Soil improvement and soil quality optimization comprehensive management method based on big data analysis

Abstract

The invention discloses a soil improvement and soil quality optimization comprehensive management method based on big data analysis, and the method comprises the following steps: S1, collecting and preprocessing multi-source soil data, and constructing a soil data input matrix; s2, initializing parameters of the fuzzy C-means clustering model, and constructing a model basic structure; s3, introducing a hosting crab optimization algorithm, and optimizing clustering model parameters; s4, executing clustering analysis, and generating a fuzzy soil type distribution map; s5, constructing an improved strategy scheme space, and fusing crop types, historical records and resource information; s6, making a multi-target scoring standard, and selecting an optimal improvement strategy; s7, deploying an optimal improvement strategy to a farmland operation system, and collecting soil feedback data in real time; and S8, updating model input and parameters based on feedback data, and re-executing optimization to realize a closed loop. According to the invention, big data and an intelligent optimization technology are fused, and accurate matching and efficient configuration of soil improvement strategies are realized.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to a comprehensive management method for soil improvement and soil quality optimization based on big data analysis. Background Art

[0002] As agricultural production evolves toward higher efficiency, intensification, and precision, farmland soil, as the core foundation of cultivated land ecosystems, has a crucial impact on crop yield, quality, and ecological sustainability. However, long-term irrational tillage practices, excessive fertilization, and a single management model have led to soil compaction, organic matter degradation, acid-base imbalances, and weakened water retention in many parts of my country, severely impacting agricultural production efficiency and resource utilization. Therefore, promoting soil improvement and optimizing soil quality has become a key component of modern agricultural management.

[0003] Existing soil improvement technologies primarily rely on expert systems based on soil testing data and field experience, or on rule bases derived from historical soil improvement records, for crop management. While these approaches can achieve targeted management to a certain extent, they also have significant limitations. First, traditional methods often rely on single-point or small-sample data sampling, making it difficult to reflect the spatial variability and dynamic trends of soil types across large and diverse farmland areas, leading to uncertainty in the application of these strategies. Second, soil properties exhibit highly nonlinear coupling, such as the complex relationships between structure, fertility level, water retention capacity, and pH. Existing methods lack systematic modeling to fully understand the synergistic mechanisms between these variables. Furthermore, traditional management approaches often struggle to integrate multiple constraints, such as the varying needs of target crops, regional resource availability, and economic implementation costs. This results in soil improvement strategies failing to achieve true local adaptation and economic adaptability.

[0004] In recent years, some studies have introduced data mining and machine learning methods to analyze and process soil information. Methods such as support vector machines and random forests have been tried for soil classification and crop yield prediction. However, these methods generally focus on the classification modeling level and lack a comprehensive management framework for improvement strategies based on multi-source soil attributes. At the same time, most of these models focus on the output of static analysis results and lack a closed-loop management path. They cannot form a complete cycle of "perception-analysis-decision-feedback", and cannot adapt to complex situations such as fuzzy transitions between different plots and unclear boundaries of soil types. In addition, most models ignore the resource constraints and economic cost assessments of actual agricultural operations, resulting in poor implementation and weak adaptability of the recommended strategies in actual implementation, which cannot meet production management needs.

[0005] In terms of strategy screening, existing systems often use rule-based, linear scoring, or expert scoring to evaluate the pros and cons of strategies, making it difficult to simultaneously consider the multi-objective balance between soil improvement effects, crop response potential, and input costs. Furthermore, they lack the ability to model uncertainty and ambiguity, making it difficult to handle the state of plots in real soil space with multiple attributes and blurred boundaries. At the same time, most studies only address the one-way allocation method of "selecting one strategy for a certain type of soil" at the strategy level, and are unable to design integrated strategies for mixed plots with multiple memberships, which seriously restricts the level of intelligent management of large-scale farmland heterogeneity.

[0006] Therefore, how to provide a comprehensive management method for soil improvement and soil quality optimization based on big data analysis is an urgent problem that technicians in this field need to solve. Summary of the Invention

[0007] One objective of the present invention is to propose a comprehensive management method for soil improvement and soil quality optimization based on big data analysis. This method integrates multi-source heterogeneous data processing technology, fuzzy C-means clustering modeling, an improved hermit crab optimization algorithm, and a multi-objective scoring strategy to systematically construct an intelligent identification and refined decision-making mechanism for soil characteristics of different plots. The method describes in detail the entire process, from soil data collection, clustering modeling, strategy screening, to personalized integration and feedback optimization. This method boasts strong adaptability, high strategy matching accuracy, controllable improvement effects, and a strong intelligent feedback closed-loop capability. It can significantly improve the scientific, economic, and sustainable nature of farmland soil management and has broad prospects for agricultural application.

[0008] A comprehensive management method for soil improvement and soil quality optimization based on big data analysis according to an embodiment of the present invention includes the following steps:

[0009] S1. Collect multi-source soil data in the farmland area, pre-process the multi-source soil data, and construct a soil data input matrix;

[0010] S2. Initialize the key parameters of the fuzzy C-means clustering model and build the basic structure of the fuzzy C-means clustering model;

[0011] S3. Introducing the hermit crab optimization algorithm to collaboratively optimize the key parameters of the fuzzy C-means clustering model to obtain the optimized fuzzy C-means clustering model;

[0012] S4, inputting the soil data input matrix into the optimized fuzzy C-means clustering model, performing clustering operation, outputting the membership degree of each plot corresponding to each cluster category, and generating a fuzzy soil type distribution map;

[0013] S5. Based on the fuzzy soil type distribution map, construct an improvement strategy space for each type of soil, and generate a set of candidate improvement strategies based on the target crop type, historical improvement records, and resource availability;

[0014] S6. Based on the membership distribution of each plot in the fuzzy soil type distribution map, a multi-objective scoring standard is formulated, and the optimal improvement strategy corresponding to each soil type is selected from the candidate improvement strategy set;

[0015] S7. Deploy the optimal improvement strategy to the farmland operation system and collect soil status feedback data in real time during the execution process, including the time series change values of the target soil parameters and the crop growth response;

[0016] S8. Based on the soil state feedback data, the input data and parameter distribution of the fuzzy C-means clustering model are updated, and the hermit crab optimization algorithm is re-executed to reconstruct the model to complete the closed-loop optimization.

[0017] Optionally, the multi-source soil data specifically includes soil physical and chemical properties, historical modification records and spatial location data, which are used to construct a soil data input matrix.

[0018] Optionally, the multi-source soil data is preprocessed, specifically including missing value filling, outlier removal, standardization and spatial information alignment, to improve the quality and consistency of the multi-source soil data.

[0019] Optionally, the S2 specifically includes:

[0020] S21, setting the number of cluster categories to guide the fuzzy C-means clustering model to divide the soil data into multiple soil type categories, wherein the number of cluster categories determines the number of cluster centers;

[0021] S22, setting a fuzzy factor, which is used to adjust the degree of membership distribution of each sample data to different cluster centers during the clustering process. The fuzzy factor is greater than 1 to control the fuzziness of the clustering result;

[0022] S23, based on the constructed soil data input matrix, extracting the eigenvector of each sample as input data of the fuzzy C-means clustering model, and establishing the fuzzy C-means clustering model input layer;

[0023] S24, initializing the initial positions of multiple cluster centers in the soil feature space according to the set number of cluster categories, each cluster center corresponding to a representative feature configuration of a soil type as the basis of the clustering target of the fuzzy C-means clustering model;

[0024] S25. Initialize a sample membership matrix, where the membership matrix is used to store the degree of membership of each sample to all cluster centers. The membership value is a real number, and the sum of all memberships of each sample is 1.

[0025] S26. According to the set number of cluster categories, fuzzy factors, initial cluster center positions and sample membership matrix, the basic structure of the fuzzy C-means clustering model is constructed, including the input layer of the model, the cluster center parameter set and the membership calculation mechanism.

[0026] Optionally, the S3 specifically includes:

[0027] S31, initializing the population of the hermit crab optimization algorithm, dividing the population into a main population and multiple sub-populations, wherein the sub-populations include a boundary exploration population, a local refinement population, and an elite memory population, which are used to perform global search, local search, and optimal solution retention respectively;

[0028] S32. Randomly assign a set of key parameters of the fuzzy C-means clustering model to each hermit crab individual in the main population to form a parameter group:

[0029] [C i ,m i ,V i ,μ i ];

[0030] Among them, C i is the number of cluster categories, m i is the fuzzy factor, V i is the cluster center set, μ i is the sample membership matrix;

[0031] S33, construct the corresponding fuzzy C-means clustering model based on each parameter group, and calculate the clustering error objective function value J i :

[0032]

[0033] Among them, x n is the nth soil sample, v j is the jth cluster center, μ nj Indicates the membership value of the nth sample belonging to the jth cluster center, C i represents the number of cluster categories set by the i-th individual, and N represents the total number of soil samples;

[0034] S34. Calculate the comprehensive fitness function value F of all candidate individuals in the main group i , including the initial generated individuals, sub-population migration individuals and new individuals generated by shell fusion:

[0035]

[0036] Among them, DBI s is the Davies-Bouldin index, Si is the silhouette coefficient, α and β are weighting factors;

[0037] S35. Deploy several scout crabs in the boundary area of the solution space. Each scout crab samples multiple candidate points in the boundary area, evaluates the sampled points based on the fitness function, and selects the best candidate point as the boundary optimal point x. s and feed back the optimal point of the boundary to the main group;

[0038] S36. Execute the search strategy of each sub-group: the boundary exploration group performs large perturbation updates on the parameters, the local refinement group performs small perturbation updates on the neighborhood of the current optimal solution, and the elite memory group saves the optimal solutions of previous generations and participates in rotation and reorganization;

[0039] S37. In the main population, select any two individuals A and B and calculate the complementary synergy Comp(A,B):

[0040] Comp(A,B)=γ1·‖x A -x B ‖+γ2·|F A -F B |+γ3·D local ;

[0041] Among them, x A 、x B is the parameter vector, F A is the comprehensive fitness function value of individual A, F B is the comprehensive fitness function value of individual B, γ1 is the weighted coefficient of the solution vector difference term, γ2 is the weighted coefficient of the fitness difference term, γ3 is the weighted coefficient of the local density term, D local is the local solution density. When Comp(A,B)>θ, shell fusion is performed to generate a new individual, where θ represents the triggering threshold of shell fusion:

[0042] x new =λ o ·x A +(1-λ)·x B ;

[0043] Among them, x new represents the solution vector of the new individual generated after shell fusion, λ o is the fusion coefficient;

[0044] S38. For all candidate individuals, including the new individuals generated after shell fusion, combine the boundary optimal point x of the scout crab feedback s , perform a guided location update:

[0045]

[0046] in, represents the new position of the i-th candidate individual after the t+1-th iteration, is the position of the i-th candidate individual in the current main population at the t-th iteration, δ1, δ2 and δ3 are update coefficients, Represents the sample membership matrix corresponding to the individual The changing trend of is the nonlinear interaction function between individuals and population centers, is the average value of all individual parameter vectors of the current main group;

[0047] S39, recalculating the fitness function values of all individuals whose positions have been updated, performing individual replacement, retention, or migration operations on the main group and each sub-group according to the fitness changes, and updating the global optimal solution;

[0048] S310, when the number of iterations reaches the set maximum number of iterations T max Finally, the parameter combination with the highest fitness is output [C * ,m * ,V * ,μ * ] and construct the optimized fuzzy C-means clustering model.

[0049] Optionally, the S4 specifically includes:

[0050] S41. Input the constructed soil data into the matrix X = {x1, x2, ..., x N} is input into the optimized fuzzy C-means clustering model, x i represents the eigenvector of the i-th soil sample, i = 1, 2, …, N, where N represents the total number of soil samples;

[0051] S42, using the parameter combination with the highest fitness [C * ,m * ,V * ,μ * ], for each soil sample x i Calculate for each cluster center v j The membership degree μ ij :

[0052]

[0053] Among them, μ ij Indicates the membership value of the i-th sample belonging to the j-th category, C * is the final number of cluster categories, m * is the fuzzy factor, v j ∈V * represents the j-th cluster center vector, ‖‖ represents the Euclidean distance;

[0054] S43, build dimension is N×C * The sample membership matrix μ={μ ij}, where each row represents the degree of membership of a soil sample in each cluster category;

[0055] S44. Based on the membership matrix μ, a fuzzy soil type distribution map is generated, where the fuzzy soil type distribution map is used to express the fuzzy belonging relationship distribution of different soil types for each plot in the farmland.

[0056] Optionally, the S5 specifically includes:

[0057] S51. Classify the soil samples according to their cluster categories according to the fuzzy soil type distribution map to form multiple soil type groups;

[0058] S52. For each soil type, construct a preliminary improvement strategy space, wherein the preliminary improvement strategy space includes a combination of soil improvement measures, covering physical structure conditioning, nutrient regulation, organic matter supplementation, acid-base neutralization, and microbial activation;

[0059] S53. Introduce target crop type information and analyze crop-specific soil requirement parameters as constraints for the preliminary improvement strategy space;

[0060] S54. Call historical improvement record data to conduct correlation analysis on the operation combinations, implementation effects and crop responses of similar soil types in past improvement practices, for reference and optimization of the current strategy construction process;

[0061] S55. Consider the resources available within the current timeframe, including available materials, technical equipment, operational capacity, and budget, to identify the feasibility of implementing various improvement measures in actual farmland, thereby limiting infeasible or prohibitively costly strategic options.

[0062] S56. Taking into account the target crop type, historical improvement records, and resource availability constraints, the preliminary strategy plan space is filtered, screened, and reconstructed to form a set of candidate improvement strategies suitable for the current agricultural scenario.

[0063] Optionally, the crop's specific soil requirement parameters specifically include structural adaptability, fertility level, and water retention capacity, which are used to guide the adaptation and matching of improvement strategies with target crop growth conditions.

[0064] Optionally, the S6 specifically includes:

[0065] S61, extracting the membership distribution information corresponding to each plot in the fuzzy soil type distribution map, and obtaining the membership weights of the plots under different soil type categories;

[0066] S62. Develop a multi-objective scoring standard for comprehensively evaluating the quality of candidate improvement strategies for each soil type.

[0067] S63. Evaluate each candidate soil improvement strategy set item by item based on the established multi-objective scoring criteria to generate corresponding strategy scoring results;

[0068] S64. Selecting the strategy with the best strategy score result from the strategy set for each soil type as the optimal improvement strategy for the soil type category;

[0069] S65. Combine the membership weights of each plot under different soil type categories, perform weighted integration on the selected optimal improvement strategies for each category, and generate an integrated personalized improvement strategy at the plot level.

[0070] Optionally, the multi-objective scoring criteria specifically include soil improvement effectiveness, target crop response performance and implementation cost, which comprehensively measure the adaptability and superiority of the candidate improvement strategies.

[0071] The beneficial effects of the present invention are:

[0072] The present invention realizes the comprehensive management and intelligent decision-making of farmland soil improvement and soil quality optimization by integrating big data analysis technology, fuzzy clustering modeling method and intelligent optimization algorithm, which significantly overcomes the shortcomings of existing technologies in terms of rough soil classification, single strategy matching, lack of personalization and weak feedback mechanism. By collecting a large amount of multi-source soil data in the farmland area and performing standardized preprocessing, a high-quality soil data input matrix is constructed, which effectively breaks through the information loss problem caused by the traditional technology relying on small samples and point data, and provides a comprehensive and detailed data foundation for subsequent analysis. The fuzzy C-means clustering model is introduced and combined with the hermit crab optimization algorithm to coordinately optimize the clustering parameters, so that the soil type division has stronger spatial continuity and boundary fuzzy recognition ability, which significantly improves the accuracy and stability of soil classification, and is particularly suitable for actual scenarios with complex land transitions and staggered distribution of soil properties.

[0073] The present invention further integrates key factors such as crop growth requirements, historical improvement experience, and resource availability, constructs a multi-factor driven strategy set, and comprehensively evaluates various strategies by setting a multi-objective scoring standard that includes soil improvement effectiveness, target crop response performance, and implementation cost. This mechanism effectively solves the defects of traditional methods that use a single dimension or rule-based scoring to select the best, and achieves a balance between improvement effect, crop yield, and economic feasibility. At the same time, based on the membership distribution of each plot in the fuzzy clustering results, the present invention innovatively weights and fuses the optimal strategies to generate personalized soil improvement plans for specific plots, fully embodying the modern agricultural management concept of "adapting to local conditions, classifying policies, and precise regulation."

[0074] Through the above improvements, the present invention has established a complete closed-loop management process covering data collection, intelligent identification, strategy construction, optimization and screening, and personalized integration. It has significant advantages such as full information utilization, high intelligence, accurate strategy matching, and strong execution feasibility. This method can be widely used in scenarios such as farmland soil improvement, improving arable land quality, and optimizing the allocation of agricultural production resources. It has significant practical value and promotion potential, and can effectively support the realization of high-quality agricultural development and green production goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0076] Figure 1 This is a flow chart of a comprehensive management method for soil improvement and soil quality optimization based on big data analysis proposed by the present invention;

[0077] Figure 2 This is a schematic diagram of the processing flow of constructing a soil type distribution map based on a fuzzy C-means clustering model in a comprehensive management method for soil improvement and soil quality optimization based on big data analysis proposed in the present invention;

[0078] Figure 3 This is a collaborative optimization structure diagram of the fuzzy clustering parameters optimized by the hermit crab optimization algorithm in combination with the comprehensive management method of soil improvement and soil quality optimization based on big data analysis proposed by the present invention. DETAILED DESCRIPTION

[0079] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0080] refer to Figure 1 、 Figure 2 and Figure 3A comprehensive management method for soil improvement and soil quality optimization based on big data analysis includes the following steps:

[0081] S1. Collect multi-source soil data in the farmland area, pre-process the multi-source soil data, and construct a soil data input matrix;

[0082] S2. Initialize the key parameters of the fuzzy C-means clustering model and build the basic structure of the fuzzy C-means clustering model;

[0083] S3. Introducing the hermit crab optimization algorithm to collaboratively optimize the key parameters of the fuzzy C-means clustering model to obtain the optimized fuzzy C-means clustering model;

[0084] S4, inputting the soil data input matrix into the optimized fuzzy C-means clustering model, performing clustering operation, outputting the membership degree of each plot corresponding to each cluster category, and generating a fuzzy soil type distribution map;

[0085] S5. Based on the fuzzy soil type distribution map, construct an improvement strategy space for each type of soil, and generate a set of candidate improvement strategies based on the target crop type, historical improvement records, and resource availability;

[0086] S6. Based on the membership distribution of each plot in the fuzzy soil type distribution map, a multi-objective scoring standard is formulated, and the optimal improvement strategy corresponding to each soil type is selected from the candidate improvement strategy set;

[0087] S7. Deploy the optimal improvement strategy to the farmland operation system and collect soil status feedback data in real time during the execution process, including the time series change values of the target soil parameters and the crop growth response;

[0088] S8. Based on the soil state feedback data, the input data and parameter distribution of the fuzzy C-means clustering model are updated, and the hermit crab optimization algorithm is re-executed to reconstruct the model to complete the closed-loop optimization.

[0089] The present invention provides a comprehensive management method for soil improvement and soil quality optimization based on big data analysis. By constructing an intelligent analysis system with a fuzzy C-means clustering model at its core and incorporating a hermit crab optimization algorithm for parameter optimization, this system achieves in-depth mining and detailed modeling of multi-source soil data, enabling accurate identification of the distribution characteristics of different soil types in farmland. By generating a fuzzy soil type distribution map, it reflects the membership information of soil types in each plot, providing a basis for subsequent strategy matching. At the strategy formulation level, the method integrates target crop needs, historical improvement records, and resource availability information to establish a multi-objective scoring mechanism, comprehensively assessing the feasibility and superiority of candidate strategies and enabling personalized recommendations tailored to local conditions. The system possesses closed-loop optimization capabilities and can dynamically adjust model parameters and improvement plans based on soil status feedback, continuously improving management accuracy and responsiveness. Compared with traditional methods, the present invention offers significant advantages, including high modeling accuracy, scientific strategy matching, strong execution flexibility, and a complete decision-making closed-loop. It significantly improves soil improvement efficiency and crop yields, promoting the development of intelligent and sustainable agriculture.

[0090] In this embodiment, the multi-source soil data specifically includes soil physical and chemical properties, historical soil modification records, and spatial location data, which are used to construct a soil data input matrix.

[0091] In this embodiment, the multi-source soil data is preprocessed, including missing value filling, outlier removal, standardization and spatial information alignment, so as to improve the quality and consistency of the multi-source soil data.

[0092] In this embodiment, S2 specifically includes:

[0093] S21, setting the number of cluster categories to guide the fuzzy C-means clustering model to divide the soil data into multiple soil type categories, wherein the number of cluster categories determines the number of cluster centers;

[0094] S22, setting a fuzzy factor, which is used to adjust the degree of membership distribution of each sample data to different cluster centers during the clustering process. The fuzzy factor is greater than 1 to control the fuzziness of the clustering result;

[0095] S23, based on the constructed soil data input matrix, extracting the eigenvector of each sample as input data of the fuzzy C-means clustering model, and establishing the fuzzy C-means clustering model input layer;

[0096] S24. Initializing the initial positions of multiple cluster centers in the soil feature space according to the set number of cluster categories, each cluster center corresponds to a representative feature configuration of a soil type, which serves as the basis for clustering targets of the fuzzy C-means clustering model;

[0097] S25. Initialize a sample membership matrix, where the membership matrix is used to store the degree of membership of each sample to all cluster centers. The membership value is a real number, and the sum of all memberships of each sample is 1.

[0098] S26. According to the set number of cluster categories, fuzzy factors, initial cluster center positions and sample membership matrix, the basic structure of the fuzzy C-means clustering model is constructed, including the input layer of the model, the cluster center parameter set and the membership calculation mechanism.

[0099] The steps of initializing and constructing the fuzzy C-means clustering model in the present invention, through the systematic setting of key parameters such as the number of cluster categories, fuzzy factors, sample input features, initial cluster centers and membership matrix, effectively improve the scientific modeling and clustering accuracy of the clustering model in the soil type identification process. The reasonable setting of the number of cluster categories ensures that the model has sufficient expression ability and can cover the diverse distribution of soil types; the introduction of fuzzy factors enhances the model's flexible expression ability for boundary samples and avoids the misjudgment problem caused by traditional hard classification; by initializing the sample membership matrix and cluster center position, a reasonable clustering starting point is constructed, which improves the stability and convergence speed of the model in the iterative solution process. The model initialization process ensures that the subsequent clustering process has stronger robustness and adaptability in the complex, high-dimensional soil feature space, can accurately identify the fuzzy transition characteristics of soil types in farmland, and lays a solid data foundation and model structure guarantee for subsequent intelligent decision-making, strategy matching and dynamic optimization, and has wide application value.

[0100] In this embodiment, S3 specifically includes:

[0101] S31, initializing the population of the hermit crab optimization algorithm, dividing the population into a main population and multiple sub-populations, wherein the sub-populations include a boundary exploration population, a local refinement population, and an elite memory population, which are used to perform global search, local search, and optimal solution retention respectively;

[0102] S32. Randomly assign a set of key parameters of the fuzzy C-means clustering model to each hermit crab individual in the main population to form a parameter group:

[0103] [C i ,m i ,V i ,μ i ];

[0104] Among them, C i is the number of cluster categories, m i is the fuzzy factor, V i is the cluster center set, μ i is the sample membership matrix;

[0105] S33, construct the corresponding fuzzy C-means clustering model based on each parameter group, and calculate the clustering error objective function value J i :

[0106]

[0107] Among them, x n is the nth soil sample, v j is the jth cluster center, μ nj Indicates the membership value of the nth sample belonging to the jth cluster center, C i represents the number of cluster categories set by the i-th individual, and N represents the total number of soil samples;

[0108] S34. Calculate the comprehensive fitness function value F of all candidate individuals in the main group i , including the initial generated individuals, sub-population migration individuals and new individuals generated by shell fusion:

[0109]

[0110] Among them, DBI s is the Davies-Bouldin index, S i is the silhouette coefficient, α and β are weighting factors;

[0111] S35. Deploy several scout crabs in the boundary area of the solution space. Each scout crab sample multiple candidate points in the boundary area, evaluate the sampled points based on the fitness function, and select the best candidate point as the boundary optimal point x. s and feed back the optimal point of the boundary to the main group;

[0112] S36. Execute the search strategy of each sub-group: the boundary exploration group performs large perturbation updates on the parameters, the local refinement group performs micro-perturbation updates on the neighborhood of the current optimal solution, and the elite memory group saves the optimal solutions of previous generations and participates in rotation and reorganization;

[0113] S37. In the main population, select any two individuals A and B and calculate the complementary synergy Comp(A,B):

[0114] Comp(A,B)=γ1·‖x A -x B ‖+γ2·|F A -F B |+γ3·D local ;

[0115] Among them, x A 、x B is the parameter vector, F A is the comprehensive fitness function value of individual A, F Bis the comprehensive fitness function value of individual B, γ1 is the weighted coefficient of the solution vector difference term, γ2 is the weighted coefficient of the fitness difference term, γ3 is the weighted coefficient of the local density term, D local is the local solution density. When Comp(A,B)>θ, shell fusion is performed to generate a new individual, where θ represents the triggering threshold of shell fusion:

[0116] x new =λ o ·x A +(1-λ)·x B ;

[0117] Among them, x new represents the solution vector of the new individual generated after shell fusion, λ o is the fusion coefficient;

[0118] S38. For all candidate individuals, including the new individuals generated after shell fusion, combine the boundary optimal point x of the scout crab feedback s , perform a guided location update:

[0119]

[0120] in, represents the new position of the i-th candidate individual after the t+1-th iteration, is the position of the i-th candidate individual in the current main population at the t-th iteration, δ1, δ2 and δ3 are update coefficients, Represents the sample membership matrix corresponding to the individual The changing trend of is the nonlinear interaction function between individuals and population centers, is the average value of all individual parameter vectors of the current main group;

[0121] S39, recalculating the fitness function values of all individuals whose positions have been updated, performing individual replacement, retention, or migration operations on the main group and each sub-group according to the fitness changes, and updating the global optimal solution;

[0122] S310, when the number of iterations reaches the set maximum number of iterations T max Finally, the parameter combination with the highest fitness is output [C * ,m * ,V * ,μ * ] and construct the optimized fuzzy C-means clustering model.

[0123] The present invention introduces an improved hermit crab optimization algorithm to collaboratively optimize the key parameters of the fuzzy C-means clustering model, significantly improving the adaptability and clustering accuracy of the clustering model in complex soil feature spaces. The algorithm innovatively constructs a structure for the co-evolution of a main population and multiple subpopulations, with the subpopulations responsible for boundary exploration, local refinement, and elite retention, respectively, thereby achieving a dynamic balance between global search and local convergence capabilities. By defining a comprehensive fitness function and integrating clustering quality indicators such as the Davies-Bouldin index and silhouette coefficient, the clustering effect of individual solutions is comprehensively evaluated. A scout crab mechanism is further introduced to deploy feedback points at the boundary of the solution space, effectively preventing the algorithm from falling into local optimality. A complementary synergy judgment mechanism is proposed, combining solution vector differences, fitness differences, and local density to dynamically trigger shell fusion operations, enhance population diversity, and improve population evolution quality. In the individual position update strategy, boundary optimal point guidance, individual historical trends, adaptive membership perturbations, and nonlinear interaction mechanisms are integrated to achieve fine control and rapid optimization of the parameter space. Finally, the iterative control strategy stably outputs the global optimal parameter group. The optimized clustering model has stronger stability and generalization ability, can more accurately characterize the fuzzy type distribution of farmland soil, and provide a high-quality modeling basis for strategy matching and personalized improvement.

[0124] In this embodiment, the S4 specifically includes:

[0125] S41. Input the constructed soil data into the matrix X = {x1, x2, ..., x N} is input into the optimized fuzzy C-means clustering model, x i represents the eigenvector of the i-th soil sample, i = 1, 2, ..., N, where N represents the total number of soil samples;

[0126] S42, using the parameter combination with the highest fitness [C * ,m * ,V * ,μ * ], for each soil sample x i Calculate for each cluster center v j The membership degree μ ij :

[0127]

[0128] Among them, μ ij Indicates the membership value of the i-th sample belonging to the j-th category, C * is the final number of cluster categories, m * is the fuzzy factor, v j ∈V * represents the j-th cluster center vector, ‖‖ represents the Euclidean distance;

[0129] S43, build dimension is N×C * The sample membership matrix μ={μ ij}, where each row represents the degree of membership of a soil sample in each cluster category;

[0130] S44. Based on the membership matrix μ, a fuzzy soil type distribution map is generated, where the fuzzy soil type distribution map is used to express the fuzzy belonging relationship distribution of different soil types for each plot in the farmland.

[0131] The present invention applies the optimized fuzzy C-means clustering model to a high-dimensional soil data input matrix to accurately calculate the membership distribution of each plot sample under each soil type, significantly improving the expressiveness and practical adaptability of soil type classification. Compared with traditional hard clustering or single-category classification methods, the present invention introduces the concept of fuzzy membership in the clustering process, which not only reflects the transitional and mixed characteristics between soil types, but also effectively avoids the problem of inaccurate classification of boundary samples. By constructing a membership matrix, the system comprehensively expresses the fuzzy degree of belonging of each sample under different cluster centers, enhancing the modeling ability of soil spatial continuity characteristics. Furthermore, the generated fuzzy soil type distribution map not only has high resolution and stability, but also can serve as the basis for subsequent improvement strategy allocation, crop planting structure optimization and field management decision-making. The map presents the multidimensional attribute fusion characteristics of soil classification in a data-driven manner, realizing dynamic expression from point data to plot level, effectively improving the accuracy and intelligence level of agricultural management, and has significant practical value and scalability.

[0132] In this embodiment, the S5 specifically includes:

[0133] S51. Classify the soil samples according to their cluster categories according to the fuzzy soil type distribution map to form multiple soil type groups;

[0134] S52. For each soil type, construct a preliminary improvement strategy space, wherein the preliminary improvement strategy space includes a combination of soil improvement measures, covering physical structure conditioning, nutrient regulation, organic matter supplementation, acid-base neutralization, and microbial activation;

[0135] S53. Introduce target crop type information and analyze crop-specific soil requirement parameters as constraints for the preliminary improvement strategy space;

[0136] S54. Call historical improvement record data to conduct correlation analysis on the operation combinations, implementation effects and crop responses of similar soil types in past improvement practices, for reference and optimization of the current strategy construction process;

[0137] S55. Consider the resources available within the current timeframe, including available materials, technical equipment, operational capacity, and budget, to identify the feasibility of implementing various improvement measures in actual farmland, thereby limiting infeasible or prohibitively costly strategic options.

[0138] S56. Taking into account the target crop type, historical improvement records, and resource availability constraints, the preliminary strategy plan space is filtered, screened, and reconstructed to form a set of candidate improvement strategies suitable for the current agricultural scenario.

[0139] The present invention constructs a space of improvement strategies for multiple types of soil based on a fuzzy soil type distribution map, systematically integrating target crop needs, historical improvement experience, and resource allocation, significantly improving the scientific nature, pertinence, and feasibility of soil improvement strategy construction. By grouping soil samples under different clustering categories, the basis for differences between soil types is clarified, and a multi-dimensional, combinable strategy space is constructed to effectively support local policy implementation and classified regulation. The introduction of crop-specific demand parameters allows strategy construction to focus not only on the physical and chemical characteristics of the soil itself, but also on biological coupling relationships such as crop root structure and nutrient absorption mechanism, thereby improving the matching degree of strategy to crop growth adaptability. At the same time, combined with historical improvement record data, successful experiences and failed lessons are discovered in a data-driven manner, achieving experience accumulation and knowledge transfer in strategy design, and reducing trial and error costs. Furthermore, various resource factors such as materials, equipment, labor, and budget are incorporated into the feasibility filtering logic to effectively avoid theoretical solutions that cannot be implemented, thereby enhancing the practicality and execution of the strategy. The final output set of candidate improvement strategies combines agricultural scientific basis with engineering reality constraints, forming a strategic foundation with reasonable structure, multiple trade-offs, and strong adaptability, supporting subsequent intelligent evaluation and precise deployment, and has significant application value and implementation potential.

[0140] In this embodiment, the specific soil requirement parameters of the crop include structural adaptability, fertility level, and water retention capacity, which are used to guide the adaptation and matching of the improvement strategy with the growth conditions of the target crop.

[0141] In this embodiment, S6 specifically includes:

[0142] S61, extracting the membership distribution information corresponding to each plot in the fuzzy soil type distribution map, and obtaining the membership weights of the plots under different soil type categories;

[0143] S62. Develop a multi-objective scoring standard for comprehensively evaluating the quality of candidate improvement strategies for each soil type.

[0144] S63. Evaluate each candidate soil improvement strategy set item by item based on the established multi-objective scoring criteria to generate corresponding strategy scoring results;

[0145] S64. Selecting the strategy with the best strategy score result from the strategy set for each soil type as the optimal improvement strategy for the soil type category;

[0146] S65. Combine the membership weights of each plot under different soil type categories, perform weighted integration on the selected optimal improvement strategies for each category, and generate an integrated personalized improvement strategy at the plot level.

[0147] By introducing a multi-objective scoring mechanism and a fuzzy membership weighting strategy, this paper achieves a precise decision-making process from soil type identification to personalized improvement strategy generation, significantly improving the scientific nature of soil improvement and its adaptability at the plot level. By extracting the membership weights of each plot in the fuzzy soil type distribution map, it comprehensively reflects the mixed characteristics of different plots with respect to multiple soil attributes, breaking through the traditional rigid "one plot, one type" classification approach and effectively addressing the continuity and boundary ambiguity of soil type distribution. In the candidate improvement strategy evaluation phase, a multi-objective scoring criteria centered on soil improvement effect, crop response potential, and implementation cost is constructed to ensure that the selected strategy combines ecological adaptability, economic feasibility, and production efficiency. On this basis, the optimal strategies for different soil types are weighted and integrated, and the membership of each plot is combined to achieve personalized strategy reconstruction for the plot, ensuring that the improvement plan obtained for each plot is highly consistent with its soil state characteristics and crop growth requirements. This method fully embodies the concept of "precision agriculture", ensuring the scientific nature of strategy allocation and improving the implementation effectiveness of actual deployment, providing strong decision-making support for intelligent agricultural management.

[0148] In this embodiment, the multi-objective scoring criteria specifically include soil improvement effectiveness, target crop response performance and implementation cost, which comprehensively measure the adaptability and superiority of candidate improvement strategies.

[0149] Example 1:

[0150] To verify the feasibility of the present invention in practice, it was applied to a certain farmland experimental plot. Due to long-term improper application of chemical fertilizers and excessive tillage, the soil generally suffers from acid-base imbalance, low organic matter content, and poor water retention capacity, resulting in a poor crop growth environment and average crop yields far below expected levels. To improve this situation, this embodiment adopts the comprehensive management method for soil improvement and soil quality optimization based on big data analysis proposed by the present invention. The method first collects a large amount of soil data from the experimental plot. Using a variety of sensors, the soil pH, organic matter content, moisture content, electrical conductivity, and other key physical and chemical indicators are monitored in real time. Crop planting information and historical improvement records are also collected. After data preprocessing, a unified input matrix is obtained. Based on this data, a fuzzy C-means clustering method is used to construct a soil type distribution map, dividing the farmland soil into multiple fuzzy categories, thereby reflecting the gradual and interlaced characteristics between plots. Subsequently, the present invention uses an improved hermit crab optimization algorithm to collaboratively optimize the key parameters of the clustering model, making the clustering results more precise and accurate, and accurately estimating the membership of each soil type on each plot.

[0151] On this basis, a set of candidate improvement strategies was constructed based on the soil type distribution map. This approach comprehensively considered the specific needs of crops for soil, such as structural adaptability, fertility level, and water retention capacity. Furthermore, it incorporated multiple factors, including historical improvement experience, resource availability, and actual input costs, to develop a multi-objective scoring standard for a comprehensive evaluation of candidate improvement strategies. Ultimately, the optimal improvement strategy was selected, and a personalized soil improvement plan was generated by combining the weighted fusion of the membership of each plot under different soil types. Experimental results showed that this integrated management approach can significantly improve the physical and chemical properties of soil, while promoting healthy crop growth and increasing yields.

[0152] Table 1 Comparison of key soil indicators and crop yields before and after farmland improvement

[0153]

[0154] As can be seen from Table 1, the method of the present invention has achieved remarkable results in improving soil quality and crop yield. First, in terms of soil acidity and alkalinity, the pH value increased from the original 5.2 to 6.8, an improvement of 30.8%, effectively alleviating the acidic soil problem. The soil environment tends to be neutral, which is more suitable for the growth of most crops. Secondly, the organic matter content increased from 1.9% to 3.2%, an increase of 68.4%, indicating that the application of organic improvement materials such as humus and organic fertilizers has greatly enhanced the soil fertility level, providing a more stable and continuous nutrient supply for crops.

[0155] In terms of water retention capacity, soil moisture content increased from 12% before improvement to 18%, a 50.0% increase. This indicates that optimizing tillage structure and applying biochar or mulch have effectively enhanced soil water retention and reduced the occurrence of soil drought stress. Electrical conductivity decreased from 300 μS / cm to 180 μS / cm, a 40.0% improvement. This reduction in conductivity indicates a reduced soil salt burden, a more stable electrolyte environment, and improved soil aeration and root growth conditions.

[0156] The most direct benefit is reflected in crop yields. Before implementation, the crop yield per unit area was 2,800 kg / hectare. After applying the method of the present invention, the yield increased to 4,200 kg / hectare, an increase of 50.0%. This data not only reflects the positive feedback of improved soil physical and chemical properties, but also fully verifies the scientific nature and adaptability of the improvement strategy. The present invention successfully achieves the differentiated allocation of improvement resources according to local conditions through the fuzzy clustering model, intelligent strategy scoring and optimization mechanism constructed through big data analysis, making farmland management more refined and efficient.

[0157] In summary, the data presented in this table fully demonstrates the comprehensive advantages of this invention in improving soil basic conditions, enhancing the farming environment, and promoting crop yield increases. It provides a highly practical, intelligent, and economical soil improvement and soil quality optimization management method for agricultural production. This method also has broad application value and practical significance in larger-scale and more complex environments.

[0158] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A comprehensive management method for soil improvement and soil quality optimization based on big data analysis, characterized in that: The steps include: S1. Collect multi-source soil data in the farmland area, pre-process the multi-source soil data, and construct a soil data input matrix; S2. Initialize the key parameters of the fuzzy C-means clustering model and build the basic structure of the fuzzy C-means clustering model; S3. Introducing the hermit crab optimization algorithm to collaboratively optimize the key parameters of the fuzzy C-means clustering model to obtain the optimized fuzzy C-means clustering model; S4, inputting the soil data input matrix into the optimized fuzzy C-means clustering model, performing clustering operation, outputting the membership degree of each plot corresponding to each cluster category, and generating a fuzzy soil type distribution map; S5. Based on the fuzzy soil type distribution map, construct an improvement strategy space for each type of soil, and generate a set of candidate improvement strategies based on the target crop type, historical improvement records, and resource availability; S6. Based on the membership distribution of each plot in the fuzzy soil type distribution map, a multi-objective scoring standard is formulated, and the optimal improvement strategy corresponding to each soil type is selected from the candidate improvement strategy set; S7. Deploy the optimal improvement strategy to the farmland operation system and collect soil status feedback data in real time during the execution process, including the time series change values of the target soil parameters and the crop growth response; S8. Based on the soil state feedback data, the input data and parameter distribution of the fuzzy C-means clustering model are updated, and the hermit crab optimization algorithm is re-executed to reconstruct the model to complete the closed-loop optimization.

2. The method for comprehensive management of soil improvement and soil quality optimization based on big data analysis according to claim 1, characterized in that: The multi-source soil data specifically includes soil physical and chemical properties, historical soil modification records, and spatial location data, which are used to construct a soil data input matrix.

3. The method for comprehensive management of soil improvement and soil quality optimization based on big data analysis according to claim 1, characterized in that: The multi-source soil data preprocessing specifically includes missing value filling, outlier removal, standardization and spatial information alignment, which are used to improve the quality and consistency of multi-source soil data.

4. The method for comprehensive management of soil improvement and soil quality optimization based on big data analysis according to claim 1, characterized in that: The S2 specifically includes: S21, setting the number of cluster categories to guide the fuzzy C-means clustering model to divide the soil data into multiple soil type categories, wherein the number of cluster categories determines the number of cluster centers; S22, setting a fuzzy factor, which is used to adjust the degree of membership distribution of each sample data to different cluster centers during the clustering process. The fuzzy factor is greater than 1 to control the fuzziness of the clustering result; S23, based on the constructed soil data input matrix, extracting the eigenvector of each sample as input data of the fuzzy C-means clustering model, and establishing the fuzzy C-means clustering model input layer; S24. Initializing the initial positions of multiple cluster centers in the soil feature space according to the set number of cluster categories, each cluster center corresponds to a representative feature configuration of a soil type, which serves as the basis for clustering targets of the fuzzy C-means clustering model; S25. Initialize a sample membership matrix, where the membership matrix is used to store the degree of membership of each sample to all cluster centers. The membership value is a real number, and the sum of all memberships of each sample is 1. S26. According to the set number of cluster categories, fuzzy factors, initial cluster center positions and sample membership matrix, the basic structure of the fuzzy C-means clustering model is constructed, including the input layer of the model, the cluster center parameter set and the membership calculation mechanism.

5. The method for comprehensive management of soil improvement and soil quality optimization based on big data analysis according to claim 1, characterized in that: The S3 specifically includes: S31, initializing the population of the hermit crab optimization algorithm, dividing the population into a main population and multiple sub-populations, wherein the sub-populations include a boundary exploration population, a local refinement population, and an elite memory population, which are used to perform global search, local search, and optimal solution retention respectively; S32. Randomly assign a set of key parameters of the fuzzy C-means clustering model to each hermit crab individual in the main population to form a parameter group: [C i ,m i ,V i ,μ i ]; Among them, C i is the number of cluster categories, m i is the fuzzy factor, V i is the cluster center set, μ i is the sample membership matrix; S33, construct the corresponding fuzzy C-means clustering model based on each parameter group, and calculate the clustering error objective function value J i : Among them, x n is the nth soil sample, v j is the jth cluster center, μ nj Indicates the membership value of the nth sample belonging to the jth cluster center, C i represents the number of cluster categories set by the i-th individual, and N represents the total number of soil samples; S34. Calculate the comprehensive fitness function value F of all candidate individuals in the main group i , including the initial generated individuals, sub-population migration individuals and new individuals generated by shell fusion: Among them, DBI s is the Davies-Bouldin index, S i is the silhouette coefficient, α and β are weighting factors; S35. Deploy several scout crabs in the boundary area of the solution space. Each scout crab samples multiple candidate points in the boundary area, evaluates the sampled points based on the fitness function, and selects the best candidate point as the boundary optimal point x. s and feed back the optimal point of the boundary to the main group; S36. Execute the search strategy of each sub-group: the boundary exploration group performs large perturbation updates on the parameters, the local refinement group performs small perturbation updates on the neighborhood of the current optimal solution, and the elite memory group saves the optimal solutions of previous generations and participates in rotation and reorganization; S37. In the main population, select any two individuals A and B and calculate the complementary synergy Comp(A,B): Comp(A,B)=γ1·‖x A -x B ‖+γ2·|F A -F B |+γ3·D local ; Among them, x A 、x B is the parameter vector, F A is the comprehensive fitness function value of individual A, F B is the comprehensive fitness function value of individual B, γ1 is the weighted coefficient of the solution vector difference term, γ2 is the weighted coefficient of the fitness difference term, γ3 is the weighted coefficient of the local density term, D local is the local solution density. When Comp(A,B)>θ, shell fusion is performed to generate a new individual, where θ represents the triggering threshold of shell fusion: x new =λ o ·x A +(1-λ)·x B ; Among them, x new represents the solution vector of the new individual generated after shell fusion, λ o is the fusion coefficient; S38. For all candidate individuals, including the new individuals generated after shell fusion, combine the boundary optimal point x of the scout crab feedback s , perform a guided location update: in, represents the new position of the i-th candidate individual after the t+1-th iteration, is the position of the i-th candidate individual in the current main population at the t-th iteration, δ1, δ2 and δ3 are update coefficients, Represents the sample membership matrix corresponding to the individual The changing trend of is the nonlinear interaction function between individuals and population centers, is the average value of all individual parameter vectors of the current main group; S39, recalculating the fitness function values of all individuals whose positions have been updated, performing individual replacement, retention, or migration operations on the main group and each sub-group according to the fitness changes, and updating the global optimal solution; S310, when the number of iterations reaches the set maximum number of iterations T max Finally, the parameter combination with the highest fitness is output [C * ,m * ,V * ,μ * ] and construct the optimized fuzzy C-means clustering model.

6. The method for comprehensive management of soil improvement and soil quality optimization based on big data analysis according to claim 1, characterized in that: The S4 specifically includes: S41. Input the constructed soil data into the matrix X = {x1, x2, ..., x N } is input into the optimized fuzzy C-means clustering model, x i represents the eigenvector of the i-th soil sample, i = 1, 2, …, N, where N represents the total number of soil samples; S42, using the parameter combination with the highest fitness [C * ,m * ,V * ,μ * ], for each soil sample x i Calculate for each cluster center v j The membership degree μ ij : Among them, μ ij Indicates the membership value of the i-th sample belonging to the j-th category, C * is the final number of cluster categories, m * is the fuzzy factor, v j ∈V * represents the j-th cluster center vector, ‖‖ represents the Euclidean distance; S43, build dimension is N×C * The sample membership matrix μ={μ ij }, where each row represents the degree of membership of a soil sample in each cluster category; S44. Based on the membership matrix μ, a fuzzy soil type distribution map is generated, where the fuzzy soil type distribution map is used to express the fuzzy belonging relationship distribution of different soil types for each plot in the farmland.

7. The method for comprehensive management of soil improvement and soil quality optimization based on big data analysis according to claim 1, characterized in that: The S5 specifically includes: S51. Classify the soil samples according to their cluster categories according to the fuzzy soil type distribution map to form multiple soil type groups; S52. For each soil type, construct a preliminary improvement strategy space, wherein the preliminary improvement strategy space includes a combination of soil improvement measures, covering physical structure conditioning, nutrient regulation, organic matter supplementation, acid-base neutralization, and microbial activation; S53. Introduce target crop type information and analyze crop-specific soil requirement parameters as constraints for the preliminary improvement strategy space; S54. Call historical improvement record data to conduct correlation analysis on the operation combinations, implementation effects and crop responses of similar soil types in past improvement practices, for reference and optimization of the current strategy construction process; S55. Consider the resources available within the current timeframe, including available materials, technical equipment, operational capacity, and budget, to identify the feasibility of implementing various improvement measures in actual farmland, thereby limiting infeasible or prohibitively expensive strategic options. S56. Taking into account the target crop type, historical improvement records, and resource availability constraints, the preliminary strategy plan space is filtered, screened, and reconstructed to form a set of candidate improvement strategies suitable for the current agricultural scenario.

8. The method for comprehensive management of soil improvement and soil quality optimization based on big data analysis according to claim 7, characterized in that: The specific soil requirement parameters of the crops include structural adaptability, fertility level, and water retention capacity, which are used to guide the adaptation and matching of improvement strategies with the growth conditions of the target crops.

9. The method for comprehensive management of soil improvement and soil quality optimization based on big data analysis according to claim 1, characterized in that: The S6 specifically includes: S61, extracting the membership distribution information corresponding to each plot in the fuzzy soil type distribution map, and obtaining the membership weights of the plots under different soil type categories; S62. Develop a multi-objective scoring standard for comprehensively evaluating the quality of candidate improvement strategies for each soil type. S63. Evaluate each candidate soil improvement strategy set item by item based on the established multi-objective scoring criteria to generate corresponding strategy scoring results; S64. Selecting the strategy with the best strategy score result from the strategy set for each soil type as the optimal improvement strategy for the soil type category; S65. Combine the membership weights of each plot under different soil type categories, perform weighted integration on the selected optimal improvement strategies for each category, and generate an integrated personalized improvement strategy at the plot level.

10. The method for comprehensive management of soil improvement and soil quality optimization based on big data analysis according to claim 9, characterized in that: The multi-objective scoring criteria specifically include soil improvement effectiveness, target crop response performance and implementation cost, which comprehensively measure the adaptability and superiority of candidate improvement strategies.

Images