Winter wheat field soil fertility evaluation data processing method based on comprehensive weight calculation

Abstract

The invention discloses a winter wheat field soil fertility evaluation data processing method based on comprehensive weight calculation, and relates to the technical field of data processing, and the method comprises the steps: S1, collecting nutrient content and moisture retention data from field soil through a sensor network, performing grouping processing on the acquired nutrient content and moisture retention data by adopting decision node splitting and entropy calculation to obtain a soil fertility layering result; s2, according to a soil fertility layering result, acquiring root depth and leaf moisture content indexes of the winter wheat in different layers of soil from crop monitoring equipment, analyzing association strength of the root depth and leaf moisture content indexes and fertility layering by adopting leaf node classification and feature selection standards, and determining a fertility contribution coefficient; according to the winter wheat field soil fertility evaluation data processing method based on comprehensive weight calculation, efficient resource allocation and stable yield are realized, and the drought resistance and resource utilization efficiency of winter wheat in a drought environment are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a data processing method for evaluating winter wheat field soil fertility based on comprehensive weight calculation. Background Art

[0002] Winter wheat is one of the world's most important food crops, and its yield stability is directly linked to food security and sustainable agricultural development. In the context of frequent droughts, winter wheat's drought resistance has become a crucial factor in ensuring yields. In particular, how to enhance winter wheat's drought resistance through scientific management while optimizing resource allocation to achieve high yields in soils with varying fertility levels has become a key topic in agricultural research.

[0003] Currently, research on winter wheat drought resistance has largely focused on a single environment or management practice, such as the independent effects of irrigation or fertilizer application. However, these approaches often overlook the complex interactions between soil fertility and drought resistance. Existing solutions typically employ fixed patterns for fertility management and irrigation scheduling, lacking dynamic analysis of differences in drought resistance under varying soil fertility conditions. This results in difficulties in accurately matching resource allocation to actual field soil conditions in actual production, which in turn impacts winter wheat yield stability. In this context, soil fertility stratification has become a core technical factor influencing drought resistance assessment. Soil fertility stratification involves classifying field soils into different fertility levels based on differences in nutrient status, structure, and water retention capacity. Because fertility levels directly influence crop root development and water use efficiency, soils of varying fertility levels exhibit significant differences in their responses to irrigation and fertilizer application. However, current research has struggled to accurately quantify the specific contribution of fertility levels to drought resistance when assessing fertility stratification. This is due to the complex relationship between the dynamics of fertility stratification and the drought resistance requirements of crops at different growth stages, and a lack of effective technical tools to capture this dynamic interaction. The core difficulty of drought resistance assessment based on fertility stratification lies in how to accurately quantify the contribution of fertility level to drought resistance. Differences in fertility stratification will lead to differences in soil water retention and nutrient supply capacity, and these differences further affect the physiological response of winter wheat under drought conditions. For example, in low-fertility soils, the root system of winter wheat may be poorly developed due to insufficient nutrients, resulting in a decrease in water absorption capacity; while in high-fertility soils, excessive fertilization may cause water waste or soil compaction, which also reduces the drought resistance effect. This nonlinear relationship between fertility and drought resistance makes quantifying fertility contribution a complex technical problem. Summary of the Invention

[0004] The purpose of the present invention is to provide a data processing method for evaluating winter wheat field soil fertility based on comprehensive weight calculation, to quantify the differences in drought resistance of winter wheat under different fertility levels through scientific methods in field experiments, and to optimize fertility management and irrigation strategies based on these differences to improve yield stability.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for processing winter wheat field soil fertility evaluation data based on comprehensive weight calculation, the method comprising the following steps:

[0006] S1. Collect nutrient content and water retention data from field soil through a sensor network, group the collected nutrient content and water retention data using decision node splitting and entropy calculation to obtain soil fertility stratification results.

[0007] S2. Based on the soil fertility stratification results, obtain the root depth and leaf moisture content indicators of winter wheat in different soil stratifications from crop monitoring equipment. Use leaf node classification and feature selection criteria to analyze the correlation strength between root depth and leaf moisture content indicators and fertility stratification. Calculate the contribution weight of each soil fertility stratification to root depth and leaf moisture content, and determine the fertility contribution coefficient.

[0008] S3. If the fertility contribution coefficient exceeds the preset threshold, the water use efficiency change of winter wheat under drought conditions is calculated by combining the simulation module with the pruning method to obtain a quantitative value of drought resistance;

[0009] S4. Based on the quantitative values ​​of drought resistance, cross-validation evaluation and classification accuracy are used to classify and optimize fertilizer application rates and irrigation scheduling. A weighted average method is used to calculate the contribution weights of fertilizer application rates and irrigation scheduling to water use efficiency, determine the adjustment direction of the resource allocation plan, and obtain preliminary optimized configuration parameters.

[0010] S5. Extract relevant variables from the preliminary optimized configuration parameters, integrate root development data and nutrient supply data using initial parameter settings and error function definitions, and determine the dynamic balance model of resource allocation.

[0011] It can be seen from the above technical solution that the present invention has the following beneficial effects:

[0012] This winter wheat field soil fertility evaluation data processing method based on comprehensive weight calculation is aimed at the business scenario problem of insufficient stability of winter wheat yield under arid environment. It collects multi-layer soil nutrient and moisture data through a sensor network, adopts decision node splitting and entropy calculation to perform fertility stratification, analyzes the relationship between root depth and leaf moisture content and fertility stratification, and determines the fertility contribution coefficient. When the coefficient exceeds the threshold, the present invention calculates water use efficiency through a simulation module combined with pruning processing, quantifies drought resistance, and then optimizes fertilizer application and irrigation scheduling using cross-validation and classification accuracy, extracts highly correlated variables to establish a dynamic balance model. If the yield stability does not meet the target, the present invention uses gradient descent to adjust the irrigation interval and fertilizer ratio, combines historical drought data to verify the applicability of the scheme, and updates the contribution coefficient of the fertility stratification database. The present invention achieves efficient resource allocation and yield stability through multi-level data fusion and dynamic optimization, significantly improving the drought resistance and resource utilization efficiency of winter wheat under arid environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a flow chart of the data processing method for winter wheat field soil fertility evaluation based on comprehensive weight calculation of the present invention. DETAILED DESCRIPTION

[0014] 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 the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0015] like Figure 1 As shown, the present invention provides a technical solution: a method for processing winter wheat field soil fertility evaluation data based on comprehensive weight calculation, the method comprising the following steps:

[0016] S1. Nutrient content and water retention data are collected from field soils through a sensor network. The collected nutrient content and water retention data are grouped and processed using decision node splitting and entropy calculation to obtain soil fertility stratification results.

[0017] S2. Based on the soil fertility stratification results, obtain the root depth and leaf moisture content indicators of winter wheat in different soil stratifications from crop monitoring equipment. Use leaf node classification and feature selection criteria to analyze the correlation strength between root depth and leaf moisture content indicators and fertility stratification to determine the fertility contribution coefficient;

[0018] S3. If the fertility contribution coefficient exceeds the preset threshold, the water use efficiency change of winter wheat under drought conditions is calculated by combining the simulation module with the pruning method to obtain a quantitative value of drought resistance;

[0019] S4. Based on the quantitative values ​​of drought resistance, cross-validation evaluation and classification accuracy are used to classify and optimize fertilizer application and irrigation scheduling, determine the adjustment direction of resource allocation plans, and obtain preliminary optimized configuration parameters;

[0020] S5. extracting relevant variables from the preliminary optimized configuration parameters, integrating root development data and nutrient supply data using initial parameter settings and error function definitions, and determining a dynamic equilibrium model for resource allocation;

[0021] S6. If the dynamic equilibrium model shows that the yield stability index is lower than the target level, the irrigation interval and fertilizer ratio are adjusted through gradient descent update and convergence condition judgment to obtain the final resource allocation plan;

[0022] S7. Based on the final resource allocation plan, obtain drought environment verification data from the historical database, combine variable step size adjustment and simulation feedback loop to determine the applicability of the plan, and obtain yield stability assessment results;

[0023] S8. Update the contribution coefficient in the soil fertility stratification database based on the yield stability evaluation results, and determine the input data for the next evaluation cycle in combination with the proportional balance constraint and the final solution output.

[0024] This implementation utilizes a distributed sensor network to collect real-time data on field soil nutrients and moisture levels. The collected data is effectively grouped using split nodes and entropy functions within a decision tree to achieve preliminary stratification of soil fertility. To obtain crop response data, growth characteristics such as root depth and leaf moisture content are monitored, and a feature selection algorithm is applied to analyze their correlation with fertility stratification. This allows the system to derive a fertility contribution coefficient, which reflects nutrient utilization and water absorption capacity. When this coefficient exceeds a preset threshold, the system uses a simulation module and pruning strategy to calculate the water use efficiency of winter wheat under drought conditions, thereby quantifying drought resistance. Furthermore, cross-validation methods and classification accuracy metrics employed in machine learning are used to optimize and classify fertilizer application rates and irrigation strategies, initially generating a reasonable resource allocation plan. Based on this, a dynamic equilibrium model is constructed, combining crop root development characteristics with soil nutrient availability. If the model indicates yield stability is suboptimal, the resource allocation parameters are adjusted using a gradient descent algorithm to optimize the solution. Finally, the system simulates and compares the generated allocation plan with historical drought data, and uses a feedback mechanism to assess its feasibility and stability. The evaluation results are then used to update the contribution coefficient and database content, forming a closed-loop optimization path.

[0025] This implementation method significantly improves the scientificity and accuracy of fertilization and irrigation in winter wheat planting by constructing a soil fertility evaluation and resource allocation system driven by contribution coefficients. The data-driven decision-making process is used to achieve hierarchical management, enhance the system's adaptability to field heterogeneity, and improve the efficiency of soil resource utilization. At the same time, the use of dynamic adjustment and feedback mechanisms ensures the timeliness and reliability of the plan, thereby optimizing the water and fertilizer use structure and reducing resource waste while ensuring yield stability. Especially under drought conditions, the simulation and prediction of water use efficiency can help improve crop stress resistance. In addition, this method has strong adaptability and scalability, which is convenient for promotion and application in different regions and seasons.

[0026] S1 includes acquiring nutrient content and moisture data from multiple layers of field soil depth through a sensor network, storing them as a structured data set, and obtaining original soil data; preprocessing the original soil data using data cleaning technology to remove outliers and missing values ​​to obtain cleaned soil data; extracting the characteristic values ​​of nutrient content and moisture data based on the cleaned soil data, constructing a characteristic vector, and obtaining a soil feature set; splitting the soil feature set using a decision tree algorithm, determining the splitting nodes based on entropy calculation, and obtaining a preliminary grouping result; if the entropy value of the preliminary grouping result is lower than a preset threshold, merging the groups and optimizing the grouping structure to obtain an optimized grouping result; sorting the optimized grouping results through hierarchical analysis, combining the nutrient content and moisture data to determine the soil fertility stratification and obtain the final stratification result; generating soil fertility stratification data based on the final stratification result, storing it in a queryable format, and obtaining a soil fertility stratification data set.

[0027] The core of this implementation method is to obtain data for soil fertility stratification from soil at different depths through a sensor network, and process the data according to a strict process, perform feature extraction, decision tree splitting, group optimization, hierarchical sorting and data structured storage. First, the sensor network is deployed on the field plots, and three depth layers are set in each field, namely 10 cm, 30 cm and 50 cm below the surface. Each depth layer is equipped with a sensor to collect the concentrations of three nutrients: nitrogen, phosphorus and potassium, as well as the moisture content. The time interval for each collection is 1 hour, and the collection results are automatically uploaded to the central server. Each record consists of 5 fields, namely time, depth, nitrogen concentration, phosphorus concentration, potassium concentration and moisture content, which constitute a structured original soil dataset.

[0028] After the data collection is completed, the system performs data cleaning on the original soil data. The cleaning process includes two steps. The first step is to eliminate outliers. The normal range of nitrogen concentration is set to 0 to 100 mg / kg, phosphorus concentration is set to 0 to 50 mg / kg, potassium concentration is set to 0 to 150 mg / kg, and moisture content is set to 0 to 50%. All values ​​outside the above range will be directly eliminated. The second step is to fill in missing values. If a field in a record is empty and the data missing rate of the field in the depth layer is less than 5%, the field will be filled with the average value of the two adjacent time points at the same depth; if the missing rate of the field is higher than 5%, the entire record will be deleted and will no longer be included in subsequent processing.

[0029] The cleaned data is used for feature extraction. The data in each depth layer is aggregated by time, and the average values ​​of nitrogen, phosphorus, potassium and water in that time period are calculated to form a set of eigenvalues. Each set of eigenvalues ​​constitutes a eigenvector, which includes 4 specific values, namely the average value of nitrogen, the average value of phosphorus, the average value of potassium and the average value of water. After all the eigenvectors are summarized, they are input into the decision tree model for splitting. The splitting process uses information entropy as the splitting criterion. Each split first calculates the distribution ratio of each group in the current sample set under different target categories, and then calculates the information entropy based on this. The information entropy is calculated by traversing each category, taking its proportion multiplied by its logarithm, and taking the weighted sum of the results. Then evaluate the degree to which each feature reduces the information entropy, select the feature with the largest information gain as the splitting node, and split it.

[0030] The soil feature set, after data cleaning and feature extraction, was used as input data. This feature set consists of several feature vectors, each containing four values: average nitrogen concentration, average phosphorus concentration, average potassium concentration, and average moisture content. To achieve classification and splitting of this feature set, a decision tree classification algorithm was employed. The key to this algorithm is to select an optimal feature as the basis for splitting and then partition the sample based on the specific value range of this feature. This ensures that the purity of the subsample set is significantly improved with each partition. The splitting process is as follows: All samples are considered as a whole, and their classification information is calculated. The classification is based on pre-set soil fertility grade labels. If no labels are initially available, clustering is used to divide the data into several initial clusters. The total information entropy of the entire sample set is then calculated. Information entropy is calculated by calculating the proportion of each class in the sample set, multiplying this proportion by its base-2 logarithm, and then taking the negative of the weighted sum across all classes. The result is the entropy of the current sample set, which measures its degree of heterogeneity. Each splittable feature (i.e., nitrogen, phosphorus, potassium, and moisture) is tested separately. All possible values ​​of the feature are sorted, and the adjacent midpoints of the feature value are selected as candidate split thresholds. For example, if the nitrogen concentration values ​​are 10, 15, 20, 25, and 30 mg / kg, the candidate split points are 12.5, 17.5, 22.5, and 27.5. For each candidate split point, the data is divided into two groups: one less than or equal to the value and one greater than the value. The sub-information entropy of these two groups is calculated, and a weighted average is calculated based on the sample size ratio to obtain the expected entropy value at that split point. The expected entropy corresponding to the split point is subtracted from the original total entropy to obtain the information gain. This process is repeated for each candidate feature and all its candidate split points, ultimately finding the feature with the maximum information gain and its corresponding split point. This is then determined as the node and threshold for the splitting of the decision tree in this round. Using this optimal splitting feature and split point as the boundary, the soil feature set is divided into two subsets, representing the subsets of data that fall into different preliminary soil categories under the given feature. The system now obtains a valid preliminary grouping result. Repeat steps 1 through 4 for these two subsets, continuing to search for the optimal splitting feature and executing the next level of splitting until the termination criteria are met. Termination criteria include, but are not limited to, if the entropy of the current subset falls below a preset threshold (e.g., 0.2), the number of samples falls below a minimum number of samples (e.g., 10), or the information gain falls below a minimum gain threshold (e.g., 0.01), at which point further splitting ceases and the current node becomes a leaf node. This entire process is recursively executed, ultimately constructing a preliminary grouping structure consisting of multiple leaf nodes. Each leaf node represents a soil characteristic cluster, representing the preliminary results of a soil stratification cluster with similar nitrogen, phosphorus, potassium, and moisture characteristics.

[0031] After the split is complete, the system recalculates the entropy value of each group to determine its classification purity. If the entropy value of a group is lower than the set threshold of 0.2, the group is considered to have relatively concentrated information and can be considered a stable category. If there are multiple groups with entropy values ​​less than 0.2, these groups will be merged into a new group to form an optimized grouping result. The threshold of 0.2 is determined by selecting the average entropy value before the information gain significantly decreased in the historical data as the standard. 200 historical split nodes are counted, and their information entropy distribution range is 0.1 to 0.35. The dividing point of 0.2, which has strong information stability, is selected as the empirical split threshold.

[0032] The optimized grouping structure is sorted. The sorting process is divided into three steps: the first step is to standardize the four characteristic values ​​of nitrogen, phosphorus, potassium, and water in each group, and unify the numerical range to 0 to 1. The standardization method is to subtract the overall average value from the characteristic value and then divide it by its range value; the second step is to assign weights to the four standardized values, with the weights being 30% for nitrogen, 20% for phosphorus, 30% for potassium, and 20% for water; the third step is to multiply the four standardized values ​​of each group by the corresponding weights, and the sum is obtained to obtain the comprehensive score of the group, which is sorted from high to low according to the comprehensive score. The corresponding level of the sorted groups is the hierarchical level of soil fertility, which is marked as the first layer, the second layer, and the third layer from high to low.

[0033] Finally, the results of each stratification are consolidated into soil fertility stratification data. Each data entry includes the stratification level, depth range covered, average nitrogen, phosphorus, potassium, and moisture values, sampling time period, and geographic location. All resulting data is stored in a structured data format and stored in a table within the database system. Each record is indexed by time, location, and stratification level, ensuring rapid access and reuse for subsequent queries, simulations, or optimization processes.

[0034] S2 includes obtaining root depth and leaf moisture content data of winter wheat from different soil fertility layers through crop monitoring equipment, storing them as structured data sets, and obtaining original crop data; using data preprocessing technology to clean the original crop data, remove outliers and missing values, and obtain cleaned crop data; based on the cleaned crop data, extracting the characteristic values ​​of root depth and leaf moisture content, constructing characteristic vectors, and obtaining a crop feature set; using the random forest algorithm to classify the crop feature set, calculating the feature importance based on information gain, determining the correlation strength between root depth and leaf moisture content and soil fertility layers, and obtaining preliminary correlation results; if the classification accuracy of the preliminary correlation result is lower than the preset threshold, the feature vector is re-screened through the feature selection criteria, and the random forest model is optimized to obtain an optimized correlation result; by analyzing the optimized correlation results, the contribution weight of each soil fertility layer to root depth and leaf moisture content is calculated to obtain the fertility contribution coefficient; based on the fertility contribution coefficient, the growth adaptability data of winter wheat in different soil fertility layers is generated and stored in a queryable format to obtain the final growth adaptability data set.

[0035] Crop monitoring equipment installed within the soil fertility stratification zones of the field simultaneously collects root depth and leaf moisture content data for winter wheat across different strata. Root depth is measured in millimeters. The monitoring equipment records the maximum root depth once every 24 hours based on either electrical resistance response or image recognition technology, with a set effective data collection range of 10 to 500 mm. Leaf moisture content is measured as a percentage, converted from leaf infrared reflectivity, with a set data collection range of 20% to 90%. Each raw data entry contains the sampling time, corresponding soil stratum number, root depth, and leaf moisture value, forming a structured table to form the raw crop dataset. Data preprocessing is a two-step process. The first step is outlier removal: records with root depths less than 10 mm or greater than 500 mm are removed; records with leaf moisture contents below 20% or above 90% are also removed. The second step is missing value processing, checking the data table for null values. If the missing percentage of a field is less than 5%, it is filled using the mean of the two adjacent time points within the same stratum number. If the missing percentage exceeds 5%, the entire record is deleted. After processing, the data is aggregated by stratum number and sampling time interval, and the mean root depth and leaf moisture content are calculated for each group. A feature vector containing the stratum number, time interval, average root depth, and average leaf moisture content is obtained for each group, forming the crop feature set.

[0036] The crop feature set was input into a random forest classification model, with the classification target being the soil fertility stratification code. The model construction process was as follows: 100 decision trees were first constructed, each with a maximum depth of 10 layers. The number of features considered for each split was the square root of the total number of features, meaning that when two features were currently shared, one feature was used at a time. During model training, each tree was trained using bootstrap sampling with replacement from the original sample set, and the tree structure was constructed. At each split node, the model calculated the information gain of all possible split points for the currently available features. This information gain was calculated by first calculating the initial information entropy of the current sample set, then calculating the weighted average entropy of all subsets after partitioning with that feature. The difference between the two was the feature gain. The feature with the highest information gain and its corresponding split point were selected as the splitting criteria for the current node. After all trees were constructed, the system calculated the frequency of each feature being used as a split node across all trees and the weighted sum of its corresponding information gain. This was normalized to obtain the feature's importance score, known as the feature contribution value.

[0037] The characteristic contribution values ​​of root depth and leaf moisture content are their respective correlation strengths for soil stratification classification. The system then evaluates the overall performance of the model, with the evaluation indicator being classification accuracy. The accuracy is calculated by using samples outside the training set for prediction, and recording the ratio of the number of correctly predicted samples to the total number of samples. If the overall accuracy is lower than 85, the feature optimization process is re-entered. This 85 is the system's preset accuracy threshold, which is determined based on the classification stability inflection point in the historical 200 sets of model training results. A value lower than this indicates that the model is underfitting or has feature redundancy. After entering the optimization process, the system sorts the feature contribution values, retains the features ranked in the top 50 percentile, eliminates the low-contribution features, and rebuilds the random forest model. The parameters remain unchanged, and the above training process is repeated until the classification accuracy is greater than or equal to 85.

[0038] Once the model accuracy reaches the target, the system extracts the average information gain value of the corresponding features of all samples under each soil fertility stratum number, and obtains the average contribution values ​​of root depth and leaf moisture content to the stratum classification. The two contribution values ​​are weighted and averaged to obtain the fertility contribution coefficient for that stratum. This coefficient is used to represent the comprehensive impact of the stratum on crop growth. The weighting method is set according to the actual physiological dependence of the crop, with the root depth weighted at 60% and the leaf moisture content weighted at 40%. The fertility contribution coefficient values ​​corresponding to each stratum number are uniformly normalized to the range of 0 to 1 to facilitate horizontal comparative analysis. Finally, the stratum number, contribution coefficient, mean root depth, mean leaf moisture content, and sampling time records are integrated into a standardized data structure to construct growth adaptability data records. All records are uniformly stored in the database system in a structured format and indexed by stratum number and time to support on-demand query and model call.

[0039] The previous step obtained the fertility contribution coefficient for each soil fertility stratum. This coefficient is calculated by calculating the average information gain of root depth and leaf moisture content for the soil fertility stratum classification results, combining them with the weighted average of the feature weights, and then normalizing them to a range of 0 to 1. Root depth is weighted at 60%, and leaf moisture content at 40%, reflecting their dominant and supporting roles in the growth response, respectively. Each stratum number is assigned a unique fertility contribution coefficient. The system uses this coefficient to determine a growth adaptability score. The score is equivalent to the coefficient, with scores closer to 1 indicating greater crop adaptability to that fertility stratum. Next, the system constructs growth adaptability data records. Each record consists of the following fields: soil stratum number, fertility contribution coefficient, average root depth, average leaf moisture content, data collection time period, sampling area number, and growth adaptability score. All data is derived from previous model output or monitoring data processing results, requiring no additional manual input. Each field uses standardized units: root depth in millimeters, leaf moisture content in percentage, time period expressed as a date range, and area number using standard plot code format. In order to realize subsequent calls and system linkage, the system stores the above growth adaptability dataset in a structured format. The data format is a relational table structure, each record is a row, the field name is the column name, and it is stored using a standard database management system. The system sets a joint index for the layer number field and the time period field to support fast query and screening based on soil level or time range. At the same time, the system provides a unified data interface for other modules to call, including drought resistance analysis module, resource allocation module and subsequent evaluation cycle module. The final completed growth adaptability dataset can be linked and called within this system, and can also provide external interface support through standardized export formats, such as exporting to CSV format, JSON format or SQL structure, for use as a data input source for agricultural intelligent analysis systems, plot management systems or agricultural policy decision support systems, to achieve efficient sharing and multi-system collaboration.

[0040] Crop monitoring equipment is used to continuously obtain data on the growth status of winter wheat in different soil fertility layers. It mainly includes two types of equipment for monitoring root depth and leaf moisture content. For root depth monitoring, a root scanning imager can be used. This device combines a transparent root box with a high-resolution image acquisition device. Through periodic imaging and image recognition algorithms, it accurately measures the maximum root penetration depth, with measurement accuracy reaching the millimeter level. Another option is to use a buried resistance root sensor. This device is based on the principle of resistance changes in soil layers at different depths, and reflects the root distribution in real time, making it suitable for long-term field deployment. Alternatively, a segmented root monitoring probe can be used. By setting electrical contacts at different depths, the time and depth of the root contacting the sensor area are recorded, thereby inferring the root expansion. To monitor leaf moisture content, an infrared spectroscopy leaf moisture meter can be used. This device obtains non-contact moisture data by measuring changes in leaf reflectance in the near-infrared band, making it suitable for rapid measurement over large areas. Alternatively, a portable leaf moisture meter can be used. This device uses a clamp-on capacitive sensor structure to directly read leaf moisture percentage, providing stable and reliable measurement results. Furthermore, multispectral remote sensing devices can be used. Sensors mounted on drones or fixed platforms can capture red, green, blue, and near-infrared imagery, and algorithmically process it to obtain information on crop leaf moisture distribution over a large area. For situations requiring simultaneous monitoring of multiple parameters, a combined crop growth monitoring system can be used. This system integrates root probes, moisture sensors, environmental factor monitoring modules, and data acquisition terminals. Data is transmitted to a server via wired or wireless means, enabling the combined monitoring of root development and leaf status. All devices can automatically sample at a set interval, and the resulting data is stored in a structured manner for easy subsequent model calculation and analysis.

[0041] S3 includes: if the fertility contribution coefficient exceeds a preset threshold, soil fertility data and environmental factor data are obtained from the winter wheat planting area through the data acquisition module to obtain an initial environmental data set; based on the initial environmental data set, the principal component analysis method is used to extract the main feature vectors of soil fertility and environmental factors to obtain an environmental feature set; the environmental feature set is loaded through the simulation module, combined with the pruning processing method, the water use efficiency change trend of winter wheat under a drought environment is calculated to obtain an efficiency change data set; if the fluctuation amplitude of the efficiency change data set exceeds the preset range, the efficiency change data set is classified through the support vector machine algorithm to determine the key influencing factors of water use efficiency to obtain a classification result; based on the classification result, the weighted average method is used to calculate the contribution weight of the influencing factors to water use efficiency to obtain a quantitative value of drought resistance performance; through the quantitative value of drought resistance performance, combined with soil fertility data, the adaptability distribution data of winter wheat under different drought environments are generated and stored in a structured format to obtain a final adaptability data set; based on the final adaptability data set, the cluster analysis method is used to grade the drought resistance performance of winter wheat to obtain drought resistance performance grade data.

[0042] In this solution, the drought resistance analysis process is triggered when the fertility contribution coefficient for a particular soil layer is greater than or equal to 0.6. This fertility contribution coefficient is calculated by weighting the average information gain of the fertility stratification results by winter wheat root depth and leaf moisture content during the analysis process. The value ranges from 0 to 1, with 0.6 being the preset threshold. The specific value is based on a statistical analysis of the drought resistance performance of winter wheat under different fertility conditions in 200 historical samples, selecting the cutoff point where yield stability is significantly controlled by soil fertility as the critical value. When a stratification coefficient reaches or exceeds 0.6, the system uses the data acquisition module deployed in the area to collect soil nutrient data, including nitrogen, phosphorus, and potassium concentrations, and environmental factor data from the field at a daily sampling frequency for 30 consecutive days. Environmental factors include temperature, humidity, wind speed, light intensity, and rainfall. The unit of nitrogen concentration is milligrams per kilogram, with a range of 0 to 100; phosphorus is 0 to 50; and potassium is 0 to 150. The unit of temperature is degrees Celsius, with a range of 5 to 40; the unit of humidity is percentage, with a range of 20 to 100; the unit of wind speed is meters per second, with a range of 0 to 15; the unit of light intensity is watts per square meter, with a range of 100 to 1000; and the unit of rainfall is millimeters, with a range of 0 to 50. These data constitute the initial environmental dataset.

[0043] After data collection, all data were standardized and preprocessed by subtracting the mean and dividing by the standard deviation for each metric to eliminate the effects of different dimensions. Principal component analysis (PCA) was then used to reduce the dimensionality of the standardized data. The covariance matrix was calculated, and the eigenvalues ​​and corresponding eigenvectors were extracted. The eigenvalues ​​were sorted from largest to smallest, and the top principal components with a cumulative contribution of 85% or more were selected as representative variables. Typically, three to five principal components were selected to form the environmental feature set. This environmental feature set was loaded as input into the crop water use simulation module. This module uses a daily-scale crop response simulation structure to calculate the impact of each input feature combination on the water use efficiency of winter wheat. Water use efficiency is defined as the reciprocal of water consumed per unit of dry matter yield, expressed in kilograms per cubic meter. The system calculates net water use efficiency by simulating daily root water uptake and leaf evapotranspiration. The simulation was repeated for 30 consecutive days, and the output was a time series containing daily efficiency values, forming an efficiency change dataset.

[0044] To calculate the changing trend of water use efficiency of winter wheat under drought environment, the specific implementation process is as follows: first, after completing the principal component analysis, the environmental features with a cumulative contribution rate of 85% or more are numbered from high to low according to the variance contribution as the first main feature, the second main feature, and finally the Nth main feature, where N is 3 to 5, and the basis for determining N is the number of features corresponding to the first time the cumulative contribution rate reaches or exceeds 85% in the historical samples; then the above environmental features are loaded into the simulation module as the only input, and the time step is set to 1 day, the simulation period is 30 days, and the drought scenario is a daily sequence with soil moisture content lower than 40% of the field water holding capacity and rainfall less than or equal to 2 mm. The field water holding capacity threshold and rainfall threshold are both obtained from the statistics of 200 groups of historical samples, where 40% water content is the critical point at which crop leaves significantly wilt and the root water absorption rate significantly decreases, and 2 mm rainfall is the upper limit at which the effect on effective soil water replenishment is not significant; before the simulation begins, all input features are normalized on a daily scale, and the characteristic values ​​of each day are calculated by the mean and extreme values ​​of the same period. The difference is dimensionless to ensure that the effects of different dimensions in the model are comparable; on each simulation day, the simulation module executes four sub-steps in sequence. Sub-step 1 is the estimation of root water absorption. The effective root water absorption rate is updated daily based on the comprehensive changes of temperature, humidity, light, wind speed, and rainfall corresponding to the 1st to Nth main features, and the volume of water absorbed by the root system per unit time is output; Sub-step 2 is the estimation of evapotranspiration. The comprehensive effects of leaf evaporation and stomatal conductance are updated daily based on the same characteristics, and the evapotranspiration water volume per unit time is output. ; Sub-step 3 is the calculation of net available water, which is obtained by subtracting evapotranspiration from root water absorption. When this value is less than 0, it is recorded as 0 to prevent the non-physical situation of negative water supply; Sub-step 4 is the calculation of water use efficiency, which is the reciprocal of the ratio of the daily dry matter increment to the daily total water consumption. The calculation result is recorded as the daily efficiency value in kilograms per cubic meter; the above four sub-steps are executed in a daily cycle until the 30th day, and an efficiency time series with a length of 30 is obtained.In order to reduce redundancy and improve stability, the environmental features involved in the calculation are screened by combining the pruning method. The pruning rule is to implement a positive perturbation of 10% and a negative perturbation of 10% on each main feature, record the average change ratio of water use efficiency, and judge the features with an average change ratio less than 5% as non-critical features and remove them from the next simulation day. The 5% threshold is derived from the distribution statistics of the perturbation sensitivity in 500 efficiency time series, which is at the upper limit of the acceptable error range. In the re-simulation after pruning, only the remaining features are retained, the above four sub-steps are repeated, and the 30-day efficiency series are recalculated. Then, the consistency test of the two rounds of series is performed. If the daily average difference of the two rounds of series is less than 0.05 and the range difference is less than 0.1, the pruned result is used. Otherwise, the last unpruned result is retained to avoid information loss caused by oversimplification. ; Calculate the fluctuation indicators of the final efficiency time series, including the range, standard deviation and the daily average of the change rate between two adjacent days, and write the daily efficiency value, range, standard deviation, daily average change rate, time index, corresponding main feature number and its daily value into the efficiency change dataset. The dataset is stored in a structured table format. The fields include date, layer number, daily efficiency value, range, standard deviation, daily average change rate, number of retained main features and daily value of each main feature. The layer number comes from the fertility layer identifier that triggers this process. The daily efficiency value is used for subsequent classification and quantitative calculation. The range and standard deviation are used to determine the fluctuation amplitude. The daily average change rate is used to identify the continuous trend. The number of retained main features is used to restore the pruning scale. The daily value of each main feature is used to trace the cause of the efficiency change. The generation of the efficiency change dataset is completed.

[0045] Subsequently, the system performs a fluctuation amplitude analysis on the data set and calculates its range, standard deviation, and rate of change. If the range exceeds 0.4 or the standard deviation exceeds 0.2, it means that the water use efficiency response is unstable, and the system needs to further determine the main influencing factors. These two thresholds are derived from the statistical mean setting of the model for 500 sets of historical efficiency data to ensure timely identification of abnormal fluctuations. At this time, the system calls the support vector machine classification algorithm, takes environmental characteristics as input and efficiency fluctuation categories as output, and performs binary classification modeling. The model uses a radial basis kernel function, sets the kernel width to 0.5, the penalty factor to 1, and the training samples are all sampling points in the efficiency change data set. After training, the classification boundary is output and the degree of support of each feature for the boundary construction is identified. The support is calculated as the product of the average boundary distance of the feature in the support vector set and the classification accuracy. After normalization, the feature importance ranking is obtained.

[0046] The top three environmental factors were extracted as key influencing factors, and their weighted impact values ​​on changes in water use efficiency were calculated. The weighting method is to add the contribution values ​​of each factor and divide them by the total number so that the normalized sum of the contribution values ​​is 1. The resulting weighted total value is the quantitative value of drought resistance, ranging from 0 to 1. The closer the quantitative value of drought resistance is to 1, the stronger the adaptability of winter wheat to drought under the current soil fertility and environmental characteristics. The system then integrates the quantitative value of drought resistance corresponding to each soil layer with the nutrient content and main environmental factor values ​​of the layer to generate a complete record including layer number, nitrogen, phosphorus and potassium concentrations, key environmental factor values, simulation time period, efficiency curve trend and quantitative value of drought resistance, constructs adaptive distribution data, and writes it into the database table in a structured data format.

[0047] All recorded data were clustered using the K-means clustering algorithm. The number of clusters was set to 3, and the initial cluster centers were set to 0.2, 0.5, and 0.8, representing low, medium, and high drought resistance levels, respectively. The algorithm used Euclidean distance as the distance metric, and convergence was determined when the centroid movement distance was less than 0.001 or a maximum of 100 iterations. After clustering, the system assigned a grade based on the size of the cluster center, dividing the quantitative drought resistance value into three intervals. The system then output the corresponding results, mapping the stratification number to the drought resistance grade. This final drought resistance grade dataset was constructed, serving as an input for subsequent fertilization and irrigation optimization and resource scheduling strategies.

[0048] S4 includes classifying fertilizer application amount and irrigation scheduling by cross-validation method based on the quantitative value of drought resistance to obtain a classification accuracy data set; if the fluctuation range of the classification accuracy data set exceeds the preset threshold, the combination of fertilizer application amount and irrigation scheduling is ranked by feature importance using the random forest algorithm to determine the key resource allocation factors and obtain a factor ranking data set; based on the factor ranking data set, the weighted average method is used to calculate the contribution weight of fertilizer application amount and irrigation scheduling to water use efficiency to obtain a weight distribution data set; if the fertilizer application amount weight in the weight distribution data set is higher than the irrigation scheduling weight, the linear regression method is used to analyze the relationship between fertilizer application amount and water use efficiency to obtain fertilizer optimization adjustment parameters; based on the fertilizer optimization adjustment parameters, combined with environmental factor data, the data standardization method is used to process soil fertility data to generate a resource allocation optimization data set; based on the resource allocation optimization data set, the irrigation scheduling scheme is grouped using the cluster analysis method to obtain irrigation scheduling optimization parameters; based on the irrigation scheduling optimization parameters and the fertilizer optimization adjustment parameters, preliminary optimization configuration parameters are generated and stored in a structured format to obtain the final optimization configuration data set.

[0049] Using the quantitative value of drought resistance as the source of classification labels and fertilizer application rate and irrigation scheduling as the independent variables to be classified, a sample set for cross-validation was first constructed. Each sample contained the quantitative value of drought resistance, fertilizer application rate, irrigation cycle, single irrigation water volume, irrigation start threshold, irrigation stop threshold, sampling time and soil stratification number, and corresponding water use efficiency. The fertilizer application rate was measured in kilograms per hectare, the irrigation cycle was measured in days, the single irrigation water volume was measured in millimeters, the irrigation start and stop thresholds were measured in percentages of field water holding capacity, and the quantitative value of drought resistance was a quantitative indicator between 0 and 1 calculated in the previous step. The quantitative value of drought resistance was divided into high adaptability and low adaptability according to the threshold of 0.6, which was consistent with the previous step. The significant adaptability dividing point was determined by the yield stability comparison statistics of 200 groups of historical samples.

[0050] Then, cross-validation classification is performed, and the samples are randomly divided into 10 parts and rotated in sequence. In each round, one part is taken as the validation set and the remaining 9 parts are taken as the training set. The classifier adopts a deterministic threshold rule: the median of the fertilizer application amount and the median of the irrigation cycle within each soil layer are used as the dividing line. When the fertilizer application amount of a single record is not lower than the dividing line and the irrigation cycle is not longer than the dividing line, the high adaptability class is output; otherwise, the low adaptability class is output. After completing one round of training and validation, the accuracy of this round is calculated. The accuracy is calculated by dividing the number of correctly predicted samples in the validation set by the total number of samples in the validation set and converting it into a percentage. The ratio is recorded in the classification accuracy data set; after 10 rounds of repetition, 10 accuracy values ​​are obtained, and the fluctuation amplitude indicators are calculated, including the range and standard deviation, where the range is the maximum value minus the minimum value, and the standard deviation is the average dispersion of the accuracy of each round from the mean; if the range is greater than 0.1 or the standard deviation is greater than 0.05, it is determined that the fluctuation exceeds the threshold and enters the feature sorting link, otherwise skip this link and directly enter the subsequent weight calculation. The above two thresholds are determined by statistically analyzing the distribution of the cross-validation results of the most recent 200 batches and taking the value at the upper limit of the stable interval to ensure sufficient sensitivity to abnormal fluctuations.

[0051] When feature sorting is triggered, a random forest classification model is trained. The input features are limited to fertilizer application amount, irrigation cycle, single irrigation water volume, irrigation start threshold, and irrigation stop threshold. The targets are the aforementioned high-fitness and low-fitness classes. The number of trees in the forest is set to 200 to reduce variance, the maximum depth is set to 12 layers to avoid overfitting, and the minimum number of leaf node samples is set to 10 to ensure statistical stability. The number of features considered for each split is the square root of the total number of features, rounded down to introduce randomness. The importance metric uses the average contribution based on the reduction in node purity. After training, the importance score of each feature is obtained and sorted from high to low to form a factor-sorted dataset.

[0052] In the weight calculation stage, the scores belonging to the fertilizer category are summed to obtain the total fertilizer score, and the scores belonging to the irrigation category are summed to obtain the total irrigation score. The minimum and maximum values ​​of the two total scores are mapped so that they fall between 0 and 1. The fertilizer application weight and irrigation scheduling weight are obtained according to the rule that the sum of the two is equal to 1, forming a weight distribution data set. At the same time, the feature sorting snapshot based on which the weight is generated is recorded for traceability.

[0053] If the fertilizer application weight in the weight distribution data set is greater than the irrigation scheduling weight, a linear regression analysis is performed to analyze the relationship between fertilizer application and water use efficiency. Specifically, water use efficiency is used as the dependent variable and fertilizer application amount is used as the independent variable. The samples are divided into a training set and a validation set in an eight-to-two ratio, and the random seed is fixed to ensure reproducibility. After fitting a straight line in the training set, the slope, intercept, and determination coefficient are output. If the slope is positive and the determination coefficient is not less than 0.6, the minimum application amount increment required to achieve a 5% increase in water use efficiency is calculated and recorded as a positive adjustment parameter. If the slope is negative and the determination coefficient is not less than 0.6, the minimum application amount increment required to achieve a 5% increase in water use efficiency is calculated and recorded as a positive adjustment parameter. The maximum possible reduction range under the condition that the risk of water use efficiency decline does not exceed 2% is recorded as a negative adjustment parameter. If the determination coefficient is less than 0.6, no adjustment range is produced and a suggestion to maintain the status quo is recorded. The above 0.6 threshold comes from the lower limit statistics of the historical fitting goodness of fit to ensure that the regression explanatory power meets the standard; then a resource allocation optimization data set is generated, and the fertilizer optimization adjustment parameters obtained by regression are merged with the environmental factor data of the same time window, and the soil fertility data is standardized. The standardization method is to calculate the mean and standard deviation of the nitrogen, phosphorus and potassium concentrations respectively, and then subtract the mean from the measured value of each record and divide it by the standard deviation. To ensure the comparability of different dimensions in subsequent clustering, each generated record contains the soil layer number, standardized nitrogen, phosphorus and potassium, environmental factor values, fertilizer optimization adjustment parameters, current irrigation parameters and corresponding water use efficiency; in the irrigation grouping link, the irrigation cycle, single irrigation water volume, irrigation threshold and irrigation stop threshold in the resource allocation optimization dataset are used as clustering features. Cluster analysis is performed and the number of clusters is set to 3. The initial centroid takes the 25th, 50th and 75th percentiles of each feature as the starting point. The maximum number of iterations is set to 100 times. Convergence is determined by the moving distance of the centroid between two adjacent iterations. The clustering is stopped when the distance is less than 0.001. After clustering is completed, the four central values ​​of each cluster are output as the irrigation scheduling optimization parameters and the intra-cluster variance is recorded to measure the consistency within the group. Finally, the irrigation scheduling optimization parameters and the fertilizer optimization adjustment parameters are aligned and merged according to the soil layer number and time. A structured record including the layer number, fertilizer optimization direction, fertilizer adjustment range, irrigation cycle, single water volume, irrigation start threshold, irrigation stop threshold, generation time and description fields based on weight is generated and stored in the final optimized configuration data set. At the same time, a joint index of layer number and generation time is established on the data table to support efficient query and version tracing.

[0054] S5 includes using the principal component analysis method to extract correlation variables from the preliminary optimization parameters to generate a variable screening data set; if the number of variables in the variable screening data set exceeds a preset threshold, the correlation coefficient matrix between the variables is calculated by the correlation analysis method to obtain a key variable combination; based on the key variable combination, the root development data and nutrient supply data are integrated to generate a comprehensive feature data set; through the comprehensive feature data set, a dynamic balance model of resource allocation is constructed using the support vector machine method to obtain a preliminary dynamic model; if the prediction error of the preliminary dynamic model exceeds a preset threshold, the model parameters are optimized by the gradient descent method to obtain an optimized dynamic model; based on the optimized dynamic model, combined with the root development characteristics and nutrient supply characteristics, resource allocation adjustment parameters are generated to determine the final dynamic balance plan.

[0055] In this step, all numerical fields in the preliminary optimization parameters are first standardized. The processing method is to remove the mean and scale each field using the mean and standard deviation of the field in the current data window to eliminate dimensional differences and provide stable input for subsequent principal component extraction.

[0056] The principal component analysis method is then used to extract correlation variables from the standardized parameters. The specific process is to calculate the covariance matrix and sort it from large to small by eigenvalue, and then accumulate the variance contribution rate. When the cumulative variance contribution rate reaches or exceeds 85 for the first time, the extraction is stopped and the corresponding set of high-load original variables of the principal component at this time is included in the variable screening data set. The high-load judgment standard is that the absolute load value on the current principal component ranks among the top 3 of all variables of the principal component and the absolute load value is not less than 0.5. This dual standard is determined by obtaining the balance point between explanatory power and redundancy in the interpretability evaluation of 300 sets of historical parameter sets; after completing the variable screening, the number of variables is counted. If the number exceeds the preset threshold of 5, the correlation analysis method is performed to further compress redundancy. The preset threshold of 5 is determined based on the cross-validation results of the number of variables and model stability in 300 sets of historical samples. When there are more than 5 variables, the generalization performance of the support vector machine begins to decline significantly, so 5 is set as the upper limit.

[0057] The correlation analysis was measured using the Pearson correlation coefficient. The correlation coefficient matrix between variables was calculated and screened by absolute value. When the absolute correlation coefficient between any two variables was not less than 0.7, they were determined to be highly correlated pairs. Only the variable with the larger combined absolute load on the principal component set was retained. The threshold of 0.7 was derived from the critical point of stable redundancy removal in the structural test of 300 groups of samples. After obtaining the key variable combination, the root development data and nutrient supply data were strictly integrated according to the time index and soil stratification number. The root development data fields were fixed as average root length, maximum root penetration depth and root density, with units of centimeters, millimeters and roots per square centimeter respectively. The nutrient supply data fields were fixed as nitrogen supply rate, phosphorus supply rate and potassium supply rate, with units of millimeters. Grams per kilogram per day, the alignment rule is to take the time-weighted average when there are multiple records with the same layer number and the same day, and fill in the missing record ratio with the average of the adjacent days before and after when it is less than 5, and remove the record when it exceeds 5. Finally, a comprehensive feature data set including a combination of key variables and six physiological and nutrient indicators is generated; the comprehensive feature data set is used as the only training input, and the support vector machine method is used to construct a dynamic balance model of resource allocation. The model output is the discriminant value of whether the resource allocation has reached dynamic balance and the corresponding balance score. The kernel function uses the radial basis kernel function, the kernel width is set to 0.5, and the penalty factor is set to 1. Both are selected by grid search on a discrete grid with a kernel width of 0.1 to 1 and a penalty factor of 0.1 to 10 through 10-fold cross validation. The minimum point of verification error is determined, the ratio of training and verification sets is 8 to 2, and the random seed is fixed to ensure reproducibility. After training, the mean square error is calculated using the verification set as the prediction error; if the prediction error exceeds the preset threshold of 0.1, the gradient descent method is started to optimize the model parameters. The threshold of 0.1 comes from the minimum upper bound that separates the stable performance segment from the unstable performance segment in the error distribution statistics of 200 groups of samples. The learning rate of gradient descent is fixed to 0.01 and is determined as a compromise point after comparing the convergence speed and overshoot risk through cross-validation in the range of 0.001 to 0.05. The maximum number of iterations is set to 500 to limit the computational cost. The convergence criterion is that the verification error of two consecutive iterations decreases by less than 0.001, which stops, and the direction of parameter update is based on the verification error. The verification error is applied in the opposite direction of the numerical gradient of the kernel width and the penalty factor. The numerical gradient is calculated using small perturbation differences without changing the model structure. Once the optimization is complete and the optimized dynamic model is obtained, the resource allocation adjustment parameter generation step begins. The balance score of the optimized dynamic model for the comprehensive feature dataset under the current stratification number and time index is used as the target. The sensitivity of the score to local changes in three adjustable resource quantities: fertilizer application rate, irrigation cycle, and single irrigation water volume is calculated using a numerical perturbation method. The perturbation amplitude is fixed at 5% of the respective historical median values. Three sensitivities are then generated based on the consistency of the sensitivity and directionality. The correction value is determined by raising the balance score to no less than 0 without crossing the physiological safety boundary.The minimum step size is 8, the physiological safety boundary is that the fertilizer application rate does not exceed the 90th percentile and is not lower than the 10th percentile of the current stratum history, the irrigation cycle is no shorter than 3 days and no longer than 30 days, and the single irrigation water volume is no less than 5 mm and no higher than 40 mm. The score target of 0.8 is determined by the minimum compliance score obtained in the joint quantile analysis of yield stability and water use efficiency of the qualified samples. The three correction values ​​are combined with the average root length, maximum root penetration depth, root density, and three nutrient supply rates to form a resource allocation adjustment parameter record. The model's judgment of balance or imbalance is recorded together with the balance score. Finally, the final dynamic balance plan is aggregated by stratum number and time index. The plan is saved in a structured format and a joint index is established for the stratum number and time index to support fast query and traceability.

[0058] S6 includes obtaining a yield stability index through a dynamic equilibrium model, calculating the deviation between the index and a preset threshold using a statistical analysis method, and obtaining a deviation data set; based on the deviation data set, updating the irrigation interval parameter using a gradient descent method, and generating an adjusted irrigation parameter set in combination with environmental adaptability data; obtaining a fertilizer ratio parameter through the adjusted irrigation parameter set, and predicting the parameter adjustment effect using a linear regression method to obtain a prediction error value; if the prediction error value exceeds a preset threshold, adjusting the gradient descent step size based on the convergence condition to generate an optimized irrigation parameter set; obtaining agricultural resource allocation data based on the optimized irrigation parameter set, dividing resource allocation priorities using a cluster analysis method, and obtaining a resource allocation plan; generating a final resource allocation plan based on the resource allocation plan and combining it with environmental adaptability data; obtaining a yield stability index based on the final resource allocation plan, judging whether it reaches the preset threshold, and obtaining a verification result.

[0059] In this step, for each layer number and time index, a 30-day daily production sequence is extracted from the dynamic equilibrium model, and the daily mean, daily standard deviation, coefficient of variation, and low-yield abnormality ratio of the sequence are calculated. The coefficient of variation is defined as the ratio of the standard deviation to the mean, and the low-yield abnormality ratio is defined as the proportion of days with a yield lower than the mean minus two standard deviations. The coefficient of variation is weighted by 70% and the low-yield abnormality ratio is weighted by 30% to obtain a yield stability index score between 0 and 1, and 0.8 is set as the preset threshold. The value of 0.8 is based on the 20 The minimum compliance score is obtained by the joint quantile statistics of yield stability and subsequent realization rate in the 0 group of historical samples; then the absolute difference between the yield stability index score and the threshold is calculated for each record as the deviation value, and a deviation data set containing four fields: layer number, time index, yield stability index score, and deviation value is generated; based on the deviation data set, the gradient descent update process of the irrigation interval parameter is started, and the initial irrigation interval is the interval number of days in the current execution plan, with the value boundary of the shortest 3 days and the longest 30 days, the initial step size is 2 days, and the direction is determined by the finite difference sensitivity method. Under the condition of not crossing the boundary, the irrigation interval is divided into the following categories: The irrigation interval was shortened by one day and extended by one day respectively, and the dynamic balance model was called to recalculate the new 30-day yield stability index score. The difference between the two scores and the original score was compared, and the direction that could increase the score was selected as the update direction. In order to reflect the constraints of environmental differences on the adjustment range, the quantitative value of drought resistance in the environmental adaptability data was introduced to map the step size into three-level correction coefficients. When the quantitative value of drought resistance is less than 0.4, the correction coefficient is 1.2; when it is between 0.4 and 0.7, the correction coefficient is 1.0; when it is greater than or equal to 0.7, the correction coefficient is 0.8. The three-level thresholds are derived from the 200 A stratified statistical analysis of the stability of historical samples at different drought resistance levels strikes a balance between avoiding excessive fluctuations and ensuring effective improvement. The initial step size is multiplied by a correction coefficient to obtain the actual update step size, and the irrigation interval is adjusted in the determined direction. If the value boundary is reached, the irrigation interval is truncated at the boundary. After each update, the yield stability index score and deviation are immediately recalculated and recorded in the adjustment log. Iterations continue until any convergence condition is met. Convergence conditions include a deviation decrease of less than 0.01 for three consecutive iterations or a cumulative number of iterations of 20. If the deviation increases for two consecutive times, the current step size is multiplied by 0.5 and then continue to iterate. The above step size, boundary and convergence threshold are determined by the convergence speed and overshoot risk comparison test of historical samples, which can achieve stable improvement without increasing too much computational cost. After obtaining the adjusted irrigation parameter set, the fertilizer ratio parameters are obtained according to the layer number and production season. The rule for determining the fertilizer ratio parameters is to give priority to the median value of the nitrogen fertilizer ratio, phosphorus fertilizer ratio and potassium fertilizer ratio of the historical execution records of the same layer and the same season. If the historical records of the layer are less than 10, the latest ratio in the final optimized configuration data set of the previous sequence is used as the benchmark and the existing execution ratio of the season is weighted smoothed with weights of 70% and 30% to reduce accidental fluctuations. Subsequently, a linear regression model is established to predict the effect of parameter adjustment, with water use efficiency as the The dependent variable is the irrigation interval and the proportion of nitrogen fertilizer, phosphate fertilizer and potash fertilizer as the independent variables. The training and validation are divided into 8 to 2 and the random seed is fixed to ensure the reproducibility of the results. The prediction error value is calculated on the validation set. The prediction error value is defined as the average of the absolute value of the difference between the predicted water use efficiency and the actual observed water use efficiency. 0.1 is used as the preset error threshold, which comes from the upper bound statistics of the stable interval of 200 groups of historical validation error distributions. If the prediction error value does not exceed 0.1, the current parameters are accepted and the resource allocation stage is entered. If the prediction error value exceeds 0.1, the adaptive step size adjustment link is entered, and the gradient descent step size is modified according to the convergence condition. The convergence condition is that the prediction error after two adjacent regressions decreases by no less than 0.01 and The condition is satisfied three times in a row. If it is not satisfied and the error increases, the step size is multiplied by 0.5. If the error does not decrease or increase, the step size is kept unchanged. The step size corresponds to the daily adjustment range of the irrigation interval and is limited to 1 to 7 days. After the adaptation is completed, the irrigation interval update and regression evaluation are re-executed. A maximum of 5 rounds are performed to limit the computational overhead, and the optimized irrigation parameter set is output. The agricultural resource allocation data is obtained based on the optimized irrigation parameter set. The agricultural resource allocation data includes two main category fields, the number of irrigation interval days and the ratio of three fertilizers, and two management fields, the layer number and the time index. After the above fields are normalized with the range, cluster analysis is performed to divide the resource allocation priority. The number of clusters is fixed at 3 categories, and the 25th and 50th percentiles of each field are used for initialization. The 75th percentile was used as the initial center of the three groups. The standardized Euclidean distance was used as the similarity metric, with a maximum number of iterations of 100 and a stopping threshold of 0.001 for center movement. After clustering, the three clusters were mapped to priority 1, priority 2, and priority 3 according to the weighted scores of the average yield stability index score and average water use efficiency of each cluster from high to low, with weights of 70% of the yield stability index score and 30% of the water use efficiency. This formed a resource allocation plan. The resource allocation plan was combined with the environmental adaptability data to generate the final resource allocation plan. The combination rule was that when the quantitative value of the drought resistance performance of the layer was less than 0.4, the priority of the corresponding cluster of the layer was increased by one level but not exceeding priority 1. When the quantitative value of the drought resistance performance of the layer was between 0.4 and 0.7, the cluster priority remains unchanged, but a 10% resource bias is applied to irrigation intervals and nitrogen fertilizer proportions. When the quantified drought resistance value of the layer is greater than or equal to 0.7, the original plan is maintained without additional bias. Finally, this plan is fed back into the dynamic equilibrium model to generate a new 30-day yield sequence. The new yield stability index score is calculated using the same method as the first step and compared with the threshold of 0.8. If the score is greater than or equal to 0.8, the verification result is recorded as passed. If the score is less than 0.8, it is recorded as failed, and the current parameter set and corresponding deviation value are retained for the next round of iteration. The pass and fail records are stored in a structured form and a joint index is established based on the layer number and time index.

[0060] S7 includes obtaining verification data under a drought environment from a historical database, using a data screening method to extract a feature data set related to the current resource allocation plan to obtain a feature data set; using the feature data set, combined with a variable step adjustment method, to calculate the matching degree between the resource allocation plan and the verification data to obtain a matching analysis result; if the matching analysis result is lower than a preset threshold, the variable step is adjusted through a simulation feedback loop to generate an optimized matching analysis result; based on the optimized matching analysis result, a clustering analysis method is used to divide the priority of resource allocation to obtain a resource allocation priority data set; based on the resource allocation priority data set, combined with the constraints of the drought environment, an adjusted resource allocation plan is generated; based on the adjusted resource allocation plan, a simulation feedback loop is used to evaluate yield stability to obtain a yield stability evaluation result; based on the yield stability evaluation result, it is determined whether the preset threshold is met to generate a final resource allocation plan.

[0061] First, verification data under drought conditions were retrieved from the historical database. The drought conditions were screened with daily rainfall less than or equal to 2 mm and soil moisture content less than 40% of the field water holding capacity. This threshold adopted the unified caliber of the previous system to ensure caliber consistency. To ensure comparative equivalence, only records of the same crop growth stage, the same soil fertility layer, and the same unit as the current resource allocation plan were retained in the last three years. The screened fields were fixed as the total fertilizer application amount, nitrogen fertilizer ratio, phosphorus fertilizer ratio, potassium fertilizer ratio, irrigation interval days, single irrigation water volume, root depth mean, leaf moisture content mean, water use efficiency and time index. Data exceeding the agricultural production limit were deleted. The abnormal values ​​within the reasonable range of technology are as follows: total fertilizer application amount 0 to 200 kg per hectare, nitrogen fertilizer ratio 0 to 60, phosphorus fertilizer ratio 0 to 40, potassium fertilizer ratio 0 to 40, irrigation interval 3 to 30 days, single irrigation water volume 5 to 40 mm, root depth 10 to 500 mm, leaf moisture content 20 to 90, water use efficiency greater than 0; for fields with a missing ratio not higher than 5, the average value of the adjacent two days is used to fill in the missing ratio, and records with more than 5 are removed. After completion, a feature data set is generated; then the matching degree between the current resource allocation plan and each record of the feature data set is calculated. The calculation order is to first calculate the absolute value of the difference of each pair of parameters, and then The single parameter deviation value is normalized by the range of the parameter in the feature data set. The closer the single parameter deviation value is to 0, the better the match is. All single parameter deviation values ​​are weighted and averaged to obtain the comprehensive deviation. Then, the comprehensive deviation is subtracted from 1 to obtain the matching degree. The matching degree range is 0 to 1. The weights are given by the weight distribution data set mentioned above. The fertilizer application-related weights and irrigation-related weights are determined according to the data set. If the fertilizer-related weights need to be distributed among nitrogen, phosphorus and potassium, they are distributed in equal proportion to avoid bias. After completing all the matching degree calculations, the matching degree analysis results are obtained and judged with 0.7 as the preset threshold. The threshold is derived from the minimum upper bound separating reliable and unreliable data from the distribution statistics of 200 historical matching records. If the matching degree of any record falls below 0.7, a simulation feedback loop is entered to adaptively adjust the variable step size to improve the matching degree. The adaptive rule sets the initial step size of each adjustable parameter equal to 5 of the parameter's historical value range, such as irrigation interval in days, single water volume in millimeters, and three fertilizer ratios in percentages, and limits the boundaries to the above agronomic ranges. If the matching degree does not improve for two consecutive rounds, the parameter step size is multiplied by 0.5. If the matching degree improves by 0.05 or more for two consecutive rounds, the step size is multiplied by 1.2. Each round of update is executed one by one according to the parameters. The execution order is irrigation interval, single irrigation water volume, nitrogen fertilizer ratio, phosphorus fertilizer ratio, and potassium fertilizer ratio. The update direction is based on reducing the comprehensive deviation, that is, try one step in each positive and negative direction within the boundary and calculate the new matching degree. Select the step with higher matching degree as the actual update. If it touches the edge, it will be truncated at the boundary. After completing a round, the matching degree of all records is recalculated and the round, step size and matching degree are recorded. The cycle is repeated until the matching degree of all records is not less than 0.7 or reaches the upper limit of 20 rounds, forming the optimized matching degree analysis result. After obtaining the optimized matching degree analysis result, the resource allocation priority is divided, and the mean iteration clustering method is used to cluster the resource allocation. All schemes are clustered according to two types of features: matching degree and comprehensive value of key parameter deviation. The comprehensive value of key parameter deviation is the weighted average of the deviation values ​​of irrigation interval and single water volume, and the weights of the two are equal. The number of clusters is fixed at 3 to correspond to the three priorities of high, medium and low. The initial centers are the 25th, 50th and 75th percentiles of the above two types of features. The distance measurement is the Euclidean distance after the two types of features are normalized by the range. The maximum number of iterations is 100 times. The convergence criterion is that the distance between two adjacent center moves is less than 0.001. After clustering is completed, the average matching degree of each cluster is assigned priority 1, priority 2 and priority 3 from high to low, and the resource allocation priority dataset is output. On this basis, the drought environment constraint conditions are superimposed to generate an adjusted resource allocation plan. The constraint conditions are fixed as the single irrigation water volume is not more than 40 mm, the irrigation interval is not less than 3 days, the nitrogen fertilizer ratio is not higher than 60, and the total fertilizer application amount does not exceed 200 kg per hectare. If any parameter exceeds the limit, the parameter is adjusted to the allowable boundary and the other parameters remain unchanged to reduce the linkage error; the adjusted resource allocation plan is then evaluated for yield stability. The evaluation method is to input the plan back into the dynamic equilibrium model to generate a 30-day daily scale yield sequence and calculate the yield stability index, which is obtained by weighting the coefficient of variation and the low-yield abnormality ratio with inverse weights of 70% and 30%. The yield stability evaluation result is obtained. The yield stability index for any scheme is compared with a threshold of 0.8, which is derived from the joint quantile statistics of stability and achievement rates for 200 historical samples. If the yield stability index for any scheme falls below 0.8, fine-tuning is performed within the simulation feedback loop using a daily equivalent adjustment amplitude of 2 variable steps. Prioritizing one positive and one negative minimum step for each irrigation interval and single water volume, if the target is still not met, a minimum step adjustment is then made for each of the three fertilizer ratios. The yield stability index is recalculated immediately after each fine-tuning, with a maximum of 10 rounds of fine-tuning allowed. If the target is still not met, the scheme is marked as failed and its deviation and step size trajectory are retained for the next evaluation cycle. When the yield stability index for all schemes is greater than or equal to 0, the scheme is considered to have failed.At 8:00 AM, the system determines that the preset threshold has been met and outputs the final resource allocation plan in a structured format. The record fields include layer number, time index, nitrogen fertilizer ratio, phosphorus fertilizer ratio, potassium fertilizer ratio, irrigation interval, single water volume, matching degree, priority, yield stability index, and the values ​​and source descriptions of all key thresholds and parameters. A joint index is also established on the layer number and time index to support fast query and audit traceability.

[0062] S8 includes obtaining the contribution coefficient in the soil fertility stratification database through the yield stability assessment results, updating the contribution coefficient by using the data fusion method, and obtaining an updated contribution coefficient data set; according to the updated contribution coefficient data set, combined with the proportional balance constraint, the linear regression algorithm is used to calculate the weight distribution of the resource allocation scheme to obtain the weight distribution result; if the weight distribution result meets the preset constraint conditions, the candidate input data set for the next assessment cycle is extracted by the data screening method to obtain the candidate input data set; according to the candidate input data set, the clustering analysis method is used to divide the data priority to obtain the priority sorting data set; if the matching degree between the priority sorting data set and the resource allocation scheme is lower than the preset threshold, the input data set is adjusted by the iterative optimization method to obtain the optimized input data set; according to the optimized input data set, combined with the constraint conditions of the assessment cycle, the input data set for the next assessment cycle is generated.

[0063] In this step, the yield stability index score is first read for each layer number and time index. The score comes from the previous step weighted by the coefficient of variation and the low-yield abnormality ratio, and the value range is 0 to 1; the existing contribution coefficient of the same layer is read from the soil fertility layer database at the same time, and the value range is 0 to 1; then data fusion is performed to update the contribution coefficient. Specifically, the yield stability index score is first mapped to the evidence coefficient and the fusion weight is set accordingly. The mapping rule is that when the yield stability index score is greater than or equal to 0.9, the evidence coefficient is 1.0 and the fusion weight is 0.7; when the yield stability index score is between 0.8 and 0.9, the evidence coefficient is 0.9 and the fusion weight is 0. The fusion weight is 0.6. When the yield stability index score is between 0.7 and 0.8, the evidence coefficient is 0.8 and the fusion weight is 0.5. When the yield stability index score is between 0.6 and 0.7, the evidence coefficient is 0.6 and the fusion weight is 0.4. When the yield stability index score is less than 0.6, the evidence coefficient is 0.5 and the fusion weight is 0.3. The above five values ​​are based on the reliability stratification statistics of 200 groups of historical samples, so that high stability records have a higher weight in the fusion and low stability records have limited disturbance to the historical coefficients. For each layer, a more accurate fusion coefficient is generated by "weighted averaging of the evidence coefficient according to the fusion weight and the old contribution coefficient according to the residual weight". The new contribution coefficient is formed to form an updated contribution coefficient data set, and all stratified coefficients are truncated from 0 to 1 to avoid crossing the boundary; then the weight distribution of the resource allocation scheme is calculated under the "proportional balance constraint". The proportional balance constraint is that the sum of the weights of each resource is equal to 1 and the single weight is between 0.2 and 0.6. This interval is determined by the marginal contribution distribution statistics of water and fertilizer to water use efficiency in the implementation plan of the past three seasons. The lower and upper limits are set to avoid extreme resource bias; the weight distribution is calculated using a linear regression algorithm, with water use efficiency as the dependent variable and three types of resource parameters as independent variables. The three types of parameters are fertilizer application amount, irrigation interval, and single irrigation water volume. To enhance the coupling with soil conditions, the updated contribution coefficients were used as sample weights in the regression model before modeling, i.e., samples with higher updated contribution coefficients were given a higher influence. After the modeling was completed, the absolute values ​​of the three regression coefficients obtained from the regression were taken as the initial importance and mapped from 0 to 1 using a minimum-maximum mapping. The three importances were then normalized and truncated according to the proportional balance constraint to obtain the weight distribution results consisting of fertilizer weight, irrigation interval weight, and single water weight. If the weight distribution result met the proportional balance constraint and the range of the three weights did not exceed 0.4, the next step of data screening was performed. Otherwise, the most important item was reduced to a value not exceeding 0.6. The upper limit is retracted and the remaining two items are expanded in equal proportion until the constraints are met before proceeding to the next step; data screening is then performed to form a candidate input data set for the next evaluation cycle. The screening rule is to select records from the last two seasons that are consistent with the current stratification, the units are consistent, and all fields are complete, and to eliminate abnormal values ​​that exceed the agronomic reasonable range. The range is fixed as total fertilizer application of 0 to 200 kg per hectare, nitrogen fertilizer ratio of 0 to 60%, phosphorus fertilizer ratio of 0 to 40%, potassium fertilizer ratio of 0 to 40%, irrigation interval of 3 to 30 days, single irrigation water volume of 5 to 40 mm, and water use efficiency greater than 0; fields with a missing ratio of no more than 5% are filled with the average of the two consecutive days, and records with more than 5% are deleted. , after completion, the candidate input data set is obtained; based on the candidate input data set, the cluster analysis method is used to divide the data priority, and the clustering characteristics are the comprehensive value of the updated contribution coefficient, water use efficiency and the parameter difference of the current resource allocation plan. The comprehensive value of the parameter difference is the equal weighted average of the irrigation interval difference and the single water volume difference; the number of clusters is set to 3, and the initial center is the 25% quantile, 50% quantile and 75% quantile of the above three characteristics. The distance measurement is the Euclidean distance after range standardization, the maximum iteration is 100 times, and the convergence threshold is that the center movement is less than 0.001; after clustering is completed, the weighted scores of the average updated contribution coefficient and the average water use efficiency of each cluster are assigned priority 1 from high to low , priority 2 and priority 3, where the weights are 70% of the updated contribution coefficient and 30% of the water use efficiency, to obtain the priority ranking data set; then the matching degree between the priority ranking data set and the current resource allocation plan is calculated. The matching degree is defined as the result of "the inverse of the comprehensive value of parameter difference and the weighted value of the updated contribution coefficient", ranging from 0 to 1, and the threshold is set to 0.75. The threshold is derived from the reliable boundary determined by the sample matching distribution of the 200 sets of historical plans through yield stability review; if the matching degree is lower than 0.75, iterative optimization is performed on the candidate input data set to improve the matching degree. The optimization method is to fine-tune the input data one by one according to the parameters, and the initial step size is 5% of the historical range of the parameter. If the matching degree does not improve for two consecutive rounds, the step size is multiplied by 0.5. If the matching degree improves by 0.05 or more for two consecutive rounds, the step size is multiplied by 1.2. The adjustment order is irrigation interval, single water volume, fertilizer application rate, and the ratio of the three fertilizers. The direction is selected based on improving the matching degree and is truncated within the agronomic boundary. It is iterated for a maximum of 20 rounds, and the optimized input dataset is output. Finally, the input dataset for the next evaluation cycle is generated based on the optimized input dataset and the constraints of the evaluation cycle. The evaluation cycle constraints are a fixed time window length of 30 days, a minimum of 10 records per stratum, a maximum to minimum ratio of sample size for each stratum of no more than 3, and a contribution coefficient of any updated stratum of less than 0.At 4 o'clock, records from adjacent time periods in the same layer must be supplemented until the minimum number of records is met. If this cannot be supplemented, the layer will be marked as the priority sampling layer for the next cycle. Each record in the generated dataset contains the layer number, time index, updated contribution coefficient, total fertilizer application amount and nitrogen, phosphorus and potassium ratio, irrigation interval and single water volume, water use efficiency and data source identification. A joint index is established on the layer number and time index to support rapid retrieval and traceability.

[0064] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A data processing method for evaluating winter wheat field soil fertility based on comprehensive weight calculation, characterized in that: The method comprises the following steps: S1. Nutrient content and water retention data are collected from field soils through a sensor network. The collected nutrient content and water retention data are grouped and processed using decision node splitting and entropy calculation to obtain soil fertility stratification results. S2. Based on the soil fertility stratification results, obtain the root depth and leaf moisture content indicators of winter wheat in different soil stratifications from crop monitoring equipment. Use leaf node classification and feature selection criteria to analyze the correlation strength between root depth and leaf moisture content indicators and fertility stratification. Calculate the contribution weight of each soil fertility stratification to root depth and leaf moisture content, and determine the fertility contribution coefficient. S3. If the fertility contribution coefficient exceeds the preset threshold, the water use efficiency change of winter wheat under drought conditions is calculated by combining the simulation module with the pruning method to obtain a quantitative value of drought resistance; S4. Based on the quantitative values ​​of drought resistance, cross-validation evaluation and classification accuracy are used to classify and optimize fertilizer application rates and irrigation scheduling. A weighted average method is used to calculate the contribution weights of fertilizer application rates and irrigation scheduling to water use efficiency, determine the adjustment direction of the resource allocation plan, and obtain preliminary optimized configuration parameters. S5. Extract relevant variables from the preliminary optimized configuration parameters, integrate root development data and nutrient supply data using initial parameter settings and error function definitions, and determine the dynamic balance model of resource allocation.

2. The method for processing winter wheat field soil fertility evaluation data based on comprehensive weight calculation according to claim 1, characterized in that: Said S1 comprises: Nutrient content and moisture data are obtained from multiple layers of soil depth in the field through a sensor network and stored as a structured data set to obtain raw soil data; Data cleaning technology is used to preprocess the original soil data, remove outliers and missing values, and obtain cleaned soil data; Based on the cleaned soil data, the characteristic values ​​of nutrient content and moisture data are extracted, and characteristic vectors are constructed to obtain the soil feature set; The decision tree algorithm was used to split the soil feature set, and the split nodes were determined based on entropy calculation to obtain preliminary grouping results. If the entropy value of the preliminary grouping result is lower than the preset threshold, the groups are merged and the grouping structure is optimized to obtain the optimized grouping result; The optimized grouping results are sorted through hierarchical analysis, and the soil fertility stratification is determined by combining nutrient content and moisture data to obtain the final stratification results; Based on the final stratification results, soil fertility stratification data is generated and stored in a queryable format to obtain a soil fertility stratification dataset.

3. The method for processing winter wheat field soil fertility evaluation data based on comprehensive weight calculation according to claim 1, characterized in that: The S2 includes: The root depth and leaf moisture content data of winter wheat were obtained from different soil fertility layers using crop monitoring equipment and stored as a structured dataset to obtain raw crop data. Data preprocessing technology is used to clean the original crop data, remove outliers and missing values, and obtain cleaned crop data; Based on the cleaned crop data, the characteristic values ​​of root depth and leaf moisture content are extracted, and the characteristic vector is constructed to obtain the crop feature set; The random forest algorithm was used to classify the crop feature set, and the feature importance was calculated based on information gain. The correlation strength between root depth and leaf water content and soil fertility stratification was determined, and preliminary correlation results were obtained. If the classification accuracy of the preliminary association result is lower than the preset threshold, the feature vectors are re-screened using the feature selection criteria, and the random forest model is optimized to obtain the optimized association result; By analyzing and optimizing the correlation results, the contribution weights of each soil fertility layer to the root depth and leaf moisture content were calculated to obtain the fertility contribution coefficient; Based on the fertility contribution coefficient, the growth adaptability data of winter wheat in different soil fertility layers are generated and stored in a queryable format to obtain the final growth adaptability dataset.

4. The method for processing winter wheat field soil fertility evaluation data based on comprehensive weight calculation according to claim 1, characterized in that: The S3 includes: If the fertility contribution coefficient exceeds the preset threshold, the soil fertility data and environmental factor data are obtained from the winter wheat planting area through the data acquisition module to obtain the initial environmental data set; Based on the initial environmental data set, principal component analysis was used to extract the main feature vectors of soil fertility and environmental factors to obtain the environmental feature set. The environmental feature set was loaded into the simulation module and combined with the pruning method to calculate the water use efficiency change trend of winter wheat under drought conditions, and the efficiency change data set was obtained. If the fluctuation range of the efficiency change data set exceeds the preset range, the efficiency change data set is classified using the support vector machine algorithm to determine the key influencing factors of water use efficiency and obtain the classification results; According to the classification results, the weighted average method was used to calculate the contribution weights of the influencing factors to water use efficiency and obtain the quantitative value of drought resistance. By combining the quantitative value of drought resistance with soil fertility data, we generate the adaptability distribution data of winter wheat under different drought environments and store it in a structured format to obtain the final adaptability dataset. Based on the final adaptability data set, the cluster analysis method was used to grade the drought resistance of winter wheat and obtain the drought resistance grade data.

5. The method for processing winter wheat field soil fertility evaluation data based on comprehensive weight calculation according to claim 1, characterized in that: The S4 includes: Based on the quantitative value of drought resistance, the cross-validation method was used to classify fertilizer application and irrigation scheduling, and a classification accuracy data set was obtained; If the fluctuation range of the classification accuracy dataset exceeds the preset threshold, the random forest algorithm is used to rank the feature importance of the combination of fertilizer application amount and irrigation scheduling, determine the key resource allocation factors, and obtain the factor ranking dataset; Based on the factor ranking data set, the weighted average method is used to calculate the contribution weights of fertilizer application and irrigation scheduling to water use efficiency, and the weight distribution data set is obtained; If the fertilizer application amount weight in the weight distribution data set is higher than the irrigation scheduling weight, the relationship between fertilizer application amount and water use efficiency is analyzed by linear regression method to obtain the fertilizer optimization adjustment parameters; Based on fertilizer optimization adjustment parameters, combined with environmental factor data, and using data standardization methods to process soil fertility data, a resource allocation optimization dataset was generated; Through the resource allocation optimization data set, the cluster analysis method is used to group the irrigation scheduling schemes and obtain the irrigation scheduling optimization parameters; According to the irrigation scheduling optimization parameters and fertilizer optimization adjustment parameters, preliminary optimization configuration parameters are generated and stored in a structured format to obtain the final optimization configuration data set.

6. The method for processing winter wheat field soil fertility evaluation data based on comprehensive weight calculation according to claim 1, characterized in that: The S5 includes: The principal component analysis method is used to extract relevant variables from the preliminary optimized parameters and generate a variable screening data set; If the number of variables in the variable screening data set exceeds the preset threshold, the correlation coefficient matrix between the variables is calculated using the correlation analysis method to obtain the key variable combination; Integrate root development data and nutrient supply data to generate a comprehensive characterization dataset based on a combination of key variables; By integrating the feature data set, the support vector machine method is used to build a dynamic balance model of resource allocation and obtain a preliminary dynamic model; If the prediction error of the preliminary dynamic model exceeds a preset threshold, the model parameters are optimized by the gradient descent method to obtain an optimized dynamic model; According to the optimized dynamic model, combined with the root development characteristics and nutrient supply characteristics, resource allocation adjustment parameters are generated to determine the final dynamic balance plan.

7. The method for processing winter wheat field soil fertility evaluation data based on comprehensive weight calculation according to claim 1, characterized in that: It also includes S6. If the dynamic equilibrium model shows that the yield stability index is lower than the target level, the irrigation interval and fertilizer ratio are adjusted through gradient descent update and convergence condition judgment to obtain the final resource allocation plan.

8. The method for processing winter wheat field soil fertility evaluation data based on comprehensive weight calculation according to claim 7, characterized in that: The S6 specifically includes: The yield stability index is obtained through the dynamic equilibrium model, and the deviation between the index and the preset threshold is calculated using statistical analysis methods to obtain the deviation data set; Based on the deviation data set, the gradient descent method is used to update the irrigation interval parameters, and the adjusted irrigation parameter set is generated by combining the environmental adaptability data; The fertilizer ratio parameters are obtained through the adjusted irrigation parameter set, and the linear regression method is used to predict the parameter adjustment effect and obtain the prediction error value; If the prediction error value exceeds the preset threshold, the gradient descent step size is adjusted according to the convergence condition to generate the optimized irrigation parameter set; According to the optimized irrigation parameter set, agricultural resource allocation data is obtained, and the resource allocation priority is divided using the cluster analysis method to obtain the resource allocation plan; Generate the final resource allocation plan through resource allocation plan and environmental adaptability data; According to the final resource allocation plan, the output stability index is obtained, and it is determined whether the preset threshold is reached to obtain the verification result.

9. The method for processing winter wheat field soil fertility evaluation data based on comprehensive weight calculation according to claim 8, characterized in that: The method further includes S7, obtaining drought environment verification data from a historical database based on the final resource allocation plan, combining variable step size adjustment and simulation feedback loop to determine the applicability of the plan, and obtaining a yield stability assessment result. The S7 specifically includes: Acquire verification data under drought environment from historical database, use data screening method to extract characteristic data sets related to current resource allocation scheme, and obtain characteristic data sets; The matching degree between the resource allocation plan and the verification data is calculated by combining the feature data set with the variable step adjustment method to obtain the matching degree analysis result; If the matching analysis result is lower than the preset threshold, the variable step size is adjusted through the simulation feedback loop to generate an optimized matching analysis result; According to the optimized matching analysis results, the cluster analysis method is used to divide the resource allocation priorities and obtain the resource allocation priority data set; Generate an adjusted resource allocation plan by combining the resource allocation priority dataset with the constraints of the arid environment; According to the adjusted resource allocation plan, the simulation feedback loop is used to evaluate the output stability and obtain the output stability evaluation results; The production stability assessment results are used to determine whether the preset threshold is met and generate the final resource allocation plan.

10. The method for processing winter wheat field soil fertility evaluation data based on comprehensive weight calculation according to claim 9, characterized in that: The process further includes S8, updating the contribution coefficient in the soil fertility stratification database based on the yield stability evaluation results, and determining the input data for the next evaluation cycle in combination with the proportional balance constraint and the final solution output. The process specifically includes: The contribution coefficients in the soil fertility stratification database are obtained through yield stability assessment results, and the contribution coefficients are updated using data fusion methods to obtain an updated contribution coefficient dataset. Based on the updated contribution coefficient data set and combined with the proportional balance constraint, the linear regression algorithm is used to calculate the weight distribution of the resource allocation plan to obtain the weight distribution result; If the weight distribution result meets the preset constraints, the candidate input data set for the next evaluation cycle is extracted through the data screening method to obtain the candidate input data set.

11. The method for processing winter wheat field soil fertility evaluation data based on comprehensive weight calculation according to claim 10, characterized in that: The S8 further includes: Based on the candidate input data sets, cluster analysis method is used to divide the data priorities and obtain the priority ranking data sets; If the matching degree between the priority-ranked dataset and the resource allocation plan is lower than the preset threshold, the input dataset is adjusted through an iterative optimization method to obtain an optimized input dataset; Based on the optimized input data set and the constraints of the evaluation cycle, the input data set for the next evaluation cycle is generated.

Images