Method and system for automatically generating a large model-based farmland planting management scheme

Abstract

The present application relates to the technical field of farmland planting data processing, and provides a farmland planting management scheme automatic generation method and system based on a large model to solve the problem that the prior art cannot customize individualized planting management schemes and accurately judge farming time, comprising: obtaining corresponding multi-source heterogeneous data based on a unique plot identifier; converting the multi-source heterogeneous data into semantic feature blocks, and then constructing dynamic prompt words; using a large language model to reason based on the dynamic prompt words to generate a preliminary planting management scheme; structurally analyzing and logically checking the preliminary planting management scheme to generate a final planting management scheme, which can automatically generate individualized and full-process farmland planting management schemes.

Description

Technical Field

[0001] This invention relates to the field of farmland planting data processing technology, and in particular to a method and system for automatically generating farmland planting management plans based on a large model. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Farmland planting management solutions include crop planting recommendations, farmland environmental parameters, and product forecasts, providing a data foundation for improving farmland planting decisions. Existing farmland planting management solutions primarily rely on the experience of agricultural experts or rule-based expert systems. These systems provide planting suggestions for specific crops based on pre-set rule bases built on individual experience, but they cannot be personalized according to the soil characteristics, historical climate conditions, and geographical location of specific plots, nor can they accurately determine the timing of agricultural activities based on real-time weather changes and crop growth status. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for automatically generating farmland planting management plans based on a large model, which can automatically generate personalized, full-process farmland planting management plans.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of the present invention provides a method for automatically generating farmland planting management schemes based on large models.

[0006] In one or more embodiments, a method for automatically generating farmland planting management plans based on a large model is provided, including: Configure a unique identifier for each land parcel based on its boundary coordinates. Based on the unique identifier of the plot, corresponding multi-source heterogeneous data is obtained; the multi-source heterogeneous data includes soil data, meteorological data, and crop variety characteristic data. The multi-source heterogeneous data is transformed into semantic feature blocks; Based on the semantic feature blocks, dynamic prompt words are constructed; the dynamic prompt words include role positioning information, reasoning logic guidance information, and output format constraint information. Based on the dynamic prompts, a preliminary planting and management plan is generated using a large language model for reasoning. The preliminary planting management plan is subjected to structured analysis and logical verification to generate the final planting management plan, which includes a timeline of the entire growth cycle, agricultural operation standards, and a list of agricultural inputs.

[0007] As one implementation method, converting the multi-source heterogeneous data into semantic feature blocks includes: The multi-source heterogeneous data is subjected to land parcel feature encoding to generate land parcel context feature blocks; Environmental deviation analysis is performed on the multi-source heterogeneous data to generate a constraint factor feature block; The crop variety characteristic data is vectorized into crop demand to generate target constraint feature blocks.

[0008] As one implementation method, the multi-source heterogeneous data is subjected to parcel feature encoding to generate parcel context feature blocks, including: The geographic coordinates of the land parcels are transformed into geographic description information that includes regional boundary delineation and terrain features through geocoding algorithms; The geographic description information and soil type information are fused to generate the context feature block of the land parcel.

[0009] As one implementation method, environmental deviation analysis is performed on the multi-source heterogeneous data to generate a constraint factor feature block, including: The real-time soil data is compared with the standard soil data in the crop standard requirement database. When the real-time soil data deviates from the standard soil data by more than a preset threshold, a description of the soil limiting factor is generated. Historical meteorological data are compared with standard meteorological data in the crop standard requirement database. When the fluctuation of historical meteorological data exceeds a preset threshold, a meteorological constraint factor description is generated.

[0010] As one implementation, the reasoning logic guidance information includes a diagnostic step, a deduction step, a decision-making step, and a verification step executed sequentially; wherein, the diagnostic step is used to guide the large language model to analyze the characteristics of the plot and the crop characteristics corresponding to the recommended optimal crop, and to identify key limiting factors; The deduction step is used to guide the large language model to calculate the reproductive period time nodes based on the accumulated temperature model and the growth model; The decision-making steps are used to guide the large language model to match agricultural operations and adjust operation parameters according to the key constraint factors at the reproductive time nodes. The verification step is used to guide the large language model to check logical consistency before generation.

[0011] As one implementation method, the preliminary planting management plan is subjected to structured analysis and logical verification, including: The preliminary planting management plan is formatted and parsed to extract structured agricultural operation objects; The agricultural operation objects are subjected to hard validation by the rule engine; When the agricultural operation object fails the hard validation of the rule engine, the error information is fed back to the large language model, triggering re-inference until a valid planting management plan is generated.

[0012] As one implementation method, the final planting management scheme also includes dynamic triggering rules, which define threshold conditions for agricultural operations based on real-time monitoring data.

[0013] A second aspect of the present invention provides an automatic generation system for farmland planting management schemes based on a large model.

[0014] In one or more embodiments, an automatic generation system for farmland planting management schemes based on large models includes: The unique identifier configuration module for land parcels is used to configure unique identifiers for land parcels based on their boundary coordinates. A multi-source heterogeneous data acquisition module is used to acquire corresponding multi-source heterogeneous data based on the unique identifier of the plot; the multi-source heterogeneous data includes soil data, meteorological data, and crop variety characteristic data; A semantic feature block conversion module is used to convert the multi-source heterogeneous data into semantic feature blocks; A dynamic prompt word construction module is used to construct dynamic prompt words based on the semantic feature blocks; the dynamic prompt words include role positioning information, reasoning logic guidance information, and output format constraint information; The preliminary planting management plan generation module is used to generate a preliminary planting management plan based on the dynamic prompt words and by using a large language model for reasoning. The final planting management plan generation module is used to perform structured analysis and logical verification on the preliminary planting management plan to generate the final planting management plan, which includes a timeline of the entire growth cycle, agricultural operation standards, and a list of agricultural inputs.

[0015] A third aspect of the present invention provides a computer-readable storage medium.

[0016] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described method for automatically generating farmland planting management schemes based on large models.

[0017] A fourth aspect of the present invention provides an electronic device.

[0018] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the above-described method for automatically generating farmland planting management schemes based on large models.

[0019] Compared with the prior art, the beneficial effects of the present invention are: This invention obtains corresponding multi-source heterogeneous data based on the unique identifier of the land parcel; transforms the multi-source heterogeneous data into semantic feature blocks, and then constructs dynamic prompt words; based on the dynamic prompt words, it uses a large language model for reasoning to generate a preliminary planting management plan; it performs structured analysis and logical verification on the preliminary planting management plan, and finally provides a personalized, full-process farmland planting management plan, which can accurately determine the timing of agricultural activities based on actual field conditions, thereby avoiding yield reduction or resource waste caused by inappropriate timing and improving agricultural production efficiency. Attached Figure Description

[0020] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0021] Figure 1 This is a flowchart of an automatic generation method for farmland planting management schemes based on a large model, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of the automatic generation system for farmland planting management scheme based on a large model, according to an embodiment of the present invention. Figure 3 This is a schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0024] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0025] Figure 1 A schematic diagram illustrating the principle of an automatic generation method for farmland planting management schemes based on a large model, according to an embodiment of the present invention, is provided. Figure 1 The automatic generation method for farmland planting management scheme based on large model in this embodiment may include the following steps S101 to S106.

[0026] The specific implementation process of steps S101 to S106 is as follows: Step S101: Configure a unique identifier for the land parcel based on the parcel boundary coordinates.

[0027] In one embodiment, a unique plot identifier is a digital code used to distinguish different farmland plots, and the identifier is associated with the plot's geographical location, boundary range, and basic attribute information. The purpose of the unique plot identifier is to establish an index relationship between the plot and multi-source data, enabling accurate retrieval and association of soil data, meteorological data, and historical planting records for a specific plot. The unique plot identifier can be generated by an automatically assigned serial number or based on geocoding rules, ensuring that each plot is unique and traceable.

[0028] In other alternative embodiments, the unique identifier for a land parcel is generated using electronic fence technology. Specifically, the coordinate sequence of land parcel boundaries is obtained on an electronic map through user drawing or automatic identification. These coordinate points are then connected sequentially to form a closed polygon, which constitutes the electronic fence. Based on the geometric attributes of the electronic fence, the area, perimeter, and center point coordinates of the land parcel are calculated, and a unique land parcel identification code is generated according to the land parcel's geographical location and creation time. The land parcel identification code uses a mixed alphanumeric encoding, including a region code, land parcel serial number, and timestamp, ensuring the uniqueness and traceability of the identifier. The advantage of electronic fence technology is that it can accurately delineate farmland boundaries, avoiding overlapping or omissions of land parcels, and providing an accurate spatial reference for subsequent data matching and scheme generation.

[0029] Step S102: Based on the unique identifier of the plot, obtain the corresponding multi-source heterogeneous data; the multi-source heterogeneous data includes soil data, meteorological data and crop variety characteristic data.

[0030] Soil data includes, but is not limited to, physicochemical properties such as soil type, texture, organic matter content, nitrogen, phosphorus and potassium content, and pH value. These data reflect the soil fertility status and nutrient supply capacity of the plot.

[0031] Meteorological data includes, but is not limited to, historical temperature, rainfall, sunshine duration, humidity, and other time-series data, which reflect the climate characteristics and meteorological risks of the area where the plot is located.

[0032] Crop variety characteristic data includes, but is not limited to, biological characteristic parameters such as crop growth cycle, accumulated temperature requirement, light requirement, water requirement, and disease resistance. These data define the crop's growth patterns and environmental adaptability. The acquisition of multi-source heterogeneous data relies on the data layer's storage and indexing mechanisms. Corresponding data records are retrieved from the database using unique plot identifiers, and scattered data is integrated into a unified dataset.

[0033] In some alternative embodiments, the acquisition of multi-source heterogeneous data employs an automatic matching method. Specifically, based on the unique identifier of the land parcel, data records associated with the land parcel are retrieved in the data layer.

[0034] For soil data, the nearest neighbor soil grid is queried in the soil grid database based on the coordinates of the plot center point, and parameters such as soil type, texture, organic matter content, nitrogen, phosphorus and potassium content, and pH value are extracted from the grid.

[0035] For meteorological data, historical meteorological station data for the region is queried in the meteorological database according to the administrative division where the land is located, and time-series data such as temperature, rainfall, sunshine duration, and humidity for the most recent set historical time period (e.g., 5 years) are extracted.

[0036] For crop variety characteristic data, the system queries the crop knowledge graph for parameters such as growth cycle, accumulated temperature requirement, light requirement, water requirement, and disease resistance based on the crop variety selected by the user. The automatic matching logic is as follows: it prioritizes matching data sources within a radius (e.g., 5 km) of the plot's center point; if no data source is found within this radius, the search radius is expanded (e.g., 10 km) to ensure the timeliness and representativeness of the data. The advantage of automatic matching is that it reduces manual intervention, improves data acquisition efficiency, and ensures data accuracy and consistency.

[0037] Step S103: Convert multi-source heterogeneous data into semantic feature blocks.

[0038] Semantic feature blocks refer to the process of transforming raw numerical data into textual descriptions with agricultural semantics, enabling large language models to understand the meaning of the data.

[0039] Specifically, multi-source heterogeneous data is transformed into semantic feature blocks, including: Step S1031: Encode the land parcel features of the multi-source heterogeneous data to generate land parcel context feature blocks.

[0040] Multi-source heterogeneous data is used to encode land parcel features, generating land parcel context feature blocks, including: Step S10311: Transform the geographic coordinates of the land parcel into geographic description information that includes regional boundary delineation and terrain features through a geocoding algorithm; for example, transform latitude and longitude coordinates into a description of "located in the North China Plain, with flat terrain".

[0041] The geocoding algorithm first queries an administrative division database based on the coordinates of the plot's center point to obtain information such as the province, city, county, and township where the plot is located. Then, it queries a terrain database based on the plot's center point coordinates to obtain the terrain type of the area, which includes categories such as plains, hills, mountains, and basins. The administrative division information and terrain type are then combined to form a geographic description, such as "located in a certain district of a certain city in a certain province, with a plain terrain."

[0042] Geographic description information is fused with soil type information, which is extracted from soil data, such as "soil type is black soil". The fusion operation uses text concatenation to generate a plot context feature block, such as "located in a certain district of a certain city in a certain province, with a plain terrain and black soil type".

[0043] The contextual feature block of the land parcel provides the macro-environmental background of the land parcel for the large language model, enabling the model to understand the geographical location and natural conditions of the land parcel, laying the foundation for subsequent reasoning.

[0044] Step S10312: Integrate geographic description information with soil type information to generate a parcel context feature block. For example, "Located in the North China Plain, with flat terrain and black soil type".

[0045] The logical formula for encoding the context feature block of a land parcel is: ; Coordinates represents the geographic coordinates of the land parcel. Represents geocoding algorithms, Represents soil type, Representative feature splicing operation, This represents the generated land parcel context feature block.

[0046] Step S1032: Perform environmental deviation analysis on multi-source heterogeneous data to generate constraint factor feature blocks.

[0047] Specifically, environmental bias analysis is performed on multi-source heterogeneous data to generate limiting factor feature blocks, including: Step S10321: Compare the real-time soil data with the standard soil data in the crop standard requirement database. When the real-time soil data deviates from the standard soil data by more than a preset threshold, generate a soil limiting factor description; for example, if the detected nitrogen content N is greater than... If so, the feature block "Nitrogen-rich soil" will be generated; Step S10322: Compare historical meteorological values ​​with standard meteorological values ​​in the crop standard requirement database. When the fluctuation of historical meteorological values ​​exceeds a preset threshold, generate a meteorological constraint factor description. For example, if historical temperature fluctuations are large, generate "high risk of late spring cold snap".

[0048] The constraint factor feature block provides the large language model with the risk factors and constraints of the land parcel, enabling the model to identify adverse environmental factors and make targeted adjustments when generating solutions.

[0049] Step S1033: Vectorize the crop demand data for crop variety characteristics to generate target constraint feature blocks. For example, transform "wheat growth requires 2500℃·day of accumulated temperature" into the constraint "target accumulated temperature requirement of 2500℃·day".

[0050] The plot context feature block provides basic environmental information about the plot, the constraint factor feature block provides risk factors and constraints, and the target constraint feature block provides crop growth requirements and optimization objectives. These three types of feature blocks together constitute the input data for the large language model's inference, enabling the model to comprehensively understand plot conditions and crop needs, and generate accurate planting plans.

[0051] Step S104: Construct dynamic prompt words based on semantic feature blocks; dynamic prompt words include role positioning information, reasoning logic guidance information, and output format constraint information.

[0052] Dynamic prompts are structured instructions used to guide large language models in reasoning. These instructions encompass three dimensions: role positioning information, reasoning logic guidance information, and output format constraints. Role positioning information ensures that the model reasons based on professional knowledge; reasoning logic guidance information ensures that the model's reasoning process is logically rigorous; and output format constraints ensure that the model's output can be parsed by the machine.

[0053] Role positioning information is used to define the identity of the large language model as an agronomic expert, activate the agricultural knowledge graph parameters inside the model, and enable the model to reason based on agricultural expertise.

[0054] Role positioning information refers to setting a role definition statement at the beginning of the dynamic prompt, such as "You are a senior agronomy expert with 20 years of experience in farmland planting and management". The statement is used to define the identity of the large language model and activate the agricultural knowledge graph parameters inside the model.

[0055] Reasoning logic guidance information refers to setting reasoning logic guidance statements in the middle of dynamic prompts. The statements are arranged in the order of diagnosis, deduction, decision-making, and verification, forcing the model to think in a preset order.

[0056] Specifically, the reasoning logic guidance information includes a diagnostic step, a deduction step, a decision-making step, and a verification step executed sequentially; wherein, the diagnostic step is used to guide the large language model to analyze the characteristics of the plot and the characteristics of the crop corresponding to the recommended optimal crop, and to identify key limiting factors; The deduction steps are used to guide the large language model to calculate the reproductive period time nodes based on the accumulated temperature model and the growth model; The decision-making steps are used to guide the large language model to match agricultural operations with key constraint factors and adjust operation parameters at key points in the reproductive period. The verification step is used to guide the large language model to check logical consistency before generation.

[0057] For example, the guiding statement for the diagnostic steps is: "First, please analyze the input plot context feature block and limiting factor feature block to identify key limiting factors affecting crop growth, such as whether soil nutrients are sufficient or whether there is a weather risk."

[0058] The guiding statement for the deduction steps is: "Next, based on the accumulated temperature model and the crop growth model, calculate the start and end times of each growth period of the crop. The formula for the accumulated temperature model is that the accumulated temperature is equal to the daily average air temperature minus the basic temperature for crop growth. When the accumulated temperature reaches the accumulated temperature required for the crop growth period, the growth period ends."

[0059] The guiding statement for the decision-making process is: "Again, please match the corresponding agricultural operations according to the identified key limiting factors, and adjust the operation parameters according to the limiting factors. For example, if the soil is rich in nitrogen, reduce the amount of base fertilizer; if there is a risk of late spring frost, postpone the sowing time."

[0060] The guiding statement for the verification step is: "Finally, before generating the plan, please check the logical consistency of the plan, ensure that the sowing time is earlier than the harvest time, and ensure that the amount of agricultural inputs used is within a reasonable range."

[0061] The above four guiding statements are embedded with dynamic prompts in sequence, forcing the model to reason according to the logical chain of diagnosis, deduction, decision-making, and verification, thus avoiding the model from jumping to answers or omitting key steps.

[0062] Output format constraints refer to the statement that sets the output format requirements at the end of the dynamic prompt, such as "Please output the planting management plan in JSON format, including the timeline, operation standards and material list fields". The statement is used to constrain the model to output a structured format plan.

[0063] Output format constraints are used to constrain the large language model to output a structured planting management plan. The format includes fields such as time axis, operation criteria, and threshold conditions to ensure that the plan output by the model can be parsed and executed by the machine.

[0064] Step S105: Based on dynamic prompts, use a large language model to perform reasoning and generate a preliminary planting management plan.

[0065] The large language model used here is a fine-tuned large language model based on agricultural data. The fine-tuning process is as follows: A large amount of professional agricultural literature, planting manuals, agricultural technology Q&A, and historical planting records were collected to construct a dedicated agricultural corpus. The corpus includes crop growth knowledge, agricultural operation standards, pest and disease control plans, and meteorological disaster response strategies. Supervised learning fine-tuning of the general large language model was then performed using this corpus. The goal of fine-tuning was to enable the model to learn agricultural terminology, reasoning logic, and decision-making patterns. The fine-tuned model outperformed the general model in agricultural tasks, generating planting plans that better conform to agricultural laws and actual needs. The advantage of fine-tuning the model lies in improving its professionalism and accuracy in the agricultural field, and reducing the probability of the model generating illusions or erroneous reasoning.

[0066] Dynamic prompts are input into a large language model, which then performs reasoning to generate a preliminary planting management plan. The reasoning process of the large language model is implemented based on the attention mechanism of the Transformer architecture.

[0067] First, the large language model converts the concatenated prompt word sequence into a high-dimensional vector space, and then transforms the text into a vector representation through an input embedding layer. Second, the large language model uses a self-attention mechanism to calculate the dot product of the query matrix Q, the key matrix K, and the value matrix V, establishing long-distance dependencies.

[0068] The formula for calculating the attention mechanism is: ; Where Q represents the query matrix, K represents the key matrix, and V represents the value matrix. Representing the scaling factor, the softmax function is used to convert correlation into weight allocation.

[0069] The large language model automatically learns the high-weight negative correlation by calculating the attention weight between soil nitrogen content characteristics and fertilizer application decisions, that is, the amount of fertilizer should be reduced when the soil nitrogen content is high.

[0070] The large language model generates time series data based on positional encoding, simulates the accumulated temperature process, and outputs the start and end times of each growth stage. Utilizing a logic gating mechanism, the model outputs agricultural triggering conditions in an IF-THEN structure, such as IF soil temperature greater than or equal to 10℃ for 3 consecutive days THEN sowing, rather than fixed time points. Finally, the model generates a preliminary planting management plan including a timeline of the entire growth cycle, agricultural operation standards, and a list of agricultural inputs.

[0071] Step S106: Perform structured analysis and logical verification on the preliminary planting management plan to generate the final planting management plan, which includes a timeline of the entire growth cycle, agricultural operation standards, and a list of agricultural inputs.

[0072] Specifically, the preliminary planting management plan undergoes structured analysis and logical verification, including: The preliminary planting management plan is formatted and parsed to extract structured agricultural operation objects; Perform hard validation of agricultural operation objects using a rule engine; When an agricultural operation fails the hard validation of the rule engine, the error message is fed back to the large language model, triggering re-inference until a valid planting management plan is generated.

[0073] Structured parsing refers to using regular expressions or a JSON parser to extract structured agricultural operation objects from the text stream output by the model. These agricultural operation objects contain attributes such as time, type, and parameters. For example, a sowing operation object includes attributes such as sowing start time, sowing end time, sowing depth, and row / plant spacing. The JSON parser extracts attribute values ​​such as time, type, and parameters from the text based on predefined field names and data types, and then constructs the agricultural operation objects.

[0074] Logical validation includes three steps: format parsing, hard validation by the rule engine, and feedback retry.

[0075] Formatted parsing refers to extracting structured data from a preliminary plan according to a predetermined format.

[0076] Hard validation in a rule engine refers to performing temporal and numerical validation on structured data. Temporal validation prevents the generation of time-related logical errors, while numerical validation prevents the generation of data that wastes resources or pollutes the environment.

[0077] Timing verification refers to checking whether the time sequence of agricultural operations conforms to the crop growth pattern. It involves extracting the end time of sowing operations and the start time of harvesting operations, and determining whether the start time of harvesting operations is later than the end time of sowing operations. If the start time of harvesting operations is earlier than or equal to the end time of sowing operations, the timing verification is deemed to have failed, and the error message is marked as a logical exception.

[0078] The specific process of timing verification is as follows: Extract the end time of the sowing operation and the start time of the harvesting operation. Record the end time of the sowing operation as... The start time of the harvest task is recorded as follows: System judgment Is it later than The judgment condition is If the judgment condition is met, the time sequence check passes, indicating that the crop is undergoing a positive growth cycle.

[0079] If the judgment condition is not met, that is Earlier than or equal to If the timing check fails, the system will mark the error message as "Time logic error: Harvest time is not later than sowing time" and trigger the feedback retry mechanism. The logic of the timing check is as follows: crops need to go through a certain growth cycle from sowing to harvest, and the harvest time must be later than the sowing time. If the opposite occurs, it will violate the crop growth law.

[0080] Numerical verification refers to checking whether the agricultural input usage values ​​are within the physically reasonable range. It involves extracting the agricultural input usage values ​​from the agricultural operation objects and determining whether the agricultural input usage values ​​are within the preset physically reasonable range. If the agricultural input usage values ​​are not within the physically reasonable range, the numerical verification is deemed to have failed, and the error message is marked as an illusion.

[0081] The specific process of numerical verification is as follows: Extract the agricultural input usage values ​​from the agricultural operation objects; these values ​​are recorded as "dosage". Determine whether the dosage is within a preset physical reasonable range; the determination criteria are as follows. ,in This represents the preset physical limit value or maximum safety threshold. If the judgment condition is met, the numerical verification passes, indicating that the amount of agricultural input used is within a reasonable range. If the judgment condition is not met, i.e., dosage is less than or equal to 0 or dosage is greater than or equal to... If the numerical verification fails, the system will mark the error message as "numerical anomaly: agricultural input usage exceeds reasonable range" and trigger the feedback retry mechanism. The logic of numerical verification is as follows: agricultural input usage must be a positive number and cannot exceed the maximum absorption capacity of crops or the carrying capacity of the environment. If a zero value, a negative value, or an excessive value appears, it indicates that the model is malfunctioning or that there is a calculation error.

[0082] Feedback retry refers to the process where, when an agricultural operation fails the hard validation of the rule engine, the error information is fed back to the large language model as new context, triggering the model's self-correcting inference chain to re-determine the timeline or adjust the values ​​until a valid planting management plan is generated.

[0083] Formatted parsing transforms unstructured text into structured data, while the rule engine's hard validation ensures the data's correctness and rationality. Feedback retries enable the model's self-correction and iterative optimization.

[0084] The full growth cycle timeline refers to a dynamic timetable generated based on accumulated temperature projections. The timetable lists the start and end times of each growth stage of the crop from sowing to harvest, as well as the execution time windows for key agricultural operations.

[0085] Agricultural operation standards refer to operational specifications that include specific parameters such as sowing depth, row and plant spacing, and irrigation quotas. These specifications clearly define the execution standards and quality requirements for each agricultural operation.

[0086] The agricultural input matching list refers to a list of seeds, fertilizers and pesticides that are accurate to the amount used per acre. The list specifies the types, specifications and amounts of agricultural inputs required for each agricultural operation.

[0087] The method provided by this invention enables personalized customization and dynamic adjustment of farmland planting management plans. Based on the unique identifier of each plot and deep fusion of multi-source data, and through reasoning using a large language model, the generated planting plan can accurately adapt to the actual conditions of a specific plot, including soil characteristics, historical climate conditions, and geographical location, thereby solving the problem of existing systems offering a one-size-fits-all approach. It defines precise start conditions and judgment criteria for each agricultural operation, rather than simple time points, allowing farmers to accurately judge the timing of agricultural activities based on actual field conditions, thus avoiding yield reduction or resource waste due to inappropriate timing. Utilizing a large language model fine-tuned for the agricultural field, it simulates the decision-making process of professional agronomists, comprehensively considering multiple dimensions such as soil characteristics, historical climate, and crop requirements, resulting in planting plans with high scientific and professional qualities, thereby lowering the decision-making threshold for ordinary farmers.

[0088] In some alternative embodiments, the final planting management scheme also includes dynamic triggering rules, which define threshold conditions for agricultural operations based on real-time monitoring data.

[0089] Dynamic trigger rules define threshold conditions for agricultural operations based on real-time monitoring data, using an IF-THEN structure. For example, a sowing trigger rule might be "IF soil temperature greater than or equal to 10℃ for three consecutive days and soil moisture content between 60% and 80% THEN execute sowing." The condition part of the rule includes two monitoring indicators: soil temperature and soil moisture content, while the action part is to execute sowing. An irrigation trigger rule might be "IF soil moisture content lower than 50% THEN execute irrigation, with an irrigation volume of 30 cubic meters per acre." The condition part of the rule includes the soil moisture content monitoring indicator, and the action part is to execute irrigation and specify the irrigation volume. The purpose of dynamic trigger rules is to transform fixed time points into conditional judgments based on real-time monitoring, enabling agricultural operations to be dynamically adjusted according to actual field conditions, thus improving the adaptability and accuracy of the plan.

[0090] like Figure 2 As shown, the automatic generation system for farmland planting management schemes based on large models provided in this embodiment of the invention can be implemented in software. The automatic generation system for farmland planting management schemes based on large models includes the following software modules: a plot unique identifier configuration module 201, a multi-source heterogeneous data acquisition module 202, a semantic feature block conversion module 203, a dynamic prompt word construction module 204, a preliminary planting management scheme generation module 205, and a final planting management scheme generation module 206.

[0091] The following is an introduction to the functions of each software module in the automatic generation system of farmland planting management plans based on large models: The unique identifier configuration module 201 is used to configure a unique identifier for a land parcel based on the boundary coordinates of the land parcel. The multi-source heterogeneous data acquisition module 202 is used to acquire corresponding multi-source heterogeneous data based on the unique identifier of the plot; the multi-source heterogeneous data includes soil data, meteorological data and crop variety characteristic data; Semantic feature block conversion module 203 is used to convert the multi-source heterogeneous data into semantic feature blocks; The dynamic prompt word construction module 204 is used to construct dynamic prompt words based on the semantic feature blocks; the dynamic prompt words include role positioning information, reasoning logic guidance information, and output format constraint information; The preliminary planting management plan generation module 205 is used to generate a preliminary planting management plan based on the dynamic prompt words and by using a large language model for reasoning. The final planting management plan generation module 206 is used to perform structured analysis and logical verification on the preliminary planting management plan to generate the final planting management plan, which includes a full growth cycle timeline, agricultural operation standards, and a list of agricultural inputs.

[0092] It should be noted that each module in the automatic generation system of farmland planting management scheme based on large model in the embodiments of the present invention corresponds one-to-one with each step in the automatic generation method of farmland planting management scheme based on large model in the above embodiments, and their specific implementation processes are the same, so they will not be repeated here.

[0093] The structure of the electronic device according to an embodiment of the present invention will be described in detail below. Figure 3 This is a schematic diagram of the composition structure of an electronic device provided in an embodiment of the present invention. It can be understood that... Figure 3 The diagram shows only an exemplary structure of the electronic device, not the entire structure. Some or all of the structures shown may be implemented as needed.

[0094] The electronic device provided in this embodiment of the invention includes: at least one processor 301, a memory 302, a user interface 303, and at least one network interface 304. The various components in the automatic generation system for farmland planting management schemes based on large models are coupled together via a bus system 305. It can be understood that the bus system 305 is used to realize the connection and communication between these components. In addition to a data bus, the bus system 305 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 3 The general designated all buses as Bus System 305.

[0095] The user interface 303 may include a monitor, keyboard, mouse, trackball, click wheel, buttons, touchpad, or touch screen.

[0096] It is understood that memory 302 can be volatile memory or non-volatile memory, or both. In this embodiment of the invention, memory 302 is capable of storing data to support the operation of the terminal. Examples of this data include any computer programs used to operate on the terminal, such as operating systems and applications. The operating system includes various system programs, such as framework layers, core library layers, driver layers, etc., used to implement various basic services and handle hardware-based tasks. Applications can include various applications.

[0097] In some embodiments, the automatic generation system for farmland planting management schemes based on large models provided in this invention can be implemented using a combination of hardware and software. As an example, the automatic generation system for farmland planting management schemes based on large models provided in this invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the automatic generation method for farmland planting management schemes based on large models provided in this invention. For example, the processor in the form of a hardware decoding processor can employ one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0098] As an example, processor 301 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., wherein the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0099] As an example of the hardware implementation of the automatic generation system for farmland planting management scheme based on a large model provided in this embodiment of the invention, the device provided in this embodiment of the invention can be directly executed by a processor 301 in the form of a hardware decoding processor. For example, it can be executed by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components to implement the automatic generation method for farmland planting management scheme based on a large model provided in this embodiment of the invention.

[0100] The memory 302 in this embodiment of the invention is used to store various types of data to support the operation of the automatic generation system for farmland planting management schemes based on large models, or to store data for execution. Figure 1 The program code for the method shown. Examples of this data include: any executable instructions for operation on a large-model-based automatic generation system for farmland planting management schemes, such as executable instructions that can be included in the executable instructions, implementing the large-model-based automatic generation method for farmland planting management schemes according to embodiments of the present invention.

[0101] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including functions for executing... Figure 1 The program code for the method shown. In such an embodiment, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by the central processing unit, it performs the various functions defined in the apparatus of this application.

[0102] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0103] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for automatically generating farmland planting management plans based on a large model, characterized in that, include: Configure a unique identifier for each land parcel based on its boundary coordinates. Based on the unique identifier of the plot, corresponding multi-source heterogeneous data is obtained; the multi-source heterogeneous data includes soil data, meteorological data, and crop variety characteristic data. The multi-source heterogeneous data is transformed into semantic feature blocks; Based on the semantic feature blocks, dynamic prompt words are constructed; the dynamic prompt words include role positioning information, reasoning logic guidance information, and output format constraint information. Based on the dynamic prompts, a preliminary planting and management plan is generated using a large language model for reasoning. The preliminary planting management plan is subjected to structured analysis and logical verification to generate the final planting management plan, which includes a timeline of the entire growth cycle, agricultural operation standards, and a list of agricultural inputs.

2. The method for automatically generating farmland planting management plans based on a large model as described in claim 1, characterized in that, Transforming the multi-source heterogeneous data into semantic feature blocks includes: The multi-source heterogeneous data is subjected to land parcel feature encoding to generate land parcel context feature blocks; Environmental deviation analysis is performed on the multi-source heterogeneous data to generate a constraint factor feature block; The crop variety characteristic data is vectorized into crop demand to generate target constraint feature blocks.

3. The method for automatically generating farmland planting management plans based on a large model as described in claim 2, characterized in that, The multi-source heterogeneous data is subjected to land parcel feature encoding to generate land parcel context feature blocks, including: The geographic coordinates of the land parcels are transformed into geographic description information that includes regional boundary delineation and terrain features through geocoding algorithms; The geographic description information and soil type information are fused to generate the context feature block of the land parcel.

4. The method for automatically generating farmland planting management plans based on a large model as described in claim 2, characterized in that, Environmental bias analysis is performed on the multi-source heterogeneous data to generate a constraint factor feature block, including: The real-time soil data is compared with the standard soil data in the crop standard requirement database. When the real-time soil data deviates from the standard soil data by more than a preset threshold, a description of the soil limiting factor is generated. Historical meteorological data are compared with standard meteorological data in the crop standard requirement database. When the fluctuation of historical meteorological data exceeds a preset threshold, a meteorological constraint factor description is generated.

5. The method for automatically generating farmland planting management plans based on a large model as described in claim 1, characterized in that, The reasoning logic guidance information includes a diagnostic step, a deduction step, a decision-making step, and a verification step executed sequentially; wherein, the diagnostic step is used to guide the large language model to analyze the characteristics of the plot and the crop characteristics corresponding to the recommended optimal crop, and to identify key limiting factors; The deduction step is used to guide the large language model to calculate the reproductive period time nodes based on the accumulated temperature model and the growth model; The decision-making steps are used to guide the large language model to match agricultural operations and adjust operation parameters according to the key constraint factors at the reproductive time nodes. The verification step is used to guide the large language model to check logical consistency before generation.

6. The method for automatically generating farmland planting management plans based on a large model as described in claim 1, characterized in that, The preliminary planting management plan is subjected to structured analysis and logical verification, including: The preliminary planting management plan is formatted and parsed to extract structured agricultural operation objects; The agricultural operation objects are subjected to hard validation by the rule engine; When the agricultural operation object fails the hard validation of the rule engine, the error information is fed back to the large language model, triggering re-inference until a valid planting management plan is generated.

7. The method for automatically generating farmland planting management plans based on a large model as described in claim 1, characterized in that, The final planting management plan also includes dynamic triggering rules, which define threshold conditions for agricultural operations based on real-time monitoring data.

8. An automatic generation system for farmland planting management plans based on a large model, characterized in that, The method for automatically generating farmland planting management plans based on large models, as described in any one of claims 1-7, includes: The unique identifier configuration module for land parcels is used to configure unique identifiers for land parcels based on their boundary coordinates. A multi-source heterogeneous data acquisition module is used to acquire corresponding multi-source heterogeneous data based on the unique identifier of the plot; the multi-source heterogeneous data includes soil data, meteorological data, and crop variety characteristic data; A semantic feature block conversion module is used to convert the multi-source heterogeneous data into semantic feature blocks; A dynamic prompt word construction module is used to construct dynamic prompt words based on the semantic feature blocks; the dynamic prompt words include role positioning information, reasoning logic guidance information, and output format constraint information; The preliminary planting management plan generation module is used to generate a preliminary planting management plan based on the dynamic prompt words and by using a large language model for reasoning. The final planting management plan generation module is used to perform structured analysis and logical verification on the preliminary planting management plan to generate the final planting management plan, which includes a timeline of the entire growth cycle, agricultural operation standards, and a list of agricultural inputs.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the method for automatically generating farmland planting management schemes based on large models as described in any one of claims 1-7.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the method for automatically generating farmland planting management schemes based on large models as described in any one of claims 1-7.

Images