Intelligent application method and system of biological fertilizer for rapidly fertilizing black soil

Abstract

The invention discloses an intelligent application method and system of a biological fertilizer for rapidly fertilizing black soil, and relates to the technical field of intelligent fertilization. The method comprises the following steps: by taking lipid molecules, a black soil environment, functional microorganisms and a target plant as main body nodes, analyzing an association relationship, fusing target black soil characteristics, and constructing a fertility improvement graph structure; carrying out planting full-period analysis on the target black soil, constructing a period planting chain, and configuring deviation parameters according to the soil demand relationship of the planting period; according to the fertilization graph structure and the deviation parameters, fertilization main body difference mapping is carried out on the periodic planting chain, and node fertilization difference is obtained; and by taking the node fertility difference as an application compensation target, performing target compensation analysis according to an application time sequence variable relationship, obtaining a compensation application strategy with optimal input minimum compensation, and generating application control information. The technical problem that in the prior art, fertilization is not accurate, and consequently the black soil fertilization period is long is solved, and the technical effects of shortening the fertilization period and improving the fertilization accuracy are achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent fertilization technology, specifically to an intelligent application method and system for bio-fertilizer that rapidly enriches black soil. Background Technology

[0002] Black soil, as an important agricultural soil resource, significantly impacts crop yield and quality due to its fertility level. However, with the continuous increase in agricultural planting intensity and the long-term use of chemical fertilizers in a singular and extensive manner, black soil degradation has become increasingly prominent, manifested in decreased organic matter content and weakened nutrient supply capacity. Current fertilization methods largely rely on manual experience or rough planting patterns, lacking precise perception and adjustment of soil conditions and crop growth cycles. This leads to significant haphazardness in the fertilization process, resulting in low fertilizer efficiency, high input costs, and long soil recovery cycles. Particularly in the process of black soil fertilization, the mismatch between application timing and actual needs often hinders effective fertilization, limiting the sustainable utilization of black soil and green agricultural development. Summary of the Invention

[0003] This application provides an intelligent application method and system for bio-fertilizer to rapidly enrich black soil, which solves the technical problem of inaccurate fertilization in the prior art, resulting in a long black soil enrichment cycle.

[0004] The first aspect of this application provides an intelligent method for rapidly enriching black soil with bio-fertilizer, the method comprising:

[0005] Lipid molecules, black soil environment, functional microorganisms, and target plants are used as the main nodes. The relationships between these nodes are analyzed, and the characteristics of the target black soil are integrated to construct a fertility improvement diagram structure, including positive and negative relationships. The entire planting cycle of the target black soil is analyzed to construct a cyclical planting chain, and deviation parameters are configured according to the soil demand relationship of the planting cycle. Based on the fertility improvement diagram structure and the deviation parameters, the cyclical planting chain is mapped to the fertility improvement subject difference to obtain the node fertility improvement difference. Using the node fertility improvement difference as the application compensation target, target compensation analysis is performed based on the application time-series variable relationship to obtain the compensation application strategy with minimum input and optimal compensation, and application control information is generated.

[0006] A second aspect of this application provides an intelligent application system for bio-fertilizer that rapidly enriches black soil, the system comprising:

[0007] The graph construction module uses lipid molecules, black soil environment, functional microorganisms, and target plants as main nodes, analyzes the relationships between these nodes, integrates the characteristics of the target black soil, and constructs a fertility graph structure, including positive and negative relationships. The analysis module performs a full-cycle analysis of the target black soil, constructs a cyclical planting chain, and configures deviation parameters based on the soil demand relationships of the planting cycle. The mapping module maps the fertility differences of the main nodes in the cyclical planting chain based on the fertility graph structure and the deviation parameters, obtaining the node fertility differences. The compensation analysis module uses the node fertility differences as the application compensation target, performs target compensation analysis based on the application time-series variables, obtains the optimal compensation application strategy with minimum input and compensation, and generates application control information.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] First, lipid molecules, black soil environment, functional microorganisms, and target plants are used as the main nodes. The relationships between these nodes are analyzed, and the characteristics of the target black soil are integrated to construct a fertility improvement diagram structure, including positive and negative relationships. Next, the entire planting cycle of the target black soil is analyzed, a cyclical planting chain is constructed, and deviation parameters are configured based on the soil demand relationships during the planting cycle. Then, based on the fertility improvement diagram structure and the deviation parameters, the cyclical planting chain is mapped to the fertility improvement differences of the main nodes, obtaining the node fertility improvement differences. Finally, using the node fertility improvement differences as the application compensation target, target compensation analysis is performed based on the application time-series variables to obtain the optimal compensation application strategy with minimum input and generate application control information. This solves the technical problem of inaccurate fertilization in existing technologies, which leads to long black soil fertilization cycles, achieving the technical effects of shortening the fertilization cycle and improving fertilization accuracy. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A schematic diagram of a smart application method for rapidly enriching black soil with bio-fertilizer provided in an embodiment of this application;

[0012] Figure 2 This is a schematic diagram of an intelligent application system for bio-fertilizer that rapidly enriches black soil, provided in an embodiment of this application.

[0013] Figure labeling: Graph construction module 11, parsing module 12, mapping module 13, compensation analysis module 14. Detailed Implementation

[0014] This application provides an intelligent application method and system for bio-fertilizer to rapidly enrich black soil, which solves the technical problem of inaccurate fertilization in the prior art, resulting in a long black soil enrichment cycle.

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0016] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0017] Example 1, as Figure 1 As shown, this application provides an intelligent application method for bio-fertilizer to rapidly enrich black soil, wherein the method includes:

[0018] Lipid molecules, black soil environment, functional microorganisms and target plants are used as the main nodes. The relationships between the main nodes are analyzed, and the characteristics of the target black soil are integrated to construct a fertilization diagram structure, including positive and negative relationships.

[0019] In this embodiment, lipid molecules, black soil environment, functional microorganisms, and target plants are used as the main nodes in the graph structure. Specifically, based on the physicochemical properties, biological attributes, and mechanisms of action of each main node in the black soil fertilization process, key feature indicators are extracted, and each main node is structurally modeled. Then, the relationships between the main nodes are analyzed, including but not limited to the metabolic regulation relationship between lipid molecules and functional microorganisms, the interaction relationship between functional microorganisms and the black soil environment, and the influence of target plant root exudates on lipid molecules and microbial community structure. Based on the above relationships, specific soil characteristic data of the target black soil area (such as organic matter content, aggregate structure, enzyme activity level, etc.) are integrated to construct a graph structure network (fertilization graph structure) reflecting the dynamic interaction between the main entities. The type and direction of the edges in the graph are defined to represent positive promoting relationships (such as lipid molecules enhancing microbial activity) and negative inhibiting relationships (such as microbial competition leading to a decrease in soil enzyme activity).

[0020] Furthermore, by using lipid molecules, black soil environment, functional microorganisms, and target plants as key nodes, the relationships between these key nodes are analyzed, including:

[0021] Each subject node is assigned a subset of subjects based on its type characteristics. Then, based on a graph attention network, weights are coupled between the subject subsets to generate a multi-scale edge weight matrix, which quantifies the contribution intensity of positive and negative relationships.

[0022] In constructing the fertilization graph structure, the types of features are categorized based on the intrinsic attributes of each principal node, and principal subsets are constructed accordingly. For example, lipid molecules are divided into different subsets based on their carbon chain length, polar functional group configuration, and lipid solubility; the black soil environment is established as an environmental state subset based on parameters such as aggregate structure, organic matter content, and water retention capacity; functional microorganisms are constructed as a microbial subset based on their metabolic functions, community composition, and functional gene expression levels; and target plants are divided into a plant subset based on root system configuration, secretion spectrum characteristics, and nutritional requirements at different growth stages. Based on this, a graph attention network is used to model the interactions between different principal subsets, dynamically assigning weights to each edge in the graph through an attention mechanism.

[0023] Specifically, a cross-subset graph structure is constructed, and edge weights are modeled at multiple scales: within the same subject subset, a first-scale graph is built to capture micro-interactions between similar elements; between different subject subsets, a second-scale graph is constructed to capture coupling effects across types. Attention weights are calculated in each scale graph to obtain the contribution intensity of the corresponding edges, with positive edge weights and negative edge weights. Finally, the edge weight matrices from all scales are merged to form a multi-scale edge weight matrix, which serves as the basis for graph connections when constructing the fertilization graph structure, thereby achieving quantitative modeling and structural representation of the complex relationships between different subjects in the black soil fertilization system.

[0024] Furthermore, for each subject node, a subject subset is constructed according to the subject's type characteristics, including:

[0025] For the main nodes of lipid molecules, subsets of lipid type, carbon chain length, and functional group abundance are constructed; for the main nodes of black soil environment, subsets of organic matter content, aggregate structure, and enzyme activity level are constructed; for the main nodes of functional microorganisms, subsets of bacterial community, fungal community, and functional gene are constructed; and for the main nodes of target plants, subsets of root system configuration, root secretion spectrum, and nutrient requirements are constructed.

[0026] For the main nodes of lipid molecules, lipid type subsets (such as fatty acids, phospholipids, sterols, etc.) are constructed according to differences in molecular structure and function. They are further subdivided into carbon chain length subsets according to their carbon chain length (such as short chain, medium chain, long chain). At the same time, functional group abundance subsets are constructed based on the statistical analysis of functional group types and quantities to characterize their biological activity potential and microbial response characteristics.

[0027] For the main nodes of the black soil environment, based on soil analysis data, they are divided into subsets of organic matter content (such as high, medium and low content gradient distribution), subsets of aggregate structure (such as the proportion characteristics of large aggregates, medium aggregates and small aggregates), and subsets of enzyme activity levels (such as soil enzyme indicators such as urease, peroxidase, sucrase, etc.) to reflect soil quality and biological reactivity.

[0028] For the functional microbial main nodes, bacterial community subsets (such as nitrogen-fixing bacteria and phosphate-solubilizing bacteria), fungal community subsets (such as arbuscular mycorrhizal fungi and white-rot fungi), and functional gene subsets (such as the expression levels of key metabolic genes for nitrogen fixation, phosphate solubilization, and stress resistance) are constructed according to microbial taxonomy and functional characteristics, so as to model their role mechanism in the black soil improvement process.

[0029] For the main nodes of the target plant, a subset of root system configurations (such as fibrous root type, taproot type, complex type, etc.) is constructed based on the morphological characteristics of the plant root system. A subset of root exudate profiles (such as the abundance of exudates such as organic acids, phenols, alcohols, etc.) is established based on the composition and metabolic profile of root exudates. At the same time, a subset of nutrient requirements (such as the time-concentration requirement function distribution of nitrogen, phosphorus, potassium and trace elements) is established based on the nutrient element absorption characteristics of the plant at different growth stages.

[0030] Furthermore, by integrating the characteristics of the target black soil, a soil improvement map structure is constructed, including:

[0031] Collect multi-source heterogeneous data of the target black soil, including at least hyperspectral images, temporal water tension, redox potential, and microbial metagenomics; perform spatiotemporal alignment and missing filling on the multi-source heterogeneous data to obtain a unified spatiotemporal feature tensor; use a Transformer encoder to extract global and local features from the feature tensor and couple them, and use a graph convolutional decoder to map them to the node attributes and edge attributes of a graph attention network structure to obtain the fertilization graph structure.

[0032] Specifically, firstly, multi-source heterogeneous data of the target black soil region are collected, including but not limited to hyperspectral imagery (for extracting information such as surface vegetation index and spatial distribution of organic matter), temporal water tension (reflecting soil water-holding capacity and wet-dryness characteristics), redox potential (characterizing the state of electron acceptors and the microbial activity environment in the soil), and microbial metagenomic data (for describing soil microbial community composition and functional gene abundance). Next, the multi-source heterogeneous data are preprocessed, including temporal interpolation, spatial resampling, missing value imputation, and noise correction, thereby constructing a fused feature tensor on a unified coordinate system and time scale. Then, a Transformer encoder is used to extract deep features from this multi-dimensional spatiotemporal feature tensor, capturing global dependencies and local contextual coupling characteristics between features through its self-attention mechanism, achieving fusion encoding of features at different modalities and scales. Finally, the encoding results are input into a graph convolutional decoder, further decoding them into node attributes and edge attributes through the topological information in the graph structure, thus completing the mapping process from multi-source raw data to a fertilization graph structure. In the fertilization graph structure, each node corresponds to a subject subset (such as a subset of lipid molecules, a subset of functional microorganisms, etc.), and the edges between nodes carry weight information reflecting their positive or negative correlations, which can be dynamically updated as the state of the black soil environment evolves.

[0033] Furthermore, based on graph attention networks, weights are coupled to each subject subset to generate a multi-scale edge weight matrix, which quantifies the contribution intensity of positive and negative relationships, including:

[0034] A first scale is constructed within the same principal node, and a second scale is constructed between different principal nodes to form a multi-scale hierarchical graph. A multi-head attention mechanism is applied to any first scale to obtain a first-scale edge weight matrix. For any second scale, node-level cross-scale projection is calculated to obtain a second-scale weight matrix. The first-scale edge weight matrix and the second-scale weight matrix are fused to obtain a multi-scale edge weight matrix, where positive values ​​represent positive relationships and their strengths, and negative values ​​represent negative relationships and their strengths. The multi-scale edge weight matrix is ​​injected into the graph attention network as adjacency weights, so that each edge simultaneously carries a relationship type label, a multi-scale coupling weight value, and a scale source marker.

[0035] Based on the source category and structural hierarchy of the main nodes, two scales of the graph structure are constructed: the first scale is to establish internal edges between main nodes of the same type (e.g., nodes of various species within a functional microbial subset) to express synergistic or inhibitory relationships between similar factors; the second scale is to establish cross-category edges between main nodes of different types (e.g., connecting specific lipid molecule nodes with specific plant root nodes) to reflect the functional linkage or antagonistic effects between heterogeneous factors, forming a multi-scale hierarchical graph structure with nested upper and lower layers.

[0036] For edges within the first scale, a multi-head attention mechanism is applied to calculate the attention scores of adjacent nodes from different perspectives. The attention weights of multiple heads are combined to form the first-scale edge weight matrix, which expresses the strength and direction of the association between different nodes in the same category. For cross-class edges in the second scale, a node-level cross-scale projection mechanism is designed. By projecting the representation of a node in its own category feature space to the heterogeneous category feature space, and calculating the semantic matching degree after projection, the second-scale weight matrix is ​​obtained, which is used to characterize the coupling association strength between different categories of subjects.

[0037] The first-scale edge weight matrix and the second-scale weight matrix are merged according to the node connection relationship to construct a unified multi-scale edge weight matrix. In the matrix, positive values ​​indicate that the relationship represented by the edge has a positive effect (such as promoting absorption, co-activation, etc.), and the magnitude of the value represents the strength of the positive effect; negative values ​​indicate that the edge has a negative effect (such as resource competition, metabolic inhibition, etc.), and the absolute value of the value represents the strength of the negative effect.

[0038] Injecting the multi-scale edge weight matrix as adjacency weight information into the graph attention network allows each edge in the network to simultaneously carry: a relation type label (positive or negative); a multi-scale coupling weight value (a floating-point real number used to participate in attention updates); and a scale source label (indicating whether the edge belongs to the scale within the same type of node or the scale between cross-type nodes).

[0039] A full-cycle analysis of the target black soil was conducted, a cyclical planting chain was constructed, and deviation parameters were configured based on the soil demand relationship of the planting cycle.

[0040] By analyzing the entire planting cycle of the target black soil and combining the planting time and growth stage of the target plants, a cyclical planting chain with multiple time nodes is constructed. Based on the growth requirements of plants at each time node, the changes in soil nutrients and environmental conditions are analyzed, and corresponding deviation parameters are configured to reflect the differences and changing trends of soil requirements at different stages.

[0041] Furthermore, a full-cycle analysis of the target black soil planting process is conducted to construct a cyclical planting chain, and deviation parameters are configured based on the soil demand relationship during the planting cycle, including:

[0042] The target plants for the entire planting cycle of the target black soil are obtained. The cyclical planting chain is constructed according to the planting time and growth cycle of the target plants, which includes multiple time nodes. Each time node has a plant name and its growth cycle label. Soil requirements are analyzed for the plant names and growth cycles of each time node in the cyclical planting chain. The difference is calculated based on the soil requirements of the nodes to obtain the deviation parameters of each time node, including the deviation requirement tensor and the difference in requirement value.

[0043] Specifically, historical and current crop data for the target black soil region are collected to identify the target plant species throughout the entire planting cycle. Based on information such as planting time, growth cycle, and crop rotation logic of different plants, a cyclical planting chain is constructed. This chain is divided into multiple time nodes, with each node associated with a specific plant name and its growth cycle label, representing the current agricultural growth stage (e.g., sowing, tillering, heading). For each time node, based on the plant name and growth cycle, and combined with crop nutrient requirements and black soil fertility characteristics, node-level soil requirement analysis is performed, extracting key requirement parameters such as nutrient type and quantity, organic matter level, moisture status, aggregate structure, and microbial activity required by the crop at that stage. A difference analysis is conducted between the actual black soil environmental indicators and the standard requirement parameters at the current node. Using tensor modeling, a deviation requirement tensor with a multi-dimensional indicator structure is constructed, outputting the difference in requirement values ​​for each dimension to quantify the degree of deviation between the soil condition and the target requirements at that stage.

[0044] Furthermore, the deviation parameters at each time point are obtained, including:

[0045] For each time node, a ternary demand tensor of soil, crop, and meteorology is established, in which soil, crop, and meteorology each have multi-dimensional index quantification rules; based on sample data, demand analysis is performed at each time node according to plant name, planting time, and growth cycle to obtain soil demand information of plants at each time node, and the ternary demand tensor is projected and transformed; the demand tensors are subtracted according to the continuous relationship of time nodes to obtain the deviation demand tensor and the demand value difference.

[0046] For each time point in the cyclical planting chain, a ternary demand tensor of soil, crop, and meteorology is constructed. This tensor uses soil, crop, and meteorology as its three-dimensional axes, with each dimension quantified using multidimensional indicators: the soil dimension includes parameters such as organic matter content, available nutrient concentration, water tension, and enzyme activity levels; the crop dimension includes factors such as plant variety, growth stage, root characteristics, and nutrient requirement intensity; and the meteorological dimension includes environmental conditions such as temperature, humidity, rainfall, and sunshine duration. Using a sample dataset from the target region, time-node matching analysis is performed on the ternary demand tensor according to labels such as plant name, planting time, and growth cycle to obtain soil demand information for crops at each time point. Soil demand information is mapped to the observed multi-source index space of black soil through tensor projection transformation, thereby constructing a ternary tensor of the actual state at each time node. By comparing the changing trends of the demand tensor between different time nodes, the ternary tensors of adjacent time nodes are differentially calculated according to the continuity of time to obtain the deviation demand tensor of each time period, and from this, the specific demand value difference is derived, which is used to quantify the direction and magnitude of soil compensation or adjustment between the current node and the next node.

[0047] Based on the fertilization diagram structure and the deviation parameters, the fertilization difference of the cyclic planting chain is mapped to obtain the fertilization difference of each node.

[0048] Based on the fertilization graph structure and deviation parameters, the fertilization difference of the main body is mapped to the cyclic planting chain. By using the deviation parameters as a perturbation signal to act on the edge weights of the graph structure, time coupling deduction is initiated, and the adjustment direction and adjustment amount of each main node at different time nodes are iteratively calculated. Finally, the node fertilization difference reflecting the fertilization compensation required by each node is obtained.

[0049] Furthermore, based on the fertilization diagram structure and the deviation parameters, the fertilization difference of the cyclic planting chain is mapped to obtain the node fertilization difference, including:

[0050] The deviation parameter is used as a perturbation signal and superimposed on the edge weights of the fattening graph structure, causing the fattening graph structure to change dynamically over time. Time-coupled extrapolation is initiated on the dynamic fattening graph structure. Each node collects the demand gap and available adjustment amount from its neighboring nodes, and then iteratively calculates the adjustment direction and adjustment amount of its own node in combination with its own attributes to meet the adjustment requirements of the perturbation signal. This process continues until the deviation requirements of each time node are met. At the end of the iteration, the fattening difference of each main node is obtained as the node fattening difference.

[0051] By using the deviation parameter as a perturbation signal and superimposing it on the edge weights of the fertility graph structure, the fertility graph structure can dynamically change over time, reflecting the changes in soil demand at different time points and the mutual influence between the main nodes.

[0052] A time-coupled extrapolation process is initiated on a dynamically changing fertilization graph structure. Specifically, each main node collects demand gap information and available adjustment amounts from its neighboring nodes, and calculates the adjustment direction and magnitude to satisfy the disturbance signal by combining its own attributes and state. Through multiple iterations, the nodes continuously adjust their own fertilization amount to respond to soil demand deviations until the deviation demand at each time node is met or the preset iteration termination condition is reached. Finally, at the end of the iteration process, each main node obtains the corresponding fertilization difference, i.e., the node fertilization difference, which reflects the fertilization intensity and direction that the node needs to compensate for in the current cycle.

[0053] Using the difference in fertilization at the nodes as the application compensation target, target compensation analysis is performed based on the relationship between application time-series variables to obtain the optimal compensation application strategy with minimum input and generate application control information.

[0054] Based on the time-series decomposition of the fertilization difference at each node, the fertilizer release curve, and the crop growth requirement curve, a compensation time-series chain is formed. Each link in the chain is marked with the start, peak, and depletion times of the fertilizer effect, ensuring a precise match between the fertilization effect and the peak stage of crop demand. Simultaneously, combining the dynamic evolution patterns of the target black soil, a time-series chain of soil characteristic changes is constructed and aligned with the compensation time-series chain, ensuring the overlap between the peak fertilizer effect and the soil nutrient trough and the peak crop fertilization time. Based on the aligned dual time-series chains, using minimum input and optimal compensation effect as evaluation indicators, an optimization algorithm searches for the optimal application sequence, determining the best fertilization operation time, spatial location, and combination scheme of lipid molecules, microbial agents, and amendments in terms of formulation and dosage. Finally, application control information is generated to guide actual fertilization operations, achieving precise compensation, improving fertilization efficiency, and shortening the black soil fertilization cycle.

[0055] Furthermore, using the fertilization difference at the aforementioned nodes as the application compensation target, target compensation analysis is performed based on the relationship between application time-series variables to obtain the optimal compensation application strategy with minimum input, generating application control information, including:

[0056] Using the node fertilization difference as the application compensation target, the node fertilization difference is decomposed temporally according to the fertilizer release curve and crop demand curve to form a compensation time-series chain, where each link is marked with the fertilizer effect initiation, peak, and depletion time. Based on the target black soil characteristics and the cyclical planting chain, a soil characteristic change time-series chain is constructed according to the evolution law of the target black soil characteristics with the crop growth stage. Using the node fertilization difference as a shared anchor point, the compensation time-series chain is aligned with the soil characteristic change time-series chain, so that the fertilizer effect peak overlaps with the crop demand peak and the soil nutrient trough in time. With the minimum input and optimal compensation as the evaluation objective, the optimal application sequence is searched on the aligned double chain to obtain the application control information, including the application operation time, spatial location, lipid-microbial agent-amendant formula and dosage.

[0057] First, based on the nutrient gap information represented by the node fertilization difference, and combining the fertilizer release curves of inputs such as lipid molecules, microbial agents, and amendments with the growth requirement curves of the target crop, this difference is decomposed in the time dimension to obtain a compensation time-series chain. The compensation time-series chain covers key moments such as application initiation, peak fertilizer effect, and fertilizer depletion, clearly defining the time nodes and fertilization intensity of each link. On this basis, combining the environmental response characteristics of the target black soil and the crop growth stage information in the cyclical planting chain, a soil characteristic change time-series chain is established. This chain is used to dynamically characterize the temporal evolution of key attributes such as soil organic matter, aggregate structure, and nutrient release capacity. Subsequently, using the node fertilization difference as a shared anchor point, the compensation time-series chain and the soil characteristic change time-series chain are aligned, ensuring that the peak period of fertilizer release overlaps with the peak of crop nutrient demand and the trough of soil nutrients in the time dimension, thereby improving the accuracy and timeliness of application. Ultimately, using "minimum input and optimal compensation" as the multi-objective optimization criterion, and based on the aligned dual-chain structure, the optimal application sequence is determined through a spatiotemporal dynamic search strategy, outputting application control information. This application control information includes the specific application time, application location, and the corresponding lipid-microbial agent-amendant compound formulation and precise dosage for each application location.

[0058] In summary, the embodiments of this application have at least the following technical effects:

[0059] First, lipid molecules, black soil environment, functional microorganisms, and target plants are used as the main nodes. The relationships between these nodes are analyzed, and the characteristics of the target black soil are integrated to construct a fertility improvement diagram structure, including positive and negative relationships. Next, the entire planting cycle of the target black soil is analyzed, a cyclical planting chain is constructed, and deviation parameters are configured based on the soil demand relationships during the planting cycle. Then, based on the fertility improvement diagram structure and the deviation parameters, the cyclical planting chain is mapped to the fertility improvement differences of the main nodes, obtaining the node fertility improvement differences. Finally, using the node fertility improvement differences as the application compensation target, target compensation analysis is performed based on the application time-series variables to obtain the optimal compensation application strategy with minimum input and generate application control information. This solves the technical problem of inaccurate fertilization in existing technologies, which leads to long black soil fertilization cycles, achieving the technical effects of shortening the fertilization cycle and improving fertilization accuracy.

[0060] Example 2, based on the same inventive concept as the intelligent application method of bio-fertilizer for rapidly enriching black soil in the foregoing examples, such as... Figure 2 As shown, this application provides an intelligent application system for bio-fertilizer to rapidly enrich black soil, wherein the system includes:

[0061] Graph Construction Module 11: Using lipid molecules, black soil environment, functional microorganisms, and target plants as main nodes, analyzes the relationships between the main nodes, integrates the characteristics of the target black soil, and constructs a fertility graph structure, including positive and negative relationships; Analysis Module 12: Performs full-cycle analysis of the target black soil planting, constructs a cyclical planting chain, and configures deviation parameters according to the soil demand relationship of the planting cycle; Mapping Module 13: Based on the fertility graph structure and the deviation parameters, maps the fertility difference of the main nodes in the cyclical planting chain to obtain the node fertility difference; Compensation Analysis Module 14: Using the node fertility difference as the application compensation target, performs target compensation analysis based on the application time-series variable relationship to obtain the compensation application strategy with minimum input and optimal compensation, and generates application control information.

[0062] Furthermore, the graph construction module 11 is used to perform the following methods:

[0063] Each subject node is assigned a subset of subjects based on its type characteristics. Then, based on a graph attention network, weights are coupled between the subject subsets to generate a multi-scale edge weight matrix, which quantifies the contribution intensity of positive and negative relationships.

[0064] Furthermore, the graph construction module 11 is used to perform the following methods:

[0065] For the main nodes of lipid molecules, subsets of lipid type, carbon chain length, and functional group abundance are constructed; for the main nodes of black soil environment, subsets of organic matter content, aggregate structure, and enzyme activity level are constructed; for the main nodes of functional microorganisms, subsets of bacterial community, fungal community, and functional gene are constructed; and for the main nodes of target plants, subsets of root system configuration, root secretion spectrum, and nutrient requirements are constructed.

[0066] Furthermore, the graph construction module 11 is used to perform the following methods:

[0067] Collect multi-source heterogeneous data of the target black soil, including at least hyperspectral images, temporal water tension, redox potential, and microbial metagenomics; perform spatiotemporal alignment and missing filling on the multi-source heterogeneous data to obtain a unified spatiotemporal feature tensor; use a Transformer encoder to extract global and local features from the feature tensor and couple them, and use a graph convolutional decoder to map them to the node attributes and edge attributes of a graph attention network structure to obtain the fertilization graph structure.

[0068] Furthermore, the graph construction module 11 is used to perform the following methods:

[0069] A first scale is constructed within the same principal node, and a second scale is constructed between different principal nodes to form a multi-scale hierarchical graph. A multi-head attention mechanism is applied to any first scale to obtain a first-scale edge weight matrix. For any second scale, node-level cross-scale projection is calculated to obtain a second-scale weight matrix. The first-scale edge weight matrix and the second-scale weight matrix are fused to obtain a multi-scale edge weight matrix, where positive values ​​represent positive relationships and their strengths, and negative values ​​represent negative relationships and their strengths. The multi-scale edge weight matrix is ​​injected into the graph attention network as adjacency weights, so that each edge simultaneously carries a relationship type label, a multi-scale coupling weight value, and a scale source marker.

[0070] Furthermore, the parsing module 12 is used to perform the following methods:

[0071] The target plants for the entire planting cycle of the target black soil are obtained. The cyclical planting chain is constructed according to the planting time and growth cycle of the target plants, which includes multiple time nodes. Each time node has a plant name and its growth cycle label. Soil requirements are analyzed for the plant names and growth cycles of each time node in the cyclical planting chain. The difference is calculated based on the soil requirements of the nodes to obtain the deviation parameters of each time node, including the deviation requirement tensor and the difference in requirement value.

[0072] Furthermore, the parsing module 12 is used to perform the following methods:

[0073] For each time node, a ternary demand tensor of soil, crop, and meteorology is established, in which soil, crop, and meteorology each have multi-dimensional index quantification rules; based on sample data, demand analysis is performed at each time node according to plant name, planting time, and growth cycle to obtain soil demand information of plants at each time node, and the ternary demand tensor is projected and transformed; the demand tensors are subtracted according to the continuous relationship of time nodes to obtain the deviation demand tensor and the demand value difference.

[0074] Furthermore, the mapping module 13 is used to perform the following method:

[0075] The deviation parameter is used as a perturbation signal and superimposed on the edge weights of the fattening graph structure, causing the fattening graph structure to change dynamically over time. Time-coupled extrapolation is initiated on the dynamic fattening graph structure. Each node collects the demand gap and available adjustment amount from its neighboring nodes, and then iteratively calculates the adjustment direction and adjustment amount of its own node in combination with its own attributes to meet the adjustment requirements of the perturbation signal. This process continues until the deviation requirements of each time node are met. At the end of the iteration, the fattening difference of each main node is obtained as the node fattening difference.

[0076] Furthermore, the compensation analysis module 14 is used to perform the following method:

[0077] Using the node fertilization difference as the application compensation target, the node fertilization difference is decomposed temporally according to the fertilizer release curve and crop demand curve to form a compensation time-series chain, where each link is marked with the fertilizer effect initiation, peak, and depletion time. Based on the target black soil characteristics and the cyclical planting chain, a soil characteristic change time-series chain is constructed according to the evolution law of the target black soil characteristics with the crop growth stage. Using the node fertilization difference as a shared anchor point, the compensation time-series chain is aligned with the soil characteristic change time-series chain, so that the fertilizer effect peak overlaps with the crop demand peak and the soil nutrient trough in time. With the minimum input and optimal compensation as the evaluation objective, the optimal application sequence is searched on the aligned double chain to obtain the application control information, including the application operation time, spatial location, lipid-microbial agent-amendant formula and dosage.

[0078] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0079] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0080] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A smart application method for bio-fertilizer to rapidly enrich black soil, characterized in that, The method includes: Lipid molecules, black soil environment, functional microorganisms and target plants are used as the main nodes. The relationships between the main nodes are analyzed, and the characteristics of the target black soil are integrated to construct a fertility map structure, including positive and negative relationships. A full-cycle analysis of planting in the target black soil was conducted, a cyclical planting chain was constructed, and deviation parameters were configured based on the soil demand relationship of the planting cycle. Based on the fertilization diagram structure and the deviation parameters, the fertilization difference of the cyclic planting chain is mapped to obtain the node fertilization difference. Using the difference in fertilization at the nodes as the application compensation target, target compensation analysis is performed based on the relationship between application time-series variables to obtain the optimal compensation application strategy with minimum input and generate application control information.

2. The intelligent application method of bio-fertilizer for rapidly enriching black soil according to claim 1, characterized in that, Using lipid molecules, black soil environment, functional microorganisms, and target plants as key nodes, the relationships between these key nodes were analyzed, including: Each main node is used to construct a subset of the main entity based on its type characteristics. Based on graph attention networks, weights are coupled to each subject subset to generate a multi-scale edge weight matrix, which quantifies the contribution intensity of positive and negative relationships.

3. The intelligent application method of bio-fertilizer for rapidly enriching black soil according to claim 2, characterized in that, Each subject node is used to construct a subset of the subject based on its type characteristics, including: Lipid type subsets, carbon chain length subsets, and functional group abundance subsets are constructed for the main nodes of lipid molecules; Subsets of organic matter content, aggregate structure, and enzyme activity level were constructed for the main nodes of the black soil environment; Construct bacterial community subsets, fungal community subsets, and functional gene subsets for the main functional microbial nodes; For the main nodes of the target plant, construct a subset of root system configuration, a subset of root secretion spectrum, and a subset of nutrient requirements.

4. The intelligent application method of bio-fertilizer for rapidly enriching black soil according to claim 2, characterized in that, The fusion of target black soil features to construct a soil improvement map structure includes: Collect multi-source heterogeneous data of the target black soil, including at least hyperspectral images, temporal water tension, redox potential, and microbial metagenomics; Spatiotemporal alignment and missing padding are performed on the multi-source heterogeneous data to obtain a unified spatiotemporal feature tensor. The global and local features extracted from the feature tensor are extracted using a Transformer encoder and coupled together. Then, a graph convolutional decoder is used to map the node and edge attributes of the graph attention network structure to obtain the fattening graph structure.

5. The intelligent application method of bio-fertilizer for rapidly enriching black soil according to claim 3, characterized in that, Based on graph attention networks, weights are coupled to each subject subset to generate a multi-scale edge weight matrix, which quantifies the contribution intensity of positive and negative relationships, including: The first scale is constructed within the same main node, and the second scale is constructed between different main nodes to form a multi-scale hierarchical graph. A multi-head attention mechanism is applied to any first scale to obtain the first scale edge weight matrix; For any second scale, calculate the node-level cross-scale projection to obtain the second scale weight matrix; The first-scale edge weight matrix and the second-scale weight matrix are fused to obtain a multi-scale edge weight matrix, where positive values ​​represent positive relationships and their strength, and negative values ​​represent negative relationships and their strength. The multi-scale edge weight matrix is ​​injected into the graph attention network as adjacency weights, so that each edge carries a relation type label, a multi-scale coupling weight value, and a scale source label.

6. The intelligent application method of bio-fertilizer for rapidly enriching black soil according to claim 1, characterized in that, A full-cycle planting analysis of the target black soil was conducted, a cyclical planting chain was constructed, and deviation parameters were configured based on the soil demand relationship of the planting cycle, including: Obtain the target plants for the entire planting cycle of the target black soil, and construct the cycle planting chain according to the planting time and growth cycle of the target plants, including multiple time nodes, each time node having a plant name and its growth cycle label; Soil requirements are analyzed for the plant names and growth cycles at each time point in the cyclical planting chain. Based on the soil requirements at each time point, differential calculations are performed to obtain the deviation parameters for each time point, including the deviation requirement tensor and the difference in requirement value.

7. The intelligent application method of bio-fertilizer for rapidly enriching black soil according to claim 6, characterized in that, Obtain the deviation parameters at each time point, including: For each time point, a ternary demand tensor of soil-crop-meteorology is established, in which soil, crop, and meteorology each have multi-dimensional index quantification rules; Based on the sample data, we conducted a time-node demand analysis according to plant name, planting time, and growth cycle to obtain soil demand information for plants at each time node, and then performed a ternary demand tensor projection transformation. The demand tensor is subtracted according to the continuous relationship of time nodes to obtain the deviation demand tensor and the demand value difference.

8. The intelligent application method of bio-fertilizer for rapidly enriching black soil according to claim 6, characterized in that, Based on the fertilization diagram structure and the deviation parameters, the fertilization difference of the cyclic planting chain is mapped to obtain the node fertilization difference, including: The deviation parameter is used as a perturbation signal and superimposed on the edge weights of the fattening graph structure, so that the fattening graph structure changes dynamically over time. A time-coupled extrapolation is initiated on the dynamic fattening graph structure. Each node collects the demand gap and available adjustment amount from its neighboring nodes, and then iteratively calculates the adjustment direction and adjustment amount of its own node in combination with its own attributes to meet the adjustment requirements of the disturbance signal. The iteration continues until the deviation requirements of each time node are met. At the end of the iteration, the fattening difference of each main node is obtained as the node fattening difference.

9. The intelligent application method of bio-fertilizer for rapidly enriching black soil according to claim 8, characterized in that, Using the difference in fertilization at the nodes as the application compensation target, target compensation analysis is performed based on the relationship between application time-series variables to obtain the optimal compensation application strategy with minimum input, and application control information is generated, including: Using the fertilization difference at the nodes as the application compensation target, the fertilization difference at the nodes is decomposed into a time series based on the fertilizer release curve and crop demand curve to form a compensation time series chain, wherein each link of the chain is marked with the fertilizer effect start-up, peak and depletion time. Based on the target black soil characteristics and the cyclical planting chain, and according to the evolution of the target black soil characteristics with the crop growth stage, a temporal chain of soil characteristic changes is constructed. Using the difference in fertilization at nodes as a shared anchor point, the compensation time series chain is aligned with the soil characteristic change time series chain, so that the peak of fertilizer effect overlaps with the peak of crop demand and the trough of soil nutrients in time. With the goal of minimizing input and optimizing compensation, the optimal application sequence is searched on the aligned double chain to obtain the application control information, including the application time, spatial location, lipid-bacterial agent-amendant formulation and dosage.

10. An intelligent application system for bio-fertilizer that rapidly enriches black soil, characterized in that, A smart application method for a bio-fertilizer for rapidly enriching black soil according to any one of claims 1-9, the system comprising: Graph construction module: Lipid molecules, black soil environment, functional microorganisms and target plants are used as main nodes. The relationships between the main nodes are analyzed, the characteristics of the target black soil are integrated, and the fertilization graph structure is constructed, including positive and negative relationships. Analysis module: Performs full-cycle analysis of the target black soil, constructs a cycle planting chain, and configures deviation parameters based on the soil demand relationship of the planting cycle; Mapping module: Based on the fertilization diagram structure and the deviation parameters, the fertilization difference of the cyclic planting chain is mapped to obtain the node fertilization difference; Compensation Analysis Module: Taking the difference in fertilization at the nodes as the application compensation target, the module performs target compensation analysis based on the relationship between application time-series variables to obtain the optimal compensation application strategy with minimum input and generate application control information.

Images