A method for intelligent optimization of crop planting structure in a major grain production function zone

Abstract

This invention discloses an intelligent optimization method for crop planting structure in major grain-producing functional areas, comprising: Step 1, acquiring historical and current crop types at a 10km×10km raster pixel scale to clarify the evolution of planting structure and the current distribution of major crop types in major grain-producing functional areas; Step 2, combining the climate-soil environmental attributes of the 10km×10km raster pixels, using the APSIM model and data-driven model to train and generate an optimal surrogate model, and generating a locally optimal soil conservation management measure scheme through the NSGA-II algorithm; Step 3, under the constraints of no reduction in total grain output and no decrease in farmers' planting income, using the area proportion of each crop type in each raster as a research variable, and combining the parameters of yield and soil organic carbon content change in the local optimization matching scheme, using GAMS software programming to solve and generate a global planting structure optimization allocation scheme. This invention achieves the scientific, precise, and systematic nature of planting structure optimization at the global scale.

Description

Technical Field

[0001] This invention belongs to the fields of agricultural geography and land use information, and specifically relates to an intelligent optimization method for crop planting structure in major grain-producing functional areas. Background Technology

[0002] Optimizing the planting structure is a key measure to improve the efficiency of agricultural resource utilization, ensure food security, promote ecological environmental protection, and enhance economic benefits. Its core objective is to scientifically adjust the proportion and spatial layout of crop planting areas based on the natural resource endowment and socio-economic conditions of different regions, thereby achieving the coordinated advancement of various objectives and improving the overall efficiency of agricultural production.

[0003] Current research on planting structure optimization mainly focuses on the allocation of crop types, with less attention paid to the regional implementation of soil conservation management measures and the synergistic optimization of crop types and soil conservation management measures. From the perspective of the methodological frameworks used, previous research on planting structure optimization can be broadly divided into two categories: local optimization and global optimization. Local optimization refers to first selecting the crop type that maximizes the optimization benefits within each unit based on its attribute characteristics, and then combining these optimal patterns in a bottom-up manner to form a global optimization scheme. Common methods for determining locally optimal crop types include using the MaxEnt model to evaluate crop suitability, using crop growth process mechanism models to simulate the yield growth rate of each crop, and linking process mechanism models with data-driven models to search for the optimal pattern for each unit.

[0004] Global optimization refers to the top-down coordination of resource allocation across various crop types within a larger regional system to maximize regional benefits. Global optimization can be broadly categorized into two optimization processes. One involves first defining the parameters (such as grain yield and ecological benefits) for different crop types in each unit, and then using linear programming, multi-objective interval parametric programming, or multi-objective optimization algorithms like NSGA-II to find the optimal solution for the entire region. The other approach involves first planning the area and quantity of each crop type using linear programming or multi-objective interval parametric programming, and then allocating the crop types spatially based on their suitability (obtained through the analytic hierarchy process or the MaxEnt model).

[0005] The shortcomings of the two planting structure optimization methods mentioned above are as follows: Although the local optimization technology can solve the problem of crop types in a single region, it ignores the synergistic effect between regions and the comprehensive effect of the system, resulting in uneven allocation of overall resources; while the global optimization technology can allocate resources across the entire region, it ignores the differences in local conditions, thus affecting the actual application effect. Summary of the Invention

[0006] To overcome these technical limitations, this invention innovatively developed a planting structure optimization technology process, ranging from "local crop-livestock model optimization and matching" to "global planting structure optimization and allocation." First, using a process mechanism and data-driven fusion model, the optimal management measures under different climate and soil conditions were explored in depth, and tailored soil conservation solutions for different crop types in various regions. Then, using a multi-objective optimization algorithm, a global optimization model was designed to meet different objective requirements, achieving optimized allocation of planting structure at the global scale. Compared with traditional research methods, this approach provides optimal solutions at the local scale and achieves integrated resource allocation at the global scale, enhancing both the scientific rigor of planting structure optimization and its operability in practical applications.

[0007] This invention discloses an intelligent optimization method for crop planting structure in major grain-producing functional areas, which includes the following steps:

[0008] Step 1: Obtain historical and current crop types at a 10km × 10km grid pixel scale to clarify the evolution of planting structure and distribution of current major crop types in major grain-producing functional areas;

[0009] Step 2: Combining the climate-soil environmental attributes of 10km × 10km raster pixels, the optimal surrogate model is generated by training the APSIM model and the data-driven model, and the local optimal land conservation management measures are generated by the NSGA-II algorithm.

[0010] Step 3: Under the constraints of no decrease in total grain output and no decrease in farmers' planting income, the area ratio of each type of crop in each grid is used as the research variable. Combined with the parameters of yield and soil organic carbon content change in the local optimization matching scheme, the whole-domain planting structure optimization and allocation scheme is generated by programming with GAMS software.

[0011] Furthermore, step 1 also includes the following steps:

[0012] Step 11: Generate crop distribution data with a resolution of 100m × 100m based on the crop distribution data with a resolution of 10m × 10m using a resampling method;

[0013] Step 12: Generate crop type map encoding based on the crop type encoding of the plot at a resolution of 100m × 100m;

[0014] Step 13: Calculate the number of planting times and rotation times based on the crop type map encoding to generate various crop types;

[0015] Step 14: Generate the main crop types based on the most widespread implementation patterns within a 10km × 10km raster cell, according to various crop types.

[0016] Furthermore, in step 12, taking into account both classification accuracy and computational difficulty, a 5-year analysis time window is set. Map algebra operations are performed on the crop type distribution data for each period within the window to achieve spatial overlay of crop type information.

[0017] ;

[0018] In the formula: PC represents the information code of crop type change on cultivated land plot, t represents the time sequence number of crop type data, the earlier the year, the larger the t value; This indicates the type of crop planted on this plot of land. The codes are 1, 2, 3, and 4, with corn, soybeans, and rice being coded as 1, 2, and 3 respectively, while other crops, fallow land, or non-arable land are uniformly coded as 4;

[0019] The various crop types include continuous corn cropping, continuous soybean cropping, continuous rice cropping, and rice-soybean rotation. Among them, rice-soybean rotation includes balanced rotation, corn-preferred rotation, and soybean-preferred rotation.

[0020] Furthermore, step 2 also includes the following steps:

[0021] Step 21: Collect publicly available climate and soil data, and generate local climate-soil environmental attributes with a spatial resolution of 10km × 10km raster pixels using bilinear interpolation.

[0022] Step 22: For various types of crops, construct a combination of management parameters for soil conservation management measures, specifically including nitrogen fertilizer application rate, straw return ratio and tillage depth;

[0023] Step 23: Further combine the local climate-soil environmental attributes of each raster cell with the management parameters of soil conservation management measures, and extract a sufficient number of representative simulation scenarios for APSIM model simulation; generate the changes in local crop yield and local soil organic carbon content for each year of the next 30 years through APSIM model planting simulation; use soil, climate attributes and management parameters as predictive factors of three types of driving inputs to train various data-driven models so that they can simulate the output results of APSIM simulation, and select the data-driven model with the best performance as the optimal surrogate model;

[0024] Step 24: Using the NSGA-II algorithm, select the optimal match between crop type and soil conservation management measures for each raster cell from the optimal surrogate model and all management measures combination as the local optimization matching scheme.

[0025] Furthermore, step 3 also includes the following steps:

[0026] Step 31: Based on the main crop types, crop yields, changes in soil organic carbon content, and farmer income parameters, a set of basic parameters is generated by constructing a global optimization and allocation model for planting structure.

[0027] Step 32: Generate a global model constraint framework based on the basic parameter set by setting constraint conditions;

[0028] Step 33: Based on the global model constraint framework and combined with the optimization scenario settings, a nonlinear programming model to be solved is generated through the objective function;

[0029] Step 34: Based on the nonlinear programming model to be solved, the optimal area ratio of each type of crop in each grid is generated by programming with GAMS software.

[0030] Furthermore, in step 31, the set of basic parameters includes the yield of various crops, the total grain yield and farmers' planting income generated by each raster cell under different crop type ratios, and the average SOC change of each raster cell under different crop type ratios.

[0031] To measure the grain production capacity of multi-year crop rotation systems, it is necessary to analyze the yields of various crops within a multi-year analysis time window:

[0032]

[0033] Represents raster cells medium crops production, Represents raster cells The area of ​​cultivated land in China Indicates the type of crop planted in this raster cell. area percentage Indicates the type of crop planted medium crops The planting ratio, Represents raster cells Inland crop types Crops The average annual yield over many years;

[0034] Further calculations were performed on the total grain yield and farmers' planting income generated by each raster cell under different crop type proportions:

[0035] ;

[0036] ;

[0037] Represents raster cells Total grain crop output in China Represents raster cells Farmers' planting income Indicates in raster cells Chinese crops The average net profit over many years Indicates in raster cells Crop types adopted in China Available unit area policy subsidies;

[0038] Average SOC change of each raster cell under different crop type proportions:

[0039]

[0040] raster pixels The average change in SOC Represents raster cells The change in SOC after five years of implementing crop type p.

[0041] Furthermore, in step 32, under the premise that the total grain output and the output of each type of grain do not decrease, the output constraint formula is expressed as:

[0042]

[0043]

[0044] This represents the total grain output of the current study area. Indicates the crops in the current study area Total output;

[0045] To ensure that farmers' planting income does not decrease within each raster cell, the income constraint formula is expressed as follows:

[0046]

[0047] Indicates the current raster cell Farmers' planting income;

[0048] The formula for the constraint condition of crop type is expressed as follows:

[0049]

[0050] And if raster pixels This type of crop has never appeared in Chinese history. ,set up .

[0051] Furthermore, in step 33, three optimization scenarios are set: prioritizing food production, prioritizing arable land protection, and a trade-off between the two.

[0052] The goal of the grain production priority scenario is to optimize grain yield and pursue... That is, to maximize the total grain output of the entire study area;

[0053] The goal of the farmland protection priority scenario is to optimize farmland quality and pursue... That is, to maximize the change in the average soil organic carbon content across the entire study area;

[0054] Scenario of trade-off between farmland protection and food production The goal is to achieve significant improvements in both dimensions, while minimizing the differences between them to avoid sacrificing one dimension to advance the other. The specific calculation formula is as follows:

[0055]

[0056]

[0057]

[0058]

[0059]

[0060] For the extent of improvement in the dimension of food production, To assess the extent of improvement in farmland protection, This represents the current actual value of the total grain output in the study area. This represents the change in the average soil organic carbon content across the entire study area. This represents the current actual value of the change in the average soil organic carbon content across the entire study area. This represents the average improvement across the two dimensions. The variance represents the magnitude of improvement in both dimensions.

[0061] Furthermore, in step 34, the software GAMS is used to programmatically solve the constructed global planting structure optimization and allocation model. Specifically, the crop type and rotation mode of the raster cells in the model are set as the basic set, the cultivated land area, yield, income, and ΔSOC are set as the basic parameters, the area ratio of different crop types in each raster cell is set as the core decision variable, three types of constraints are set for yield, income, and crop type, and corresponding maximization objective functions are constructed for the three optimization scenarios. The functions are solved by a nonlinear programming solver, and the optimal area ratio configuration results of different crop types in each raster cell are output, thus obtaining the global planting structure optimization and allocation scheme.

[0062] The beneficial effects achieved by this invention are:

[0063] Compared with traditional research methods, this invention provides precise and optimal solutions for crop types and soil conservation management measures at the local scale, and also realizes the systematic integration and scientific allocation of agricultural resources at the global scale. It not only improves the scientific, precise and systematic nature of planting structure optimization, but also enhances its operability and applicability in actual agricultural production. It can be applied to the optimization of planting structure in different major grain-producing functional areas, and provides technical support for farmland protection, food security and high-quality agricultural development in major grain-producing functional areas. Attached Figure Description

[0064] Figure 1 This is a flowchart of an intelligent optimization method for crop planting structure in major grain-producing functional areas proposed in this invention;

[0065] Figure 2 This is a flowchart of the classification method for current and historical crop types in the intelligent optimization method for crop planting structure in major grain-producing functional areas proposed in this invention;

[0066] Figure 3 This is a flowchart of the local optimization matching of crop types and soil conservation management measures in the intelligent optimization method for crop planting structure in major grain-producing functional areas proposed in this invention.

[0067] Figure 4 This is a framework diagram for constructing a global optimization and allocation model of planting structure in a method for intelligent optimization of crop planting structure in major grain-producing functional areas proposed in this invention. Detailed Implementation

[0068] The present invention will be further described below with reference to specific embodiments, and the advantages and features of the present invention will become clearer as a result. However, these embodiments are merely exemplary and do not constitute any limitation on the scope of the present invention. Those skilled in the art should understand that modifications or substitutions can be made to the details and form of the technical solutions of the present invention without departing from the spirit and scope of the present invention, but all such modifications and substitutions fall within the protection scope of the present invention.

[0069] This invention, based on the practical needs of Northeast China and its crucial role in food security as the "granary of black soil," coupled with the current predicament of declining quantity and quality of black soil, proposes a research objective to achieve a balance between food production (making good use of land) and farmland protection (maintaining good land). It points out that this objective should be achieved by matching appropriate crop types with appropriate land conservation management measures. A multi-scale optimization process, moving from point to area, is employed, with the optimization scale continuously expanding and improving from point regions to local areas to the entire region. First, based on the crop distribution types at the 10m × 10m field scale, the main crop types in the region are summarized in a 10km × 10km raster grid through long-term sequence overlay. Second, optimal land conservation management measures are sought for each raster grid crop type to achieve the balance between farmland protection and food production, thus optimizing the planting and breeding patterns at the local scale. Finally, from the perspective of the entire Northeast region, the allocation of crop types is carried out across the entire region to achieve optimization of the planting structure at the global scale.

[0070] Overall technical flowchart as follows Figure 1 As shown, the process begins with basic data collection and organization. Crop distribution data is then used to categorize crop types. Field trial data is used to calibrate and validate the parameters of the process mechanism model. The validated model is then run in batches using meteorological and soil data from each raster cell, combined with gradient-set soil conservation parameters. A data-driven model is then used to learn the model and encapsulate it to obtain the optimal surrogate model. The NSGA-II algorithm is then used to match optimal soil conservation management measures for different crop types in each raster cell. Finally, a crop type allocation model is constructed at a global scale. By setting optimization scenarios such as prioritizing farmland protection, prioritizing food production, and the trade-off between the two, along with various constraints, linear programming is performed to achieve global optimization and clarify the "zoning-based policy" scheme.

[0071] The specific implementation plan consists of the following three steps:

[0072] Step 1: Classify current and historical crop types.

[0073] To facilitate subsequent research on planting structure optimization, it is necessary to simplify the classification rules for crop types as much as possible, grouping crop types with similar characteristics into one category. Maize rotation, soybean rotation, rice rotation, and maize-soybean rotation (rice-soybean rotation) are the four major crop types in Northeast China. Specific classification rules and procedures are as follows: Figure 2 As shown, it includes the following steps:

[0074] Step 11: First, to improve the data processing speed, the crop distribution data at 10 m resolution is resampled to 100 m resolution. The resampling method is to take the crop type with the widest planting area for each 100 m × 100 m raster cell.

[0075] Step 12: Next, according to formula (1), the crop planting sequence is superimposed in time to obtain the crop type map code of each grid cell.

[0076] The cropping system in Northeast China is typically a single-crop-per-year system; therefore, the crop types specifically refer to the crop planting sequence of a particular plot of farmland across different years. Considering both classification accuracy and computational difficulty, a 5-year analysis window is used. Map algebra operations are performed on the crop type distribution data for each period within the window to achieve spatial overlay of crop type information. The calculation formula is as follows:

[0077] (1)

[0079] In the formula: PC represents the crop type map code of a certain cultivated land plot, and t represents the time sequence number of the crop type data. The earlier the year, the larger the t value. For example, if 2021–2025 is used as the analysis time window, then t=1–5 represent 2021–2025 respectively. This indicates the type of crop planted on this plot of land. The codes are 1, 2, 3, and 4, with corn, soybeans, and rice coded as 1, 2, and 3 respectively, while other crops, fallow land, or non-arable land are uniformly coded as 4.

[0080] Step 13: Through encoding, the number of plantings of various crops and the number of crop rotations for each raster cell within a 5-year analysis window can be further calculated. To simplify the analysis, this invention defines continuous cropping as a crop type in which the same crop is continuously planted on the same plot for at least 4 years within 5 years. That is, when a certain crop is planted 4 or more times within the analysis period, and the number of rotations is no more than 1, it can be determined that the raster cell has implemented the continuous cropping type for that crop. For Northeast China, common continuous cropping patterns include maize continuous cropping, soybean continuous cropping, and rice continuous cropping. Regarding crop rotation, rice-soybean rotation is the most common type of crop rotation in Northeast China. The types of crop rotation involved are quite complex. This invention focuses on two key parameters: the number of soybean plantings and the number of rotations. It further categorizes numerous rice-soybean rotation patterns into three types: ① Balanced rotation (corn and soybeans are rotated annually over 5 years); ② Corn-preferred rotation (corn is planted for 3 out of 5 years, or 4 years with 2 crop rotations, encompassing the traditional corn-corn-soybean three-year rotation); ③ Soybean-preferred rotation (soybeans are planted for 3 out of 5 years, or 4 years with 2 crop rotations, encompassing the traditional soybean-soybean-corn three-year rotation).

[0081] Step 14: Finally, the implementation area of ​​various crop types is further summarized and categorized across 10 km raster cells, and the most widely implemented crop type within a raster cell is identified as the primary crop type for that raster cell. This will reveal the current distribution pattern of crop types in Northeast China, providing a practical reference and optimization foundation for subsequent local optimization matching of planting and breeding patterns and overall optimization and allocation of planting structure.

[0082] Step 2: Local optimization matching of crop types and soil conservation management measures.

[0083] The crop growth process mechanism model APSIM, which has been widely validated and applied in Northeast China, was selected and fused with several commonly used data-driven models (random forest, support vector machine, XGBoost, etc.) to achieve data augmentation of the process mechanism model at a higher spatial resolution. The specific method includes the following steps:

[0084] Step 21: Collect publicly available climate and soil data for Northeast China, and use bilinear interpolation to uniformly interpolate the spatial resolution (raster pixel size) of the data to 10km × 10km. Meteorological data includes daily maximum and minimum temperatures, precipitation, and radiation parameters for historical periods (1991–2020) and future periods (2031–2060, including three climate change scenarios: SSP1-2.6, SSP2-4.5, and SSP5-8.5). Soil data includes soil parameters such as soil organic carbon (SOC), total nitrogen (TN), cation exchange capacity (CEC), pH, bulk density (BD), soil texture (Sand, Silt, Clay), field capacity (DUL), and wilting water content (LL15) at different soil depths (0–2 m).

[0085] Step 22: For various crop types, construct fine-tuning scenarios for soil conservation management measures, setting combinations of management parameters such as nitrogen fertilizer application rate, tillage depth, and straw return rate according to a gradient. Specifically, the nitrogen fertilizer application rate ranges from 0 to 400 kg N / ha, with a gradient of 10 kg N / ha. The straw return rate ranges from 0 to 100%, with a gradient of 5%. Tillage depth includes 0 (no-till), 15, 20, 25, 30, 35, and 40 cm.

[0086] Step 23: The generated management parameter combination is further combined with the climate-soil environmental attributes of each raster cell. A suitable number of representative combination scenarios are extracted and used to run a 30-year planting simulation using the APSIM model, outputting the changes in crop yield and soil organic carbon content for each year.

[0087] To improve model simulation efficiency and achieve batch transferable simulation of the region, further training of data-driven models was conducted to obtain surrogate models capable of rapidly simulating APSIM simulation results. The selected data-driven models included LASSO regression, multilayer perceptron neural network (MLP), random forest (RF), support vector machine (SVM), and XGBoost. The predictive factors used as input to the data-driven models included three categories: soil, climate attributes, and management parameters, as shown in Table 1. The APSIM output to be simulated was the 30-year average crop yield and the change in soil organic carbon content (ΔSOC) before and after planting simulation.

[0088] Finally, the simulation performance of each model was tested, and the model with the best performance was selected as the optimal surrogate model. This model was then used to further simulate the arable land quality and grain production capacity of all raster pixels that had historically been planted with this crop type, based on land conservation management measures.

[0089] Table 1. Predictors for Data-Driven Model Training Input

[0090]

[0091] The calculation formulas for various accumulated temperatures, such as CDD, GDDlow, GDDhigh, and HDD, are as follows:

[0092] (2)

[0094] (3)

[0096] (4)

[0098] (5)

[0100] In the formula: Indicates hourly temperature. , and These represent the three cardinal temperatures of a crop: the lower limit temperature, the optimum temperature, and the upper limit temperature. Based on existing research, the time divisions for different growth stages of various crops and the corresponding cardinal temperature settings are shown in Table 2.

[0101] Table 2. Cardinal temperatures at different growth stages of different crops

[0102]

[0103] Step 24: For each raster cell with a given crop type, batch simulations are performed using the trained optimal surrogate model for all combinations of management measures. Specifically, the optimization objectives are to maximize average crop yield and maximize the change in soil organic carbon content (ΔSOC) before and after the planting simulation. The NSGA-II algorithm is used to search for Pareto optimal solutions. Then, the Pareto optimal solutions are filtered using 95% of the maximum average crop yield and ΔSOC value among all combinations as thresholds. If multiple solutions still meet the conditions, the optimal solution is further selected based on the principle of maximizing yield and minimizing nitrogen fertilizer application, thus achieving optimal matching between crop types and soil conservation management measures for each raster cell. Each raster cell may have multiple crop types; for example, a raster cell may simultaneously have maize and soybean continuous cropping. For maize continuous cropping, there is an optimal soil management scheme (e.g., nitrogen application of 200 kg / ha, straw return rate of 50%, and tillage depth of 20 cm); for soybean continuous cropping, there is an optimal soil management scheme (e.g., nitrogen application of 30 kg / ha, straw return rate of 20%, and tillage depth of 25 cm). For each raster cell, it is necessary to search for the optimal soil management scheme for all crop types present in that raster. Finally, the local optimization matching process for the optimal nitrogen fertilizer application rate, tillage depth, and straw return rate under each crop type is as follows: Figure 3 As shown, this will enable precise customization of soil conservation measures for different regions, clarify the optimal soil conservation schemes and regional adaptability for different crop types, and provide refined local technical parameter support for subsequent optimization of the overall planting structure.

[0104] Step 3: Optimize and allocate the planting structure across the entire area.

[0105] The main goal of optimizing the planting structure in Northeast China is to achieve a balance between farmland protection and grain production. To this end, the following planting structure optimization and allocation model was constructed to further improve the quality of farmland and grain yield across the entire Northeast China region.

[0106] Step 31, setting parameter variables:

[0107] For each 10km×10km grid cell Its interior is not entirely arable land, and often contains multiple types of crops. Therefore, the first step is to clarify the cultivated land area within each raster cell. Area proportion of various crop types Furthermore, to measure the grain production capacity of multi-year crop rotation patterns, it is necessary to analyze various crops over a longer analytical time window (such as 5 years in this case). The output. The specific calculation formula is:

[0108] (6)

[0110] In the formula: Represents raster cells medium crops production, Represents raster cells The area of ​​cultivated land in China Indicates the type of crop planted in this raster cell. The area ratio will also be the main target for subsequent optimization and allocation. Indicates the type of crop planted medium crops The planting ratio, for example, in a corn-corn-soybean rotation pattern, the corn planting ratio is... The soybean planting ratio is . Represents raster cells Inland crop types Crops The average annual yield over many years.

[0111] Based on this, the total grain output and farmers' planting income generated by each raster cell under different crop types and proportions can be calculated:

[0112] (7)

[0114] (8)

[0116] In the formula: Represents raster cells Total grain crop output in the country. Represents raster cells Farmers' planting income Indicates in raster cells Chinese crops The average net profit over the years (unit: yuan / kg). Indicates in raster cells Crop types adopted in China Available policy subsidies per unit area (unit: yuan / hectare).

[0117] Similarly, the average SOC change of each raster cell can also be measured under different crop types and proportions:

[0118] (9)

[0120] In the formula: raster pixels The average SOC change represents the change in farmland quality for that raster cell. Represents raster cells The change in SOC after five years of implementing crop type p.

[0121] Step 32, Setting Constraints

[0122] Global optimization and allocation must be carried out under certain constraints. First, it is necessary to ensure that the total grain output and the output of all types of grain in the entire Northeast region do not decrease. The formula for the output constraint is expressed as:

[0123] (10)

[0125] (11)

[0127] Secondly, it is necessary to ensure that the planting income of farmers within each raster cell does not decrease. The income constraint formula is expressed as:

[0128] (12)

[0130] Meanwhile, to respect farmers' planting habits and crop adaptability, each raster cell should only consider crop types that have appeared in historical periods. The formula for the crop type constraint is expressed as follows:

[0131] (13)

[0133] And if raster pixels This type of crop has never appeared in Chinese history. ,set up .

[0134] Step 33, Optimize scenario settings

[0135] Three optimization scenarios were set up: prioritizing grain production, prioritizing arable land protection, and a balance between the two.

[0136] The food production priority scenario primarily considers maximizing total food output, i.e., pursuing... .

[0137] The farmland protection priority scenario primarily considers the optimization of farmland quality, namely, maximizing the change in the average soil organic carbon content across the entire area, and pursuing... .

[0138] Balancing farmland protection and food production requires a comprehensive consideration of both dimensions. The goal is to achieve the greatest possible improvement in each dimension while minimizing the differences between the two, avoiding situations where one dimension is sacrificed to advance the other. .in The formula for calculating the improvement in a specific dimension is as follows:

[0139] (14)

[0141] (15)

[0143] (16)

[0145] (17)

[0147] (18)

[0149] Step 34: Further utilize the professional operations research and planning analysis software GAMS to programmatically solve the constructed global planting structure optimization and allocation model. Specifically, set the crop types and crop rotation methods of the raster cells in the model as the basic set, and set cultivated land area, yield, income, ΔSOC, etc. as basic parameters. Set the area proportion of different crop types in each raster cell as the core decision variable. Embed the three types of constraints set in Step 32—yield, income, and crop type—into the software, and construct corresponding maximization objective functions for the three optimization scenarios in Step 33. Select an appropriate nonlinear programming solver to solve the problem, and output the optimal area proportion configuration results for different crop types in each raster cell, thus obtaining the global planting structure optimization and allocation scheme. The overall framework for constructing the global planting structure optimization and allocation model is as follows: Figure 4 As shown.

[0150] The above are merely preferred embodiments of the present invention and do not constitute any limitation on the scope of protection of the present invention; all technical solutions formed by equivalent transformations or equivalent substitutions fall within the scope of protection of the present invention; the parts of the present invention not described in detail are well-known technologies to those skilled in the art.

Claims

1. A method for intelligent optimization of crop planting structure in major grain-producing functional areas, characterized in that, The intelligent optimization method for crop planting structure in major grain-producing functional areas includes the following steps: Step 1: Obtain historical and current crop types at a 10km × 10km grid pixel scale to clarify the evolution of planting structure and distribution of current major crop types in major grain-producing functional areas; Step 2: Combining the climate-soil environmental attributes of 10km × 10km raster pixels, the optimal surrogate model is generated by training the APSIM model and the data-driven model, and the local optimal land conservation management measures are generated by the NSGA-II algorithm. Step 3: Under the constraints of maintaining total grain output and farmers' planting income, the area proportion of each type of crop in each grid is used as the research variable. Combined with the parameters of yield and soil organic carbon content changes in the local optimization matching scheme, a global planting structure optimization and allocation scheme is generated through programming using GAMS software.

2. The intelligent optimization method for crop planting structure in major grain-producing functional areas according to claim 1, characterized in that, Step 1 also includes the following steps: Step 11: Generate crop distribution data with a resolution of 100m × 100m based on the crop distribution data with a resolution of 10m × 10m using a resampling method; Step 12: Generate crop type map encoding based on the crop type encoding of the plot at a resolution of 100m × 100m; Step 13: Calculate the number of planting times and rotation times based on the crop type map encoding to generate various crop types; Step 14: Generate the main crop types based on the most widespread implementation patterns within a 10km × 10km raster cell, according to various crop types.

3. The intelligent optimization method for crop planting structure in major grain-producing functional areas according to claim 2, characterized in that, In step 12, taking into account both classification accuracy and computational difficulty, a 5-year analysis time window is set. Map algebra operations are performed on the crop type distribution data for each period within the window to achieve spatial overlay of crop type information. ； In the formula: PC represents the information code of crop type change on cultivated land plot, t represents the time sequence number of crop type data, the earlier the year, the larger the t value; This indicates the type of crop planted on this plot of land. The codes are 1, 2, 3, and 4, with corn, soybeans, and rice being coded as 1, 2, and 3 respectively, while other crops, fallow land, or non-arable land are uniformly coded as 4; The various crop types include continuous corn cropping, continuous soybean cropping, continuous rice cropping, and rice-soybean rotation. Among them, rice-soybean rotation includes balanced rotation, corn-preferred rotation, and soybean-preferred rotation.

4. The intelligent optimization method for crop planting structure in major grain-producing functional areas according to claim 1, characterized in that, Step 2 also includes the following steps: Step 21: Collect publicly available climate and soil data, and generate local climate-soil environmental attributes with a spatial resolution of 10km × 10km raster pixels using bilinear interpolation. Step 22: For various types of crops, construct a combination of management parameters for soil conservation management measures, specifically including nitrogen fertilizer application rate, straw return ratio and tillage depth; Step 23: Further combine the local climate-soil environmental attributes of each raster cell with the management parameters of soil conservation management measures, and extract a sufficient number of representative simulation scenarios for APSIM model simulation; generate the changes in local crop yield and local soil organic carbon content for each year of the next 30 years through APSIM model planting simulation; use soil, climate attributes and management parameters as predictive factors of three types of driving inputs to train various data-driven models so that they can simulate the output results of APSIM simulation, and select the data-driven model with the best performance as the optimal surrogate model; Step 24: Using the NSGA-II algorithm, select the optimal match between crop type and soil conservation management measures for each raster cell from the optimal surrogate model and all management measures combination as the local optimization matching scheme.

5. The intelligent optimization method for crop planting structure in major grain-producing functional areas according to claim 1, characterized in that, Step 3 also includes the following steps: Step 31: Based on the main crop types, crop yields, changes in soil organic carbon content, and farmer income parameters, a set of basic parameters is generated by constructing a global optimization and allocation model for planting structure. Step 32: Generate a global model constraint framework based on the basic parameter set by setting constraint conditions; Step 33: Based on the global model constraint framework and combined with the optimization scenario settings, a nonlinear programming model to be solved is generated through the objective function; Step 34: Based on the nonlinear programming model to be solved, the optimal area ratio of each type of crop in each grid is generated by programming with GAMS software.

6. The intelligent optimization method for crop planting structure in major grain-producing functional areas according to claim 5, characterized in that, In step 31, the set of basic parameters includes the yield of various crops, the total grain yield and farmers' planting income generated by each raster cell under different crop type ratios, and the average SOC change of each raster cell under different crop type ratios. To measure the grain production capacity of multi-year crop rotation systems, it is necessary to analyze the yields of various crops within a multi-year analysis time window: ； Represents raster cells medium crops production, Represents raster cells The area of ​​cultivated land in China Indicates the type of crop planted in this raster cell. area percentage Indicates the type of crop planted medium crops The planting ratio, Represents raster cells Inland crop types Crops The average annual yield over many years; Further calculations were performed on the total grain yield and farmers' planting income generated by each raster cell under different crop type proportions: ； ； Represents raster cells Total grain crop output in China Represents raster cells Farmers' planting income Indicates in raster cells Chinese crops The average net profit over many years Indicates in raster cells Crop types adopted in China Available unit area policy subsidies; Average SOC change of each raster cell under different crop type proportions: ； raster pixels The average change in SOC Represents raster cells The change in SOC after five years of implementing crop type p.

7. The intelligent optimization method for crop planting structure in major grain-producing functional areas according to claim 5, characterized in that, In step 32, under the premise that the total grain output and the output of each type of grain do not decrease, the output constraint formula is expressed as: ； ； This represents the total grain output of the current study area. Indicates the crops in the current study area Total output; To ensure that farmers' planting income does not decrease within each raster cell, the income constraint formula is expressed as follows: ； Indicates the current raster cell Farmers' planting income; The formula for the constraint condition of crop type is expressed as follows: ； And if raster pixels This type of crop has never appeared in Chinese history. ,set up .

8. The intelligent optimization method for crop planting structure in major grain-producing functional areas according to claim 5, characterized in that, In step 33, three optimization scenarios are set: prioritizing food production, prioritizing arable land protection, and a trade-off between the two. The goal of the grain production priority scenario is to optimize grain yield and pursue... That is, to maximize the total grain output of the entire study area; The goal of the farmland protection priority scenario is to optimize farmland quality and pursue... That is, to maximize the change in the average soil organic carbon content across the entire study area; Scenario of trade-off between farmland protection and food production The goal is to achieve significant improvements in both dimensions, while minimizing the differences between them to avoid sacrificing one dimension to advance the other. The specific calculation formula is as follows: ； ； ； ； ； For the extent of improvement in the dimension of food production, To assess the extent of improvement in farmland protection, This represents the current actual value of the total grain output in the study area. This represents the change in the average soil organic carbon content across the entire study area. This represents the current actual value of the change in the average soil organic carbon content across the entire study area. This represents the average improvement across the two dimensions. The variance represents the magnitude of improvement in both dimensions.

9. The intelligent optimization method for crop planting structure in major grain-producing functional areas according to claim 5, characterized in that, In step 34, the software GAMS is used to programmatically solve the constructed global planting structure optimization and allocation model. Specifically, the crop type and rotation mode of the raster cells in the model are set as the basic set, the cultivated land area, yield, income, and ΔSOC are set as basic parameters, the area ratio of different crop types in each raster cell is set as the core decision variable, and three types of constraints are set: yield, income, and crop type. Corresponding maximization objective functions are constructed for the three optimization scenarios, and solved by a nonlinear programming solver. The optimal area ratio configuration results of different crop types in each raster cell are output, and the global planting structure optimization and allocation scheme is obtained.

Images