A method and system for returning corn stalks to the field for fertilization

Abstract

This application discloses a method and system for returning corn straw to the field for fertilization, relating to the field of agricultural technology. The application includes basic data preprocessing, digital twin construction, dynamic decomposition model embedding, optimization decision generation, and fertilization prescription output. Based on multi-source data fusion and spatial interpolation algorithms, a high-fidelity virtual farmland environment is constructed, overcoming the limitations of traditional methods in characterizing field heterogeneity. By embedding a dynamic decomposition model of straw and soil microorganisms, quantitative prediction of straw decomposition rate and nutrient release is achieved, providing a scientific basis for fertilization timing and dosage. Combining state space construction and strategy learning algorithms, dynamic matching of nutrient supply and demand and adaptive optimization of fertilization strategies are realized, significantly improving fertilizer utilization efficiency, reducing environmental pollution risks, generating precise fertilization prescriptions and outputting them to a display terminal, thus enhancing the intelligence level and ease of operation of agricultural production.

Description

Technical Field

[0001] This application relates to the field of agricultural technology, specifically to a method and system for returning corn stalks to the field for fertilization. Background Technology

[0002] Traditional agricultural fertilization management relies heavily on experience-based judgment or fixed formulas, making it difficult to achieve dynamic matching between nutrient release and crop demand under straw return conditions. Existing straw return management lacks digital twins and intelligent decision support, making it difficult to achieve closed-loop optimization of "data-model-decision," which restricts the efficiency of straw resource utilization and sustainable farmland development.

[0003] The prior art, such as the invention application patent with announcement number CN119090668B, discloses a fertilization method and device based on corn straw returning to the field, which includes: receiving planting data from corn straw returning growers; acquiring multiple historical soil samples based on the straw returning area; detecting the multiple historical soil samples using a pre-constructed soil data detection method to obtain historical soil data; acquiring target soil data based on the target crop; calculating the amount of straw per unit area based on the planting data; extracting a set of straw decomposition curves from a pre-constructed database based on the amount of straw per unit area and the corn type; acquiring data on the amount of mixed fertilizer per unit area using the set of straw decomposition curves; and performing fertilization operations on the straw returning area based on the data on the amount of mixed fertilizer per unit area to complete the fertilization based on corn straw returning to the field.

[0004] Regarding the above-mentioned solutions, the inventors of this application have discovered that the above technologies have at least the following technical problems: 1. Currently, there is a lack of weighted fusion and standardization to improve data quality and consistency, which cannot avoid the problem of data source error accumulation in traditional methods, and cannot ensure the reliability of subsequent model inputs, thus failing to improve overall prediction accuracy. 2. The authenticity and reference value of the analysis results cannot be guaranteed, and reliable basis for subsequent decision-making cannot be provided. 3. There is a lack of high-fidelity virtual farmland simulation environment constructed based on standardized datasets and hyperspectral data through spatial interpolation algorithms; the accurate digital mapping of the farmland environment has not been achieved, and a high-resolution simulation environment cannot be provided, which cannot enhance the realism and predictive ability of the model, nor support comprehensive and objective analysis.

[0005] 2. Currently, the lack of a combination of mechanistic models and data-driven approaches makes it impossible to accurately predict nutrient dynamics, avoid the idealized assumptions of traditional single models, improve the accuracy and adaptability of models, ensure the scientific nature of fertilization decisions, reduce waste and pollution caused by blind fertilization, achieve closed-loop optimization and real-time decision-making for complex agronomic processes, and adjust fertilization strategies in a timely manner, thus failing to improve fertilizer utilization and crop yield.

[0006] 3. Currently, there is a lack of fertilizer application calculations based on optimization decisions, which can transform decisions into actionable prescriptions. It cannot provide variable fertilization recommendations and optimal window periods, and cannot achieve precise and automated fertilization to generate accurate prescriptions and output them to display terminals. It cannot guarantee the pertinence and balance of field operations, and cannot improve the efficiency and sustainability of agricultural resource utilization. Summary of the Invention

[0007] To address the aforementioned technical shortcomings, the purpose of this application is to provide a method and system for returning corn stalks to the field for fertilization.

[0008] To solve the above-mentioned technical problems, this application adopts the following technical solution: In the first aspect, this application provides a method for returning corn straw to the field for fertilization, which includes the following steps: Step 1, basic data preprocessing: Based on the pre-acquired basic data, data cleaning and weighted fusion analysis are performed to generate a standardized dataset.

[0009] Step 2: Digital Twin Construction: Based on the standardized dataset, the preset GIS map, and hyperspectral image data, a digital twin of the land parcel is constructed using a spatial interpolation algorithm to form a high-fidelity virtual farmland simulation environment.

[0010] Step 3: Dynamic decomposition model embedding: Embed the preset straw soil microbial dynamic decomposition model into the plot digital twin, and input basic data to simulate the straw decomposition rate and nutrient release, so as to output the decomposition rate prediction and nutrient release prediction.

[0011] Step 4: Optimize decision generation: Based on the predicted decomposition rate and nutrient release, construct the state space to integrate nutrient status, and then learn the optimal strategy to generate optimized fertilization decisions.

[0012] Step 5: Fertilizer prescription output: Based on the optimized fertilization decision, calculate the amount of fertilizer and generate a precise fertilization prescription, which is then output to the display terminal.

[0013] Preferably, the basic data includes soil data, straw data, and environmental data.

[0014] Preferably, the step of performing data cleaning and weighted fusion analysis based on pre-acquired basic data to generate a standardized dataset includes: performing data cleaning operations based on soil data, straw data, and environmental data to obtain a cleaned dataset, wherein the data cleaning operations include missing value handling, outlier removal, and data standardization; and performing weighted fusion analysis on the cleaned dataset based on preset weight coefficients for data source accuracy to generate a standardized dataset, wherein the weighted fusion analysis is performed based on a weighted average formula.

[0015] Preferably, the step of constructing a digital twin of a land parcel using a spatial interpolation algorithm to form a high-fidelity virtual farmland simulation environment includes: performing data alignment and coordinate unification processing based on the standardized dataset, a preset geographic information system (GIS) map, and hyperspectral image data to generate an integrated spatial dataset; performing interpolation calculations on missing or low-resolution areas using the Kriging spatial interpolation method based on the integrated spatial dataset to generate a high-resolution spatial distribution map; and generating a digital twin of the land parcel using 3D modeling technology based on the high-resolution spatial distribution map and the preset GIS map, including real-time soil data, straw data, and environmental data to form a high-fidelity virtual farmland simulation environment.

[0016] Preferably, the step of embedding a preset straw soil microbial dynamic decomposition model into the plot digital twin and inputting basic data to simulate the straw decomposition rate and nutrient release, so as to output the decomposition rate prediction and nutrient release prediction, includes: A1. Initializing model parameters: Based on the plot digital twin, loading the preset model parameters of the straw soil microbial dynamic decomposition model, wherein the model parameters include microbial activity coefficient, straw carbon-nitrogen ratio threshold, temperature sensitivity coefficient and humidity influence factor.

[0017] A2. Input basic data: Extract basic data parameters from the standardized dataset, including soil temperature, soil moisture, soil pH, straw carbon-nitrogen ratio, and microbial biomass, and input the basic data parameters into the dynamic decomposition model.

[0018] A3. Simulate decomposition rate and nutrient release: Based on the dynamic decomposition model of straw and soil microorganisms, calculate the straw decomposition rate using a decomposition rate algorithm that integrates temperature, humidity, and microbial activity parameters; based on the decomposition rate prediction, calculate the nutrient release prediction using a nutrient release algorithm that correlates straw nutrient content and decomposition rate.

[0019] A4. Prediction of output decomposition rate and nutrient release.

[0020] Preferably, the step of calculating the straw decomposition rate using the decomposition rate algorithm includes: using the calculation formula... The straw decomposition rate was obtained ,in It is expressed as the baseline decomposition rate constant, which is preset based on straw type and soil properties; Represented as a temperature effect function, Represented as a humidity effect function, It is represented as a microbial activity function.

[0021] Preferably, the step of predicting the nutrient release amount based on the decomposition rate prediction and using a nutrient release algorithm includes: calculating the nutrient release amount using a formula. The amount of nitrogen released was determined. Phosphorus nutrient release and potassium nutrient release , , and These are represented as the nitrogen, phosphorus, and potassium nutrient content of the straw, respectively. This represents the time period corresponding to nutrient release.

[0022] Preferably, the step of constructing a state space based on the predicted decomposition rate and nutrient release to integrate nutrient status, and then learning the optimal strategy to generate optimized fertilization decisions includes: B1. Constructing a state space based on the predicted decomposition rate and nutrient release, wherein the state variables include real-time soil nutrient content and predicted nutrient release to integrate nutrient status.

[0023] B2. Define an action space, where actions include fertilizer type selection, fertilizer amount setting, and fertilizer time period determination, as fertilization decision options; design a reward function, which is optimized based on nutrient utilization rate, economic benefits, and environmental benefits to calculate the reward value of the simulation; use a reinforcement learning algorithm to perform simulations in the digital twin of the land parcel based on the state space, action space, and reward function to learn the optimal fertilization strategy.

[0024] B3. Based on the learned optimal fertilization strategy, generate optimized fertilization decisions, including fertilizer formula, fertilizer amount, and fertilization time period.

[0025] Preferably, the step of calculating the fertilization amount and generating a precise fertilization prescription based on the optimized fertilization decision, and then outputting it to the display terminal, includes: extracting the target nutrient amount and crop nutrient requirement parameters based on the optimized fertilization decision; calculating the fertilization amount by combining the existing soil nutrient amount and the nutrient amount released by straw to obtain the actual fertilization amount; determining the fertilizer formula based on the actual fertilization amount and the fertilizer database; determining the fertilization time period based on the crop growth model and real-time environmental data; and finally integrating and generating a precise fertilization prescription to output to the display terminal.

[0026] In its second aspect, this application provides a system for a corn straw fertilization method, comprising: a basic data preprocessing module, which performs data cleaning and weighted fusion analysis based on pre-acquired basic data to generate a standardized dataset.

[0027] The digital twin construction module, based on the standardized dataset, preset geographic information system (GIS) map, and hyperspectral image data, constructs digital twins of land parcels through spatial interpolation algorithms to form a high-fidelity virtual farmland simulation environment.

[0028] The dynamic decomposition model embedding module is used to embed a preset straw soil microbial dynamic decomposition model into the plot digital twin, and input basic data to simulate the straw decomposition rate and nutrient release, so as to output the decomposition rate prediction and nutrient release prediction.

[0029] The optimization decision generation module constructs a state space based on the predicted decomposition rate and nutrient release to integrate nutrient status, thereby learning the optimal strategy and generating optimized fertilization decisions.

[0030] The fertilizer prescription output module calculates the amount of fertilizer and generates a precise fertilizer prescription based on the optimized fertilization decision, and then outputs it to the display terminal.

[0031] The beneficial effects of this application are as follows: 1. The corn straw returning to the field fertilization method and system provided in this application, based on multi-source data fusion and spatial interpolation algorithm, constructs a high-fidelity virtual farmland environment, breaking through the limitations of traditional methods in failing to characterize field heterogeneity; by embedding a dynamic decomposition model of straw and soil microorganisms, it realizes quantitative prediction of straw decomposition rate and nutrient release, providing a scientific basis for fertilization timing and dosage; combined with state space construction and strategy learning algorithm, it realizes dynamic matching of nutrient supply and demand and adaptive optimization of fertilization strategy, significantly improving fertilizer utilization efficiency, reducing environmental pollution risks, generating precise fertilization prescriptions and outputting them to the display terminal, thereby improving the level of intelligence and ease of operation of agricultural production.

[0032] 2. This application improves data quality and consistency through weighted fusion and standardization, avoiding the problem of data source error accumulation in traditional methods. This ensures the reliability of subsequent model inputs, thereby improving overall prediction accuracy. It guarantees the authenticity and reference value of the analysis results, providing a reliable basis for subsequent decision-making. Based on standardized datasets and hyperspectral data, a high-fidelity virtual farmland simulation environment is constructed through spatial interpolation algorithms; it achieves accurate digital mapping of the farmland environment, provides a high-resolution simulation environment, and enhances the model's realism and predictive capabilities. It supports comprehensive and objective analysis, which is helpful for subsequent model embedding and optimization, and reduces the cost of field experiments.

[0033] 3. This application, by combining mechanistic models and data-driven approaches, accurately predicts nutrient dynamics, avoiding the idealized assumptions of traditional single models. This improves the accuracy and adaptability of the model, ensuring the scientific nature of fertilization decisions and reducing waste and pollution caused by indiscriminate fertilization. It also achieves closed-loop optimization and real-time decision-making for complex agronomic processes, enhancing the system's intelligence and efficiency, and allowing for timely adjustments to fertilization strategies, thereby increasing fertilizer utilization and crop yield while reducing environmental impact.

[0034] 4. This application calculates fertilizer application based on optimized decision-making, transforms decisions into actionable prescriptions, provides variable fertilization recommendations and optimal window periods, and realizes the precision and automation of fertilization. It generates precise prescriptions and outputs them to the display terminal, ensuring the pertinence and balance of field operations, reducing human intervention, and improving the efficiency and sustainability of agricultural resource utilization. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a flowchart illustrating the implementation steps of the method described in this application.

[0037] Figure 2 This is a schematic diagram of the system structure connection of this application. Detailed Implementation

[0038] 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 some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0039] Please see Figure 1 As shown, this application provides a method for returning corn straw to the field for fertilization in the first aspect, including: Step 1, basic data preprocessing: based on the pre-acquired basic data, data cleaning and weighted fusion analysis are performed to generate a standardized dataset.

[0040] In a specific instance, the basic data includes soil data, straw data, and environmental data.

[0041] It should be noted that soil data includes soil moisture, soil pH, soil organic matter content, soil nitrogen content, soil phosphorus content, soil potassium content, soil carbon content, soil porosity, and soil temperature; straw data includes straw biomass, straw carbon-to-nitrogen ratio, straw moisture content, and straw decomposition rate; environmental data includes air temperature, air humidity, light intensity, rainfall, wind speed, and carbon dioxide concentration.

[0042] Furthermore, soil moisture and temperature sensors collect field data in real time; soil sampling and analysis determine pH, organic matter content, soil nitrogen content, soil phosphorus content, soil potassium content, and soil carbon content; a soil porosity meter measures soil porosity parameters; a field sampling and weighing method is used to determine straw biomass; an elemental analyzer determines the straw carbon-nitrogen ratio; a moisture meter measures straw moisture content; and an in-situ decomposition experiment determines the straw decomposition rate. These data are obtained through: automatic monitoring of air temperature, humidity, rainfall, and wind speed at a weather station; measurement of light intensity using a photosynthetically active radiation sensor; and monitoring of atmospheric carbon dioxide concentration using a carbon dioxide concentration sensor. Content; all environmental data are transmitted to the data center in real time via IoT devices.

[0043] It should be noted that data acquisition adopts a multi-source heterogeneous acquisition method, which combines real-time sensor monitoring, laboratory analysis, and field measurement; and all acquisition equipment is regularly calibrated to ensure data accuracy and consistency, and the acquisition frequency is dynamically adjusted according to the data type and growing season.

[0044] In a specific example, the process of performing data cleaning and weighted fusion analysis based on pre-acquired basic data to generate a standardized dataset includes: performing data cleaning operations based on soil data, straw data, and environmental data to obtain a cleaned dataset, wherein the data cleaning operations include missing value handling, outlier removal, and data standardization; and performing weighted fusion analysis on the cleaned dataset based on preset weight coefficients for data source accuracy to generate a standardized dataset, wherein the weighted fusion analysis is performed based on a weighted average formula.

[0045] It should be noted that in the data cleaning operation, missing value handling uses mean imputation, specifically: the imputed value equals the sum of all valid data point values ​​divided by the total number of valid data points; the physical meaning is to fill missing values ​​by averaging to ensure data integrity; outlier removal is based on the 3σ principle, specifically: when the absolute deviation of a data point value from the mean exceeds three times the standard deviation, the data point value is considered an outlier and deleted; the physical meaning is to identify and remove significantly deviating outliers to improve data quality; data standardization uses the Min-Max standardization method, specifically: the standardized value equals the original value minus the minimum value of the dataset divided by the difference between the maximum and minimum values; the physical meaning is to scale the data to the mean. Intervals, eliminating the influence of dimensions.

[0046] Furthermore, The principle is expressed as the Raida criterion. It first assumes that a set of test data contains only random errors, calculates and processes the data to obtain the standard deviation, determines an interval with a certain probability, and considers that any error exceeding this interval is not a random error but a gross error, and data containing such errors should be discarded.

[0047] It should be noted that the weighted fusion analysis is based on the weighted average formula, which is calculated as follows: the fusion value in the standardized dataset is equal to the sum of the cleaned data values ​​of each data source multiplied by the corresponding preset weight coefficients. The weight coefficients are based on the preset accuracy of the data sources, and the sum of all weight coefficients is 1. The physical meaning is to reflect the reliability of the data sources through weights and improve the accuracy of fusion.

[0048] Furthermore, based on historical data collection records, the accuracy indicators of each data source are statistically analyzed, including data collection error rate, data integrity, and sampling frequency. According to these accuracy indicators, the initial weight coefficients of each data source are calculated using a weight allocation formula: data collection error rate multiplied by 1 minus the weight factor corresponding to the data collection error rate, data integrity multiplied by the weight factor corresponding to data integrity, and sampling frequency multiplied by the weight factor corresponding to the sampling frequency. Based on real-time data quality monitoring results, the initial weight coefficients are dynamically adjusted to obtain the final preset weight coefficients.

[0049] It should be noted that the weight factors corresponding to the data acquisition error rate, data integrity, and sampling frequency are obtained through factor analysis. First, the information of acquisition error rate, data integrity, and sampling frequency is condensed, and then the variance explained after rotation is obtained. The weights are obtained by dividing the cumulative variance explained.

[0050] It should be noted that factor analysis is a well-known technique. It is a multivariate statistical analysis method that starts by studying the internal dependencies of variables and reduces some variables with complex relationships to a few comprehensive factors. Information condensation is expressed as the calculation of the median. The variance explained rate is the amount of information extracted by the factors. Variance explained rate = eigenvalues ​​ / total number of analysis terms. The rotated variance explained rate is expressed as the variance explained by the factors after maximum variance rotation.

[0051] Step 2: Digital Twin Construction: Based on the standardized dataset, the preset GIS map, and hyperspectral image data, a digital twin of the land parcel is constructed using a spatial interpolation algorithm to form a high-fidelity virtual farmland simulation environment.

[0052] It should be noted that hyperspectral image data is a type of spectral image data containing hundreds of narrow bands, which can capture detailed reflectance characteristics of ground objects at different wavelengths for fine classification and analysis, such as vegetation health monitoring and soil property identification. Hyperspectral image data is acquired through remote sensing platforms, such as satellite sensors or hyperspectral cameras mounted on drones, to provide high-resolution spectral information.

[0053] In a specific example, the construction of a digital twin of a land parcel using a spatial interpolation algorithm to form a high-fidelity virtual farmland simulation environment includes: performing data alignment and coordinate unification processing based on the standardized dataset, a preset GIS map, and hyperspectral image data to generate an integrated spatial dataset; performing interpolation calculations on missing or low-resolution areas using the Kriging spatial interpolation method based on the integrated spatial dataset to generate a high-resolution spatial distribution map; and generating a digital twin of the land parcel using 3D modeling technology based on the high-resolution spatial distribution map and the preset GIS map, including real-time soil data, straw data, and environmental data to form a high-fidelity virtual farmland simulation environment.

[0054] It should be noted that data alignment involves matching soil, straw, and environmental parameters in the standardized dataset with the spatial coordinates of the GIS map and the pixel coordinates of the hyperspectral imagery, ensuring that all data are in the same coordinate system. Coordinate unification uses the WGS84 coordinate system, converting data from different sources into unified coordinates using a coordinate transformation formula: ,in Represented as original coordinates, Represented as the transformed unified coordinates, Represented as a coordinate transformation function, it signifies a linear transformation based on affine transformation; its physical meaning is to eliminate spatial bias in data by unifying the coordinate system, thereby ensuring the accuracy of subsequent interpolation.

[0055] Furthermore, the specific implementation of the coordinate transformation function includes translation, rotation, and scaling operations. Least square fitting is performed based on control points to minimize the transformation error, thereby completing the coordinate transformation and generating an integrated spatial dataset.

[0056] It should be noted that the WGS-84 coordinate system (World Geodetic System-1984 Coordinate System) is an internationally adopted geocentric coordinate system.

[0057] It should be noted that Kriging space interpolation is a statistically based optimal unbiased estimation method; it is calculated using the formula... Determine the prediction point The interpolation results, where Represented as the prediction point, Represented as each known point, Represented as a known point Weighting coefficients for prediction points, It is a known point The measured value, Represented as the number corresponding to the known point. , This represents the total number of known points.

[0058] Furthermore, the physical meaning of Kriging spatial interpolation is to predict unknown point values ​​by weighted averaging of known point values, while also considering spatial autocorrelation to improve interpolation accuracy.

[0059] It should be noted that interpolation results, such as soil nutrient values, represent simulated values ​​for that point in the digital twin; measured values ​​are extracted from a standardized dataset; and the weighting coefficients of known points to predicted points are obtained through factor analysis and satisfy the following conditions: To ensure unbiasedness.

[0060] It should be noted that the 3D modeling technology uses a voxel-based rendering method to convert the interpolated spatial distribution map into a 3D mesh model, where each voxel contains parameter values ​​(such as humidity and temperature), and simulates the farmland environment through a real-time rendering engine; the physical meaning is to create a virtual copy that is consistent with the real plot, supporting dynamic simulation and interaction.

[0061] Furthermore, the high-fidelity virtual farmland simulation environment includes a visualization interface that allows users to browse soil profiles at different depths and real-time environmental changes, implemented based on OpenGL or similar graphics libraries.

[0062] This application improves data quality and consistency through weighted fusion and standardization, avoiding the problem of data source error accumulation in traditional methods. This ensures the reliability of subsequent model inputs, thereby improving overall prediction accuracy. It guarantees the authenticity and reference value of the analysis results, providing a reliable basis for subsequent decision-making. Based on standardized datasets and hyperspectral data, a high-fidelity virtual farmland simulation environment is constructed through spatial interpolation algorithms; it achieves accurate digital mapping of the farmland environment, provides a high-resolution simulation environment, and enhances the model's realism and predictive capabilities. It supports comprehensive and objective analysis, which is helpful for subsequent model embedding and optimization, reducing the cost of field experiments.

[0063] Step 3: Dynamic decomposition model embedding: Embed a preset straw soil microbial dynamic decomposition model into the plot digital twin, and input basic data to simulate the straw decomposition rate and nutrient release, so as to output the decomposition rate prediction and nutrient release prediction.

[0064] In a specific example, the step of embedding a preset straw soil microbial dynamic decomposition model into the plot digital twin and inputting basic data to simulate the straw decomposition rate and nutrient release, so as to output the decomposition rate prediction and nutrient release prediction, includes: A1. Initializing model parameters. Based on the plot digital twin, loading the preset model parameters of the straw soil microbial dynamic decomposition model, wherein the model parameters include microbial activity coefficient, straw carbon-nitrogen ratio threshold, temperature sensitivity coefficient and humidity influence factor.

[0065] A2. Input basic data: Extract basic data parameters from the standardized dataset, including soil temperature, soil moisture, soil pH, straw carbon-nitrogen ratio, and microbial biomass, and input the basic data parameters into the dynamic decomposition model.

[0066] A3. Simulate decomposition rate and nutrient release: Based on the dynamic decomposition model of straw and soil microorganisms, calculate the straw decomposition rate using a decomposition rate algorithm that integrates temperature, humidity, and microbial activity parameters; based on the decomposition rate prediction, calculate the nutrient release prediction using a nutrient release algorithm that correlates straw nutrient content and decomposition rate.

[0067] A4. Prediction of output decomposition rate and nutrient release.

[0068] In a specific example, calculating the straw decomposition rate using the decomposition rate algorithm includes: using the calculation formula... The straw decomposition rate was obtained ,in It is expressed as the baseline decomposition rate constant, which is preset based on straw type and soil properties; Represented as a temperature effect function, Represented as a humidity effect function, It is represented as a microbial activity function.

[0069] It should be noted that the straw decomposition rate is expressed as the percentage reduction in straw mass per unit time. The baseline decomposition rate constant is preset based on straw type and soil properties, including: extracting the carbon-nitrogen ratio of corn straw from historical records, analyzing straw decomposition rate data under various soil properties; various soil properties such as soil pH and organic matter content; soil properties are classified into multiple categories or ranges based on soil pH and organic matter content; for example, soil properties are classified as acidic, neutral, and alkaline according to pH, and low, medium, and high according to organic matter content. The average historical decomposition rate is calculated by combining the carbon-nitrogen ratio of corn straw and soil properties as the baseline decomposition rate constant; for example, if the carbon-nitrogen ratio of corn straw is approximately 25, and the neutral soil has a pH of 6.5-7.5 and an organic matter content of 3%, the baseline decomposition rate constant may be preset to 0.05% / day. The accuracy of the baseline decomposition rate constant is verified through controlled experiments, such as measuring the decomposition rate under specific conditions and comparing it with the preset value, adjusting the baseline decomposition rate constant to ensure reliability.

[0070] It should be noted that the temperature effect function is dimensionless and reflects the modulation of the decomposition rate by temperature. This includes: temperature effect function. middle, Expressed as activation energy parameter, Indicated as reference temperature, The input soil temperature, It is expressed as a natural constant.

[0071] Furthermore, the activation energy parameter is preset based on the metabolic characteristics of microorganisms. The specific process is as follows: extract the metabolic rate data of each type of microorganism (such as bacteria and fungi) at a specific temperature from the database; calculate the activation energy parameter by linear regression based on the metabolic rate data of each type of microorganism at a specific temperature; finally, preset the categorized parameter according to the microbial classification (such as aerobic bacteria and anaerobic bacteria, etc.) (e.g., the activation energy parameter of mesophilic aerobic bacteria is preset to 50 kJ / mol), and dynamically adjust the parameter value based on real-time microbial biomass data.

[0072] It should be noted that the humidity effect function is dimensionless and reflects the modulation of the decomposition rate by humidity. This includes: the humidity effect function. middle This represents the input soil moisture. This is a reference value for soil moisture, based on a preset soil type.

[0073] Furthermore, the reference value for soil moisture is based on the preset soil type and is expressed as the optimal moisture for the corresponding soil type; the physical meaning of the moisture influence function is to simulate the nonlinear effect of moisture on microbial activity, where excessively high or low moisture inhibits decomposition.

[0074] It should be noted that the microbial activity function is dimensionless and reflects the modulation of the decomposition rate by microbial biomass. This includes: microbial activity function. middle, This is expressed as the input microbial biomass. This represents the maximum value of microbial biomass.

[0075] Furthermore, the maximum value of microbial biomass is based on the preset soil health status, representing the allowable value of microbial biomass in the soil while maintaining a healthy condition; the physical meaning of the microbial activity function is that the number of microorganisms directly affects the decomposition rate.

[0076] In a specific example, the step of calculating the nutrient release prediction based on the decomposition rate prediction using a nutrient release algorithm includes: calculating using a formula... The amount of nitrogen released was determined. Phosphorus nutrient release and potassium nutrient release , , and These are represented as the nitrogen, phosphorus, and potassium nutrient content of the straw, respectively. This represents the time period corresponding to nutrient release.

[0077] It should be noted that nutrient release amount represents the amount of nutrients released within the corresponding time period; the nitrogen, phosphorus, and potassium nutrient contents of straw are extracted from straw data; the physical meaning is that nutrient release amount is directly proportional to the decomposition rate and nutrient content, simulating the release process accumulated over time.

[0078] Step 4: Optimize decision generation: Based on the predicted decomposition rate and nutrient release, construct the state space to integrate nutrient status, and then learn the optimal strategy to generate optimized fertilization decisions.

[0079] In a specific example, the construction of a state space based on the predicted decomposition rate and nutrient release to integrate nutrient status, and then learning the optimal strategy to generate optimized fertilization decisions, includes: B1. Constructing a state space based on the predicted decomposition rate and nutrient release, wherein the state variables include real-time soil nutrient content and predicted nutrient release to integrate nutrient status.

[0080] It should be noted that the construction of the state space includes real-time soil nutrient content and predicted nutrient release in the state variables, such as soil nitrogen content, soil phosphorus content, soil potassium content, predicted nitrogen release, predicted phosphorus release, and predicted potassium release. These variables are updated in real time based on the predicted decomposition rate and nutrient release. The physical meaning is to comprehensively characterize the dynamics of farmland nutrients by integrating prediction and real-time data, providing environmental state input for reinforcement learning.

[0081] Furthermore, the update frequency of the state variables is synchronized with the data acquisition to ensure that the state space reflects the latest nutrient status and is displayed in real time through the digital twin visualization module.

[0082] B2. Define an action space, where actions include fertilizer type selection, fertilizer amount setting, and fertilizer time period determination, as fertilization decision options; design a reward function, which is optimized based on nutrient utilization rate, economic benefits, and environmental benefits to calculate the reward value of the simulation; use a reinforcement learning algorithm to perform simulations in the digital twin of the land parcel based on the state space, action space, and reward function to learn the optimal fertilization strategy.

[0083] It should be noted that the defined action space includes actions such as fertilizer type selection, fertilizer amount setting, and fertilizer time period determination represented by growth stage; fertilizer type selection includes nitrogen fertilizer, phosphorus fertilizer, potassium fertilizer, or compound fertilizer, fertilizer amount setting is in kilograms per hectare, and fertilizer time period determination is represented by growth stage; the physical meaning of defining the action space is to convert the fertilization operation into continuous decision options for reinforcement learning algorithms to explore and optimize.

[0084] Furthermore, the design of the action space is based on agronomic knowledge and historical data to ensure the feasibility and practicality of decision-making, and to prevent over-fertilization through constraints.

[0085] It should be noted that the reward function is expressed as follows: ,in Represented as in state Take action below The reward value, Represented as a nutrient utilization rate function, This can be represented as an economic return function. Represented as an environmental benefit function, , and These are respectively represented as the weighting factors corresponding to the nutrient utilization rate function, the economic benefit function, and the environmental benefit function; the physical meaning of the reward function is to guide the reinforcement learning algorithm towards optimizing decisions with high nutrient utilization, high economic benefits, and low environmental impact by weighting and combining multiple objective indicators.

[0086] Furthermore, , , , We used factor analysis to obtain the weighting factors corresponding to the nutrient utilization rate function, the economic benefit function, and the environmental benefit function. First, we condensed the information of the nutrient utilization rate function, the economic benefit function, and the environmental benefit function, and obtained the weights by summing the variance explained. The nutrient utilization rate function is calculated as the ratio of the amount of nutrients absorbed by the crop to the amount of nutrients applied by fertilizer. The economic benefit function is calculated as the crop yield value minus the fertilizer cost. The environmental benefit function is calculated as a negative environmental pollution index (such as nitrogen and phosphorus loss).

[0087] It should be noted that the calculation of the simulation reward value uses the Q-learning algorithm for simulation, and the update formula is as follows: ,in Represented as a state-action value function, Represented as the learning rate, Represented as a discount factor, Represented as the next state; its physical meaning is achieved through iterative updates. The algorithm learns the long-term expected reward of taking an action in a given state, and eventually converges to obtain the optimal policy.

[0088] Furthermore, the simulation is conducted in the digital twin of the land parcel, utilizing its high-fidelity environment to simulate state transitions and reward calculations. The learning rate and discount factor are preset based on historical data, such as a learning rate of 0.1 and a discount factor of 0.9, and the algorithm's stability is ensured through multiple simulations.

[0089] B3. Based on the learned optimal fertilization strategy, generate optimized fertilization decisions, including fertilizer formula, fertilizer amount, and fertilization time period.

[0090] It should be noted that the optimal strategy learned based on the Q-learning algorithm selects the action combination with the highest reward value, and the output includes fertilizer formula, fertilizer amount, and fertilization time period. The fertilizer formula is such as the ratio of nitrogen, phosphorus, and potassium, the fertilizer amount is a specific value, and the fertilization time period is a specific date. This achieves precise fertilization and optimizes fertilization decisions. The output of the control system guides the on-site fertilization operation and can be dynamically adjusted according to real-time data.

[0091] This application, by combining mechanistic models and data-driven approaches, accurately predicts nutrient dynamics, avoiding the idealized assumptions of traditional single models. It improves the accuracy and adaptability of the model, ensuring the scientific nature of fertilization decisions and reducing waste and pollution caused by indiscriminate fertilization. It achieves closed-loop optimization and real-time decision-making for complex agronomic processes, improving the system's intelligence and efficiency, and allowing for timely adjustments to fertilization strategies, thereby increasing fertilizer utilization and crop yield while reducing environmental impact.

[0092] Step 5: Fertilizer prescription output: Based on the optimized fertilization decision, calculate the amount of fertilizer and generate a precise fertilization prescription, which is then output to the display terminal.

[0093] In a specific example, the step of calculating the fertilization amount and generating a precise fertilization prescription based on the optimized fertilization decision, and then outputting it to the display terminal, includes: extracting the target nutrient amount and crop nutrient requirement parameters based on the optimized fertilization decision; calculating the fertilization amount by combining the existing soil nutrient amount and the nutrient amount released by straw to obtain the actual fertilization amount; determining the fertilizer formula based on the actual fertilization amount and the fertilizer database; determining the fertilization time period based on the crop growth model and real-time environmental data; and finally integrating and generating a precise fertilization prescription to be output to the display terminal.

[0094] It should be noted that the specific process for calculating the actual fertilizer application rate is as follows: The actual fertilizer application rate equals the crop nutrient requirement parameter minus the existing soil nutrients minus the nutrients released from straw, divided by the fertilizer utilization rate. The unit of the actual fertilizer application rate is kilograms per hectare, representing the mass of fertilizer to be applied. Its physical meaning is the amount of fertilizer calculated based on the nutrient balance principle to ensure the crop's growth needs. The unit of the crop nutrient requirement parameter is kilograms per hectare, representing the total amount of nutrients required by the crop throughout its growth cycle, obtained from a preset database based on crop type and expected yield. The unit of the existing soil nutrients is kilograms per hectare, representing the amount of nutrients currently available in the soil, extracted from the standardized dataset. The nutrient release from straw represents the amount of nutrients released from straw decomposition, with the unit being kilograms per hectare, obtained from nutrient release prediction. The fertilizer utilization rate is a dimensionless value, preset based on soil properties and fertilizer type, ranging from 0 to 1, and its physical meaning reflects the actual absorption efficiency of fertilizer in the soil.

[0095] Furthermore, crop nutrient requirements parameters are obtained by querying a crop nutrient requirement table, which is constructed based on historical experimental data and includes nitrogen, phosphorus, and potassium requirements for different crop types (such as corn and wheat) at different yield levels. Existing soil nutrients are determined through soil sampling and analysis, including nitrogen, phosphorus, and potassium content, and stored in the standardized dataset. The nutrient release from straw is predicted based on the nutrient release output of the straw-soil microbial dynamic decomposition model, simulating the dynamic release of nutrients during straw decomposition. Fertilizer utilization rate is calibrated through field trials, taking into account the effects of soil pH, organic matter content, and fertilizer type (such as urea and compound fertilizer).

[0096] It should be noted that the fertilizer formula is determined based on the actual amount of fertilizer applied and the nutrient ratio. The optimal fertilizer combination is matched by querying a preset fertilizer database. The fertilizer database contains the nutrient content (such as the percentage of nitrogen, phosphorus and potassium) and applicable conditions (such as soil type and pH range) of various fertilizers.

[0097] Furthermore, the fertilizer database is constructed through laboratory analysis and field trials, including fertilizer names, nutrient composition, solubility, and environmental adaptability. The matching process employs a minimum cost algorithm or a nutrient balance algorithm to ensure both economy and effectiveness.

[0098] It should be noted that the determination of the fertilization time period is based on crop growth models and real-time environmental data. The growth model predicts the peak of crop nutrient absorption, and the appropriate fertilization time is selected in combination with weather forecasts to avoid rain erosion or high temperature evaporation.

[0099] Furthermore, the crop growth model adopts a physiological development stage model, and real-time environmental data, including temperature, humidity, and rainfall forecasts, are acquired from weather stations in real time and integrated into decision-making through data fusion technology.

[0100] This application calculates fertilization amount based on optimized decision-making, transforms decisions into actionable prescriptions, provides variable fertilization recommendations and optimal window periods, and realizes the precision and automation of fertilization. It generates accurate prescriptions and outputs them to the display terminal, ensuring the pertinence and balance of field operations, reducing human intervention, and improving the efficiency and sustainability of agricultural resource utilization.

[0101] Please see Figure 2 As shown, this application provides a system for a method of returning corn stalks to the field for fertilization in a second aspect.

[0102] This application provides a method and system for returning corn straw to the field for fertilization. Based on multi-source data fusion and spatial interpolation algorithms, it constructs a high-fidelity virtual farmland environment, overcoming the limitations of traditional methods in failing to adequately characterize field heterogeneity. By embedding a dynamic decomposition model of straw and soil microorganisms, it achieves quantitative prediction of straw decomposition rate and nutrient release, providing a scientific basis for fertilization timing and dosage. Combining state space construction and strategy learning algorithms, it realizes dynamic matching of nutrient supply and demand and adaptive optimization of fertilization strategies, significantly improving fertilizer utilization efficiency, reducing environmental pollution risks, generating precise fertilization prescriptions and outputting them to a display terminal, thereby enhancing the intelligence level and ease of operation of agricultural production.

[0103] The system 100 for corn stalk return fertilization according to the present invention can be installed in an electronic device. Depending on the functions implemented, the system 100 may include a basic data preprocessing module 101, a digital twin construction module 102, a dynamic decomposition model embedding module 103, an optimization decision generation module 104, and a fertilization prescription output module 105. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0104] In this embodiment, the functions of each module / unit are as follows: The basic data preprocessing module performs data cleaning and weighted fusion analysis based on the pre-acquired basic data to generate a standardized dataset.

[0105] The digital twin construction module, based on the standardized dataset, preset geographic information system (GIS) map, and hyperspectral image data, constructs digital twins of land parcels through spatial interpolation algorithms to form a high-fidelity virtual farmland simulation environment.

[0106] The dynamic decomposition model embedding module is used to embed a preset straw soil microbial dynamic decomposition model into the plot digital twin, and input basic data to simulate the straw decomposition rate and nutrient release, so as to output the decomposition rate prediction and nutrient release prediction.

[0107] The optimization decision generation module constructs a state space based on the predicted decomposition rate and nutrient release to integrate nutrient status, thereby learning the optimal strategy and generating optimized fertilization decisions.

[0108] The fertilizer prescription output module calculates the amount of fertilizer and generates a precise fertilizer prescription based on the optimized fertilization decision, and then outputs it to the display terminal.

[0109] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0110] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0111] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0112] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0113] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for returning corn stalks to the field for fertilization, characterized in that, include: Step 1: Basic Data Preprocessing: Based on the pre-acquired basic data, perform data cleaning and weighted fusion analysis to generate a standardized dataset; Step 2: Digital Twin Construction: Based on the standardized dataset, the preset GIS map, and hyperspectral image data, a digital twin of the land parcel is constructed using a spatial interpolation algorithm to form a high-fidelity virtual farmland simulation environment. Step 3: Dynamic decomposition model embedding: Embed the preset straw soil microbial dynamic decomposition model into the plot digital twin, and input basic data to simulate the straw decomposition rate and nutrient release, so as to output the decomposition rate prediction and nutrient release prediction. The process of embedding a pre-defined dynamic decomposition model of straw and soil microorganisms into a digital twin of the land plot, and inputting basic data to simulate the straw decomposition rate and nutrient release, to output a decomposition rate prediction and a nutrient release prediction, includes: A1. Initialize the model parameters. Based on the plot digital twin, load the preset model parameters of the straw soil microbial dynamic decomposition model. The model parameters include microbial activity coefficient, straw carbon-nitrogen ratio threshold, temperature sensitivity coefficient and humidity influence factor. A2. Input basic data, extract basic data parameters from the standardized dataset, including soil temperature, soil moisture, soil pH, straw carbon-nitrogen ratio and microbial biomass, and input the basic data parameters into the dynamic decomposition model; A3. Simulate decomposition rate and nutrient release: Based on the dynamic decomposition model of straw and soil microorganisms, calculate the straw decomposition rate using a decomposition rate algorithm that integrates temperature, humidity, and microbial activity parameters; based on the decomposition rate prediction, calculate the nutrient release prediction using a nutrient release algorithm that correlates straw nutrient content and decomposition rate. A4. Prediction of output decomposition rate and nutrient release; Step 4: Optimize decision generation: Based on the predicted decomposition rate and nutrient release, construct the state space to integrate nutrient status, and then learn the optimal strategy to generate optimized fertilization decisions. The process of constructing a state space based on the predicted decomposition rate and nutrient release to integrate nutrient status, and then learning an optimal strategy to generate optimized fertilization decisions includes: B1. Based on the predicted decomposition rate and nutrient release, a state space is constructed, wherein the state variables include real-time soil nutrient content and predicted nutrient release, in order to integrate nutrient status. B2. Define an action space, where actions include fertilizer type selection, fertilizer amount setting, and fertilizer time period determination, as fertilization decision options; design a reward function, which is optimized based on nutrient utilization rate, economic benefits, and environmental benefits to calculate the reward value of the simulation; use a reinforcement learning algorithm to perform simulation on the digital twin of the land parcel based on the state space, action space, and reward function to learn the optimal fertilization strategy. B3. Based on the learned optimal fertilization strategy, generate optimized fertilization decisions, including fertilizer formula, fertilizer amount, and fertilization time period; Step 5: Fertilizer prescription output: Based on the optimized fertilization decision, calculate the amount of fertilizer and generate a precise fertilization prescription, which is then output to the display terminal.

2. The method for returning corn stalks to the field for fertilization according to claim 1, characterized in that, The basic data includes soil data, straw data, and environmental data.

3. The method for returning corn stalks to the field for fertilization according to claim 1, characterized in that, The process involves data cleaning and weighted fusion analysis based on pre-acquired basic data to generate a standardized dataset, including: Based on soil data, straw data, and environmental data, data cleaning operations are performed to obtain a cleaned dataset. The data cleaning operations include missing value handling, outlier removal, and data standardization. Based on the preset weight coefficients of the data source accuracy, a weighted fusion analysis is performed on the cleaned dataset to generate a standardized dataset. The weighted fusion analysis is performed based on a weighted average formula.

4. The method for returning corn stalks to the field for fertilization according to claim 1, characterized in that, The construction of digital twins of land parcels using spatial interpolation algorithms to form a high-fidelity virtual farmland simulation environment includes: Based on the standardized dataset, the preset GIS map, and hyperspectral imagery data, data alignment and coordinate unification are performed to generate an integrated spatial dataset. Based on the integrated spatial dataset, Kriging spatial interpolation is used to interpolate missing or low-resolution areas to generate a high-resolution spatial distribution map. Based on the high-resolution spatial distribution map and the preset GIS map, digital twins of the land parcels are generated using 3D modeling technology, including real-time soil data, straw data, and environmental data, to form a high-fidelity virtual farmland simulation environment.

5. The method for returning corn stalks to the field for fertilization according to claim 1, characterized in that, The calculation of straw decomposition rate using the decomposition rate algorithm includes: Through calculation formula The straw decomposition rate was obtained ,in It is expressed as the baseline decomposition rate constant, which is preset based on straw type and soil properties; Represented as a temperature effect function, Represented as a humidity effect function, It is represented as a microbial activity function.

6. The method for returning corn stalks to the field for fertilization according to claim 1, characterized in that, The method of predicting nutrient release based on decomposition rate prediction and calculating nutrient release amount using a nutrient release algorithm includes: Through calculation formula The amount of nitrogen released was determined. Phosphorus nutrient release and potassium nutrient release , , and These are represented as the nitrogen, phosphorus, and potassium nutrient content of the straw, respectively. This represents the time period corresponding to nutrient release.

7. The method for returning corn stalks to the field for fertilization according to claim 1, characterized in that, The process of calculating the fertilizer application rate and generating a precise fertilizer prescription based on the optimized fertilization decision, and then outputting it to the display terminal, includes: Based on the optimized fertilization decision, the target nutrient content and crop nutrient requirement parameters are extracted. Combined with the existing soil nutrient content and the nutrient content released by straw, the fertilization amount is calculated to obtain the actual fertilization amount. Then, based on the actual fertilization amount and the fertilizer database, the fertilizer formula is determined. At the same time, based on the crop growth model and real-time environmental data, the fertilization time period is determined. Finally, the precise fertilization prescription is generated and output to the display terminal.

8. A system for implementing the corn stalk return-to-field fertilization method according to any one of claims 1-7, characterized in that, include: The basic data preprocessing module performs data cleaning and weighted fusion analysis based on the pre-acquired basic data to generate a standardized dataset; The digital twin construction module, based on the standardized dataset, the preset geographic information system GIS map and hyperspectral image data, constructs a digital twin of the land parcel through a spatial interpolation algorithm to form a high-fidelity virtual farmland simulation environment; The dynamic decomposition model embedding module is used to embed a preset straw soil microbial dynamic decomposition model into the plot digital twin, and input basic data to simulate the straw decomposition rate and nutrient release, so as to output the decomposition rate prediction and nutrient release prediction. The optimization decision generation module constructs a state space based on the decomposition rate prediction and nutrient release prediction to integrate nutrient status, and then learns the optimal strategy to generate optimized fertilization decisions. The fertilizer prescription output module calculates the amount of fertilizer and generates a precise fertilizer prescription based on the optimized fertilization decision, and then outputs it to the display terminal.

Images