Search

CN-122022056-A - Method, equipment, computer equipment and storage medium for predicting crop yield in black soil area

CN122022056ACN 122022056 ACN122022056 ACN 122022056ACN-122022056-A

Abstract

The invention discloses a method, equipment, computer equipment and storage medium for predicting crop yield in a black soil area, and belongs to the technical field of agricultural big data informatization. According to the method for predicting the crop yield in the black soil region, the performance of each localized prediction model is evaluated, the contribution scores of the environmental parameters to the simulation errors are analyzed in the obvious aggregation areas of the substandard partitions, so that the dominant environmental factors and the critical threshold values are obtained, the substandard partitions are further split by utilizing the dominant environmental factors and the critical threshold values, each subarea is guaranteed to have the best matched parameter set, and the prediction errors of the complex plots in the black soil region can be stably converged to the set value by the dynamic area dividing mode, so that the accuracy of yield prediction is improved.

Inventors

  • ZHANG FAN

Assignees

  • 中国科学院地理科学与资源研究所

Dates

Publication Date
20260512
Application Date
20260206

Claims (10)

  1. 1. A method for predicting crop yield in a black soil region, comprising: Acquiring weather big data and soil big data of a black soil area, wherein the weather big data comprises a history day-to-day weather element and a future day-to-day weather element; mapping the weather big data and the soil big data to a target grid to obtain a grid daily weather sequence; Partitioning the black soil region according to the black soil region resource characteristics to obtain a plurality of current level partitions; Calibrating an initial model in each current level partition by utilizing historical observation data to obtain a localized prediction model; performing performance evaluation on each localized prediction model, and calculating simulation errors of any space unit in the corresponding current level partition under the condition that the evaluation result does not meet the preset requirement; performing autocorrelation analysis on each simulation error to obtain a significant aggregation area, and extracting environmental parameters corresponding to the significant aggregation area; Calculating a contribution score of each environmental parameter to the simulation error, and confirming a dominant environmental factor and a critical threshold in the environmental parameters based on the contribution score; dividing the corresponding current level partition based on the dominant environmental factor and the critical threshold to obtain a plurality of next level partitions, and calibrating an initial model in each next level partition by using historical observation data until the localized model meets the performance requirement; And carrying out daily mechanism calculation based on the localized model and the grid daily meteorological sequence to obtain the predicted yield of the black soil region crops.
  2. 2. The black field crop yield prediction method according to claim 1, wherein the step of identifying dominant environmental factors and critical thresholds in the environmental parameters based on the contribution scores comprises: Determining the environmental parameter with the highest contribution score as the dominant environmental factor; And selecting the critical threshold based on the dominant environmental factor.
  3. 3. The method of claim 2, wherein the step of selecting the critical threshold based on the dominant environmental factor comprises: Extracting actual measurement values of dominant environmental factors in the significant aggregation area, and arranging the actual measurement values according to a preset sequence to obtain a dominant environmental factor sequence; Acquiring the average value of two adjacent values in the dominant environmental factor sequence; dividing the significant aggregation area by using the mean value to obtain a first temporary subset and a second temporary subset; Calculating a first difference of the simulated error and the average error in the first temporary subset, and calculating a second difference of the simulated error and the average error in the second temporary subset; Constructing an objective function based on the first difference and the second difference; And selecting a target average value from the average values so as to minimize the value of the objective function, and confirming the target average value as the critical threshold value.
  4. 4. The method for predicting yield of crops in a black soil region according to claim 1, further comprising: Correcting the meteorological big data of the black soil area to obtain corrected meteorological big data; Mapping the weather big data and the soil big data to a target grid to obtain a grid daily weather sequence, which comprises the following steps: and mapping the corrected weather big data and the soil big data to a target grid to obtain a grid daily weather sequence.
  5. 5. The method for predicting crop yield in a black soil region according to claim 4, wherein the step of correcting the meteorological big data of the black soil region to obtain corrected meteorological big data comprises: And correcting the daily precipitation in the weather big data by adopting a proportional correction method, and correcting the daily air temperature in the weather big data by adopting a difference correction method to obtain the weather big data.
  6. 6. The black field crop yield prediction method according to claim 1, wherein the weather metadata includes air temperature data, the method further comprising the steps of: constructing a first scenario scheme of actual measurement of a broadcasting period year by year and a second scenario scheme of fixed broadcasting period; Setting a suitable broadcasting boundary according to the air temperature data; calculating a first potential yield corresponding to the first scenario scheme and a second potential yield corresponding to the second scenario scheme based on the multicast-adapted boundary and the localized model; and calculating a seeding period difference quantity and a relative contribution rate of the yield according to the first potential yield and the second potential yield.
  7. 7. The method of claim 6, wherein the yield relative contribution rate is derived based on the following formula: ; Wherein, the For the said relative contribution rate, For the first potential yield; for the second potential yield.
  8. 8. The utility optimization device of black soil district land is characterized in that includes: The system comprises an acquisition module, a storage module and a control module, wherein the acquisition module is used for acquiring weather big data and soil big data of a black soil area, wherein the weather big data comprises a historical daily weather element and a future daily weather element; the mapping module is used for mapping the weather big data and the soil big data to a target grid to obtain a grid daily weather sequence; The partition module is used for partitioning the black soil region according to the resource characteristics of the black soil region to obtain a plurality of current level partitions; The correction module is used for calibrating the initial model in each current level partition by utilizing the historical observation data to obtain a localized prediction model; the evaluation module is used for performing performance evaluation on each localized prediction model, and calculating the simulation error of any space unit in the corresponding current level partition under the condition that the evaluation result does not meet the preset requirement; The autocorrelation analysis module is used for carrying out autocorrelation analysis on each simulation error to obtain a significant aggregation area, and extracting environmental parameters corresponding to the significant aggregation area; a contribution analysis module for calculating a contribution score of each of the environmental parameters to the simulation error, and confirming a dominant environmental factor and a critical threshold in the environmental parameters based on the contribution scores; The segmentation module is used for segmenting the corresponding current level partition based on the dominant environmental factors and the critical threshold to obtain a plurality of next level partitions, and calibrating an initial model in each next level partition by using historical observation data until the localized model meets the performance requirement; and the prediction module is used for carrying out daily mechanism calculation based on the localized model and the grid daily meteorological sequence to obtain the predicted yield of the black soil region crops.
  9. 9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the black soil region crop yield prediction method of any one of claims 1 to 7.
  10. 10. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program when executed by a processor implements the steps of the black soil region crop yield prediction method of any of claims 1 to 7.

Description

Method, equipment, computer equipment and storage medium for predicting crop yield in black soil area Technical Field The invention relates to the technical field of agricultural big data informatization, in particular to a method, equipment, computer equipment and storage medium for predicting crop yield in a black soil area. Background Due to the complex topography of the black soil region (flat land, staggered depressions), uneven soil layer thickness (strong degradation heterogeneity of the black soil layer) and remarkable microclimate characteristics. Currently, crop yield prediction for this area is mainly achieved through statistical models and crop growth mechanism models. The mechanism model (such as DSSAT, APSIM and the like) has stronger explanation in theory by simulating photosynthesis, transpiration water consumption and nutrient migration processes day by day. In recent years, with the perfection of an agricultural observation and informatization system, the agricultural production process forms multi-source data covering meteorological, soil and management links, and an agricultural big data foundation facing the regional scale application is gradually constructed. The data at least comprises a daily meteorological element sequence, layered soil parameters, crop management parameters and the like, can provide key input for crop yield evaluation and regional agricultural management, and can be used as yield constraint or evaluation basis in land utilization and planting structure decision. However, when the existing method utilizes agricultural big data to develop regional yield prediction, unified space-time input of a directly driven mechanism model is often difficult to form, and different source data have differences in spatial resolution, time scale and missing measurement processing caliber, so that a daily driving sequence and soil/management parameters are difficult to cooperatively organize under the same grid, and stable simulation and precision performance of the mechanism model in the regional scale are further affected. However, when the existing crop yield prediction method is applied to regional scale, the model simulation is usually carried out by adopting uniform parameters or fixed partition parameters based on experience division, and spatial heterogeneity caused by complex terrain conditions, soil degradation differences and local microclimate changes of a black soil region is difficult to fully reflect. In the process of trans-regional application of the mechanism model, model parameters often depend on manual experience setting or one-time calibration results based on limited observation samples, and a dynamic adjustment mechanism for parameter optimization and partition refinement according to model prediction error spatial distribution characteristics is lacked. When the matching degree of the model parameters and the local ecological environment conditions is insufficient, systematic prediction deviation of persistent overestimation or underestimation is easy to form in a specific area, so that stable improvement of the model prediction precision on the area scale is limited, and the requirements of refined agricultural production management on the accuracy of a yield prediction result and the space applicability are difficult to meet. Furthermore, in practical wide-range (e.g., provincial or municipal) applications, the prediction accuracy of the mechanism model is still low. Disclosure of Invention In view of the above analysis, embodiments of the present invention aim to provide a method, apparatus, computer device and storage medium for predicting crop yield in a black soil region, which solve one or more of the above problems in the prior art. In order to achieve the above object, in one aspect, an embodiment of the present application provides a method for predicting crop yield in a black soil region, including the steps of: Acquiring weather big data and soil big data of a black soil area, wherein the weather big data comprises a historical day-by-day weather element and a future day-by-day weather element; mapping the weather big data and the soil big data to a target grid to obtain a grid daily weather sequence; Partitioning the black soil region according to the black soil region resource characteristics to obtain a plurality of current level partitions; Calibrating an initial model in each current level partition by utilizing historical observation data to obtain a localized prediction model; Performing performance evaluation on each localized prediction model, and calculating simulation errors of any space unit in the corresponding current level partition under the condition that the evaluation result does not meet the preset requirement; performing autocorrelation analysis on each simulation error to obtain a significant aggregation area, and extracting environmental parameters corresponding to the significant aggregation area; Calculating a contribution