Method for precise allocation of agricultural resource elements based on whole-chain ecological niche gap optimization

Abstract

The application provides an agricultural resource element accurate configuration method based on full-chain ecological niche Gap optimization, and relates to the technical field of agricultural resource configuration.The application collects multi-source basic data and plot present situation data, carries out full-chain ecological niche Gap diagnosis, converts qualitative evaluation into quantitative Gap grading labels and resource short board lists, and provides calculable quantitative input for subsequent decision-making.Through Gap-oriented multi-source feature collaborative mining and dynamic adaptive prediction, feature screening and grading differentiated prediction are carried out guided by the Gap grading labels, so that the feature extraction and prediction links are deeply bound with the optimization target.Finally, through full-chain collaborative layered optimization solution, the optimization target is set according to the Gap level difference to realize the integrated output from evaluation diagnosis to the optimal configuration scheme, solve the problem that the evaluation result is disconnected with the resource decision, and improve the scientificity, reliability and landing feasibility of the agricultural resource configuration scheme.

Description

Technical Field

[0001] This invention relates to the field of agricultural resource allocation technology, and in particular to a method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization. Background Technology

[0002] Agricultural planning and resource allocation are core decision-making processes in agricultural production management, directly impacting land use efficiency, resource input-output ratio, and agricultural operating profits. With the promotion of large-scale and intensive agricultural management models, how to scientifically plan planting structures and rationally allocate various agricultural resources under limited land, water, and fertilizer conditions has become a pressing technical problem for agricultural parks and large-scale planting bases.

[0003] On the one hand, agricultural resource allocation can be based on niche theory for suitability assessment, determining planting suitability by analyzing the degree of matching between crop growth needs and environmental factors. This type of technology collects multi-source environmental data, such as climate (e.g., average annual temperature, precipitation, sunshine duration), soil (e.g., organic matter content, nitrogen, phosphorus, and potassium content), and topography (e.g., altitude, slope), to construct crop niche suitability models and output suitability scores or levels for different crops for each plot. While this technology can effectively answer the question "What is suitable to grow on this land?", the results are typically presented as suitability maps and accordion classifications, remaining at a qualitative or semi-quantitative diagnostic and evaluation level. It cannot generate specific configuration plans such as how many acres to plant, how much fertilizer to apply, or how many agricultural machines to equip, resulting in a gap between diagnostic conclusions and actual decision-making.

[0004] On the other hand, agricultural resource allocation can also be based on spatial optimization models for planning, using spatial geometric calculations or operations research methods to solve for optimal spatial layout or planting structure schemes. For example, by calculating the minimum planting distance between crops using parameters such as plant height, canopy projection, and solar altitude angle, the problem of light competition in monoculture or intercropping scenarios can be solved. Another example is the synergistic optimization of planting structure and spatial layout based on multi-objective optimization models, which outputs suggestions for adjusting planting area. While these technologies can output layout schemes with a certain degree of operability, they only focus on a single dimension such as spatial layout or planting structure, failing to incorporate the entire chain of resource elements, including water and fertilizer input, agricultural machinery operation, labor allocation, and warehousing and logistics, into a unified optimization framework. This results in a lack of validation of the output layout schemes at the resource constraint level, potentially leading to implementation failures due to insufficient water and fertilizer supply or mismatched agricultural machinery capabilities.

[0005] Currently, there is a technical problem in agricultural resource allocation where evaluation results are disconnected from resource decision-making, resulting in insufficient reliability of resource allocation plans and their inability to be implemented. Summary of the Invention

[0006] This invention provides a method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization, which solves the problem of the disconnect between evaluation results and resource decision-making in current agricultural resource allocation and improves the reliability of agricultural resource allocation schemes.

[0007] Firstly, this invention provides a method for precise allocation of agricultural resource elements based on full-chain niche gap optimization. The method includes: collecting multi-source basic data of a target area and current status data of various plots; performing full-chain niche gap diagnosis based on the multi-source basic data and current status data of various plots to obtain gap classification labels and resource deficiency lists for each plot within the target area; performing gap-oriented multi-source feature collaborative mining based on the gap classification labels and resource deficiency lists to obtain gap-oriented core feature vectors; performing dynamic adaptation prediction based on the gap classification labels and core feature vectors to obtain prediction results for each plot; the prediction results include predicted values ​​for crop suitability, resource suitability, and potential for yield improvement; and performing full-chain collaborative hierarchical optimization based on the gap classification labels and the prediction results to obtain the optimal allocation scheme for each plot within the target area.

[0008] In a second aspect, embodiments of the present invention provide an electronic device including a memory and a processor. The memory stores a computer program, and the processor is configured to call and run the computer program stored in the memory to perform the steps of the method as described in the first aspect and any possible implementation thereof.

[0009] Thirdly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the steps of the method as described in the first aspect and any possible implementation thereof.

[0010] This invention provides a precise allocation method for agricultural resource elements based on full-chain niche gap optimization. By collecting multi-source basic data and current land parcel data, this invention performs full-chain niche gap diagnosis, transforming qualitative evaluation into quantitative gap classification labels and resource deficiency lists. This provides calculable quantitative input for subsequent decision-making, solving the technical challenge that existing evaluation results cannot be directly used for resource decision-making. Through gap-oriented multi-source feature collaborative mining and dynamic adaptation prediction, feature selection and graded differentiated prediction are guided by gap classification labels, deeply binding feature extraction and prediction stages with optimization objectives. This avoids the problems of stage disconnect and bias accumulation in traditional linear frameworks. Finally, through full-chain collaborative hierarchical optimization, optimization objectives are set according to gap level differences. Multiple constraints, such as gap potential constraints and full-chain resource constraints, are integrated for hierarchical solution, achieving an integrated output from evaluation diagnosis to the optimal allocation scheme. This solves the problem of disconnect between evaluation results and resource decisions in current agricultural resource allocation, improving the scientificity, reliability, and feasibility of agricultural resource allocation schemes. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating a method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization provided by an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0013] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0014] like Figure 1 As shown, the present invention provides a method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization, wherein the method comprises steps S101-S105.

[0015] S101. Collect multi-source basic data of the target area, as well as the current status data of each plot.

[0016] In some embodiments, the multi-source basic data includes: environmental and soil topography data, planting and breeding status data, and total resource volume data across the entire chain.

[0017] For example, environmental and soil topographic data include the park's average annual temperature, annual precipitation, sunshine duration, average humidity, altitude, slope, aspect, soil texture, organic matter content, available nitrogen, available phosphorus, and available potassium over the past 5 years.

[0018] For example, the data on the current status of planting and breeding includes the existing crops planted in the park, the current planting area of ​​each crop, the layout pattern, and the actual yield.

[0019] For example, the total data on the entire chain of resources includes the total land area of ​​the park, the upper limit of water and fertilizer quotas, the number of agricultural machinery / laborers, and the storage and logistics capacity.

[0020] In some embodiments, the multi-source basic data also includes newly added dynamic data. The newly added dynamic data includes seasonal dynamic data (average environmental parameters for each season over the past 3 years), resource dynamic consumption data (water and fertilizer consumption rates and agricultural machinery working hours consumption patterns for different growth stages of each crop), and planting and breeding anomaly record data (frequency of pests and diseases, yield reduction ratio, and cases of water and fertilizer waste in each plot over the past 3 years).

[0021] S102. Based on the multi-source basic data and the current status data of each plot, perform a full-chain niche gap diagnosis to obtain the gap classification labels and resource deficiency list of each plot in the target area.

[0022] As one possible implementation, step S102 can be specifically implemented as steps S1021-S1026.

[0023] S1021. Perform lightweight preprocessing on the multi-source basic data to obtain a standardized feature matrix.

[0024] For example, embodiments of the present invention can streamline preprocessing operations: only basic outlier screening and dimensional standardization are performed, without complex feature processing, avoiding premature filtering of core gap-related information and ensuring the accuracy of gap diagnosis. Specifically: outliers (data exceeding the mean ± 3 standard deviations) are identified using the 3σ criterion; numerical features are processed using the Z-Score standardization formula to convert them into a standard distribution with a mean of 0 and a variance of 1, eliminating dimensional differences. The standardization formula is as follows: ; in, Let j be the standardized value of the numerical feature of the i-th plot. This represents the original value of the j-th feature of the i-th plot. Let be the mean of the j-th feature. Let be the standard deviation of the j-th feature; one-hot encoding is used for categorical features to convert text categories into 0-1 numerical vectors.

[0025] S1022. Based on the standardized feature matrix, construct a three-dimensional ecological niche reference vector including environment, planting and breeding, and full-chain resources.

[0026] In some embodiments, the three-dimensional niche reference vector includes environmental reference, aquaculture reference, and resource reference.

[0027] For example, step S1022 can be specifically implemented as steps A1-A4.

[0028] A1. Environmental baselines are calculated by weighting climate parameters, soil parameters, and topographic parameters using the analytic hierarchy process (AHP).

[0029] A2. Based on the relative difference between the maximum theoretical yield and the minimum yield threshold of each crop, and combined with the average of the standardized suitability parameters of each crop, the planting and breeding benchmark is obtained.

[0030] A3. Determine the resource benchmark based on the smaller of the ratios of supply to demand for the three types of resources: water resources, fertilizer resources, and agricultural machinery resources.

[0031] A4. After assigning weights to the environmental benchmark, aquaculture benchmark, and resource benchmark through the analytic hierarchy process, the weighted sum is obtained to obtain the whole-chain niche benchmark vector.

[0032] For example, the construction of a full-chain ecological niche benchmark: This invention breaks through the limitations of traditional single ecological niche benchmarks and constructs a three-dimensional ecological niche benchmark model of environment-planting and breeding-full-chain resources, ensuring that the benchmark can comprehensively reflect the overall adaptability of the plot and solve the problem that the existing benchmarks are single and cannot reflect the adaptability of the whole chain.

[0033] For example, a three-dimensional niche reference vector Among them: (1) Environmental benchmarks The structure is constructed using a weighted average of climate, soil, and topographic parameters calculated via the analytic hierarchy process (AHP). The values ​​range from [0,1] and reflect the adaptability of the plot to its natural habitat. The calculation formula is as follows: ;in, The weight of the k-th environmental parameter is determined by the analytic hierarchy process (AHP), such as annual mean temperature with a weight of 0.3 and soil organic matter with a weight of 0.25. Let be the standardized value of the k-th environmental parameter, and n be the total number of environmental parameters.

[0034] (2) Cultivation and breeding standards Based on the relative difference between the maximum theoretical yield and the minimum yield threshold of each crop, and combined with the average of the standardized suitability parameters of each crop, the value range is [0,1]. This reflects the adaptability potential of a plot to different crops. The calculation formula is as follows: Where m is the number of crop types. This represents the maximum theoretical yield of the l-th crop. The minimum yield threshold for the l-th crop. These are the standardized suitability parameters for the l-th crop.

[0035] For example, the maximum theoretical output Y lmax Under optimal local climate, soil, water and fertilizer conditions, without stress, pests and diseases, and with standardized field management, the first The regional maximum productivity that a crop can achieve represents the theoretical upper limit of the productivity of a plot of land.

[0036] For example, the maximum theoretical output Y lmax The calculation can be performed using the following steps.

[0037] Step 1: Multi-source data collection, uniformly collecting three types of data sources: (1) High-yield crop test records in the park over the past 3-5 years; (2) Regional variety comparison test and high-yield creation test data released by local agricultural and rural bureaus, agricultural science institutes, and agricultural technology extension stations; (3) National and local crop cultivation regulations and standard values ​​of potential crop yield literature.

[0038] Step 2: Outlier Removal. The 3σ criterion is used to remove abnormal production data caused by extreme disasters, extreme weather, or human error, retaining valid samples from normal production years.

[0039] Step 3: Statistically calculate the maximum theoretical yield. Sort the effective yield samples and select the mean of the top 10% of the best high-yield samples as the maximum theoretical yield for that crop region. When data is insufficient, the optimal parameters of the WOFOST / DSSAT crop growth model are used for simulation and correction, and the final version is verified and approved by agronomic experts.

[0040] For example, the minimum production threshold Y lmin The critical yield that ensures crop planting does not result in losses and has basic economic value; yields below this level are considered to have no planting benefits and should be eliminated or the crop structure adjusted.

[0041] For example, the minimum production threshold Y lmin The calculation can be performed using the following steps.

[0042] Step 1: Construct the economic break-even yield (primary basis). Calculate the total cost per unit area for a single crop (seeds, water, fertilizer, pesticides, machinery, labor, land rent, and maintenance), and combine this with the average market price of the crop over the past 3-5 years to calculate the minimum break-even yield: In the formula: This represents the total cost of planting per unit area. This represents the average purchase price of crops over the past three years.

[0043] Step 2: Construct historical low-frequency output thresholds (auxiliary constraints). Statistically analyze the historical output series of the industrial park, using the 10th percentile output as the historical lowest stable output. .

[0044] Step 3: Determine the combined threshold. Take the maximum value of the two thresholds as the final minimum output threshold. Ensure that the threshold simultaneously meets the dual agronomical and economic constraints of "no loss and stable output".

[0045] For example, crop standardization suitability parameters The comprehensive crop suitability score is quantified based on the plot's climate, soil, topography, and habitat conditions. The value ranges from [0,1]. The higher the value, the more suitable the plot is for the growth of this type of crop.

[0046] For example, crop standardization suitability parameters The calculation can be performed using the following steps.

[0047] Step 1: Screening core crop adaptation evaluation factors. Key constraints for crop growth are uniformly selected: average annual temperature, growing season precipitation, sunshine duration, air humidity, altitude, slope, aspect, soil texture, soil organic matter, available nitrogen, available phosphorus, available potassium, and soil pH.

[0048] Step 2: Factor weight determination (AHP method). Invite 3-5 agronomy experts to construct a judgment matrix, calculate the weight of each environmental factor using the analytic hierarchy process, and conduct a consistency check to ensure that the weight allocation is scientific and free of logical bias.

[0049] Step 3: Single-factor suitability standardization scoring. Based on the optimal growth range of each crop, each indicator is scored with a standardized membership degree of 0–1: the closer the indicator is to the optimal growth range of the crop, the closer the score is to 1; if it deviates from the suitable range, the score decreases linearly to 0.

[0050] Step 4: Comprehensive suitability weighted solution. The solution is obtained by weighted summation. Comprehensive suitability parameters for crop planting plots: In the formula: Let t be the weight of the factor. These are the measured values ​​of the plot factors; These are the upper and lower limits of the suitable range for this factor.

[0051] Step 5: Expert Verification and Correction. Manual verification and fine-tuning are conducted for extreme plots, plots with special habitats, and plots exhibiting seasonal differences to ensure... It perfectly matches the actual planting patterns, and the final output value range is strictly constrained to [0,1].

[0052] (3) Resource benchmark Based on the minimum ratio of supply to demand for three types of resources—water, fertilizer, and agricultural machinery—the supply-demand ratio is standardized to the [0,1] interval, reflecting the scale of planting and breeding that a plot of land can support. The calculation formula is as follows: ;in, , , These represent the total supply of water resources, fertilizers, and agricultural machinery for the park. , , Let be the water resources, fertilizer, and agricultural machinery requirements per unit area for the l-th crop, respectively. When the supply-demand ratio is greater than 1, we take 1 to ensure that the results are constrained within the interval [0,1].

[0053] This invention can assign reasonable weights to three-dimensional benchmarks (environmental benchmark 0.4, aquaculture benchmark 0.3, and resource benchmark 0.3) using the Analytic Hierarchy Process (AHP), and form a unified whole-chain ecological niche benchmark vector through weighted calculation. To ensure the scientific validity and rationality of the benchmark, agricultural experts have verified that the benchmark is more than 95% consistent with the actual land plots.

[0054] S1023. Based on the current status data of each plot, calculate the current ecological niche vector of each plot.

[0055] S1024. Based on the current ecological niche vector of each plot and the three-dimensional ecological niche reference vector, calculate the ecological niche gap value of each plot.

[0056] S1025. Based on the niche gap value of each plot, analyze the resource potential and revenue improvement space of each plot, divide each plot into three levels: high potential, medium potential and low potential, and obtain the gap classification label of each plot.

[0057] S1026. Based on the Gap classification labels of each plot, perform a triple correlation verification of Gap classification, resource supply, and current planting and breeding status for each plot, locate the types of resource deficiencies in each plot and calculate the gap ratio, and obtain a list of resource deficiencies for each plot.

[0058] For example, this invention can calculate the difference between the current configuration of each plot and the benchmark configuration based on the constructed full-chain niche benchmark vector, thereby quantifying the full-chain niche gap. The larger the difference, the greater the optimization potential of the plot. The gap quantification formula is as follows: ;in, Let be the niche gap value of the i-th plot. This is the current ecological niche vector of the i-th plot (calculated from current environmental, aquaculture, and resource data). It serves as the baseline vector for the entire ecosystem niche, and its absolute value ensures that the gap value is non-negative. It is the first The current ecological niche vector of each plot, and the baseline ecological niche vector of the entire chain. The structures are completely identical, also consisting of three dimensions: ; : No. The current environmental niche value of each plot; : No. Current ecological niche values ​​for planting and aquaculture on each plot of land; : No. Current resource ecological niche value of each plot of land. The computational logic and It is perfectly symmetrical, except that the ideal benchmark data has been replaced with the actual data of the current state of the land parcel.

[0059] For example, the current ecological niche vector The calculation process is as follows.

[0060] 1. Current environmental ecological niche value ; Corresponding to the reference vector The calculation logic is completely consistent: ;in, : The same environmental parameter weights as the baseline vector (determined by AHP, such as annual mean temperature 0.3, soil organic matter 0.25); : No. The first plot of land Standardized values ​​of current environmental parameters (calculated from actual monitoring data). Data sources: Current measured data on the plot, including average annual temperature, precipitation, sunshine duration, soil organic matter, nitrogen, phosphorus and potassium content, slope and aspect, which correspond one-to-one with the parameter dimensions used in the baseline construction.

[0061] 2. Current ecological niche values ​​for planting and breeding ; Corresponding to the reference vector The calculation logic is completely consistent: ;in, The total number of crop types planted on the plot; : No. The first plot of land currently planted with The actual yield per acre of the crop; : The maximum theoretical yield of this crop (the same value used in the baseline vector); : No. The plot of land is for the first Standardized suitability parameters for current crop planting status (calculated based on current environmental data of the plot). Data sources: actual yield data of the plot over the past 3-5 years, and actual growth data of currently planted crops.

[0062] 3. Current resource ecological niche value ; Corresponding to the reference vector The calculation logic is completely consistent: ;in, : No. The current actual water consumption of each plot of land; : No. The water resource requirement per unit area for the crops currently grown on each plot of land. Similarly, , These are the actual consumption of chemical fertilizers and agricultural machinery, respectively. , This corresponds to the demand. Data source: Actual consumption data from water and fertilizer flow meters, agricultural machinery operation recorders, and resource ledgers.

[0063] 4. Overall current ecological niche value ; and reference vector Similarly, the overall current ecological niche value of the land parcel is obtained by weighted summation: Among them, the weights (environment 0.4, farming 0.3, resources 0.3) and the baseline vector Completely identical, ensuring that the weight systems of the two vectors are perfectly matched during subsequent gap calculations.

[0064] It should be noted that it is completely symmetrical with the reference vector structure: and Both are three-dimensional structures of environment-farming-resources, with identical weights, and are calculated in this way. In this case, the difference is the truly meaningful gap between the current situation and the ideal, and will not produce meaningless deviations due to mismatch in structure or weight.

[0065] It's important to note that all data sources are current-state data: the data across all three dimensions comes from the site's current monitoring, records, and actual production data. This ensures that the calculated gap represents the real potential for optimization, rather than a theoretical difference. This directly serves subsequent tiered optimization. Each dimension value can directly pinpoint the weakness: Low environmental habitat adaptability; Low quality → Unreasonable planting and breeding structure; Low → Low resource utilization efficiency / resource scarcity. Current ecological niche vector Adopting the whole-chain ecological niche benchmark vector The three-dimensional structure of environment, agriculture, and resources is completely symmetrical. The comprehensive ecological niche value of the plot is obtained by weighted summation. All calculations are based on the current measured data of the plot to ensure that the gap diagnosis results truly reflect the gap between the current situation and the ideal.

[0066] It should be noted that the current ecological niche vector Explanation of the source of niche baseline vector data.

[0067] I. Current Ecological Niche Vector Definition and calculation method.

[0068] 1. Definition. For the first The current ecological niche vector of each plot is calculated from the plot's current actual environmental, agricultural, and resource data, and is compared with the overall ecological niche benchmark vector. The structure is completely symmetrical and is used to quantify the current overall production suitability level of the land parcel.

[0069] 2. Calculation method (and) (Completely symmetrical). ; II. Baseline Vectors for Each Dimension Explanation of data sources for calculation 1. Environmental benchmarks ;formula: Data sources: Statistical averages of annual temperature, annual precipitation, sunshine duration, and average humidity in the park over the past 5 years; measured topographic data on plot elevation, slope, and aspect; laboratory test data on soil texture, organic matter content, available nitrogen, available phosphorus, and available potassium; weighting. It was determined by the Analytic Hierarchy Process (AHP) combined with the experience of agronomic experts.

[0070] 2. Standards for planting and breeding ;formula: ;in, The maximum theoretical yield of a crop in a given local area is derived from regional trial data, high-yield creation records, or crop growth model (such as DSSAT) simulation results. The minimum yield threshold for crops is determined by a combination of economic break-even analysis and historical yield quantiles. Standardized suitability parameters for crops are calculated using a weighted analytic hierarchy process based on key environmental factors for crop growth (temperature, precipitation, soil, etc.).

[0071] 3. Resource Benchmark ;formula: ;in, The total water resource quota, fertilizer supply limit, and agricultural machinery operation capacity limit of the park are derived from the park's resource ledger and planning data. : No. The water and fertilizer requirements per unit area and the agricultural machinery hours required for crops are derived from crop cultivation technical regulations and field trial data.

[0072] III. Current Ecological Niche Vector With the whole chain niche benchmark vector Using the exact same calculation logic and weighting system, only the data source differs (current measured vs. ideal standard) to ensure subsequent gap values. The calculation results accurately reflect the gap between the current state of the land parcel and its theoretical optimal level, providing a reliable basis for graded optimization. The quantitative calculation formulas for resource potential and revenue improvement space are as follows, refining the quantitative basis for gap grading: (1) Formula for calculating resource potential (quantifying the exploitable resource potential of a land parcel): ;in, Let i be the resource potential of the i-th plot. As a benchmark for the entire ecosystem's niche resources, The current resource utilization level of the i-th plot (calculated from the current resource consumption data) is rounded to one decimal place.

[0073] (2) Formula for calculating potential for increased revenue (quantifying the revenue growth potential of optimized land parcels): ;in, The potential for increasing revenue for the i-th plot of land. This is the optimized yield per acre based on the baseline vector prediction. Let be the current per-acre yield of the i-th plot, and the result be rounded to one decimal place. : Land parcel revenue improvement rate (%), used to quantify the revenue improvement potential brought by the optimization plan, and is one of the core indicators for Gap classification; Average yield per mu after land optimization (yuan / mu); The current yield per mu (yuan / mu) of the land is calculated based on the actual output, cost, and market price of the land.

[0074] For example, optimizing revenue Based on the whole-chain niche benchmark vector The process of obtaining the predicted theoretical optimal yield per acre consists of three steps: Step 1: Predict optimal yield based on baseline vectors. This is based on the environmental baseline of the land parcel. Standards for planting and breeding Resource benchmarks Using the tiered LightGBM model as input, the theoretical maximum yield per acre for this plot of land under suitable crop and optimal resource allocation is predicted. : The model input is a three-dimensional ecological niche baseline vector, reflecting the theoretical adaptation level of the plot; the model output is the maximum achievable yield of the plot under the optimal planting structure and resource allocation.

[0075] Step 2: Calculate the optimized yield per acre. This is done by considering the average market purchase price of the target crop over the past 3-5 years. Calculate the output value per mu after optimization : .

[0076] Step 3: Deduct the optimized unit area cost. Calculate the optimized unit area cost based on the resource allocation (water and fertilizer usage, agricultural machinery hours, etc.) in the optimized plan. The optimized yield per acre was finally obtained: ;in, The optimized plan addresses costs related to water, fertilizer, agricultural machinery, and labor, resulting in a significant reduction in costs compared to the current situation (reduced resource waste).

[0077] For example, the current yield per mu How to obtain ;in, The current actual yield per acre of the plot is obtained from historical records or actual measurement data; The actual current planting costs of the plot, including water and fertilizer, agricultural machinery, labor, land rent, etc. The average market price of crops over the past 3-5 years, eliminating the impact of price fluctuations.

[0078] It should be noted that this invention innovatively adopts a three-dimensional grading standard of "difference size + resource potential + profit improvement space" to divide the gap into three levels: high potential, medium potential, and low potential: (1) High potential gap: 1. Land parcels with resource potential ≥20% and revenue increase ≥18% are characterized by large value difference, weak resource constraints, and high potential for revenue increase. They are the key optimization targets, with the core characteristics of "sufficient resource supply, good habitat adaptability, and significant revenue increase potential"; (2) Medium potential gap: 1. Resource potential 10%-20%, revenue increase 8%-18%, these plots have medium difference, strong resource constraints, and medium revenue increase space, and focus on resource optimization; (3) Low potential gap: Land parcels with resource potential <10% and revenue increase <8% have small differences in value, strong resource constraints, and low room for revenue increase, and should be mainly focused on stabilizing production and reducing costs.

[0079] To ensure the authenticity and rationality of Gap classification results, this invention correlates and verifies the results with full-chain resource data and current planting and breeding status, constructing a triple verification mechanism of "Gap classification - resource supply - planting and breeding status". If the Gap classification of a certain plot contradicts the resource supply and planting and breeding status (e.g., a high-potential Gap plot has a water shortage that cannot support the optimized planting and breeding scale), the weights of the three-dimensional ecological niche benchmark and the Gap classification standards are readjusted until the classification results highly match the actual production. After verification, the accuracy rate of Gap diagnosis reaches over 97%.

[0080] An example is the implementation method of Gap-level full-chain correlation verification and the construction of a resource bottleneck list.

[0081] 1. Full-chain correlation verification implementation method. 1. Contradiction detection mechanism (how to identify discrepancies between grading results and reality). Contradictions are identified by constructing a triple verification matrix of "Gap grading - resource supply - current planting and breeding status" and cross-referencing quantitative indicators. The core verification rules are shown in Table 1.

[0082] Table 1 For example, the present invention can automatically compare the Gap classification label of each plot of land with the above three verification indicators; if any one of them does not meet the rules, it is marked as a "contradictory plot", and the contradiction type and specific indicator value are output; the proportion of contradictory plots is counted, and if it exceeds 5%, the benchmark weight and classification standard adjustment process is initiated.

[0083] 2. Three-dimensional niche benchmark weight adjustment method. A combination of the Analytic Hierarchy Process (AHP) and gradient descent fine-tuning is used to adjust the environmental... Planting and breeding ,resource The weights are adjusted to ensure they match the actual production scenario: Step 1: Calculate the weight adjustment gradient. For conflicting plots, calculate the contribution of each dimension to the deviation: ;in, The greater the contribution of the deviation, the higher the mismatch between the weight of that dimension and the actual scenario.

[0084] Step 2: Weight Iteration and Correction. Initial Weights: Set according to agronomic experience (e.g., Environment 0.4, Planting and Breeding 0.3, Resources 0.3); Correction Rules: Reduce the weight of dimensions with negative contribution to deviation; increase the weight of dimensions with positive contribution to deviation; Constraints: The sum of the adjusted weights must be 1, and the adjustment range of a single dimension weight shall not exceed ±0.1 to avoid sudden weight changes that could lead to system imbalance.

[0085] Step 3: Consistency Verification. After adjustment, recalculate the entire chain niche benchmark. A consistency check is performed. If the CR value is less than 0.1, the weight adjustment is considered effective; otherwise, the iterative correction is repeated.

[0086] 3. Gap grading standard adjustment method. The grading standard is based on the Gap value. Profitability Improvement Rate Resource potential The three-dimensional threshold is constructed, and the adjustment logic is as follows: the initial grading criteria are shown in Table 2.

[0087] Table 2 For example, if conflicting land parcels are concentrated in high-potential gaps, and the main issue is insufficient resource supply: increase resource potential. Increase the threshold (e.g., from 0.7 to 0.8); lower the lower limit of the Gap value (e.g., from 0.4 to 0.35) to avoid misclassifying resource-constrained plots as high-potential. If conflicting plots are concentrated in medium-potential Gap areas, and mainly have low compatibility with aquaculture: increase the yield improvement rate. The threshold (e.g., increased from 10% to 15%); tightening environmental adaptability constraints, Incorporate medium-potential land parcels into the criteria for determining their potential. The adjustment aims to ensure that, after the adjustment, high-potential land parcels have no hard resource constraints, medium-potential land parcels experience stable revenue growth, low-potential land parcels have no room for optimization, and the grading results do not logically conflict with subsequent optimization plans.

[0088] 4. Verification Termination Conditions. This invention uses dual quantitative indicators to determine the completion of verification, ensuring the authenticity and rationality of the grading results: the proportion of conflicting plots ≤ 3%; historical data backtesting accuracy ≥ 97%: backtesting is performed on production data from the past 3 years. If the matching degree between the grading results and the actual increase in revenue is ≥ 97%, then the grading standards are considered to be highly consistent with actual production; if the backtesting does not meet the standards, the weight and grading standard adjustment process is repeated until the requirements are met.

[0089] For example, the method for constructing a resource bottleneck list corresponding to Gap. The resource bottleneck list is one of the core outputs of Gap classification, used to clarify the type of resource constraints and the size of the gap for each plot of land. The construction process is as follows: 1. Identification of resource shortcomings. Based on resource niche values. The computational logic identifies the types of resources that are lacking: The resource type corresponding to the minimum value is the core resource deficiency of the plot; if multiple resource ratios are all below 0.8, they are marked as "multiple resource deficiencies" and sorted by ratio from smallest to largest.

[0090] 2. Gap Calculation. For resources with shortcomings, calculate the supply-demand gap ratio: Among them, a gap ratio of ≥30% is considered a "severe gap", 10%-30% is considered a "moderate gap", and <10% is considered a "mild gap".

[0091] 3. Standardized output of the list. The resource deficiency list for each plot of land includes the following information: plot number, gap level; type of deficiency resource (water / fertilizer / agricultural machinery); gap ratio and gap size; optimization suggestions (e.g., for plots with severe water shortages, it is recommended to adjust the irrigation method or switch to drought-resistant crops).

[0092] For example, the essence of a resource deficiency list is to identify the resource type with the greatest gap by comparing the current state of a plot with the optimal baseline state, and then use this as the focus of subsequent optimization.

[0093] For example, a list of resource shortcomings can be determined through the following steps.

[0094] Step 1: Clarify the meaning of the three-dimensional reference vector. These correspond to three types of core benchmarks: Environmental baselines (such as suitable soil organic matter, average annual temperature, rainfall, etc.); : Cultivation and breeding benchmarks (such as optimal planting patterns, yield levels, resource utilization rates, etc. for target crops); Resource benchmarks (such as water requirement per acre, fertilizer application rate, and agricultural machinery operation efficiency).

[0095] Step 2: Calculate the gap difference between the plot and the benchmark. For each plot... Calculate the difference between its current value and the baseline value: ; ; ;in, plots of land The current status of the environment, agriculture, and resources.

[0096] Step 3: Gap Classification and Prioritization. Based on the size of the gap, each plot's gap is divided into three levels: high, medium, and low: High-priority gap: 1-2 types of resources with the largest gap (e.g., if a plot's water resource supply-demand ratio is only 50% of the benchmark, then water resources are a high-priority gap); Medium-priority gap: resource types with moderate gaps; Low-priority gap: resource types with gaps close to the benchmark.

[0097] Step 4: Generate a resource deficiency list. Sort the deficiencies of each plot by priority to form a structured list, as shown in Table 3 below.

[0098] Table 3 S103. Based on the Gap hierarchical labels and resource shortcoming list, perform Gap-oriented multi-source feature collaborative mining to obtain the core feature vector of the Gap-oriented type.

[0099] As one possible implementation, step S103 can be specifically implemented as steps S1031-S1037.

[0100] S1031. Based on the standardized feature matrix, extract three types of original features: environmental, agricultural, and resource, to form a candidate feature set.

[0101] In some embodiments, the candidate feature set includes multiple candidate features.

[0102] S1032. Using the Gap classification labels and resource shortcoming list as the screening guide, calculate the Pearson correlation coefficient between each candidate feature of each plot and the niche Gap value.

[0103] S1033. Based on the Pearson correlation coefficient, the candidate features are initially screened to obtain the initial screening feature set.

[0104] S1034. Input the continuous numerical features in the initial screening feature set into the variational autoencoder, map the high-dimensional continuous features into the latent space distribution, and reconstruct the output by the decoder after reparameterization to obtain the denoised continuous feature vector.

[0105] S1035. Input the discrete category features in the initial screening feature set into the random forest model, calculate the importance score of each discrete feature through the Gini gain of multiple decision trees, and obtain the discrete feature importance vector.

[0106] S1036. Based on the niche gap value of each plot, allocate the fusion weights of the denoised continuous feature vector and the discrete feature importance vector.

[0107] S1037. Based on the denoised continuous feature vector and discrete feature importance vector, and the fusion weight, perform weighted fusion to obtain the gap-oriented core feature vector of each plot.

[0108] For example, this invention breaks through the limitations of traditional single-model feature extraction. Guided by Gap hierarchical labels and resource shortcomings, it adopts a multi-model collaboration + feature screening approach to mine core features that are highly related to Gap optimization and resource adaptation. This solves the pain points of feature extraction being disconnected from optimization targets and having large noise interference in existing technologies, and achieves synergistic linkage between feature mining and Gap diagnosis.

[0109] For example, the specific steps of Gap-oriented multi-source feature collaborative mining are as follows.

[0110] 1. Initial Feature Screening: Based on the Gap grading labels and resource deficiency list, the core features are initially screened using the correlation coefficient method. The Pearson correlation coefficient formula between features and Gap values ​​is introduced to quantify the correlation between features and Gap optimization. The formula is as follows: ;in, Let be the Pearson correlation coefficient between the j-th feature and the gap value, with a value range of [-1, 1]. The closer the absolute value is to 1, the stronger the correlation. Let be the gap value of the i-th plot. The average gap value for all plots; Let be the value of the j-th feature of the i-th plot. is the mean of the j-th feature; N is the total number of plots. Screening criteria: retain. The features are analyzed, and redundant features with weak correlations are removed.

[0111] Based on the preliminary gap diagnosis results, this invention performs collaborative feature mining on multi-source data, removes redundant and noisy features, and provides high-quality feature input for subsequent dynamic adaptation prediction, ensuring that the prediction results can accurately support gap optimization, while reducing the subsequent computational complexity.

[0112] This invention adopts a four-category feature system: three-dimensional niche baseline, current status data, gap diagnosis, and resource shortcomings, as follows: 1. Environmental baseline features ( The dimensions are shown in Table 4.

[0113] Table 4 2. Cultivation and breeding baseline characteristics ( The dimensions are shown in Table 5.

[0114] Table 5 3. Resource benchmark characteristics ( The dimensions are shown in Table 6.

[0115] Table 6 4. The diagnostic features derived from Gap are shown in Table 7.

[0116] Table 7 For example, the steps for initial feature screening are as follows.

[0117] Step 1: Correlation analysis to eliminate highly collinear features. Using the Pearson correlation coefficient matrix, highly collinear features with correlation coefficients |r|≥0.8 are eliminated, retaining the most representative indicators. When two features are highly correlated, retain the feature that is more correlated with the gap value and remove the other.

[0118] Step 2: Variance screening to eliminate low-variability features. Calculate the variance of each feature and eliminate low-variability features with variances less than a threshold (such as a soil index that is almost indistinguishable across all plots). The variance threshold is set to 0.05. Features below this value are considered noise features that have no distinguishing effect on gap classification.

[0119] Step 3: Feature importance ranking based on Gap labels. Using the Gap hierarchical labels as the target variable, a random forest model is used to calculate the importance score of each feature, and the top 30% of the core features are retained to form the final feature set. Feature importance reflects the contribution of the indicator to the gap classification; the higher the score, the stronger its supporting role for subsequent prediction models.

[0120] For example, after the above steps, the final output core feature set is as follows: Environmental features: average annual temperature, soil organic matter, slope; Agricultural and livestock features: standardized suitability parameters. Actual yield per mu; Resource-related: water supply and demand ratio, agricultural machinery operation efficiency; Gap derivatives: Gap value Profitability Improvement Rate .

[0121] For example, the calculation formula for multi-model collaborative feature extraction is as follows.

[0122] (1) Core formula of VAE model: It is specifically designed to handle continuous numerical features. The encoder maps high-dimensional features to the mean vector and log-variance vector of the latent variable distribution. It shares the bottom hidden layer and outputs two branches of distribution parameters. The decoder reconstructs the features to achieve dimensionality reduction and noise reduction. The core formula includes: Encoder output (mean vector) Log-variance vector ): ; ; Reparameterization (denoising, ensuring gradient differentiability): ,in ( This is an element-wise multiplication operation; For a random noise vector that follows a standard normal distribution, I It is the identity matrix. Depend on (Obtained through exponential calculation, ensuring the standard deviation is non-negative) The decoder reconstructs the denoised continuous features based on the latent space vectors obtained from sampling: ; The model training loss function aims to optimize the joint function of reconstruction loss and KL divergence loss. in, The reconstruction loss is measured by mean squared error (MSE) to determine the difference between the reconstructed features and the original features. KL divergence loss is used to constrain the distribution of latent variables to approximate a standard normal distribution; These are weighting coefficients used to balance the proportions of the two losses. The encoder shares the weight matrix of the hidden layer; , These are the output weight matrices for the encoder's mean branch and variance branch, respectively. This is the weight matrix of the decoder; The encoder shares the bias vectors of the hidden layer; , These are the output bias vectors for the encoder's mean branch and variance branch, respectively. This is the bias vector for the decoder; is the continuous feature vector input, and is the continuous indicators of environment, farming, and resources retained after the initial feature screening; is the mean vector of the latent variable distribution; is the standard deviation vector of the latent variable distribution; The low-dimensional latent space feature vector (latent feature vector) obtained by sampling. The denoised continuous feature vector output by the VAE model serves as a high-quality input for the subsequent dynamic adaptation prediction model.

[0123] This invention maps the original continuous features to a low-dimensional latent space using a VAE encoder, and then reconstructs the denoised features using a decoder, thereby eliminating redundant noise. The output... It not only retains the key information of the original features, but also eliminates noise caused by data acquisition errors and environmental fluctuations, providing high-quality feature input for subsequent dynamic adaptation and prediction, while reducing the computational complexity of the model.

[0124] (2) Core formula of random forest model: It is specifically designed to handle discrete categorical features. It outputs feature importance through voting by multiple decision trees. The core formula is the feature importance calculation (Gini coefficient method): ; ; in, Let K be the Gini coefficient of dataset D, and K be the total number of categories. Let be the proportion of samples of class k in dataset D; Let G be the Gini gain of feature a (the larger the gain, the higher the importance of the feature), and V be the number of values ​​that feature a can take. For a subset of samples in dataset D, take the v-th value of feature a. , These represent the number of samples in the subset and the total dataset, respectively. The discrete feature importance vector output by the random forest is... (where m is the number of discrete features) The original discrete feature vector is weighted element by element to obtain the weighted discrete feature vector. .

[0125] For example, dataset It is a dataset of discrete features and business tags for land parcels, including: Samples: all land parcels in the park; Features: discrete business features of land parcels (planting patterns, soil types, resource allocation types, etc.); Tags: core business tags of this solution (gap potential level, revenue improvement level, etc.).

[0126] (3) Attention mechanism fusion formula: Based on the gap priority, the weights of the output features of the two types of models are dynamically allocated to achieve accurate feature fusion. The core formula is: ; ; ; in, The VAE outputs the weights of continuous features. The weights of discrete features are output for the random forest. This is the weighting adjustment coefficient (taken as 0.6). This represents the current Gap value of the land parcel. The maximum value of the gap value for all plots; This represents the discrete feature vector output by the random forest. This represents the final low-dimensional core feature vector after fusion. When the Gap value is 0 (no optimization space for the plot), the formula degenerates into... This ensures that continuous features still have basic weights even without optimization potential, preventing the model from completely ignoring these features; when the gap value is at its maximum, the formula degenerates into... At this point, the continuous feature weight is the highest, reflecting the focus on the features of plots with high optimization potential; 0.5 represents the "basic weight", which is to balance the weight distribution of continuous and discrete features and avoid the complete compression of the weight of a certain type of feature.

[0127] For example, this invention employs "cross-validation + secondary correlation verification" to supplement the feature validity verification formula, ensuring the supporting role of features in prediction. The core formula is the average correlation coefficient between the feature and the predicted label: ;in, The average correlation coefficient between the core feature and all predicted labels is given by M, where M is the total number of predicted labels (in this invention, M=3, corresponding to crop suitability, resource suitability, and yield / optimization potential). The correlation coefficient between the core feature vector and the m-th predicted label. Validation criteria: Otherwise, re-select features and adjust model parameters.

[0128] 1. Definition of Validation Metric. This invention uses the average correlation coefficient between core features and predicted labels as the validity validation metric, specifically defined as follows: ;in: : The average correlation coefficient between the core features and all predicted labels; : Predict the total number of tags (in this invention) (corresponding to crop suitability, resource suitability, and yield / optimization potential). : Core feature vector and the first The Pearson correlation coefficient of each predicted label is calculated using the following formula: .

[0129] 2. Validation Standards and Procedures. Validation Standard: Average Correlation Coefficient This means that the overall correlation between the core features and all predicted labels reaches a moderately high level, ensuring that the features have sufficient support for subsequent prediction tasks. Iterative process: If If so, the feature re-screening and model parameter adjustment process will be initiated until the validation criteria are met.

[0130] S104. Based on the gap hierarchical labels and core feature vectors, perform dynamic adaptation prediction to obtain the prediction results for each land parcel.

[0131] In some embodiments, the prediction results include predicted values ​​for crop suitability, resource suitability, and potential for increased yield.

[0132] As one possible implementation, step S104 can be specifically implemented as steps S1041-S1043.

[0133] S1041. Based on the gap classification labels of each land parcel, select and determine the prediction model corresponding to each land parcel.

[0134] Each gap level corresponds to a prediction model.

[0135] S1042. Based on the core feature vectors of each land parcel and the prediction model corresponding to each land parcel, the three-dimensional prediction value of each land parcel is predicted.

[0136] In some embodiments, the three-dimensional prediction values ​​include crop suitability prediction values, resource suitability prediction values, and profit improvement potential prediction values.

[0137] S1043. Perform triple dynamic correction on the three-dimensional predicted values ​​to obtain the corrected prediction results for each land parcel.

[0138] In some embodiments, the triple dynamic correction includes multi-source fusion correction, probability interval constraint correction, and component independent correction.

[0139] For example, step S1043 can be specifically implemented as steps B1-B3.

[0140] B1. Based on the three-dimensional predicted values, niche gap values, and resource supply-demand ratios of each plot, the three-dimensional predicted values ​​are used as the main component, and the niche gap values ​​and resource supply-demand ratios are used as supplementary components for fusion correction to obtain the initial correction results.

[0141] B2. Constrain the initial correction results within the standard probability interval to obtain the corrected prediction values ​​for each component of each land parcel.

[0142] The standard probability interval is [0,1].

[0143] B3. For the three components of the corrected prediction values ​​of each component of each plot, namely the crop suitability prediction value, resource suitability prediction value, and income improvement space prediction value, the fusion correction and interval constraint are performed independently to obtain the corrected prediction results of each plot.

[0144] For example, this invention breaks away from the logic of traditional single-model prediction and undifferentiated prediction. Guided by Gap classification labels, it constructs a hierarchical multi-label prediction model to achieve differentiated adaptation prediction of land parcels with different Gap levels. At the same time, it adds a dynamic correction mechanism to solve the pain points of existing predictions being disconnected from Gap optimization and lacking accuracy.

[0145] This invention, based on Gap classification and core characteristics, accurately predicts the crop suitability, resource adaptability, and optimization potential of various plots, providing precise predictive support for subsequent hierarchical collaborative optimization. This ensures that optimization schemes meet the needs of plots with different Gap levels, enhancing the scientific rigor and relevance of the optimization. The specific implementation steps are as follows: 1. Construction of a Hierarchical Labeling System: Based on the Gap classification results, a three-dimensional composite labeling system of Gap level, crop suitability, and resource suitability is constructed to replace the traditional single label. This ensures that the labels accurately reflect the optimization needs of plots at different Gap levels, avoiding the problems of single label sources and circular reasoning in existing technologies. The specific construction content of the labeling system is designed differently according to Gap level, and the quantitative formulas for crop suitability and resource suitability are as follows: (1) Crop suitability (Quantify the suitability of the i-th plot for the l-th crop): ; Where p represents the number of key parameters affecting crop growth (such as soil organic matter, average annual temperature, and sunshine duration). The weight of the t-th parameter (determined by agricultural experts). Let t be the standardized value of the t-th parameter of the i-th plot. , Let be the minimum and maximum values ​​of the t-th parameter, respectively. The value range is [0,1], and the closer it is to 1, the higher the fit.

[0146] The list of key parameters affecting crop growth is shown in Table 8.

[0147] Table 8 It should be noted that the weight constraint is as follows: weight The non-negativity constraint must be satisfied. ), and the sum of the weights of all parameters is 1 ( ), ensuring that the suitability results fall within the range Within the range. The weights are determined by agronomic experts using the analytic hierarchy process (AHP) or the Delphi method, taking into account crop growth characteristics, regional climate conditions, and soil conditions. Different crops have different weight allocations (for example, rice has a higher weight for water resource parameters, while dryland crops have a higher weight for soil water retention parameters).

[0148] (2) Resource adaptability (Quantifying the degree of matching between resource supply and crop demand in the i-th plot): Where q represents the quantity of resource types (such as water, fertilizer, agricultural machinery, etc.). This represents the actual supply of resource type s in the i-th plot. Let be the total demand for the s-th type of resource in the i-th plot. The value range is [0,1]. The closer it is to 1, the higher the resource adaptability.

[0149] For example, this invention employs a composite approach combining historical yield grading, agronomic expert scoring, and gap potential verification to avoid circular reasoning based on current configuration as the label, ensuring the authenticity and scientific validity of the label. Specifically: Historical actual yields of each plot in the park over the past 3 to 5 years are collected, and suitability levels are determined using 30% of the crop's three-year average yield as a benchmark. For new crops lacking historical yield data, 3-5 agronomic experts provide suitability scores between 0 and 1 based on soil, climate, and seasonal dynamic data, and the average score is taken. Finally, the results are verified in conjunction with the gap grading to ensure that the label aligns with the gap potential, maintains consistency with the original label generation logic, and guarantees data continuity.

[0150] Total demand It refers to the first When planting the target crop on a plot of land, the first The theoretical demand for various resources (water, fertilizer, agricultural machinery, etc.) is obtained in the following ways: Crop cultivation regulations method: Based on the characteristics of the target crop variety, growth period, and planting density, consult the "Crop Cultivation Technical Regulations" to obtain the resource demand per unit area (such as water requirement per mu, fertilizer application per mu, and agricultural machinery operation hours per mu); Plot size conversion: Combine the actual area of ​​the plot to convert the demand per unit area into the total demand at the plot level; Scenario adaptation correction: Adjust according to the soil and climate conditions of the plot (such as increasing the water requirement for crops in arid areas and increasing the fertilizer requirement for infertile plots) to ensure that the total demand matches the actual situation of the plot.

[0151] In some embodiments, crop suitability refers to the overall suitability of a specific plot of land for planting a target crop under the current environmental and resource conditions, with a value range of [0,1]. It reflects the natural suitability of the plot for crops (such as climate and soil) and also includes the constraints of resource support capacity (water, fertilizer, and agricultural machinery). The higher the value, the greater the yield potential of planting the target crop on the plot, the higher the resource input-output ratio, and the more sufficient the optimization space.

[0152] For example, this invention adopts a composite label generation method of historical yield grading + agronomic expert scoring + Gap potential verification. The specific process is as follows: Historical yield grading: Collect the historical actual yields of each plot in the park over the past 3-5 years, and classify the suitability level based on 30% of the average yield of the crop over three years; Expert supplementary scoring: For new crops lacking historical yields, 3-5 agronomic experts give a suitability score between 0 and 1 based on soil, climate, and seasonal dynamic data, and take the average; Gap potential verification: Verify the labels in conjunction with the Gap grading results to ensure that the labels are consistent with the optimization potential of the plots, maintain continuity with the original label generation logic, and avoid conflicts between the labels and subsequent optimization goals. Construction of a graded multi-label prediction model. This invention abandons the unified prediction mode of the traditional single LightGBM model and constructs a graded prediction architecture for high / medium / low potential plots: each level of plot corresponds to an independent optimized LightGBM model; achieving differentiated prediction and improving the prediction accuracy of plots with different Gap levels; adapting to different optimization needs, providing accurate optimization solutions for high-potential plots, and providing stable yield and cost reduction strategies for low-potential plots.

[0153] 2. Construction of a hierarchical multi-label prediction model: Each level of land parcel corresponds to an independent optimized LightGBM model, achieving differentiated prediction, improving prediction accuracy, and adapting to the optimization needs of land parcels with different gap levels. The specific optimization and implementation of the model, and the core formulas are as follows: (1) Objective function of LightGBM model (incorporating L1+L2 double regularization to prevent overfitting): ;in, The objective function of the model is... Let be the loss function for the i-th sample (using cross-entropy loss, adapted for multi-label prediction). Let i be the true label of the i-th sample. These are the model's predicted values. , These are the L1 and L2 regularization coefficients (optimized through grid search). , K is the number of leaf nodes in the model's decision tree. Let be the weight of the k-th leaf node.

[0154] (2) Model input and output: The model input is a gap-oriented low-dimensional core feature vector. The output is the predicted value (0-1 continuous probability) corresponding to the three-dimensional composite label, i.e. ,in This is the predicted potential for revenue improvement for the i-th plot. The closer the probability is to 1, the higher the suitability and the greater the optimization potential.

[0155] 3. Dynamic Correction Mechanism: A new triple verification mechanism of "prediction results - Gap diagnosis - resource supply" is added to solve the problem of the disconnect between the existing prediction model and actual production, ensuring the reliability and practicality of the prediction results. At the same time, normalized Gap values ​​and resource supply-demand ratios are introduced to fuse and correct the initial prediction values ​​of the model. After correction, the values ​​are constrained within the probability interval [0,1]. The prediction value correction formula is as follows: ));in, This is the corrected predicted value. These are the initial predicted values ​​for the model. To adjust the weight (take 0.7 to ensure the dominance of the predicted value). Let be the gap value of the i-th plot. Let represent the actual available resource supply for the i-th plot. Let be the resource requirement for the i-th plot. To normalize the gap value, , This formula represents the maximum gap value for the entire park, ensuring that the revised forecast value aligns with the actual gap potential and resource supply.

[0156] For example, triple verification correction: The prediction results are compared with the gap classification and resource shortage list. If the predicted suitability contradicts the gap potential (e.g., the predicted suitability of high-potential gap plots is less than 0.5, and the predicted suitability of low-potential gap plots is greater than 0.8), the prediction results are adjusted using the above correction formula to ensure that the prediction results are consistent with the gap potential and resource supply.

[0157] S105. Based on the Gap hierarchical labels and the prediction results, perform full-chain collaborative hierarchical optimization to obtain the optimal configuration scheme for each plot in the target area.

[0158] In some embodiments, the optimal configuration scheme includes the crop planting area, resource allocation quota, and spatial layout parameters of each plot.

[0159] As one possible implementation, step S105 can be specifically implemented as steps S1051-S1055.

[0160] S1051. Based on the Gap hierarchical labels, set the optimization objective function for each plot.

[0161] Among them, high-potential plots aim to maximize profits and are subject to resource utilization efficiency constraints; medium-potential plots aim to reduce resource waste and are subject to minimum profit increase constraints; and low-potential plots aim to maintain the status quo and are subject to profit fluctuation constraints.

[0162] S1052. Based on the Gap classification labels, prediction results, and total resource data and current status data of planting and breeding in the multi-source basic data, determine the constraints.

[0163] In some embodiments, the constraints include gap potential constraints, full-chain resource constraints, spatial layout constraints, and implementation feasibility constraints; wherein, the gap potential constraints include the adjustment range of planting area, the full-chain resource constraints include total resource constraints and resource timing constraints, the spatial layout constraints include minimum planting spacing constraints and light shading constraints, and the implementation feasibility constraints include crop rotation compatibility constraints and adjustment range constraints.

[0164] S1053. Based on the gap classification labels of each land parcel, select the corresponding optimization solution algorithm for each land parcel.

[0165] Among them, high-potential plots are solved iteratively using a sequential quadratic programming algorithm, medium-potential plots are solved iteratively using a constrained particle swarm optimization algorithm, and low-potential plots are solved directly using a linear optimization algorithm.

[0166] S1054. Based on the optimization solution algorithm corresponding to each land parcel, combined with the optimization objective function and the constraints, and using the prediction results as input data, the optimization configuration parameters of each land parcel are obtained.

[0167] S1055. Perform full-chain resource linkage verification on the optimized configuration parameters, and verify the consistency of the allocation of land resources, water resources, fertilizer resources, agricultural machinery resources, labor and warehousing logistics resources in real time to obtain the optimal configuration scheme for each plot in the target area.

[0168] For example, this invention breaks away from the traditional logic of single-objective optimization and non-layered optimization. It takes Gap hierarchical labeling as the core and constructs an optimization system of hierarchical optimization + multi-constraint collaboration + full-chain resource linkage, replacing the traditional mapping + single algorithm optimization mode, and realizing the transformation from predicted value to optimal configuration scheme.

[0169] This invention, guided by Gap grading, achieves differentiated optimization of land parcels at different Gap levels while meeting various constraints. It leverages resources across the entire supply chain to maximize overall park revenue, while ensuring the optimization scheme aligns with actual production and is feasible. This completely resolves the pain point of existing optimization schemes being impractical and unenforceable. The specific implementation steps are as follows: the entire process revolves around Gap grading, leveraging resources across the entire supply chain, supplementing the core optimization formula, and ensuring quantifiable and executable results: 1. Setting Optimization Goals by Grade: Abandoning the traditional optimization model of "single goal of comprehensive benefits", we set differentiated optimization goals according to the Gap level to form a multi-goal collaborative optimization system that takes into account benefits, resource utilization, and feasibility of implementation, so as to ensure that the optimization needs of plots of different Gap levels can be met and avoid over-optimization or under-optimization.

[0170] An example is the optimization objective formula for high-potential gap plots, with the objective of maximizing revenue and improving resource utilization efficiency. ;in: : Comprehensive optimization benefits of high-potential land parcels; : A collection of high-potential gap plots; P represents the average market purchase price of the target crop over the past 3 years. : No. The optimized target yield per mu for each plot of land; : No. The planting area of ​​each plot of land; : No. Resource input costs for each plot of land (water, fertilizer, agricultural machinery, etc.); : No. Production risk costs for each plot of land (pests and diseases, extreme weather, etc.). ; Let represent the actual resource consumption of the i-th plot. Let represent the available resource supply for the i-th plot.

[0171] Example: Medium-potential gap plots (goal: optimize the current situation + reduce resource waste): ; ; in, The total amount of waste of medium-potential land resources A collection of medium-potential gap plots, The cost of wasted resources for the i-th plot of land. Let represent the current average yield per mu for the i-th plot of land. Let be the optimized per-acre yield for the i-th plot, and 1.08 be the minimum threshold for yield improvement.

[0172] (3) Low-potential gap plots (Objective: Maintain current stability + preserve resource utilization efficiency): ; ; in, The value represents the fluctuation of resource utilization efficiency for low-potential land parcels, with 0.9 being the target value for resource utilization efficiency. A revenue stability constraint ensures that the fluctuation of optimized revenue does not exceed ±5%. Let be the optimized yield per acre for the i-th plot. Let represent the current average yield per mu for the i-th plot.

[0173] At the same time, a dynamic return adjustment factor is introduced. and risk costs Improve the objective function: ( The current market price, The average price over the past 3 years, with a fluctuation range of 0.8-1.2. (P represents the average market purchase price of the target crop over the past 3 years) Let the optimized target yield per acre be the i-th plot of land. Let be the planting area of ​​the i-th plot. Let be the probability of pests and diseases occurring in the i-th plot. (Probability of natural disasters).

[0174] 2. Multi-constraint Collaborative Construction: Integrating four categories of constraints—Gap potential constraints, full-chain resource constraints, spatial layout constraints, and implementation feasibility constraints—to form a collaborative constraint system. This ensures that the optimization plan is scientifically feasible and practical, avoiding violations of production laws or resource constraints. Specific constraints and formulas are as follows: (1) Gap potential constraint: Combine the Gap classification to set adjustment range constraints to ensure that the optimization scheme fits the Gap potential.

[0175] ; in, The optimized planting area for the i-th plot. This represents the current planting area. A collection of high-potential gap sites. A collection of medium-potential gap plots, This is a collection of low-potential gap plots.

[0176] (2) Full-chain resource constraints: including total resource constraints and resource timing constraints: Total resource constraints: (s is the resource type, (Total supply of Class S resources in the park) Resource timing constraints: (t represents the crop growth period,) Let be the consumption of resource type s for the i-th plot in period t. (To correspond to the total demand) (3) Spatial layout constraints: including light shading constraints and minimum planting spacing constraints, the core formula is as follows: Minimum planting spacing constraint: ;in, Let be the minimum horizontal spacing between crops in the i-th plot. For tall-stalked crops, The local typical solar altitude angle (average value according to the season); Light occlusion constraints: ; ( Let i be the area of ​​the shading of the i-th plot. (where 0.2 is the total area of ​​the plot and 0.2 is the shading threshold).

[0177] (4) Feasibility constraints: including crop rotation compatibility constraints and adjustment range constraints, in conjunction with gap potential constraints: Crop rotation compatibility constraints: ( This refers to the planting area of ​​the first crop. This refers to the planting area for the second round of crops. A reasonable range of threshold values ​​for the area ratio of crop rotation, such as the threshold value for a wheat / corn rotation pattern. =1.1、 =1.3, adapting to the actual elastic demand of production).

[0178] 3. Full-chain linkage optimization solution: This invention adopts a "hierarchical solution + multi-algorithm fusion" mode to replace the traditional single algorithm solution, ensuring solution accuracy and efficiency. At the same time, it adds "full-chain resource linkage verification" to ensure that the allocation of various resources is coordinated and consistent, avoiding resource conflicts. The core solution formula is as follows: (1) High-potential gap plots: The Sequential Quadratic Programming (SQP) algorithm is adopted, with the core iterative formula as follows: ;in, Let be the optimization variables (planting area, resource allocation) for the k-th iteration. The step size is 0.01-0.1. The search direction (determined by the gradient of the objective function and constraints), and the iteration termination condition are: ( For the Lagrange function in (gradient at the location).

[0179] (2) Medium-potential gap plots: The constrained particle swarm optimization (PSO) algorithm is adopted, with the core update formula as follows: ; ; in, Let be the particle velocity in the k-th iteration. The inertial weight is set to 0.7. , The acceleration factor is 2 (for all). , A random number between 0 and 1. This represents the optimal solution for an individual particle. This is the globally optimal solution.

[0180] (3) Low-potential gap plots: A simple linear optimization algorithm is used, with the core solution formula as follows: ; Full-chain resource linkage verification: During the optimization process, resources across the entire chain, including land, water and fertilizer, agricultural machinery, labor, warehousing and logistics, are linked in real time to ensure consistent and coordinated allocation of resources and to constrain the formulas. (5% resource redundancy is reserved to cope with emergencies). At the same time, an iteration termination condition is set: when the difference in benefits after 5 consecutive iterations is less than 0.5%, the iteration stops, balancing solution accuracy and efficiency.

[0181] This invention provides a precise allocation method for agricultural resource elements based on full-chain niche gap optimization. By collecting multi-source basic data and current land parcel data, it performs full-chain niche gap diagnosis, transforming qualitative evaluation into quantitative gap classification labels and resource deficiency lists. This provides calculable quantitative input for subsequent decision-making, solving the technical challenge that existing evaluation results cannot be directly used for resource decision-making. Through gap-oriented multi-source feature collaborative mining and dynamic adaptation prediction, feature selection and graded differentiated prediction are guided by gap classification labels, deeply binding feature extraction and prediction with optimization objectives. This avoids the problems of disconnect and bias accumulation in traditional linear frameworks. Finally, through full-chain collaborative hierarchical optimization, optimization objectives are set according to gap level differences. Multiple constraints, such as gap potential constraints and full-chain resource constraints, are integrated for hierarchical solution, achieving an integrated output from evaluation and diagnosis to the optimal allocation scheme. This solves the problem of disconnect between evaluation results and resource decisions in current agricultural resource allocation, improving the scientificity, reliability, and feasibility of agricultural resource allocation schemes.

[0182] Optionally, the method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization provided in this embodiment of the invention further includes steps S201-S205 before step S104.

[0183] S201. Based on historical yield data and key crop growth parameters of each plot, calculate the crop suitability of each plot.

[0184] S202. Calculate the resource suitability of each plot based on the ratio of the actual supply to the demand of various resources in each plot.

[0185] S203. Using the crop suitability and resource suitability as tags, and combining them with the Gap grading tags for verification, a three-dimensional composite tag for each plot is generated.

[0186] S204. Based on the gap classification labels of each parcel, group each parcel according to its gap classification level.

[0187] S205. For each Gap level, using the core feature vector of the corresponding level plot as input and the three-dimensional composite label as the training target, a prediction model corresponding to each Gap level is trained.

[0188] Thus, this invention calculates crop suitability and resource suitability based on historical yield and resource supply data, generates three-dimensional composite labels, and trains prediction models by grouping them according to grade based on gap grade labels, so that label generation is deeply bound to gap potential. This solves the problem of single label source and easy circular reasoning in the prior art. At the same time, graded modeling realizes differentiated prediction and improves the prediction pertinence for plots with different potential grades.

[0189] Optionally, the method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization provided in this embodiment of the invention further includes steps S301-S306 after step S105.

[0190] S301. Receive real-time environmental data, resource consumption data, and crop growth data during the execution of the optimal configuration scheme.

[0191] S302. Calculate the deviation between the real-time environmental data, resource consumption data, and crop growth data and the set data.

[0192] S303. If the deviation exceeds the preset threshold, iterative correction is triggered. Based on the relationship between the actual revenue of each plot and the predicted optimized revenue and the current revenue, the niche gap value of each plot is updated.

[0193] S304. Based on the updated niche gap value, re-execute gap-oriented multi-source feature collaborative mining, dynamic adaptation prediction and full-chain collaborative hierarchical optimization solution, and output the iterative configuration scheme.

[0194] S305. Calculate the prediction accuracy improvement rate, revenue improvement rate, and resource utilization efficiency improvement rate corresponding to the configuration scheme after the iteration.

[0195] S306. When the prediction accuracy improvement rate, revenue improvement rate, and resource utilization efficiency improvement rate reach the preset target values, configure the basic parameters for the next cycle based on the iterative configuration scheme.

[0196] This invention breaks away from the traditional logic of single output and no iteration, and constructs a closed-loop system of layered output + real-time feedback + model iteration to realize the implementation and continuous optimization of configuration solutions. At the same time, it connects to gap diagnosis to form a closed loop of the whole process. In principle, it realizes a virtuous cycle of diagnosis-optimization-implementation-feedback-iteration, ensuring that the configuration solution always meets the actual production needs.

[0197] This invention delivers optimized configuration schemes precisely to different users, enabling the schemes to be implemented. At the same time, through real-time data feedback and model iteration, the configuration schemes are continuously optimized, solving the pain point of existing technologies that "lack implementation guidance and iterative optimization after outputting the schemes" and improving the long-term applicability of the schemes.

[0198] The specific implementation steps are as follows, supplementing the core iterative formula to ensure that quantization is executable: 1. Layered Closed-Loop Output: Based on the different needs of users (park managers, farmers, technicians), results are output in layers to ensure that different users can obtain appropriate implementation content. A new solution implementation tracking module has also been added to provide data support for subsequent feedback and iteration. Specific content of the layered output: (1) For park managers: Output the whole park hierarchical optimization plan (including the optimization focus and configuration parameters of each Gap level plot), comprehensive revenue forecast report (including the total revenue after optimization and the revenue distribution of each plot), risk assessment report (including market risk, natural disaster risk and countermeasures), and resource allocation plan (including annual and quarterly allocation plans of the whole chain resources).

[0199] (2) For farmers: Provide plot-level precision planting plans (clearly specifying the crops suitable for planting, planting density, and sowing time for the plot), standardized field operation procedures (such as fertilization frequency, irrigation time, and pest and disease control measures), and guidance on water and fertilizer / agricultural machinery use (such as water and fertilizer usage per acre and agricultural machinery operation time).

[0200] (3) For technical personnel: Output optimization process data (including core data of feature extraction, prediction and optimization), model parameters (including regularization coefficients and weight allocation parameters of each model), and feature weights to provide a basis for model iteration.

[0201] (4) Implementation tracking module: Records the implementation status of the plan in real time, including crop growth (such as growth and yield), resource consumption data (such as actual water and fertilizer usage and agricultural machinery hours), and income data (such as actual per-acre income), forming an implementation tracking log to provide data support for subsequent feedback and iteration.

[0202] 2. Visualization Optimization: All calculation results are converted into intuitive visual charts, allowing different users to quickly understand the solution content. Charts can also be exported and printed, enhancing the practicality of the solution. Specific visual charts include: box plots showing environmental differences under different resource modes; heat maps showing the distribution of gap levels in various plots; heat maps showing the distribution of theoretical suitability for various crops; bubble charts showing the size of potential gaps for various elements; spatial diagrams showing the optimal field layout; and comparison charts showing the differences in configuration before and after optimization and the improvement in yield.

[0203] 3. Dynamic Feedback and Iteration: A new real-time data access interface has been added to receive real-time environmental data (such as real-time rainfall and temperature), resource consumption data, and crop growth data during the implementation of the plan. Feedback trigger conditions are set: when data changes exceed 10%, or when the deviation between the plan's execution effect and the predicted value exceeds 8%, the feedback mechanism is automatically triggered. The iterative correction formula is as follows: (1) Iterative correction formula for Gap value: ;in, Take the smallest positive number, The sign function ensures that the gap is adjusted when the actual return changes in the same direction as the expected optimization direction; otherwise, the gap is adjusted when the actual return changes in the opposite direction (i.e., the expected optimization direction is reversed). and (different signs), then This means we will not adjust the gap for now, but will iterate after accumulating more data. Let g be the gap value of the i-th plot after iteration. For the actual revenue of the i-th plot, To predict and optimize returns, To ensure that the current benefits are reflected in the actual implementation results, the gap value should be aligned with the actual performance.

[0204] (2) Iterative correction formula for model parameters (taking the regularization coefficient of the LightGBM model as an example): ;in, These are the L1 regularization coefficients after iteration. The original regularization coefficient is . To verify the set-average prediction error, The average prediction error on the training set is used. When the error on the validation set is significantly higher than that on the training set, it is considered overfitting, and the regularization coefficient is increased. If the deviation between the two is small, the regularization coefficient is fine-tuned to decrease, thereby improving the model's fitting ability. The specific iterative process is as follows.

[0205] The first step, data collection and deviation analysis: Through the implementation tracking module and real-time data access interface, actual data (including actual crop yield, actual resource consumption, environmental dynamics data, and profit data) is collected synchronously during the implementation of the plan. This data is then compared with the predicted and optimized values ​​to calculate the deviation rate. The formula for calculating the deviation rate is as follows: ;in, The deviation rate of a certain type of data. These are actual data values. To predict / optimize data values, the deviation rate is rounded to two decimal places to determine whether to trigger a feedback iteration.

[0206] The second step is to trigger the judgment by comparing the deviation rate with the preset threshold (8%) and judging the magnitude of real-time data change (whether it exceeds 10%). If either condition is met, the iteration is triggered; if not, the data is continuously tracked and a deviation review is conducted every 7 days to ensure that abnormal data is captured in a timely manner.

[0207] The third step is a layered iteration across the entire process: Based on the deviation analysis results, layered iterations are performed in conjunction with previous processes to avoid logical disconnects. The specific iteration content is as follows: (1) Iterative process of Gap diagnosis: Using the iterative correction formula for Gap value mentioned above, the Gap value of each plot is recalculated, and the Gap grading standard is adjusted synchronously (if the overall deviation rate exceeds 10%, the grading threshold is finely adjusted, such as adjusting high-potential Gap to 10%). At the same time, the three-dimensional ecological niche benchmark vector was re-verified to ensure that the benchmark fits actual production.

[0208] (2) Iterative feature mining process: Based on the deviation between actual and predicted data, re-select core features (adjust the Pearson correlation coefficient selection threshold, such as from...). Adjusted to At the same time, optimize the core parameters of VAE and random forest models (such as the weight matrix of the VAE model). Distributed branch weight matrix Decoder weight matrix (The number of decision trees in the random forest) is used to re-integrate features and verify their effectiveness, ensuring that the core features accurately reflect the actual adaptation needs.

[0209] (3) Iterative dynamic adaptation prediction process: The model parameter iterative correction formula is adopted (taking the regularization coefficient of the LightGBM model as an example, the parameters of the hierarchical prediction model are adjusted, the three-dimensional composite label system is updated synchronously (combining actual crop suitability and resource suitability data, the weight allocation of label generation is corrected), the hierarchical prediction model is rerun and the dynamic correction mechanism is executed to ensure that the deviation between the predicted value and the actual production data is controlled within 8%. At the same time, according to real-time environmental data (such as seasonal climate change and soil fertility change), the input feature dimension of the prediction model is supplemented to improve the model's adaptability to dynamic scenarios and avoid the prediction results from deviating from reality.

[0210] (4) Iterative process of full-chain collaborative hierarchical optimization: Based on the iterative Gap grading results, core features and predicted values, the hierarchical optimization objective function is readjusted (e.g., the profit maximization objective of high-potential plots can introduce a real-time market price adjustment factor). Dynamic updates allow for fine-tuning of resource waste constraints on medium-potential plots based on actual resource consumption deviations, optimizing parameters in the multi-constraint system (such as minimum planting spacing in spatial layout constraints). The solar altitude angle can be adjusted according to the actual lighting conditions. (The value is taken), and the corresponding algorithm is re-applied for hierarchical solution. The resource allocation is coordinated and consistent through full-chain resource linkage verification to generate the optimal configuration scheme after iteration, ensuring that the scheme fits the actual execution deviation.

[0211] The fourth step is iterative verification and implementation: The iterated configuration scheme and model parameters are synchronously output to the corresponding users. At the same time, the implementation tracking module tracks the execution effect of the iterative scheme in real time and calculates the deviation rate after the iteration. If the deviation rate still exceeds 8%, the above iterative process is repeated until the deviation rate meets the preset threshold. If the deviation rate is controlled within 8%, the model parameters and configuration standards after this iteration are solidified as the basic parameters for the next cycle, forming a virtuous cycle of "iteration-verification-solidification-re-iteration".

[0212] 4. Iteration Effect Evaluation: Establish a quantitative evaluation system for iteration effects, quantifying the differences in effects before and after iteration from three dimensions: accuracy improvement, revenue improvement, and resource utilization optimization, to ensure that the iteration has practical value. The core evaluation indicators and formulas are as follows: (1) Prediction accuracy improvement rate: ,in The average accuracy of the prediction model after iteration. The average accuracy of the prediction model before iteration, the target .

[0213] (2) Rate of improvement in profitability: ,in This represents the actual total revenue of the land parcel after the iteration. The target is the actual total revenue of the land parcel before the iteration. .

[0214] (3) Resource utilization efficiency improvement rate: ,in This represents the average resource utilization rate of the entire park after the iteration. The target is the average utilization rate of resources across the entire park before the iteration. .

[0215] This invention is a precise allocation method for agricultural resource elements based on full-chain niche gap optimization. It breaks through the limitations of the linear framework and insufficient coordination of traditional technologies, and constructs a closed-loop collaborative system. All of these methods have been verified by principle and tested by data to ensure feasibility and advancement.

[0216] This invention breaks away from the traditional linear logic of feature extraction-classification-prediction-optimization, and constructs an integrated framework that leads the entire ecosystem niche gap diagnosis, with multi-stage collaboration and closed-loop iteration. It takes the pre-positioning of gap diagnosis as the core thread of the entire process, realizes deep linkage between each stage, and completely solves the pain point of the disconnect between diagnosis and decision-making. Compared with the existing technical framework, the collaborative efficiency is improved by more than 30%.

[0217] This invention constructs a three-dimensional ecological niche benchmark of environment-planting-full-chain resources, and innovatively adopts a three-dimensional grading standard of difference size + resource potential + benefit improvement space. It adds a full-chain correlation verification mechanism to realize dynamic grading and accurate diagnosis of gaps. The diagnostic accuracy rate reaches more than 97%, which is 22% higher than the traditional single gap diagnosis method.

[0218] This invention proposes a gap-oriented multi-model fusion feature mining mode, which adopts a VAE + random forest + attention mechanism architecture to achieve collaborative extraction and denoising of continuous and discrete features. Combined with correlation screening and effectiveness verification, it ensures that the core features are highly consistent with the optimization target.

[0219] This invention constructs a hierarchical multi-label prediction model to achieve differentiated prediction of land parcels with different Gap levels. It adds a new dynamic correction mechanism of prediction-diagnosis-resource, achieving a prediction accuracy (average R²) of 0.93, which is 20% higher than the existing single prediction model. The optimization process adopts a "hierarchical solution + multi-algorithm fusion" mode, integrating four major categories of constraints to achieve full-chain resource linkage optimization, improving the feasibility of the solution by 50%, and completely solving the pain point of optimization solutions being divorced from production reality.

[0220] This invention constructs a hierarchical output and dynamic iteration system, accurately delivering execution content according to different user needs. It introduces real-time data feedback and a full-process iteration mechanism, combined with quantitative evaluation indicators, to ensure that the configuration scheme continuously adapts to production dynamics and achieves long-term optimization. Compared with traditional non-iterative technologies, the long-term configuration accuracy is steadily improved, and the overall benefits are increased by more than 8% annually.

[0221] In summary, this invention achieves a breakthrough across the entire chain of agricultural resource elements, from diagnosis and evaluation to precise allocation, implementation, and continuous optimization. It can be widely applied in scenarios such as agricultural parks and large-scale planting bases, providing farmers and park managers with scientific and practical intelligent decision-making support, effectively improving the efficiency of agricultural resource utilization and comprehensive benefits. It has significant technological innovation, practicality, and promotional value.

[0222] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0223] Figure 2 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention; the electronic device 400 includes: a processor 401, a memory 402, and a computer program 403 stored in the memory 402 and executable on the processor 401; when the processor 401 executes the computer program 403, it implements the steps in the above method embodiments; or, when the processor 401 executes the computer program 403, it implements the functions of each module / unit in the above device embodiments.

[0224] For example, computer program 403 may be divided into one or more modules / units, one or more modules / units may be stored in memory 402 and executed by processor 401 to complete the present invention; one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, which are used to describe the execution process of computer program 403 in electronic device 400.

[0225] The processor 401 may be a central processing unit, or it may be other general-purpose processors, digital signal processors, application-specific integrated circuits, field-programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.; the general-purpose processor may be a microprocessor or any conventional processor, etc.

[0226] The memory 402 can be an internal storage unit of the electronic device 400, such as a hard disk or RAM of the electronic device 400; the memory 402 can also be an external storage device of the electronic device 400, such as a plug-in hard disk, smart memory card, secure digital card, flash memory card, etc., equipped on the electronic device 400; furthermore, the memory 402 can include both internal storage units and external storage devices of the electronic device 400. The memory 402 is used to store computer programs and other programs and data required by the terminal; the memory 402 can also be used to temporarily store data that has been output or will be output.

[0227] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization, characterized in that, include: Collect multi-source basic data of the target area, as well as current status data of each plot; Based on the aforementioned multi-source basic data and the current status data of each plot, a full-chain niche gap diagnosis is performed to obtain the gap classification labels and resource deficiency list of each plot in the target area. Based on the aforementioned Gap hierarchical labels and resource shortcoming list, Gap-oriented multi-source feature collaborative mining is performed to obtain the core feature vector of Gap-oriented features. Based on the gap classification labels and core feature vectors, dynamic adaptation prediction is performed to obtain prediction results for each plot; the prediction results include predicted values ​​for crop suitability, resource suitability, and potential for yield improvement. Based on the Gap classification labels and the prediction results, a full-chain collaborative hierarchical optimization solution is performed to obtain the optimal configuration scheme for each plot within the target area.

2. The method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization according to claim 1, characterized in that, The multi-source basic data includes: environmental and soil topography data, current planting and breeding status data, and total resource volume data across the entire chain; Based on the aforementioned multi-source basic data and the current status data of various plots, a full-chain niche gap diagnosis is performed to obtain the gap classification labels and resource deficiency lists of various plots within the target area, including: The multi-source basic data is subjected to lightweight preprocessing to obtain a standardized feature matrix; Based on the standardized feature matrix, a three-dimensional ecological niche benchmark vector is constructed, which includes environmental, aquaculture and breeding and whole-chain resources. The three-dimensional ecological niche benchmark vector includes environmental benchmark, aquaculture benchmark and resource benchmark. Based on the current status data of each plot, calculate the current ecological niche vector of each plot; Based on the current ecological niche vector of each plot and the three-dimensional ecological niche reference vector, the ecological niche gap value of each plot is calculated. Based on the niche gap value of each plot, the resource potential and revenue improvement space of each plot are analyzed, and each plot is divided into three levels: high potential, medium potential and low potential, and the gap classification label of each plot is obtained. Based on the gap classification labels of each plot, a triple correlation verification of gap classification, resource supply, and current planting and breeding status is performed on each plot to locate the types of resource deficiencies in each plot and calculate the gap ratio, thereby obtaining a list of resource deficiencies for each plot.

3. The method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization according to claim 2, characterized in that, The construction of a three-dimensional ecological niche benchmark vector based on the standardized feature matrix, including environment, aquaculture, and the entire resource chain, includes: An environmental baseline was calculated by weighting climate parameters, soil parameters, and topographic parameters using the analytic hierarchy process. The planting and breeding benchmark is obtained by taking the average of the relative difference between the maximum theoretical yield and the minimum yield threshold of each crop and the standardized suitability parameters of each crop. The resource benchmark is determined based on the smaller of the ratios of supply to demand for the three types of resources: water, fertilizer, and agricultural machinery. The environmental benchmark, aquaculture benchmark, and resource benchmark are weighted and summed using the analytic hierarchy process to obtain the whole-chain niche benchmark vector.

4. The method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization according to claim 1, characterized in that, Based on the Gap hierarchical labels and resource deficiency list, Gap-oriented multi-source feature collaborative mining is performed to obtain the core feature vector of Gap orientation, including: Based on the standardized feature matrix, three categories of original features—environment, farming, and resources—are extracted to form a candidate feature set; the candidate feature set includes multiple candidate features. Using the aforementioned Gap classification labels and resource deficiency list as screening guidelines, the Pearson correlation coefficient between each candidate feature of each plot and the niche Gap value is calculated; Based on the Pearson correlation coefficient, the candidate features are initially screened to obtain the initial screening feature set; The continuous numerical features in the initial screening feature set are input into the variational autoencoder, which maps the high-dimensional continuous features to the latent space distribution. After reparameterization, the decoder reconstructs the output to obtain the denoised continuous feature vector. The discrete category features in the initial screening feature set are input into the random forest model, and the importance score of each discrete feature is calculated by the Gini gain of multiple decision trees to obtain the discrete feature importance vector. Based on the niche gap value of each plot, the fusion weights of the denoised continuous feature vector and the discrete feature importance vector are assigned; Based on the denoised continuous feature vector and discrete feature importance vector, as well as the fusion weight, a weighted fusion is performed to obtain the gap-oriented core feature vector of each land parcel.

5. The method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization according to claim 1, characterized in that, The dynamic adaptation prediction based on gap-level labels and core feature vectors yields prediction results for each land parcel, including: Based on the Gap classification labels of each land parcel, a prediction model corresponding to each land parcel is selected and determined; wherein, each Gap level corresponds to one prediction model. Based on the core feature vectors of each plot and the corresponding prediction model, the three-dimensional prediction values ​​of each plot are predicted. The three-dimensional prediction values ​​include crop suitability prediction value, resource suitability prediction value, and yield improvement space prediction value. The three-dimensional predicted values ​​are subjected to triple dynamic correction to obtain the corrected prediction results for each land parcel; the triple dynamic correction includes multi-source fusion correction, probability interval constraint correction and component independent correction.

6. The method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization according to claim 5, characterized in that, The process of performing triple dynamic correction on the three-dimensional predicted values ​​to obtain the corrected prediction results for each land parcel includes: Based on the three-dimensional predicted values, niche gap values, and resource supply-demand ratios of each plot, the three-dimensional predicted values ​​are used as the main component, and the niche gap values ​​and resource supply-demand ratios are used as supplementary components for fusion correction to obtain the initial correction results. The initial correction results are constrained within a standard probability interval to obtain the corrected prediction values ​​for each component of each land parcel; wherein the standard probability interval is [0,1]. For each component of the corrected prediction values ​​of each plot, the three components of crop suitability prediction, resource suitability prediction, and income improvement space prediction are independently fused and constrained to obtain the corrected prediction results for each plot.

7. The method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization according to claim 5, characterized in that, Before obtaining the prediction results for each land parcel by dynamically adapting and predicting based on Gap hierarchical labels and core feature vectors, the process also includes: Based on historical yield data and key parameters affecting crop growth in each plot, the crop suitability of each plot is calculated. Based on the ratio of the actual supply to the demand of various resources in each plot, the resource suitability of each plot is calculated. The crop suitability and resource suitability are used as tags, and combined with the Gap grading tags for verification to generate three-dimensional composite tags for each plot. Based on the gap classification labels of each parcel, the parcels are grouped according to the gap classification. For each Gap level, the prediction model corresponding to each Gap level is trained by using the core feature vector of the corresponding level plot as input and the three-dimensional composite label as the training target.

8. The method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization according to claim 1, characterized in that, The optimal configuration scheme includes the crop planting area, resource allocation quota, and spatial layout parameters of each plot; The process of performing full-chain collaborative hierarchical optimization based on the Gap hierarchical labels and the prediction results to obtain the optimal configuration scheme for each plot within the target area includes: Based on the Gap classification label, the optimization objective function for each plot is set: for high-potential plots, the objective is to maximize revenue and a resource utilization efficiency constraint is set; for medium-potential plots, the objective is to reduce resource waste and a minimum revenue increase constraint is set; and for low-potential plots, the objective is to maintain the status quo and a revenue fluctuation constraint is set. Based on the Gap grading labels, prediction results, and total resource data and current planting and breeding status data from multi-source basic data, constraints are determined. These constraints include Gap potential constraints, total resource constraints, spatial layout constraints, and implementation feasibility constraints. Specifically, the Gap potential constraints include the adjustment range of planting area; the total resource constraints include total resource constraints and resource temporal constraints; the spatial layout constraints include minimum planting spacing constraints and light shading constraints; and the implementation feasibility constraints include crop rotation compatibility constraints and adjustment range constraints. Based on the gap classification labels of each land parcel, the corresponding optimization solution algorithm is selected for each land parcel: high-potential land parcels are solved iteratively using a sequential quadratic programming algorithm, medium-potential land parcels are solved iteratively using a constrained particle swarm optimization algorithm, and low-potential land parcels are solved directly using a linear optimization algorithm. Based on the optimization solution algorithm corresponding to each land parcel, combined with the optimization objective function and the constraints, and using the prediction results as input data, the optimization configuration parameters of each land parcel are obtained; The optimized configuration parameters are subjected to full-chain resource linkage verification to verify the consistency of the allocation of land resources, water resources, fertilizer resources, agricultural machinery resources, labor resources and warehousing and logistics resources in real time, so as to obtain the optimal configuration scheme for each plot in the target area.

9. The method for precise allocation of agricultural resource elements based on whole-chain niche gap optimization according to claim 1, characterized in that, After obtaining the optimal configuration scheme for each plot within the target area by performing full-chain collaborative hierarchical optimization based on the Gap hierarchical labels and the prediction results, the process further includes: Receive real-time environmental data, resource consumption data, and crop growth data during the execution of the optimal configuration scheme; Calculate the deviations between the real-time environmental data, resource consumption data, and crop growth data and the set data; If the deviation exceeds the preset threshold, iterative correction is triggered, and the niche gap value of each plot is updated according to the relationship between the actual revenue of each plot, the predicted optimized revenue, and the current revenue. Based on the updated niche gap value, we re-execute the gap-oriented multi-source feature collaborative mining, dynamic adaptation prediction and full-chain collaborative hierarchical optimization solution to output the iterative configuration scheme. Calculate the prediction accuracy improvement rate, revenue improvement rate, and resource utilization efficiency improvement rate corresponding to the iterative configuration scheme; When the prediction accuracy improvement rate, revenue improvement rate, and resource utilization efficiency improvement rate reach the preset target values, the basic parameters for the next cycle are configured based on the iterative configuration scheme.

10. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor being configured to invoke and run the computer program stored in the memory to perform the method as described in any one of claims 1 to 9.

Images