Agricultural sowing bulk blending fertilizer optimization method based on multi-source data

Abstract

The invention discloses an agricultural sowing blended fertilizer optimization method based on multi-source data, and belongs to the technical field of blended fertilizer optimization, and the method specifically comprises the steps: carrying out the denoising processing of a multi-source basic data set, removing abnormal data points, and carrying out the data standardization conversion; establishing a dynamic association map of soil nutrients and crop demands, and determining key influence weights of the nutrients on crop growth through data association analysis; determining a basic blending proportion of nitrogen, phosphorus, potassium and trace elements in the blended fertilizer by combining crop variety characteristics according to a weight distribution result of the correlation map; carrying out regionalization correction on the basic mixing proportion by referring to local climate characteristic data and aiming at the terrain difference and the moisture distribution condition of a seeding region; and in combination with the corrected proportion parameters, generating a customized blended fertilizer scheme adapted to a specific sowing scene, and determining the accurate proportion and use mode of each component.

Description

Technical Field

[0001] This invention relates to the field of fertilizer blending optimization technology, specifically to a method for optimizing agricultural sowing and fertilizer blending based on multi-source data. Background Technology

[0002] In the process of refining agricultural production, optimizing the formulation of blended fertilizers is a key technical step in improving crop yield and quality and ensuring soil fertility. Currently, the commonly used techniques in this field mainly involve obtaining basic nutrient data such as nitrogen, phosphorus, potassium, and organic matter through soil sampling and testing. This data, combined with the conventional nutrient requirements of crop varieties and with reference to local agricultural technology extension standards or historical planting experience, determines the basic component ratios of the blended fertilizer.

[0003] However, existing methods for optimizing blended fertilizers often rely on a single data dimension or fixed empirical models, failing to establish a precise mapping relationship between multi-source data and the optimization objectives of blended fertilizers. They neglect the synergistic effects of multi-dimensional data such as dynamic changes in soil fertility, crop growth cycle requirements, and fluctuations in environmental factors. By designing formulations based solely on localized data, the nutrient ratios of blended fertilizers do not match the actual production needs of specific plots. This not only fails to meet the precise nutrient supply required for crop growth but also easily leads to resource waste or soil ecological imbalance, hindering the achievement of efficient and sustainable agricultural production. Summary of the Invention

[0004] The purpose of this invention is to provide an optimization method for agricultural sowing and fertilizer blending based on multi-source data, thereby solving the problems in the background technology: The objective of this invention can be achieved through the following technical solutions: An optimization method for agricultural sowing and fertilizer blending based on multi-source data includes the following steps: S1. Collect soil nutrient data, crop variety characteristic data and local climate characteristic data of the sowing area, and integrate them to form a multi-source basic dataset; S2. Denoise the multi-source basic dataset, remove outlier data points, and then perform data standardization transformation. S3. Establish a dynamic correlation map between soil nutrients and crop requirements, and clarify the key influence weights of each nutrient on crop growth through data correlation analysis; S4. Based on the weight allocation results of the correlation graph and combined with the characteristics of crop varieties, determine the basic blending ratio of nitrogen, phosphorus, potassium and trace elements in the blended fertilizer. S5. Based on local climate characteristics data, the basic mixing ratio is adjusted regionally according to the differences in terrain and water distribution in the sowing area. S6. Based on the corrected proportion parameters, generate a customized blended fertilizer solution that is suitable for specific sowing scenarios, and clarify the precise ratio and usage of each component.

[0005] As a further aspect of the present invention: in step S1, the soil nutrient data includes nitrogen content, phosphorus content, potassium content, and organic matter content; the crop variety characteristic data includes growth cycle, fertilizer requirement pattern, and stress resistance; and the local climate characteristic data includes annual precipitation, average annual temperature, and sunshine duration.

[0006] As a further aspect of the present invention: in step S2, the process of denoising the multi-source basic dataset, removing outlier data points, and then performing data standardization transformation is as follows: The moving average method was used to process the soil nutrient data, crop variety characteristic data, and local climate characteristic data in the multi-source basic dataset separately to smooth the random fluctuations of various types of data and achieve data denoising. Based on the denoised data, the inherent attribute range of the corresponding data is compared to identify data points that exceed the range. Data points that do not conform to the attribute range are judged as abnormal data points and removed. For the multi-source basic dataset after removing outlier data points, the range standardization method is used to transform different types of data into fixed numerical ranges, thus completing the data standardization transformation.

[0007] As a further aspect of the present invention: in step S3, the process of establishing a dynamic correlation map between soil nutrients and crop requirements, and clarifying the key influence weights of each nutrient on crop growth through data correlation analysis, is as follows: The crop growth cycle is divided into four stages: germination, growth, maturity, and fruiting. Specific nutrient requirements corresponding to the nutrient requirements of each stage are extracted to clarify the soil nutrient requirements of crops at different stages. The standardized soil nutrient data are matched one by one with the crop demand indicators at each growth stage to construct a staged nutrient-demand correlation matrix, forming a basic framework for the dynamic correlation between soil nutrients and crop demand. Partial correlation analysis was performed on the correlation matrix, while keeping other nutrient data constant. The correlation strength between each soil nutrient and the corresponding crop requirement index at the same stage was calculated separately to eliminate mutual interference between nutrients. Based on the correlation strength obtained from partial correlation analysis, corresponding weights are assigned to each soil nutrient at different growth stages. The weights of each stage are integrated to form a weight system of key influences of nutrients on crop growth, thus improving the dynamic correlation map.

[0008] As a further aspect of the present invention: the specific method for performing partial correlation analysis on the correlation matrix, keeping other nutrient data constant, and calculating the correlation strength between each soil nutrient and the corresponding crop requirement index at the corresponding stage is as follows: For a specific soil nutrient in the correlation matrix, locate its corresponding data column with the crop requirement index at each growth stage, separate the data columns of other soil nutrients in the matrix and keep their values ​​constant; Using other soil nutrient data that remain constant after separation as control variables, and using target soil nutrient data and corresponding crop demand indicators as analytical variables, a partial correlation analysis model is constructed. By running the constructed partial correlation analysis model and keeping the values ​​of control variables constant, the correlation between the target soil nutrients and the corresponding crop demand indicators at the same stage is calculated, and the correlation strength of the nutrient is obtained.

[0009] As a further aspect of the present invention: in step S4, the process of determining the basic blending ratio of nitrogen, phosphorus, potassium, and trace elements in the blended fertilizer based on the weight allocation results of the correlation graph and in combination with the characteristics of the crop variety is as follows: Extract the key influence weights of each nutrient in the correlation map, and combine the growth cycle and nutrient requirement patterns of crop varieties to clarify the weight priority of key nutrients at different growth stages; Based on crop stress resistance, we analyze the synergistic enhancement or inhibition relationships among various nutrients, adjust the weight ratio of corresponding nutrients, and avoid nutrient imbalance caused by simply allocating nutrients according to their weight. The adjusted nutrient weights are matched one by one with the fertilizer requirements of crops at each growth stage, and the basic proportion range of nitrogen, phosphorus, potassium and micronutrients in the whole growth cycle is defined. By comprehensively considering the basic nutrient ratios at each growth stage, eliminating conflicting ranges, and clarifying the basic mixing ratios of nitrogen, phosphorus, potassium, and micronutrients, we can ensure that the crop varieties are adapted to their overall growth needs.

[0010] As a further aspect of the present invention: in step S5, the process of regionally adjusting the basic blending ratio based on local climate characteristic data and considering the topographical differences and moisture distribution of the sowing area is as follows: The correlation between local annual precipitation, average annual temperature, and sunshine duration and the field retention efficiency of nitrogen, phosphorus, potassium, and trace elements was established to clarify the direction of the influence of climate factors on the function of each nutrient. The planting area is divided into several modified units with uniform topographic and water distribution characteristics by overlaying and partitioning the area based on the terrain type and water distribution density. Based on the direction of climate influence in each correction unit, and combined with the effect of topography on nutrient retention and the effect of water on nutrient dissolution, the component proportion of the basic mixing ratio in the corresponding unit is adjusted. Based on the nutrient migration characteristics of adjacent correction units, cross-unit compensation correction is carried out to eliminate the nutrient imbalance problem at the partition boundary and integrate to form a regionalized correction blending ratio.

[0011] As a further aspect of the present invention: In step S6, the process of generating a customized blended fertilizer scheme adapted to specific sowing scenarios by combining the corrected proportion parameters, and clarifying the precise ratio and usage of each component, is as follows: The revised proportional parameters are matched one by one with the sowing method and cultivation cycle elements of the sowing scenario, clarifying the adaptation conditions of each parameter in the specific scenario, and laying the foundation for scenario matching for solution generation. Based on the compatibility conditions, the mixing order is determined according to the physical properties of nitrogen, phosphorus, potassium and trace elements. The corrected proportions are then converted into precise ratios of each component to ensure that the mixing logic is compatible with the characteristics of the components. By integrating precise proportions and mixing sequences, and clarifying the application time, dosage per acre, and application method (broadcast or furrow) for each component, customized blended fertilizer solutions can be formed to suit specific sowing scenarios.

[0012] The beneficial effects of this invention are: This invention integrates multi-dimensional data on soil nutrients, crop variety characteristics, and local climate features. Through standardized data purification and processing, and combined with the crop growth cycle, it establishes a dynamic correlation map between soil nutrients and crop requirements. This precisely quantifies the key influence weights of each nutrient at different growth stages, allowing the design of blended fertilizer formulations to break free from the limitations of single data or fixed empirical models. This formulation not only fully adapts to the different nutrient requirements at each stage of crop germination, growth, maturity, and fruiting, but also closely matches the actual soil fertility conditions of the planting area. It solves the problem of mismatch between nutrient ratios and actual production needs from the source, ensuring that crops receive precise nutrient supply at each growth stage, significantly improving nutrient absorption and utilization rates, and contributing to simultaneous increases in crop yield and quality.

[0013] This invention leverages local climate characteristics and regionalizes proportions based on differences in terrain and water distribution in planting areas. Combined with specific planting methods and cultivation cycles, it generates customized solutions, effectively reducing redundant nutrient input. This avoids resource waste caused by excessive nutrient loss and reduces the damage to soil structure caused by the accumulation of single nutrients, mitigating the risk of soil ecological imbalance. Simultaneously, the customized application method further optimizes nutrient efficiency, achieving intensive utilization of nutrient resources and sustainable maintenance of the soil ecosystem while ensuring high agricultural output. Attached Figure Description

[0014] The invention will now be further described with reference to the accompanying drawings.

[0015] Figure 1 This is a flowchart illustrating an agricultural sowing and fertilizer blending optimization method based on multi-source data, according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1 As shown, this invention is a method for optimizing agricultural sowing and fertilizer blending based on multi-source data, comprising the following steps: S1. Collect soil nutrient data, crop variety characteristic data and local climate characteristic data of the sowing area, and integrate them to form a multi-source basic dataset; S2. Denoise the multi-source basic dataset, remove outlier data points, and then perform data standardization transformation. S3. Establish a dynamic correlation map between soil nutrients and crop requirements, and clarify the key influence weights of each nutrient on crop growth through data correlation analysis; S4. Based on the weight allocation results of the correlation graph and combined with the characteristics of crop varieties, determine the basic blending ratio of nitrogen, phosphorus, potassium and trace elements in the blended fertilizer. S5. Based on local climate characteristics data, the basic mixing ratio is adjusted regionally according to the differences in terrain and water distribution in the sowing area. S6. Based on the corrected proportion parameters, generate a customized blended fertilizer solution that is suitable for specific sowing scenarios, and clarify the precise ratio and usage of each component.

[0018] In one embodiment of the present invention, the process of collecting soil nutrient data, crop variety characteristic data, and local climate characteristic data of the sowing area and integrating them to form a multi-source basic dataset in step S1 is as follows: Soil nutrient data, crop variety characteristics, and local climate data were collected from the planting area and integrated into a multi-source basic dataset. Soil nutrients included nitrogen, phosphorus, potassium, and organic matter content. Sampling was designed according to plot heterogeneity and samples were sent for testing to obtain analytical results. When obtaining soil nutrient data through sampling, it should be noted that the representativeness of the sampling depends on the distribution and depth of the sampling points, and spatial coverage and repeatability should be considered to reduce errors. Crop variety characteristics, including growth cycle, nutrient requirements, and stress resistance, were derived from variety data and field records. Climate characteristics, including annual precipitation, average annual temperature, and sunshine duration, were derived from meteorological observations and yearbooks. All data were paired with spatial coordinates and time labels to provide structured input for analysis.

[0019] In one embodiment of the present invention, step S2, which involves denoising the multi-source basic dataset, removing outlier data points, and then performing data standardization transformation, is as follows: For denoising of multi-source datasets, soil nutrient data, crop variety characteristic data, and local climate characteristic data were first processed using moving window smoothing to reduce measurement noise and short-term fluctuations from interfering with subsequent analysis. For data with clear time-series characteristics, such as meteorological observations and soil chemical parameters, a moving average was used to obtain a smoothed sequence by averaging several adjacent observations at each time point. For crop variety characteristics expressed as grades or categories, aggregation methods such as majority voting or proportional statistics within the window were used to obtain smoothed representative values, ensuring that the processing method matches the data attributes. It should be noted that moving window smoothing can weaken the impact of random short-term fluctuations, thus highlighting long-term or trend components. The window length should be selected based on the sampling frequency and the desired smoothing level, typically between three and seven observation points. A window that is too short will be difficult to denoise, while a window that is too long may obscure the true changes. Therefore, the window selection should be based on sampling density and crop growth characteristics, and adjusted empirically.

[0020] Based on the denoised data, outlier data points are identified and removed by comparing them with the inherent attribute ranges of the corresponding data. This process first establishes reasonable attribute boundaries for each type of indicator. These boundaries can be derived from literature reports, historical observational statistical distributions, theoretical lower and lower limits of detection methods, or expert experience summaries. Then, the denoised observations are compared with these attribute boundaries. Observations that significantly exceed the boundaries and cannot be reasonably explained are marked as outliers and removed from the analysis set. In practice, distributional statistical methods can also be used to identify outliers, such as calculating the interquartile range of the data and using its multiples as a criterion for anomaly judgment. By identifying and removing outliers by comparing with attribute ranges, the purified data is obtained. It should be noted that using distributional statistics such as the interquartile range multiplier can identify relative anomalies without relying on external thresholds. For example, 1.5 times the interquartile range can be used as the judgment threshold. In addition, when historical or literature boundaries exist, these boundaries should be given priority, and the reasons for anomaly removal and the original values ​​should be recorded simultaneously.

[0021] For the multi-source basic dataset after removing outlier data points, range standardization was used to transform different types of numerical indicators to a fixed numerical range. Specifically, for each continuous indicator, its minimum and maximum values ​​were determined, and the original values ​​were transformed into scaled values ​​within the target range using a linear mapping. Categorical or ordinal indicators were also normalized after appropriate coding to maintain dimensional consistency. Standardized data was obtained through range standardization. It should be noted that range mapping can map data of different dimensions to a uniform scale, thus facilitating comparison. However, this method is sensitive to extreme values. Therefore, before implementing the mapping, the selection method for the minimum and maximum values ​​should be confirmed, including using the minimum and maximum values ​​of the denoised samples or using historical quantiles as boundaries. At the same time, new observations exceeding the boundaries should be pruned or recalibrated, and the processing rules should be recorded to ensure repeatability.

[0022] In one embodiment of the present invention, step S3, which involves establishing a dynamic correlation map between soil nutrients and crop requirements, and clarifying the key influence weights of each nutrient on crop growth through data correlation analysis, is as follows: The crop growth cycle is divided into four stages: germination, growth, maturity, and fruiting. For each stage, corresponding nutrient requirements and specific indicators are extracted. The implementation steps first clarify the time window and indicator items for each stage. Based on the variety's growth period records, the number of days of growth or heat accumulation is divided into four stages, and typical nutrient requirements indicators such as nitrogen absorption rate, available phosphorus requirement, peak potassium absorption period, and micronutrient sensitivity period are listed for each stage, forming a staged nutrient requirement list. The nutrient requirements indicators are extracted from the growth stages and linked to the growth period. It should be noted that extracting nutrient requirements indicators by stage reflects the crop's nutritional preferences and critical demand points at different growth stages. This stage division reduces the bias of treating the entire growth period as a uniform requirement. Furthermore, the stage boundaries should be set according to the variety characteristics and local accumulated temperature to ensure the temporal consistency of the indicators.

[0023] Standardized soil nutrient data are mapped one-to-one with crop demand indicators at each growth stage to construct a staged nutrient-demand correlation matrix. In practice, the standardized nutrient vector of each spatial unit or sample point is first paired with the staged demand vector within the same spatial and temporal window. Then, the correspondence between each nutrient indicator and demand indicator at each stage is recorded in matrix form. Matrix rows represent nutrient types, matrix columns represent staged demand items, and matrix elements are the paired observation pairs or their statistical characteristics. The correlation matrix is ​​constructed through pairing. It should be noted that the principle of point-by-point or zone-by-zone pairing is to ensure spatial and temporal consistency. If the spatial sampling density does not match the stage time window, spatial resampling or intra-time window aggregation statistics should be used to ensure that each element in the matrix represents a sufficient number of observation samples.

[0024] Partial correlation analysis was performed on the correlation matrix, keeping other nutrient data constant, and the correlation strength between each soil nutrient and the corresponding crop requirement index at the same stage was calculated separately to eliminate interference between nutrients; the specific method is as follows: For the location and separation process of a certain soil nutrient in the correlation matrix, the implementation first locates the data column corresponding to the target nutrient in the staged nutrient-demand correlation matrix, and then filters the observation sequence corresponding to the nutrient by using spatial units and time windows as keys on the matrix index. Subsequently, the columns of other nutrients in the matrix are separated for use as control variables. By locating the corresponding columns, the observation sequences of the target nutrient and the demand of each growth stage are obtained. It should be noted that the location process depends on the consistency of the matrix row and column indexes and the spatial and temporal pairing rules. If the spatial density or temporal coverage of the sampling points is insufficient, spatial resampling or aggregation within the time window should be used to compensate. The requirement for the preset sample size should be based on ensuring statistical reliability. The preset method is to retain at least 10 valid pairs as the minimum example for each analysis unit. If this cannot be met, adjacent units are merged or the time window is extended to expand the sample size.

[0025] Using isolated and kept constant soil nutrient data as control variables, and target soil nutrient data and corresponding crop demand indicators as analytical variables, the specific process for constructing a partial correlation analysis model includes data consistency processing, covariance matrix calculation, regression residual calculation, and partial correlation coefficient estimation. Specifically, multiple regressions are first performed on the control variables and target variables respectively to obtain the fit residuals of the control variables to the target variable and the fit residuals of the control variables to the demand indicators. Then, the correlation coefficient between the two sets of residuals is calculated as the partial correlation coefficient. By separating and keeping the control variables constant, a partial correlation model is constructed to obtain the residual sequence used for analysis. It should be noted that this method statistically uses regression to control the influence of other variables to extract the independent association between target variables. Several judgment criteria and thresholds need to be set during model construction. The significance test threshold and the tolerance range for multicollinearity are determined based on the sample size and the number of variables. For example, the significance threshold is set to 0.05, and the variance inflation factor threshold for multicollinearity is set to 10. If the variance inflation factor exceeds the threshold, variable merging or regularized regression processing should be considered.

[0026] The partial correlation analysis model is run to determine the correlation between target soil nutrients and corresponding crop demand indicators at each stage, while keeping the values ​​of control variables constant. The implementation details involve performing regression control on each target nutrient and each stage's demand indicator and calculating the partial correlation coefficient. Simultaneously, the robustness of the results is assessed through significance testing, confidence interval estimation, and bootstrap resampling. The absolute value and direction of the obtained partial correlation coefficient are interpreted, and its statistical significance is recorded. Finally, this partial correlation coefficient is used as the correlation strength of the nutrient at that stage for subsequent weight allocation or decision-making reference. It should be noted that the correlation strength obtained by running the partial correlation model reflects the linear correlation between the target nutrient and the demand indicator under statistical control of other nutrient influences. This statistic may underestimate the true correlation when nonlinear relationships exist. Therefore, for nonlinear or abnormally distributed data, variable transformation or residual-based nonparametric correlation estimation should be performed before model running. Furthermore, regarding confidence interval estimation and robustness testing, bootstrap resampling 1000 times can be used to estimate the confidence interval; for example, a 95% confidence level can be used to assess the stability of the results.

[0027] Based on the correlation strength of each nutrient obtained from partial correlation analysis, corresponding weights are assigned to each soil nutrient at different growth stages. First, the correlation strength within each stage is positively mapped and normalized to obtain the weight allocation within the stage. The weights are set by dividing the absolute value of the partial correlation coefficient by the sum of the absolute partial correlation coefficients of all nutrients in that stage to obtain the initial ratio. Then, adjustments are made based on the synergistic or inhibitory relationship between crop stress resistance and nutrients. The adjustment rule is that if two nutrients have a synergistic effect, their respective weights are increased by the increase coefficient; if there is an antagonistic effect, the corresponding weights are decreased by the decrease coefficient. For example, the initial normalized weights for a single stage can be approximately 0.40 for nitrogen, 0.25 for phosphorus, 0.30 for potassium, and 0.05 for micronutrients. The adjustment coefficients are generally between 10% and 30% for synergistic or inhibitory correction. Finally, the weights of each stage are weighted and summarized according to the growth duration or importance of the stage to form a key influence weight system for the entire cycle, thereby improving the dynamic correlation map and providing a quantitative basis for allocation decisions.

[0028] In one embodiment of the present invention, step S4, which involves determining the basic blending ratio of nitrogen, phosphorus, potassium, and trace elements in the blended fertilizer based on the weight allocation results of the correlation graph and in combination with the characteristics of the crop variety, is as follows: The key influence weights of each nutrient in the correlation graph are extracted, and the stage priorities are determined by combining the crop variety's growth cycle and nutrient requirement patterns. First, the correlation strength values ​​of each nutrient in the four stages of germination, growth, maturity, and fruiting are read from the dynamic correlation graph. During implementation, the absolute values ​​of the coefficients are first taken, and then normalized within each stage to form the initial weight allocation. The normalization method is to sum the absolute correlation strengths of each nutrient within a certain stage and then divide the individual strength by the sum to obtain the ratio. The normalized absolute partial correlation coefficients are used as initial weight example values. For example, if the absolute partial correlation coefficients for a certain stage are nitrogen 0.40, phosphorus 0.25, potassium 0.30, and micronutrients 0.05, then the normalized weights are... The approximate weight ratios after scaling are nitrogen 0.40, phosphorus 0.25, potassium 0.30, and micronutrients 0.05. Subsequently, the importance of each stage is weighted based on the growth cycle length and nutrient requirements of the crop variety. The preset method for stage importance is declared as multiplying the stage duration percentage by the example value of the stage sensitivity coefficient. For example, the sensitivity coefficient for germination can be set to 0.8, growth to 1.2, and maturity to 1.0, resulting in a value of 0.6. These sensitivity coefficients are used to reflect the criticality of nutrients at different stages, thus obtaining a stage-based priority weight that reflects the characteristics of the variety and the nutrient requirements after weighting.

[0029] This study analyzes the synergistic or inhibitory relationships among nutrients based on crop stress resistance performance and adjusts the weight ratio of each nutrient to avoid nutrient imbalance caused by simply allocating them according to their initial weights. First, crop stress resistance performance is quantified using a 1-5 rating scale, where 1 represents weak and 5 represents strong. The rating is based on variety trial records or literature values; for example, a drought-resistant variety might be rated 4. Then, a nutrient interaction matrix is ​​established, derived from field trial results and literature review. Matrix elements are interaction coefficients, which are preset to continuous values ​​ranging from -0.3 to +0.3 to represent inhibition or synergy. For example, a nitrogen-potassium interaction coefficient of +0.15 indicates synergistic enhancement, while a phosphorus-micronutrient interaction coefficient of -0.12 indicates antagonistic effects. The interaction correction coefficients are adjusted by weighting based on crop stress resistance scores. The adjustment rule is to appropriately amplify synergistic gains or reduce antagonistic losses when the stress resistance score is high. Specifically, the adjustment thresholds are declared as follows: when the stress resistance score is ≥4, the synergistic coefficient is multiplied by 1.1 and the antagonistic coefficient is multiplied by 0.9; when the stress resistance score is ≤2, the synergistic coefficient is multiplied by 0.9 and the antagonistic coefficient is multiplied by 1.1. After adjustment, the initial weight of each nutrient is corrected item by item according to the weighted interaction effect. The range of the correction coefficient example value is declared as ±10% to ±30%. For example, the initial weight of nitrogen is 0.40. After synergy with potassium, it can be increased by 10% to 0.44. If there is antagonism with phosphorus, it may be decreased by 5% to 0.418. This ensures that the nutrient ratio takes into account both synergistic effects and stress resistance requirements.

[0030] The adjusted nutrient weights are matched one by one with the fertilizer requirements of crops at each growth stage, and the basic proportion range of nitrogen, phosphorus, potassium and trace elements in the whole growth cycle is determined accordingly. The implementation process involves first multiplying the corrected weight of each stage by the duration weight of that stage to obtain the stage contribution value. The duration weight is preset based on the proportion of that stage to the total number of days in the growth period. For example, if the growth stage accounts for 50%, the contribution value of that stage is calculated using the formula: contribution = stage weight × 0.5. Then, the contributions of each stage to a certain nutrient are summed to obtain the normalized weight sum for the entire cycle. The normalized weight is then converted into a component proportion range. The conversion uses percentage mapping and considers the feasible boundaries of construction and preparation. The mapping boundaries can be preset as 30% to 50% for nitrogen, 15% to 30% for phosphorus, 20% to 40% for potassium, and 2% to 5% for trace elements. If the calculated theoretical proportion exceeds the boundary, the deviation is truncated at the boundary value and recorded for adjustment. For example, if the theoretical nitrogen proportion after accumulation is 0.55, it is truncated to 0.50 according to the boundary, and the excess is redistributed proportionally between phosphorus and potassium. Throughout the process, the total proportion of the four components should be kept at 100% to ensure the integrity of the formulation.

[0031] By eliminating conflicts in the range of basic nutrient proportions obtained from each growth stage and clarifying the basic mixing ratio, the overall growth needs of suitable crop varieties can be ensured. First, conflict detection is performed on the proportion range of each component. The conflict identification rules and thresholds are verified in parallel with agricultural safety thresholds and formulation process constraints. For example, the agricultural safety threshold is that nitrogen should not exceed 60% of the total nutrient mass, and trace elements should not exceed 5% to avoid toxicity. An example of formulation process constraints is that the mixing ratio of two types of substances should not exceed a certain synthesis limit of 5% at the same time. If a range conflict is detected, two strategies are adopted to solve it. One is to tighten the range and find the intersection to satisfy all constraints. The other is to use linear programming to solve the minimum deviation optimization problem so that the final ratio is closest to the theoretical target under the condition of satisfying all boundaries and safety thresholds. The linear programming is to minimize the objective function, which is the sum of the squares of the actual proportion and the theoretical proportion of each component. The constraints include the upper and lower limits of each component and the equality constraint that the sum is 100%. The example solution result, after rounding, may form the final basic mixing ratio, such as nitrogen 40%, phosphorus 20%, potassium 35%, and trace elements 5%. This ratio will be used as a demonstration basic mixing ratio and verified in field trials to confirm its suitability and safety for crop growth.

[0032] In one embodiment of the present invention, step S5 involves regionally adjusting the basic blending ratio based on local climate characteristics data and considering the topographical differences and moisture distribution in the sowing area. This study establishes a correlation between local annual precipitation, average annual temperature, and sunshine duration and the field retention efficiencies of nitrogen, phosphorus, potassium, and micronutrients to clarify the influence of climatic factors on the utilization of various nutrients. The implementation process begins by compiling multi-year climatic statistics for the planting area, summarizing annual precipitation, average annual temperature, and sunshine duration by annual or growth period averages, and comparing these with nutrient utilization rate data from field measurements or literature. Retention efficiency is used as the evaluation index, representing the proportion of applied nutrients ultimately absorbed by the crop or effectively retained in the root zone. It can be set as a continuous value between 0 and 1; for example, the retention efficiency of nitrogen under moderate precipitation conditions could be 0.45, phosphorus 0.60, potassium 0.55, and micronutrients 0.65. Statistical analysis is used to determine the direction of the impact of climate change on nutrient retention efficiency. For example, increased annual precipitation increases the risk of nitrogen leaching, leading to a decrease in retention efficiency. Rising average annual temperature may accelerate nitrogen mineralization and conversion rates, thus improving short-term availability. Increased sunshine duration may enhance crop growth intensity, indirectly increasing nutrient uptake. To quantify these relationships, correlation coefficients or regression coefficients are used to characterize the direction and intensity of the impact. For example, the coefficient of precipitation's influence on nitrogen retention can be set as -0.3, and its coefficient of influence on phosphorus retention can be set as +0.1. These coefficients are used to indicate the adjustment trend of nutrient retention efficiency when climate factors change.

[0033] The planting area is divided into several correction units with uniform topographic and moisture characteristics by overlaying and partitioning based on its terrain type and moisture distribution density. During implementation, topographic data and field moisture distribution data of the planting area are first acquired. Terrain types are classified according to slope and elevation, while moisture distribution density is classified based on soil moisture monitoring or irrigation conditions. Partitioning employs a multi-factor overlay method, combining terrain classification results with moisture distribution levels to form correction units. For example, terrain can be divided into three categories: flat areas, gentle slope areas, and low-lying areas; moisture distribution can be divided into three levels: low moisture, medium moisture, and high moisture. This combination results in nine correction units. During partitioning, it is necessary to ensure that the terrain and moisture characteristics within the same correction unit are relatively consistent to reduce the interference of internal differences on the correction effect. To avoid excessive correction units leading to operational complexity, a minimum area threshold can be set, for example: the area of ​​a single correction unit should not be less than 5% of the total planting area. If it is below this threshold, it is merged with the unit with the closest adjacent characteristics. The correction units obtained through this overlay partitioning method provide a spatial basis for subsequent adjustments to the blending ratio by unit.

[0034] Based on the climate impact direction of each correction unit, combined with the effect of topography on nutrient retention and the effect of water on nutrient dissolution, the component proportions of the basic blending ratio within the corresponding correction unit are adjusted. During implementation, the climate impact direction obtained in the first step is first mapped to each correction unit, and the actual behavior of nutrients within that unit is analyzed in conjunction with topographic and water characteristics. For example, sloping units are prone to nitrogen and potassium loss due to runoff, while low-lying, high-moisture units may have a risk of micronutrient accumulation. The adjustment ratio introduces a correction coefficient based on the basic blending ratio. The correction coefficient ranges from 0.85 to 1.15, used to indicate the magnitude of adjustment. For example, in units with high precipitation and steep slopes, the correction coefficient for nitrogen can be set to 0.9 to reduce the proportion, and the correction coefficient for potassium can be set to 0.95, while in low-precipitation, flat units, the correction coefficient for nitrogen can be set to 1.05 to compensate for potential deficiencies. The effect of water on nutrient dissolution is reflected through the correction of micronutrients and phosphorus. For example, in high-moisture units, the proportion of micronutrients can be reduced to 0.9 of the original value to reduce the risk of accumulation. All corrected component percentages need to be renormalized to ensure the sum is 100%, and the correction coefficients and adjustment results of each unit should be recorded to ensure the traceability of the calculation process.

[0035] Cross-unit compensation correction is implemented to address potential nutrient migration characteristics between adjacent correction units, eliminating nutrient imbalances at zonal boundaries and integrating them into a regionalized, corrected blending ratio. Implementation begins with analyzing gradient changes in topography and moisture conditions between adjacent correction units to identify boundary areas where lateral nutrient migration may occur. For example, areas transitioning from slopes to low-lying areas are prone to nitrogen and potassium migration with runoff. Migration compensation introduces a boundary compensation coefficient for fine-tuning the ratio between adjacent units. Example values ​​for the compensation coefficient can be set between 0.95 and 1.05, representing the magnitude of outward or inward compensation. For instance, the nitrogen ratio in an upstream slope unit can be increased to 1.03 times, while the nitrogen ratio in a downstream low-lying unit can be correspondingly decreased to 0.97 times to balance overall supply. The compensation amount is determined based on the area ratio of adjacent units and the migration risk level, which is divided into low, medium, and high levels. For example, a high-risk level corresponds to a compensation coefficient close to 1.05 or 0.95. After completing cross-unit compensation, the mixing ratios of all corrected units in the entire sowing area are uniformly summarized and verified for consistency to ensure that the ratios of each unit within the regional scale are smoothly connected and meet the requirements for crop growth and environmental safety as a whole, thus obtaining the regionalized corrected mixing ratio results.

[0036] In one embodiment of the present invention, step S6, in which a customized blended fertilizer scheme adapted to a specific sowing scenario is generated by combining the corrected proportion parameters, and the process of clarifying the precise proportions and usage methods of each component is as follows: The corrected proportion parameters are mapped one-to-one with the sowing methods and tillage cycle elements of the sowing scenario, clarifying the adaptation conditions of each parameter in specific scenarios and laying the foundation for scenario matching in scheme generation. In the specific implementation process, the sowing scenario is first described in terms of elements, including sowing method categories such as mechanical row sowing, broadcast sowing, and hill sowing; tillage cycle attributes include four time windows: pre-sowing, seedling stage, inter-cultivation stage, and post-harvest maintenance. Each element and the corrected blending ratio are recorded in the scenario attribute table to achieve a one-to-one correspondence. The rule for scenario matching is set as a weighted mapping based on scenario attributes. The weights are obtained by multiplying the stage-normalized weights by the scenario sensitivity coefficient; for example, the scenario sensitivity coefficient is 1.0 for mechanical row sowing, 0.9 for broadcast sowing, and 1.1 for hill sowing, thus adjusting the contribution of each component according to the adaptation coefficient when converting the ratio to dosage. The system includes a list of preset scene element collection frequencies and minimum descriptive fields. Preset fields include sowing depth, row spacing, expected yield target, and tillage intensity. For example, sowing depth is 2 to 5 cm, row spacing is 15 to 30 cm, and tillage intensity is categorized into low, medium, and high levels. To convert proportional parameters into applicable fertilizer application rates, a conversion rule based on nutrient target per unit area is used. This rule involves multiplying the nutrient requirement for the target crop's entire growth period by the scene adaptation coefficient and then dividing by the corresponding nutrient content in each unit of fertilizer to obtain the application rate per acre or hectare. The threshold is controlled according to the soil safety upper limit, determined jointly by historical soil background and crop safety standards. For example, if the target nitrogen application rate is 150 kg / ha and the fertilizer nitrogen content is 40%, the required total amount of mixed fertilizer is calculated by dividing 150 by 0.4, resulting in 375 kg / ha. If the acceptable upper limit for soil conditions is 200 kg / ha, adjustments are made before implementation, and the reasons are recorded.

[0037] Based on compatibility conditions, the blending order is determined by prioritizing the physical properties of nitrogen, phosphorus, potassium, and trace elements. The revised proportions are then converted into precise ratios for each component to ensure the blending logic aligns with the component characteristics. During implementation, the physicochemical properties of the raw materials to be blended are first analyzed, including solubility, particle size distribution, deliquescence, acid-base reaction tendency, and compatibility with other components. The blending order is determined by prioritizing these properties to reduce adverse chemical reactions and physical agglomeration. The principle for setting the blending order is to first add solid components with low solubility and low hydrolysis, then add components with high solubility or acid-base sensitivity, and finally add trace elements or performance-enhancing additives. For example, long-acting slow-release nitrogen sources and granular phosphate fertilizers are added first, followed by nitrate nitrogen or soluble potassium salts, and finally chelated trace elements. The preset threshold for determining incompatible substances is based on chemical stability criteria using differences in solubility and pH response. The threshold is calculated using a comprehensive index that multiplies the component solubility ratio by the pH sensitivity score. For example, if the solubility ratio of two components exceeds 2 and the predicted pH change after mixing exceeds 1 unit, they are considered incompatible and should be applied in multiple applications or coated. To ensure the engineering feasibility of precise formulation, it is recommended to coat solid components with particle size differences greater than a certain percentage to avoid stratification. The preset treatment range is when the median particle size ratio exceeds 1.5 times; for example, if the median diameter of a trace element particle is 150 micrometers and that of a base fertilizer particle is 80 micrometers, the trace element should be coated or made into dispersible microparticles before mixing.

[0038] By integrating precise proportions and mixing sequences, and clarifying the application time, dosage per acre, and application method (broadcast or furrow) for each component, a customized blended fertilizer solution tailored to specific sowing scenarios is formed. Implementation details involve breaking down the target nutrient content per acre or hectare obtained based on scenario matching into a multi-application plan. The multi-application schedule follows the peak nutrient requirements of each growth stage while also considering the risk of nutrient loss. Time nodes are categorized into pre-sowing basal fertilizer, seedling topdressing, vigorous growth topdressing, and pre-irrigation supplementary fertilizer. The rules for multi-application are: prioritizing the supply to the basal layer in the early growth stage to ensure stable seedling emergence; secondly, determining the topdressing frequency based on nitrogen volatility and rainfall forecasts. The preset application ratios are: 30% to 50% for basal fertilizer, 10% to 20% for seedling fertilizer, and 30% to 50% for vigorous growth fertilizer. For example, if the total nitrogen requirement for the season is 150 kg / ha, the preset basal fertilizer ratios are 45 kg to 75 kg / ha, the seedling fertilizer ratios are 15 kg to 30 kg / ha, and the vigorous growth fertilizer ratios are 45 kg to 75 kg / ha. The final amount of application per acre is adjusted according to the unit area conversion rules and the application efficiency of the equipment. The application method is selected according to the tillage and sowing methods, such as broadcasting or furrow application. The choice between deep application or strip application is based on the crop root distribution and rainfall conditions. The application method takes into account engineering measures to prevent and mitigate nutrient loss. For example, in areas with high leaching risk, furrow application or coated slow-release products are preferred.

[0039] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for optimizing agricultural sowing and fertilizer blending based on multi-source data, characterized in that, Includes the following steps: S1. Collect soil nutrient data, crop variety characteristic data and local climate characteristic data of the sowing area, and integrate them to form a multi-source basic dataset; S2. Denoise the multi-source basic dataset, remove outlier data points, and then perform data standardization transformation. S3. Establish a dynamic correlation map between soil nutrients and crop requirements, and clarify the key influence weights of each nutrient on crop growth through data correlation analysis; S4. Based on the weight allocation results of the correlation graph and combined with the characteristics of crop varieties, determine the basic blending ratio of nitrogen, phosphorus, potassium and trace elements in the blended fertilizer. S5. Based on local climate characteristics data, the basic mixing ratio is adjusted regionally according to the differences in terrain and water distribution in the sowing area. S6. Based on the corrected proportion parameters, generate a customized blended fertilizer solution that is suitable for specific sowing scenarios, and clarify the precise ratio and usage of each component.

2. The method for optimizing agricultural sowing and blending fertilizer based on multi-source data according to claim 1, characterized in that, In step S1, the soil nutrient data includes nitrogen content, phosphorus content, potassium content, and organic matter content; the crop variety characteristic data includes growth cycle, fertilizer requirement pattern, and stress resistance; and the local climate characteristic data includes annual precipitation, average annual temperature, and sunshine duration.

3. The method for optimizing agricultural sowing and blending fertilizers based on multi-source data according to claim 1, characterized in that, In step S2, the process of denoising the multi-source basic dataset, removing outlier data points, and then performing data standardization transformation is as follows: The moving average method was used to process the soil nutrient data, crop variety characteristic data, and local climate characteristic data in the multi-source basic dataset separately to smooth the random fluctuations of various types of data and achieve data denoising. Based on the denoised data, the inherent attribute range of the corresponding data is compared to identify data points that exceed the range. Data points that do not conform to the attribute range are judged as abnormal data points and removed. For the multi-source basic dataset after removing outlier data points, the range standardization method is used to transform different types of data into fixed numerical ranges, thus completing the data standardization transformation.

4. The method for optimizing agricultural sowing and blending fertilizer based on multi-source data according to claim 1, characterized in that, In step S3, the process of establishing a dynamic correlation map between soil nutrients and crop requirements, and clarifying the key influence weights of each nutrient on crop growth through data correlation analysis, is as follows: The crop growth cycle is divided into four stages: germination, growth, maturity, and fruiting. Specific nutrient requirements corresponding to the nutrient requirements of each stage are extracted to clarify the soil nutrient requirements of crops at different stages. The standardized soil nutrient data are matched one by one with the crop demand indicators at each growth stage to construct a staged nutrient-demand correlation matrix, forming a basic framework for the dynamic correlation between soil nutrients and crop demand. Partial correlation analysis was performed on the correlation matrix, while keeping other nutrient data constant. The correlation strength between each soil nutrient and the corresponding crop requirement index at the same stage was calculated separately to eliminate mutual interference between nutrients. Based on the correlation strength obtained from partial correlation analysis, corresponding weights are assigned to each soil nutrient at different growth stages. The weights of each stage are integrated to form a weight system of key influences of nutrients on crop growth, thus improving the dynamic correlation map.

5. The method for optimizing agricultural sowing and blending fertilizer based on multi-source data according to claim 4, characterized in that, The specific method for performing partial correlation analysis on the correlation matrix, keeping other nutrient data constant, and calculating the correlation strength between each soil nutrient and the corresponding crop requirement index at the corresponding stage is as follows: For a specific soil nutrient in the correlation matrix, locate its corresponding data column with the crop requirement index at each growth stage, separate the data columns of other soil nutrients in the matrix and keep their values ​​constant; Using other soil nutrient data that remain constant after separation as control variables, and using target soil nutrient data and corresponding crop demand indicators as analytical variables, a partial correlation analysis model is constructed. By running the constructed partial correlation analysis model and keeping the values ​​of control variables constant, the correlation between the target soil nutrients and the corresponding crop demand indicators at the same stage is calculated, and the correlation strength of the nutrient is obtained.

6. The method for optimizing agricultural sowing and blending fertilizer based on multi-source data according to claim 1, characterized in that, In step S4, the process of determining the basic blending ratio of nitrogen, phosphorus, potassium, and trace elements in the blended fertilizer based on the weight allocation results of the correlation graph and in combination with the characteristics of the crop variety is as follows: Extract the key influence weights of each nutrient in the correlation map, and combine the growth cycle and nutrient requirement patterns of crop varieties to clarify the weight priority of key nutrients at different growth stages; Based on crop stress resistance, we analyze the synergistic enhancement or inhibition relationships among various nutrients, adjust the weight ratio of corresponding nutrients, and avoid nutrient imbalance caused by simply allocating nutrients according to their weight. The adjusted nutrient weights are matched one by one with the fertilizer requirements of crops at each growth stage, and the basic proportion range of nitrogen, phosphorus, potassium and micronutrients in the whole growth cycle is defined. By comprehensively considering the basic nutrient ratios at each growth stage, eliminating conflicting ranges, and clarifying the basic mixing ratios of nitrogen, phosphorus, potassium, and micronutrients, we can ensure that the crop varieties are adapted to their overall growth needs.

7. The method for optimizing agricultural sowing and blending fertilizer based on multi-source data according to claim 1, characterized in that, In step S5, the process of regionally adjusting the basic blending ratio based on local climate characteristics data and considering the topographical differences and moisture distribution in the sowing area is as follows: The correlation between local annual precipitation, average annual temperature, and sunshine duration and the field retention efficiency of nitrogen, phosphorus, potassium, and trace elements was established to clarify the direction of the influence of climate factors on the function of each nutrient. The planting area is divided into several modified units with uniform topographic and water distribution characteristics by overlaying and partitioning the area based on the terrain type and water distribution density. Based on the direction of climate influence in each correction unit, and combined with the effect of topography on nutrient retention and the effect of water on nutrient dissolution, the component proportion of the basic mixing ratio in the corresponding unit is adjusted. Based on the nutrient migration characteristics of adjacent correction units, cross-unit compensation correction is carried out to eliminate the nutrient imbalance problem at the partition boundary and integrate to form a regionalized correction blending ratio.

8. The method for optimizing agricultural sowing and blending fertilizer based on multi-source data according to claim 1, characterized in that, In step S6, the process of generating a customized blended fertilizer scheme adapted to specific sowing scenarios by combining the corrected proportion parameters, and clarifying the precise ratio and usage of each component, is as follows: The revised proportional parameters are matched one by one with the sowing method and cultivation cycle elements of the sowing scenario, clarifying the adaptation conditions of each parameter in the specific scenario, and laying the foundation for scenario matching for solution generation. Based on the compatibility conditions, the mixing order is determined according to the physical properties of nitrogen, phosphorus, potassium and trace elements. The corrected proportions are then converted into precise ratios of each component to ensure that the mixing logic is compatible with the characteristics of the components. By integrating precise proportions and mixing sequences, and clarifying the application time, dosage per acre, and application method (broadcast or furrow) for each component, customized blended fertilizer solutions can be formed to suit specific sowing scenarios.

Images