaptive yield prediction model, performing scale summarization on the region-scale yield prediction result to generate a higher-scale yield prediction result, the supply chain safety evaluation module is used for correlating the prediction result with the supply chain operation data, generating a supply and demand identification according to a preset supply and demand assessment rule, and outputting a strategy assessment instruction and demand identification information. A second aspect of the embodiment of the application provides a crop yield prediction method based on an automated modeling agent of a system of the first aspect, which comprises the steps of obtaining multi-source data related to growth and supply chain operation of target crops, performing standardized processing and time-space alignment on the multi-source data to generate a unified data set, constructing a feature set for prediction according to a stage rule and a time window rule related to a growth process of the target crops based on the unified data set, performing candidate model selection and parameter optimization based on the feature set by the automated modeling agent to generate a region ada