Soybean yield self-adaptive prediction method and system based on big data

Abstract

The present application relates to the technical field of soybean yield prediction, and particularly relates to a soybean yield self-adaptive prediction method and system based on big data, wherein the soybean yield self-adaptive prediction method based on big data comprises the following steps: obtaining fertilization management data and yield response data in a soybean planting period; obtaining a fertilization supply trajectory of a sample to be analyzed, and determining a current fertilization supply trajectory deviation of the fertilization supply trajectory relative to a fertilization supply reference trajectory; determining a fertilization deviation yield response sensitive coefficient corresponding to the current fertilization supply trajectory deviation; determining a fertilization deviation critical response intensity; and determining a soybean predicted yield. By constructing the fertilization supply trajectory deviation, the fertilization deviation yield response sensitive coefficient and the fertilization deviation critical response intensity, the present application establishes a mapping relationship from the difference in the fertilization management process to the yield prediction result, so that the prediction result can reflect the fertilization regulation effect, and the soybean yield self-adaptive prediction based on the fertilization management process is realized.

Description

Technical Field

[0001] This invention relates to the field of soybean yield prediction technology, specifically to an adaptive soybean yield prediction method and system based on big data. Background Technology

[0002] Soybeans, as an important dual-purpose crop for both grain and oil, have yields per unit area influenced by multiple factors, including planting environment, nutrient supply, and field management. With the development of agricultural big data, intelligent agricultural machinery, and information collection technologies, predicting crop yields using historical planting data has become a crucial means of guiding agricultural production management and improving planting efficiency. Among these factors, fertilization management directly determines the nutrient supply status of soybeans at different growth stages, further affecting the final yield. Therefore, constructing a yield prediction method that reflects differences in the fertilization process based on fertilization time, fertilization frequency, fertilizer application rate per unit area, nutrient type, fertilizer nutrient content, and historical yield is of great significance for achieving precision agricultural management and ensuring high and stable soybean yields.

[0003] Existing technologies typically use crop growth environment data at different time points as input to construct crop yield prediction models, including long short-term memory layers, temporal attention layers, feature attention layers, and fully connected layers. These models extract temporal variation features during crop growth by assigning weights to different periods and environmental factors, and then output crop yield prediction results. However, these methods primarily extract yield prediction features from environmental time-series data, making it difficult to characterize the dynamic nutrient supply trajectory determined by fertilization time, fertilization frequency, fertilizer application rate per unit area, and fertilizer nutrient content. In scenarios with similar environmental conditions but different fertilization management processes, even different plots with similar meteorological and basic soil conditions may exhibit different nutrient supply rhythms due to variations in the combination of fertilization time, fertilization frequency, single application rate, and fertilizer nutrient content, leading to differences in soybean yield responses. Existing technologies cannot determine the extent to which the fertilization supply trajectory deviates from the high-yield fertilization pattern, easily resulting in insufficient differentiation of prediction results and difficulty in reflecting the effects of fertilization regulation, thus failing to meet the need for adaptive soybean yield prediction based on fertilization management processes. Summary of the Invention

[0004] This invention provides a method and system for adaptive soybean yield prediction based on big data to solve the above-mentioned problems.

[0005] The soybean yield adaptive prediction method and system based on big data of the present invention adopts the following technical solution: In a first aspect, one embodiment of the present invention provides a method for adaptive soybean yield prediction based on big data. The method includes: acquiring fertilizer management data and yield response data during the soybean planting cycle; acquiring the fertilizer supply trajectory of the sample to be analyzed based on the fertilizer management data, and determining the current fertilizer supply trajectory deviation relative to the fertilizer supply baseline trajectory; determining a fertilizer deviation yield response sensitivity coefficient corresponding to the current fertilizer supply trajectory deviation based on historical fertilizer supply trajectory deviations and historical yield response data in historical samples; determining a critical response intensity of fertilizer deviation based on the fertilizer deviation yield response sensitivity coefficient and the critical position of the current fertilizer supply trajectory deviation within its respective deviation response interval; and determining the predicted soybean yield based on the critical response intensity of fertilizer deviation, the current fertilizer supply trajectory deviation, and the historical average yield of the current deviation response interval.

[0006] Further, determining the fertilizer deviation-yield response sensitivity coefficient corresponding to the current fertilizer supply trajectory deviation based on historical fertilizer supply trajectory deviation and its historical yield response data in historical samples includes: determining the historical fertilizer supply trajectory deviation of each historical sample based on historical fertilizer management data and historical yield response data of each historical sample, and establishing a correspondence between the historical fertilizer supply trajectory deviation and the historical yield response data; sorting each historical sample according to the historical fertilizer supply trajectory deviation from smallest to largest, and dividing it into several continuous deviation response intervals; determining the deviation response interval corresponding to the current fertilizer supply trajectory deviation, selecting the preceding and following deviation response intervals adjacent to the deviation response interval, and determining the average fertilizer supply trajectory deviation and average historical yield response data of historical samples in the preceding and following deviation response intervals respectively; and determining the fertilizer deviation-yield response sensitivity coefficient based on the difference between the average historical yield response data of the preceding and following deviation response intervals and the difference between the average fertilizer supply trajectory deviation of the preceding and following deviation response intervals.

[0007] Further, determining the fertilizer deviation yield response sensitivity coefficient based on the difference in average historical yield response data between the previous deviation response interval and the subsequent deviation response interval, and the difference in average fertilizer supply trajectory deviation between the previous deviation response interval and the subsequent deviation response interval, includes: calculating the difference between the average historical yield response data of the previous deviation response interval and the average historical yield response data of the subsequent deviation response interval to obtain the yield response difference; calculating the difference between the average fertilizer supply trajectory deviation of the subsequent deviation response interval and the average fertilizer supply trajectory deviation of the previous deviation response interval to obtain the deviation difference; calculating the ratio of the yield response difference to the deviation difference to obtain the initial sensitivity coefficient; and determining the fertilizer deviation yield response sensitivity coefficient based on the initial sensitivity coefficient; wherein, the fertilizer deviation yield response sensitivity coefficient is the larger of the initial sensitivity coefficient and zero.

[0008] Further, determining the critical response intensity of fertilization deviation based on the fertilization deviation-yield response sensitivity coefficient and the critical position of the current fertilization supply trajectory deviation within its respective deviation response interval includes: determining the lower and upper boundaries of the deviation response interval to which the current fertilization supply trajectory deviation belongs; determining the critical position of the interval based on the degree of approximation between the current fertilization supply trajectory deviation and the lower and upper boundaries; and determining the critical response intensity of fertilization deviation based on the fertilization deviation-yield response sensitivity coefficient and the critical position of the interval.

[0009] Further, determining the predicted soybean yield based on the fertilization deviation from the critical response intensity, the deviation of the current fertilization supply trajectory, and the historical average yield of the current deviation response interval includes: obtaining the historical average yield of the current deviation response interval; determining the critical response reduction amount based on the degree of deviation of the fertilization deviation from the critical response intensity and the current fertilization supply trajectory relative to the lower boundary of its respective deviation response interval; and determining the predicted soybean yield based on the historical average yield and the critical response reduction amount.

[0010] Further, the step of obtaining the fertilizer supply trajectory of the sample to be analyzed based on the fertilizer management data, and determining the deviation of the current fertilizer supply trajectory from the fertilizer supply benchmark trajectory, includes: dividing the soybean planting cycle into several continuous analysis windows; for each analysis window, determining the cumulative nutrient supply based on the fertilizer management data corresponding to the analysis window; obtaining the fertilizer supply trajectory based on the cumulative nutrient supply of all analysis windows; determining the benchmark nutrient supply and supply fluctuation range corresponding to each analysis window based on historical samples where the actual yield per unit area is within a preset high-yield range; obtaining the fertilizer supply benchmark trajectory based on the benchmark nutrient supply of all analysis windows; and determining the deviation of the current fertilizer supply trajectory based on the difference between the cumulative nutrient supply and the benchmark nutrient supply of each analysis window and the supply fluctuation range.

[0011] Further, dividing the soybean planting cycle into several continuous analysis windows includes: dividing the soybean planting cycle into several continuous analysis windows according to a preset fixed time interval or soybean growth stage; wherein, the soybean growth stage includes seedling stage, branching stage, flowering stage, grain filling stage, and maturity stage; for each analysis window, determining the cumulative nutrient supply of the analysis window based on the fertilization management data corresponding to the analysis window; obtaining the fertilization supply trajectory based on the cumulative nutrient supply of all analysis windows; determining the benchmark nutrient supply and supply fluctuation range corresponding to each analysis window based on historical samples where the actual yield per unit area is within a preset high-yield range; obtaining the fertilization supply benchmark trajectory based on the benchmark nutrient supply of all analysis windows; and determining the deviation of the current fertilization supply trajectory based on the difference between the cumulative nutrient supply and the benchmark nutrient supply of each analysis window and the supply fluctuation range.

[0012] Further, after determining the predicted soybean yield based on the fertilization deviation critical response intensity, the current fertilization supply trajectory deviation, and the historical average yield of the current deviation response interval, the method further includes: determining that the plot to be predicted is in a high deviation fertilization state when the current fertilization supply trajectory deviation is greater than a preset deviation threshold, or the fertilization deviation critical response intensity is greater than a preset response intensity threshold; extracting historical samples from the historical sample library that are consistent with or within a preset range of the fertilization deviation critical response intensity of the plot to be predicted, and whose actual yield per unit area is within a preset high-yield interval, and establishing a mapping relationship between the fertilization deviation critical response intensity and the fertilization supplementation parameters; wherein, the fertilization supplementation... The parameters include fertilization time, fertilization frequency, fertilizer type, nutrient type, and fertilization amount per unit area. Based on the critical response intensity of fertilization deviation of the plot to be predicted, the corresponding historical fertilization parameters are matched in the mapping relationship. The appropriate fertilizer type and nutrient type are determined from the preset candidate fertilizer schemes, and the fertilization amount per unit area is adaptively set in combination with the current monitoring data. The current monitoring data includes at least one of soil nutrient status, soil moisture, plant growth, growth stage, and environmental conditions. After the current planting cycle ends and the actual yield per unit area is obtained, the fertilization management data, fertilization parameters, and actual yield per unit area of ​​the plot to be predicted are added to the historical sample database.

[0013] Further, determining the deviation of the current fertilization supply trajectory based on the difference between the cumulative nutrient supply and the benchmark nutrient supply in each analysis window, as well as the supply fluctuation range, includes: calculating the absolute value of the difference between the cumulative nutrient supply and the benchmark nutrient supply in each analysis window to obtain the nutrient supply deviation of each analysis window; calculating the ratio of the nutrient supply deviation to the supply fluctuation range in each analysis window to obtain the standardized deviation value of each analysis window; and determining the deviation of the current fertilization supply trajectory based on the standardized deviation value of each analysis window and a preset window weight; wherein the preset window weight is determined based on the proportion of the duration of the analysis window within the soybean planting cycle.

[0014] Further, acquiring fertilization management data and yield response data within the soybean planting cycle includes: collecting fertilization management data and yield response data within the soybean planting cycle; wherein, the fertilization management data includes the fertilization time, fertilizer application rate per unit area, nutrient type, and fertilizer nutrient content for each fertilization; the yield response data includes the actual yield per unit area; preprocessing the fertilization management data and the yield response data; wherein, the preprocessing includes: associating and matching the fertilization management data and the yield response data for the same plot based on plot identifiers and planting cycle identifiers; converting the fertilization time into a relative time after sowing; unifying the fertilization management data and the yield response data into the same measurement format; and organizing the fertilization records within the same plot according to the chronological order of fertilization time.

[0015] Secondly, another embodiment of the present invention provides a soybean yield adaptive prediction system based on big data, comprising a data acquisition module, a current fertilizer supply trajectory deviation determination module, a fertilizer deviation yield response sensitivity coefficient determination module, a fertilizer deviation critical response intensity determination module, and a soybean predicted yield determination module connected in sequence, wherein: The data acquisition module is used to acquire fertilization management data and yield response data during the soybean planting cycle; The current fertilizer supply trajectory deviation determination module is used to obtain the fertilizer supply trajectory of the sample to be analyzed based on the fertilizer management data, and determine the deviation of the current fertilizer supply trajectory from the fertilizer supply benchmark trajectory. The fertilizer deviation yield response sensitivity coefficient determination module is used to determine the fertilizer deviation yield response sensitivity coefficient corresponding to the current fertilizer supply trajectory deviation based on the historical fertilizer supply trajectory deviation and its historical yield response data in the historical samples. The fertilizer deviation critical response intensity determination module is used to determine the fertilizer deviation critical response intensity based on the fertilizer deviation yield response sensitivity coefficient and the critical position of the current fertilizer supply trajectory deviation within its respective deviation response interval. The soybean yield prediction module is used to determine the soybean yield based on the fertilization deviation critical response intensity, the current fertilization supply trajectory deviation, and the historical average yield of the current deviation response interval.

[0016] The beneficial effects of the technical solution of the present invention are: In this embodiment of the invention, fertilization management data and yield response data during the soybean planting cycle are acquired; based on the fertilization management data, the fertilization supply trajectory of the sample to be analyzed is acquired, and the deviation of the current fertilization supply trajectory from the fertilization supply baseline trajectory is determined; based on the historical fertilization supply trajectory deviation and its historical yield response data in historical samples, the fertilization deviation yield response sensitivity coefficient corresponding to the current fertilization supply trajectory deviation is determined; based on the fertilization deviation yield response sensitivity coefficient and the critical position of the current fertilization supply trajectory deviation within its respective deviation response interval, the critical response intensity of the fertilization deviation is determined; based on the critical response intensity of the fertilization deviation, the current fertilization supply trajectory deviation, and the historical average yield of the current deviation response interval, the predicted soybean yield is determined.

[0017] This invention establishes a mapping relationship between differences in fertilization management processes and yield prediction results by constructing fertilization supply trajectory deviation, fertilization deviation yield response sensitivity coefficient, and fertilization deviation critical response intensity. This allows the prediction results to reflect the effect of fertilization regulation, achieving adaptive soybean yield prediction based on fertilization management processes. On the other hand, by accumulating fertilization time, fertilizer application rate per unit area, and fertilizer nutrient content according to growth stage windows to form a fertilization supply trajectory, and comparing its deviation with the baseline trajectory of historical high-yield samples, the discrete fertilization management process is transformed into a quantifiable nutrient supply rhythm deviation index, thereby distinguishing yield response differences in soybean plots with similar environmental conditions but different fertilization management processes. Furthermore, by analyzing the correspondence between fertilization supply trajectory deviation and yield response in historical samples, deviation response intervals are divided, and the yield change rate of adjacent intervals is calculated to obtain the yield response to fertilization deviation. The system uses a sensitivity coefficient to identify the degree to which yield changes in fertilization deviation occur at different deviation locations, avoiding a simple linear correlation between deviation and yield decline. Furthermore, by combining the yield response sensitivity coefficient of fertilization deviation with the critical position of the current fertilization supply trajectory deviation within its corresponding response interval, the system calculates the critical response intensity of fertilization deviation. This allows the system to determine whether the plot to be predicted is approaching a higher deviation interval and to respond in advance to the yield risk of entering a high deviation interval. In addition, by establishing a mapping relationship between the critical response intensity of fertilization deviation and fertilization parameters under high fertilization deviation conditions, and by adaptively setting the fertilization amount per unit area based on current monitoring data, fertilization decisions can be adjusted specifically based on the fertilization experience of historical high-yield samples. This corrects the deviation of the nutrient supply trajectory through fertilization intervention, and automatically corrects the fertilization amount setting results with the periodic updates of the historical sample database. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating the adaptive soybean yield prediction method based on big data provided in this embodiment of the invention; Figure 2 This is a schematic diagram of the architecture of a big data-based adaptive soybean yield prediction system provided in an embodiment of the present invention. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following detailed description, in conjunction with the accompanying drawings and preferred embodiments, provides a detailed account of the specific implementation methods, structures, features, and effects of the method proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments may be combined in any suitable form. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] The following description, in conjunction with the accompanying drawings, details the specific scheme of the soybean yield adaptive prediction method and system based on big data provided by this invention.

[0022] like Figure 1 As shown, this embodiment of the invention provides an adaptive soybean yield prediction method based on big data, including: Step S110: Obtain fertilization management data and yield response data during the soybean planting cycle.

[0023] Optionally, step S110 includes: collecting fertilization management data and yield response data during the soybean planting cycle; wherein, the fertilization management data includes the fertilization time, fertilizer application rate per unit area, nutrient type, and fertilizer nutrient content for each fertilization; and the yield response data includes the actual yield per unit area. Preprocessing the fertilization management data and yield response data includes: associating and matching the fertilization management data and yield response data of the same plot based on plot identifiers and planting cycle identifiers; converting the fertilization time to a relative time after sowing; unifying the fertilization management data and yield response data into the same measurement format; and organizing the fertilization records within the same plot according to the chronological order of fertilization time.

[0024] Using individual plots as sample units, fertilization management data and yield response data were collected throughout the soybean growing cycle. Fertilization management data included the fertilization time, fertilizer application rate per unit area, nutrient type, and fertilizer nutrient content for each fertilization. Fertilization time recorded the exact moment each fertilization occurred; fertilizer application rate per unit area reflected the intensity of a single fertilization application; nutrient type indicated the types of nutrients contained in the fertilizer, such as macronutrients like nitrogen, phosphorus, and potassium, used to differentiate the nutrient supply functions of different fertilizers; and fertilizer nutrient content reflected the concentration of the fertilizer's effective components. Yield response data included the actual yield per unit area at harvest, which could be collected using harvester yield monitoring devices or electronic weighing equipment.

[0025] Multiple fertilization records and one yield record are generated for the same plot of land within a specific planting cycle. Data association is established through plot identifiers and planting cycle identifiers to match and aggregate fertilization management data with corresponding yield response data within the same plot. Plot identifiers are used to distinguish planting units in different spatial locations, while planting cycle identifiers are used to distinguish planting batches of the same plot in different years or seasons. The combined use of the two can avoid data mismatches across plots or across cycles.

[0026] Fertilization time is usually recorded as an absolute timestamp. However, since sowing dates vary across different plots, using absolute time directly for subsequent analysis will lead to misalignment of growth stages due to different sowing times. Therefore, fertilization time can be uniformly converted to a relative time after sowing. That is, with the sowing time as the zero point, the time interval between each fertilization and the sowing time can be calculated, providing a unified time reference for fertilization events across different plots.

[0027] Fertilizer application rate per unit area, fertilizer nutrient content, and actual yield per unit area may originate from different data collection devices or manual input terminals during the data collection process, posing a risk of inconsistencies in measurement units or data formats. During preprocessing, these data can be standardized to a uniform measurement format, eliminating data incomparability issues caused by differences in dimensions or formats. Based on this, fertilization records within the same plot are organized according to the chronological order of fertilization application, forming a time-series sequence of fertilization events, providing structured input for subsequent cumulative nutrient supply calculations based on growth stage windows.

[0028] After correlation matching, time base conversion, measurement format unification and time series processing, the original sample data is transformed into a standardized structured dataset. This dataset uses plots and planting cycles as index units, and internally arranges each fertilization event and its corresponding fertilization management data in relative time order, and associates it with the final yield response results within the cycle, providing standardized sample data for the construction of subsequent fertilization supply trajectories.

[0029] Step S120: Based on fertilization management data, obtain the fertilization supply trajectory of the sample to be analyzed, and determine the deviation of the current fertilization supply trajectory from the fertilization supply baseline trajectory.

[0030] It is understandable that in actual soybean cultivation, yield per unit area is not solely determined by the total amount of fertilizer applied throughout the entire growing cycle, but is closely related to the nutrient supply rhythm at different growth stages. Even within the same or similar planting areas, different plots may have similar external environmental conditions and relatively similar total fertilizer application per unit area. However, differences in the timing, frequency, amount of fertilizer applied per application, and combination of fertilizer nutrient content can lead to variations in the cumulative nutrient supply process actually received by soybeans in the early, middle, and late growth stages, resulting in different nutrient supply rhythms and consequently, yield differentiation per unit area. Therefore, the above scheme uses the fertilization supply process formed in historical high-yield samples as a reference benchmark for high-yield fertilization scenarios. It constructs the cumulative fertilization supply trajectory of the sample to be analyzed throughout the entire growing cycle and calculates the degree of deviation of this trajectory from the high-yield fertilization supply benchmark trajectory. The resulting deviation reflects whether the fertilization process of the sample to be analyzed deviates from the nutrient supply rhythm exhibited by the high-yield sample, thereby distinguishing the differences in yield response among soybean plots with similar environmental conditions but different fertilization management processes. The following is one method for determining the deviation of the current fertilizer supply trajectory: Optionally, step S120 includes: dividing the soybean planting cycle into several continuous analysis windows according to a preset fixed time interval or soybean growth stage; determining the cumulative nutrient supply for each analysis window based on the corresponding fertilization management data; obtaining the fertilization supply trajectory based on the cumulative nutrient supply of all analysis windows; determining the benchmark nutrient supply and supply fluctuation range for each analysis window based on historical samples where the actual yield per unit area is within a preset high-yield range; obtaining the benchmark nutrient supply trajectory based on the benchmark nutrient supply of all analysis windows; and determining the deviation of the current fertilization supply trajectory based on the difference between the cumulative nutrient supply and the benchmark nutrient supply and the supply fluctuation range of each analysis window.

[0031] The length of the soybean planting cycle is denoted as Based on this cycle length, the entire soybean growth cycle is divided into... Continuous analysis window ,in, , , The following describes how the analysis window is divided: The aforementioned analysis window is the basic unit for discretizing the soybean planting cycle in terms of time. By setting continuous start and end boundaries, the complete planting cycle is divided into several consecutive time intervals, and each interval independently receives fertilization events falling within its time range.

[0032] When dividing the time according to a preset fixed time interval, equal time spans are usually used as the window length, for example, 30 days as an analysis window, to represent the soybean planting cycle. Equidistant division A continuous interval. When dividing soybeans according to their growth stages, the planting cycle can be divided into seedling stage, branching stage, flowering stage, grain filling stage, and maturity stage, with each stage serving as an independent analysis window. Since the actual duration of different growth stages varies, the time span of each window is not equal, but each window corresponds to a specific physiological development stage of soybeans, allowing the cumulative nutrient supply to directly reflect the nutrient input intensity of that key growth stage. The fixed time interval method focuses on uniform sampling in the time dimension, facilitating horizontal comparisons across cycles or plots; the growth stage method focuses on biological stage correspondence, establishing a direct link between the nutrient supply trajectory and the actual growth demand rhythm of soybeans. In practical applications, one of the division strategies can be selected based on the historical data collection density, plot management accuracy requirements, and the construction method of subsequent prediction models. Regardless of whether a fixed time interval or growth stage is used as the division basis, the resulting analysis window must meet the basic requirement of being continuous and covering the entire planting cycle, that is, the termination boundary of the previous window connects to the starting boundary of the next window, and the first window begins at the start of the planting cycle. The final window ends at the end of the planting cycle. .

[0033] For any analysis window of the currently predicted land parcel The system records all fertilization events that fall within the specified window and calculates the cumulative nutrient supply within that window. The calculation formula is as follows:

[0034] In the formula, For analysis window Event index for internal fertilization events; For analysis window The total number of internal fertilization incidents; Analysis window Inner The amount of fertilizer applied per unit area in each fertilization event; Analysis window Inner Nutrient content of the fertilizer used in this fertilization event; Analysis window Inner The timing of fertilization in this fertilization event; For indicator functions, when satisfy hour, The value is 1 if it is positive and 0 otherwise. The above cumulative nutrient supply... The calculation formula multiplies the fertilizer application rate per unit area for each application within the window by the corresponding fertilizer nutrient content, and then sums the results to obtain the total nutrient supply within that analysis window. The cumulative nutrient supply is calculated sequentially for all analysis windows. This sequence constitutes the fertilizer supply trajectory of the sample to be analyzed throughout the entire soybean growing cycle.

[0035] Based on the historical sample dataset, extract the actual yield per unit area for each historical sample. Then, historical samples with yields falling within a preset high-yield range are selected and sorted from high to low as the high-yield sample set. For example, samples from the top 20% of yields can be selected. For each historical sample in the high-yield sample set, the historical cumulative nutrient supply within each analysis window is calculated using the same method as described above. Then, the average cumulative nutrient supply corresponding to each analysis window is calculated using the mean statistical method as the baseline nutrient supply. In other words, the arithmetic mean of the cumulative nutrient supply of each sample in the high-yield sample set within this window is used as the baseline nutrient supply. Simultaneously, the standard deviation statistical method was used to calculate the supply fluctuation range for the corresponding window. This refers to the standard deviation of the cumulative nutrient supply of each sample in the high-yield sample set within that window. It is calculated from the standard deviation of the cumulative nutrient supply of all samples within the window. This forms the baseline trajectory for high-yield fertilizer supply. This is used to characterize the normal fluctuation range of nutrient supply in high-yield samples within this window.

[0036] Optionally, the above-mentioned determination of the current fertilizer supply trajectory deviation based on the difference between the cumulative nutrient supply and the benchmark nutrient supply in each analysis window, as well as the supply fluctuation range, includes: calculating the absolute value of the difference between the cumulative nutrient supply and the benchmark nutrient supply in each analysis window to obtain the nutrient supply deviation in each analysis window; calculating the ratio of the nutrient supply deviation to the supply fluctuation range in each analysis window to obtain the standardized deviation value in each analysis window; and determining the current fertilizer supply trajectory deviation based on the standardized deviation value in each analysis window and the preset window weight; wherein the preset window weight is determined based on the proportion of the duration of the analysis window within the soybean planting cycle.

[0037] For any analysis window of the current sample to be analyzed Calculate the cumulative nutrient supply for this window. The baseline nutrient supply for the corresponding window The absolute value of the difference The nutrient supply deviation of the analysis window is obtained. This deviation reflects the absolute difference between the nutrient input of the current sample at a specific growth stage and the high-yield benchmark. Its magnitude directly represents the absolute degree to which the nutrient supply of the current sample deviates from the high-yield average level within the window.

[0038] Analyze window Nutrient supply deviation divided by the supply fluctuation range of that window The standardized deviation value for this analysis window is obtained. Since the normal fertilization fluctuation range varies across different growth stages, there are significant differences in fertilization supply among high-yield samples in some stages, indicating that a certain degree of fertilization variation is permissible at these stages; conversely, fertilization supply is more consistent among high-yield samples in other stages, suggesting that the fertilization rhythm is more critical for high yields at these stages. By dividing the absolute deviation by the supply fluctuation range, the deviation of the current sample can be converted into a standardized deviation value relative to the normal fluctuation range of high-yield samples, making the degree of deviation between different analysis windows comparable. The standardized deviation value mentioned above does not simply compare the absolute difference between the current fertilization amount and the average fertilization amount for high-yield samples, but rather determines whether this difference has exceeded the normal fluctuation range of high-yield samples at this stage; the larger the standardized deviation value, the more likely the current sample is within the normal fluctuation range of the window. The more the nutrient supply deviates from the high-yield fertilizer supply pattern, the smaller the standardized deviation value indicates that the current sample is within this window. The fertilizer supply process within the region is still close to the normal fertilizer supply status of historical high-yield samples.

[0039] Since different analysis windows have different durations and scopes of influence throughout the soybean growing cycle, preset window weights are assigned to each analysis window. This weight is derived from the proportion of the current window's time period to the entire soybean planting cycle, that is: The standardized deviation of each analysis window is correlated with the corresponding window weight. After multiplying and summing, the deviation of the fertilizer supply trajectory throughout the entire soybean growing cycle is obtained. :

[0040] In the formula, The total number of analysis windows; For analyzing the window index; For window Nutrient supply deficit; For window The range of supply fluctuations; For window The preset window weight. The deviation of the above fertilizer supply trajectory. The deviation of the overall fertilizer supply trajectory from the longer-lasting or higher-proportion fertility stages is calculated using a weighted summation method. To ensure corresponding contributions and avoid simplistic equal treatment of all aspects, high-yield plots in actual soybean cultivation typically do not rely solely on high total fertilizer application. Instead, they develop a nutrient supply rhythm that matches the soybean's growth needs, such as appropriate fertilization during the seedling stage, focused fertilization during flowering and pod formation, and continuous supplemental fertilization during grain filling. While the sample to be analyzed may have a similar total fertilizer application as the high-yield sample, if the timing of fertilization is unreasonable, the distribution of single fertilization amounts is unbalanced, or fertilization is insufficient during key growth stages, its nutrient supply process will deviate from the fertilization pattern exhibited by the high-yield sample. When the fertilization supply trajectory deviates by a certain amount... When the amount is small, it indicates that the nutrient supply process of the sample under analysis at each growth stage is relatively close to that of historical high-yield samples, and its fertilization management pattern is more in line with the high-yield formation pattern; when the fertilization supply trajectory deviates from the amount When the value is large, it indicates that the nutrient supply of the current sample at multiple growth stages has deviated significantly from the high-yield fertilizer supply pattern. Even if the total amount of fertilizer is similar, the yield response may deviate from the high-yield level due to unreasonable fertilization timing or insufficient fertilizer supply at key growth stages.

[0041] Step S130: Based on the historical fertilization supply trajectory deviation and its historical yield response data in the historical samples, determine the fertilization deviation yield response sensitivity coefficient corresponding to the current fertilization supply trajectory deviation.

[0042] It is understandable that while the deviation of the current fertilization supply trajectory determined in step S120 can quantify the degree of deviation of the fertilization process of the sample under analysis from the high-yield fertilization supply mode, the degree of deviation itself is not necessarily proportional to the yield decline. In actual soybean planting, even if two samples have similar deviations, their corresponding yield responses may differ. A continued increase in deviation within certain deviation intervals may lead to a rapid decline in the actual yield per unit area, while in other deviation intervals, even if the deviation changes, the yield change may not be significant. Therefore, relying solely on the deviation cannot accurately determine the strength of the yield response to changes in fertilization deviation near the current deviation position. In view of this, this embodiment of the invention further utilizes the correspondence between deviation and yield response in historical samples to analyze whether the yield near the current fertilization supply trajectory deviation will change significantly with changes in fertilization deviation, thereby determining the fertilization deviation yield response sensitivity coefficient to characterize the sensitivity of soybean yield to changes in fertilization supply trajectory deviation under the current fertilization deviation scenario, providing a basis for the response sensitivity of the deviation position for subsequent establishment of yield prediction relationships. The following describes a scheme for determining the fertilization deviation yield response sensitivity coefficient: Optionally, step S130 includes: determining the historical fertilizer supply trajectory deviation of each historical sample based on historical fertilizer management data and historical yield response data, and establishing a correspondence between the historical fertilizer supply trajectory deviation and historical yield response data; sorting each historical sample according to the historical fertilizer supply trajectory deviation from smallest to largest, and dividing it into several continuous deviation response intervals; determining the deviation response interval corresponding to the current fertilizer supply trajectory deviation, selecting the previous and next deviation response intervals adjacent to the deviation response interval, and determining the average fertilizer supply trajectory deviation and average historical yield response data of historical samples in the previous and next deviation response intervals respectively; and determining the fertilizer deviation yield response sensitivity coefficient based on the difference between the average historical yield response data of the previous and next deviation response intervals and the difference between the average fertilizer supply trajectory deviation of the previous and next deviation response intervals.

[0043] For each historical sample in the historical sample database, the deviation of its fertilizer supply trajectory is calculated using the same method as in step S120 above, and denoted as... ,in, , This represents the number of historical samples. Simultaneously, the actual yield per unit area corresponding to this historical sample is extracted from the collected historical production response data and denoted as... This leads to the formation of a historical sample database. Group Deviation-Production Response Data Establish the correspondence between the deviation of historical fertilizer supply trajectory and historical output response data.

[0044] Deviation of fertilizer supply trajectory from historical samples Deviation from the current fertilizer supply trajectory After merging, sort the historical samples from smallest to largest, and then divide the sorted historical samples into... A series of deviations from the response range , The preset number of intervals is determined based on the total number of historical samples, with a preferred range of 10 to 20. Each deviation response interval can be divided using the same number of historical samples to ensure a balanced sample size within each interval, suitable for situations with uneven sample distribution; alternatively, intervals can be divided using deviation ranges of equal width to ensure consistent deviation range coverage, facilitating intuitive comparison of production response characteristics under different degrees of deviation.

[0045] Optionally, the determination of the fertilizer deviation yield response sensitivity coefficient based on the difference in average historical yield response data between the previous and subsequent deviation response intervals, and the difference in average fertilizer supply trajectory deviation between the previous and subsequent deviation response intervals, includes: calculating the difference between the average historical yield response data of the previous and subsequent deviation response intervals to obtain the yield response difference; calculating the difference between the average fertilizer supply trajectory deviation of the subsequent deviation response interval and the average fertilizer supply trajectory deviation of the previous deviation response interval to obtain the deviation difference; calculating the ratio of the yield response difference to the deviation difference to obtain the initial sensitivity coefficient; and determining the fertilizer deviation yield response sensitivity coefficient based on the initial sensitivity coefficient; wherein the fertilizer deviation yield response sensitivity coefficient is the larger of the initial sensitivity coefficient and zero.

[0046] The difference between the average historical yield response data of the previous deviation response interval and the average historical yield response data of the subsequent deviation response interval is the yield response difference. This difference is obtained by subtracting the average actual yield per unit area from the average actual yield per unit area in the high deviation interval, and it reflects the average change in actual yield per unit area when the deviation of the fertilizer supply trajectory transitions from the low deviation interval to the high deviation interval. The difference between the average deviation of the fertilizer supply trajectory in the subsequent deviation response interval and the average deviation of the fertilizer supply trajectory in the previous deviation response interval is the deviation difference, which characterizes the average span of the degree of deviation between two adjacent deviation response intervals.

[0047] The formula for calculating the initial sensitivity coefficient is: In the formula, and These are the average actual yield per unit area of ​​historical samples within the previous and subsequent deviation response intervals, respectively. and These represent the average deviation of the fertilizer supply trajectory for historical samples within the latter and former deviation response intervals, respectively. It is a very small constant greater than zero, with units of and . (or The units are the same, which is used to prevent the denominator from being zero.

[0048] Since the initial sensitivity coefficient may be negative because the average yield in the high deviation range is not lower than that in the low deviation range, the increase in fertilization deviation does not cause a decrease in yield, and there is no yield reduction response sensitivity. Therefore, the initial sensitivity coefficient is compared with zero, and the larger of the two is taken as the fertilization deviation yield response sensitivity coefficient:

[0049] The above non-negative constraint means that the sensitivity coefficient only represents the sensitivity to positive production reduction. When the increase in deviation does not lead to a decrease in output, the sensitivity coefficient returns to zero.

[0050] In actual soybean cultivation, the above-mentioned fertilization deviation from the yield response sensitivity coefficient Its function is to: address deviations from the current fertilizer supply trajectory Near its historical deviation position, further observe the production differences between low and high deviation intervals in the historical sample. If from... arrive If the average yield decreases significantly during the process, it indicates that the current fertilizer supply trajectory deviates from the expected level. The surrounding area is a yield-sensitive zone; if the fertilizer supply trajectory continues to deviate from the high-yield pattern, it can easily lead to a significant reduction in yield. The change is relatively large. If the average output change between the preceding and following intervals is small, it indicates that the current deviation has little impact on output. Smaller or zero. Therefore, It can determine the sensitivity of the deviation in historical production response, providing a basis for the strength of the response at the deviation location in subsequent production forecasts.

[0051] It should be noted that the above scheme is only one way to determine the sensitivity coefficient of fertilizer deviation from yield, and not the only way. The embodiments of the present invention can also use one of the following schemes to determine the sensitivity coefficient of fertilizer deviation from yield: (1) Perform local linear regression within a subset of historical samples adjacent to the current fertilizer supply trajectory deviation, and use the slope of the regression line as the sensitivity coefficient. (2) Use kernel-weighted least squares, assign kernel weights based on the distance between each historical sample deviation and the current fertilizer supply trajectory deviation, and fit a local linear relationship, using the resulting slope as the sensitivity coefficient.

[0052] Step S140: Based on the fertilization deviation yield response sensitivity coefficient and the critical position of the current fertilization supply trajectory deviation within its respective deviation response interval, determine the critical response intensity of fertilization deviation.

[0053] It is understandable that the aforementioned deviation from the fertilizer supply trajectory characterizes the degree to which the fertilization process of the analyzed sample deviates from the high-yield fertilizer supply pattern throughout the entire planting cycle. The fertilizer deviation yield response sensitivity coefficient reflects the sensitivity of yield near the current deviation position to changes in fertilization deviation. However, the fertilizer deviation yield response sensitivity coefficient itself does not distinguish the specific position of the analyzed sample within its respective deviation response interval. In actual soybean planting, even if two samples fall into the same deviation response interval, their yield risk status may differ. If the sample's deviation has just entered this interval, it indicates that its deviation is still close to a low deviation state. If the sample's deviation is close to the upper boundary of this interval, its fertilizer supply trajectory is about to enter a higher deviation interval. At this time, if the sensitivity coefficient is large, it indicates that the sample is not only in the yield-sensitive area but also close to a higher deviation state, and the impact of its fertilization process on subsequent yield prediction should be further amplified. Therefore, this invention constructs a critical response intensity for fertilization deviation based on the degree of critical approximation of the current fertilization supply trajectory deviation within its corresponding response interval. This intensity characterizes whether the current sample is in a highly sensitive response state approaching a higher deviation interval, thereby quantifying and distinguishing the differences in yield risk at different critical positions within the same interval. The following describes a scheme for determining the critical response intensity for fertilization deviation: Optionally, step S140 includes: determining the lower and upper boundaries of the deviation response interval to which the current fertilization supply trajectory deviation belongs; determining the critical position of the interval based on the degree of approximation between the current fertilization supply trajectory deviation and the lower and upper boundaries; and determining the critical response intensity of the fertilization deviation based on the fertilization deviation yield response sensitivity coefficient and the critical position of the interval.

[0054] Assuming the current fertilizer supply trajectory deviates by a certain amount The deviation response range is Within this interval, the deviation of the fertilizer supply trajectory of the sample to be analyzed satisfies The lower boundary represents the lowest degree of deviation of historical samples within the deviation response interval, while the upper boundary represents the highest degree of deviation of historical samples within the interval, and is also the critical threshold for entering a higher deviation interval.

[0055] The critical position of the interval is determined by the relative relationship between the deviation of the current fertilizer supply trajectory and the lower and upper boundaries. The specific calculation method is as follows: calculate the ratio of the difference between the deviation of the current fertilizer supply trajectory and the lower boundary to the difference between the deviation of the current fertilizer supply trajectory and the upper boundary. ,in, This is a very small constant greater than zero, with units equal to the deviation from the fertilizer supply trajectory, used to avoid a denominator of zero. The above ratio... Characterizes the degree to which the sample to be analyzed approaches the upper boundary from the lower boundary within the current deviation response interval: when Approaching the lower boundary When the numerator in the above ratio is small, the ratio approaches zero, indicating that the current sample has just entered the deviation response interval; when Approaching the upper boundary When the denominator is small and the ratio increases significantly, it indicates that the current sample is approaching a higher deviation interval.

[0056] Regarding the above ratio Taking the natural logarithm, we get Logarithmic operations, while preserving the critical approximation trend, avoid excessive numerical amplification due to proximity to the upper boundary of the interval, resulting in a more stable representation of the critical position. The above logarithmic results are then compared with the fertilization deviation yield response sensitivity coefficient. Multiplying these yields the fertilization deviation from the critical response strength. :

[0057] In the formula, The sensitivity coefficient for fertilizer application deviation from yield response; This represents the deviation from the current fertilizer supply trajectory. and These are the lower and upper boundaries of the current deviation from the response interval, respectively. It is a very small constant that is greater than zero; This represents the deviation of fertilization from the critical response strength. It can be understood that the above-mentioned sensitivity coefficient to yield response deviation... It can only characterize whether the output near the current deviation position is sensitive, but it does not distinguish whether the sample to be analyzed is at a low, middle, or near the high deviation boundary within this response interval. By introducing the interval critical position and comparing it with... Multiplication allows the critical approximation state to be amplified or weakened by incorporating the yield response sensitivity of the current deviation. When the current deviation from the fertilizer supply trajectory just enters the range, even if the fertilizer deviation is at the yield response sensitivity coefficient... The fertilization rate is relatively high, deviating significantly from the critical response strength. It won't be too high either. Currently, the deviation from the fertilizer supply trajectory is already close to the upper boundary of the range, and the sensitivity coefficient of fertilizer deviation to yield response is... When the value is large, fertilization deviates from the critical response strength. The significant increase indicates that the sample is in a state where the deviation is about to escalate and the yield response is sensitive, and the impact of its fertilization process on subsequent yield prediction is further amplified.

[0058] It should be noted that the above scheme is only one way to determine the critical response intensity of fertilization deviation, and not the only way. The embodiments of the present invention can also use one of the following schemes to determine the critical response intensity of fertilization deviation: (1) Divide the difference between the current fertilization supply trajectory deviation and the lower boundary by the interval width to obtain a linear ratio, and then multiply this linear ratio by the fertilization deviation yield response sensitivity coefficient to obtain the critical response intensity of fertilization deviation. (2) Use the ratio of the difference between the current fertilization supply trajectory deviation and the lower boundary to the difference between the current fertilization supply trajectory deviation and the upper boundary as an exponential term, and multiply this exponential term by the sensitivity coefficient to achieve exponential amplification, thus obtaining the critical response intensity of fertilization deviation.

[0059] Step S150: Determine the predicted soybean yield based on the fertilization deviation from the critical response intensity, the deviation of the current fertilization supply trajectory, and the historical average yield of the current deviation response interval.

[0060] Optionally, step S150 includes: obtaining the historical average yield of the current deviation response interval; determining the critical response reduction amount based on the intensity of the fertilization deviation from the critical response and the degree of deviation of the current fertilization supply trajectory from the lower boundary of its respective deviation response interval; and determining the predicted soybean yield based on the historical average yield and the critical response reduction amount.

[0061] Based on the deviation response interval to which the current fertilizer supply trajectory deviation belongs, the actual yield per unit area of ​​all historical samples within this interval is extracted, and their arithmetic mean is calculated as the historical average yield of this deviation response interval, denoted as . The historical average yield represents the average yield level of historical plots within the deviation interval of the plot to be predicted, providing a benchmark for subsequent predictions.

[0062] The critical response reduction is jointly determined by the intensity of the deviation from the critical response and the degree of deviation of the current fertilization supply trajectory from the lower boundary of the interval. (Current fertilization supply trajectory deviation) Its lower boundary of the deviation response interval difference This can characterize the degree to which the plot has deviated from the lower end of the current deviation response range, and correlate this degree of deviation with the critical response intensity of fertilization deviation. Multiplying them together yields the critical response reduction. .

[0063] Predicted yield per unit area of ​​the plot to be predicted The critical response reduction is obtained by subtracting the historical average output from the current deviation from the response range.

[0064] In the formula, This represents the average actual yield per unit area of ​​historical samples within the current deviation response interval; The fertilization rate deviated from the critical response strength; This represents the deviation from the current fertilizer supply trajectory. This represents the lower boundary of the current deviation from the response interval; To predict soybean yield.

[0065] Predicted output per unit area The calculation formula includes historical average yield. As a forecasting baseline, it reflects the average yield level of historical plots within the current deviation range. Critical response reduction. This is used to reflect the additional yield risk of the current sample relative to the lower end of the range. When fertilization deviates from the critical response strength... When the deviation from the current fertilization supply trajectory is significant and close to the upper boundary of the interval, the critical response reduction increases significantly, and the predicted yield is significantly lower than the historical average yield, indicating that the plot is in a high-risk state where the deviation is about to escalate and the yield response is sensitive. When fertilization deviates from the critical response intensity... Smaller or current deviation from the fertilizer supply trajectory When the critical response reduction is small as the plot approaches the lower boundary of the interval, the predicted output is close to the historical average output, indicating that the output risk of this plot is relatively controllable.

[0066] After the current planting cycle ends and the actual yield per unit area of ​​the plot is obtained, the fertilization time, fertilization frequency, fertilizer application rate per unit area, fertilizer nutrient content, and actual yield per unit area of ​​the plot throughout the entire planting cycle are added to the historical sample database, so that the sample size of the historical sample database continues to expand with the production cycle. On this basis, the high-yield sample set is periodically re-screened and the baseline nutrient supply and supply fluctuation range of each analysis window are recalculated. At the same time, the deviation response interval is re-divided and the calculation benchmarks for the interval average yield, fertilization deviation yield response sensitivity coefficient, and fertilization deviation critical response intensity are updated. This ensures that the fertilization supply baseline trajectory, deviation response interval structure, and sensitivity coefficient calculation in subsequent prediction processes can all reflect the yield response characteristics of the latest historical samples, achieving adaptive soybean yield prediction with automatic correction of prediction results based on newly added historical samples.

[0067] Understandably, under conditions of high deviation from fertilization supply, the aforementioned schemes can only identify yield risks caused by deviations in fertilization supply trajectories from high-yield patterns, and cannot directly intervene to correct the nutrient supply process. If only prediction is maintained without targeted supplemental fertilization control measures, even if the risk of yield decline is known in advance, it is difficult to recover yield losses through dynamic adjustments in fertilization management. Therefore, this invention provides an adaptive fertilization setting scheme based on yield prediction. This scheme adaptively sets the fertilization time, frequency, fertilizer type, nutrient type, and fertilization amount per unit area for plots in a high deviation state, transforming the prediction results into actual fertilization control actions. By intervening in fertilization to correct the deviation in the nutrient supply trajectory, the nutrient supply rhythm of the plot during the remaining growth stage moves closer to the high-yield fertilization supply pattern. An optional adaptive fertilization setting scheme is described below: Optionally, after step S150 above, the above-mentioned big data-based adaptive soybean yield prediction method further includes: determining that the plot to be predicted is in a high-deviation fertilizer supply state when the deviation of the current fertilizer supply trajectory is greater than a preset deviation threshold, or the critical response intensity of fertilizer deviation is greater than a preset response intensity threshold; extracting historical samples from the historical sample library that are consistent with or within a preset range of the critical response intensity of fertilizer deviation for the plot to be predicted, and whose actual yield per unit area is within a preset high-yield range, and establishing a mapping relationship between the critical response intensity of fertilizer deviation and fertilizer supplementation parameters; wherein, the fertilizer supplementation parameters include fertilizer supplementation time, fertilizer supplementation frequency, and fertilizer type. The fertilizer type, nutrient type, and fertilizer application rate per unit area are determined. Based on the fertilization deviation from the critical response intensity of the plot to be predicted, the corresponding historical fertilizer application parameters are matched in the mapping relationship. The appropriate fertilizer type and nutrient type are determined from the preset candidate fertilizer schemes, and the fertilizer application rate per unit area is adaptively set in combination with the current monitoring data. The current monitoring data includes at least one of the following: soil nutrient status, soil moisture, plant growth, growth stage, and environmental conditions. After the current planting cycle ends and the actual yield per unit area is obtained, the fertilization management data, fertilizer application parameters, and actual yield per unit area of ​​the plot to be predicted are added to the historical sample database.

[0068] Once the predicted soybean yield for the plot is determined, if the current fertilization supply trajectory deviation exceeds a preset deviation threshold, or the critical response intensity of the fertilization deviation exceeds a preset response intensity threshold, the plot is determined to be in a high-deviation fertilization supply state. The preset deviation threshold can be set based on the statistical distribution of all fertilization supply trajectory deviations in a historical sample database, for example, by taking the upper quartile or a specific percentile of historical deviations. The preset response intensity threshold can be set based on the statistical distribution of all critical response intensities of fertilization deviations in a historical sample database. It can be understood that when the current fertilization supply trajectory deviation exceeds the preset deviation threshold, or the critical response intensity of the fertilization deviation exceeds the preset response intensity threshold, it indicates that the current fertilization process of the plot has significantly deviated from the high-yield fertilization supply pattern, or is in a critical high-risk state about to enter a higher deviation range, necessitating intervention through supplemental fertilization to correct the nutrient supply trajectory.

[0069] Under conditions of high deviation from the fertilization supply, historical samples with values ​​consistent with or within a preset range of the critical fertilization deviation response intensity of the plot to be predicted are retrieved from the historical sample database. Simultaneously, samples whose actual yield per unit area falls within a preset high-yield range are selected. This preset range can be set by the user according to local planting precision requirements; for example, historical samples within one standard deviation of the current critical fertilization deviation response intensity are selected. The preset high-yield range can use the same selection ratio as the high-yield sample set. The selected historical samples not only exhibit deviation response characteristics similar to the plot to be predicted but also achieve high yields, making their fertilization experience valuable. Based on these historical samples, a mapping relationship between the critical fertilization deviation response intensity and fertilization parameters is established. This mapping relationship uses the critical fertilization deviation response intensity as the index key and the fertilization parameters of the corresponding historical samples as the mapping values.

[0070] The above-mentioned fertilization parameters include fertilization time, fertilization frequency, fertilizer type, nutrient type, and fertilization amount per unit area. Specifically, fertilization time refers to the exact moment of application relative to the timeline after sowing; fertilization frequency refers to the total number of applications (excluding basal fertilizer) within a planting cycle; fertilizer type refers to the specific variety and specifications of the commercial fertilizer; nutrient type refers to the types of nutrients contained in the fertilizer, such as macronutrients like nitrogen, phosphorus, and potassium; and fertilization amount per unit area refers to the mass of fertilizer applied per unit area during each application.

[0071] Based on the deviation of the fertilization from the critical response intensity of the plot to be predicted, the corresponding historical fertilization parameters are matched in the above mapping relationship. The fertilizer type and nutrient type are determined from a pre-configured candidate fertilizer scheme, which can be pre-entered by the user based on local agricultural input supply and soil nutrient background conditions. After determining the fertilizer type and nutrient type, the fertilization amount per unit area is adaptively adjusted based on current monitoring data. Current monitoring data includes at least one of the following: soil nutrient status, soil moisture, plant growth, growth stage, and environmental conditions.

[0072] Soil nutrient status is obtained through soil nutrient sensors or laboratory testing, reflecting the current surplus or deficiency of available nutrients in the soil; soil moisture is obtained through moisture monitoring equipment, reflecting soil water conditions; plant growth is obtained through drone remote sensing imagery or manual field observation, reflecting soybean leaf color, plant height, and biomass accumulation; growth stage is determined based on a comprehensive assessment of the relative time after sowing and plant morphological characteristics; environmental conditions include current and short-term forecasts of temperature, precipitation, and light. The above monitoring data are used to incrementally adjust fertilization parameters based on historical data. For example, when soil nutrients are abundant, the fertilization amount per unit area is appropriately reduced; when plant growth is weak, the fertilization amount per unit area is appropriately increased, thus ensuring that the fertilization amount is set to meet the actual growth needs of the current plot.

[0073] After the previous planting cycle ends and the actual yield per unit area of ​​the plot is obtained, the fertilization management data, actual fertilization parameters, and actual yield per unit area of ​​the plot throughout the entire planting cycle are added to the historical sample database. This supplementation operation allows the historical sample database to continuously accumulate complete samples containing the effects of fertilization intervention, providing a data foundation for screening high-yield samples in subsequent planting cycles, dividing deviation response intervals, calculating the sensitivity coefficient of fertilization deviation from yield response, and updating the mapping relationship of fertilization parameters. This enables the soybean yield prediction results and adaptive fertilization setting results to be automatically corrected with the addition of new historical samples.

[0074] Understandably, fertilization methods include basal application of organic fertilizer, basal application of chemical fertilizer, and foliar spraying. Basal organic fertilizer provides slow-release nutrients through soil amendment, basal chemical fertilizer provides readily available nutrients through rhizosphere application, and foliar fertilizer directly supplements micronutrients or readily available nutrients through leaf surface spraying. In the aforementioned adaptive fertilization setting scheme, the adaptively set fertilization amount corresponds to the unit area spraying amount of foliar fertilizer. When soil moisture is insufficient or root absorption capacity is limited, foliar spraying bypasses the rhizosphere absorption process, directly supplementing nutrients to the plant. At this time, based on the fertilization deviation from the critical response intensity and the plant growth and development stage in the current monitoring data, the unit area spraying amount, spraying time, and spraying frequency of foliar fertilizer are determined, achieving adaptive setting of foliar fertilization amount.

[0075] It should be noted that the embodiments of the present invention can also adopt the following schemes to adaptively set the amount of fertilizer per unit area: (1) Calculate the amount of fertilizer per unit area based on the difference between the current soil nutrient status and the nutrient requirements corresponding to the target yield, combined with the fertilizer nutrient content. (2) Compare the plant growth index with the growth benchmark of the corresponding growth stage of the high-yield sample, and adjust the amount of fertilizer per unit area according to the growth deviation according to the preset ratio. (3) Reduce and correct the amount of fertilizer per unit area obtained by historical mapping based on the precipitation forecast and temperature stress index in the current environmental conditions, so as to avoid the risk of excessive fertilization under unfavorable environmental conditions. (4) Correct the amount of fertilizer per unit area based on soil moisture data, and reduce the amount of fertilizer per unit area or irrigate simultaneously when the soil moisture is insufficient to improve fertilizer utilization. The above schemes are all based on the current monitoring data to quantitatively correct the amount of fertilizer per unit area obtained by historical mapping, and can be implemented under the condition that the corresponding monitoring data is available.

[0076] like Figure 2 As shown, based on the same inventive concept, this embodiment of the invention also provides a soybean yield adaptive prediction system 200 based on big data, including a data acquisition module 210, a current fertilizer supply trajectory deviation determination module 220, a fertilizer deviation yield response sensitivity coefficient determination module 230, a fertilizer deviation critical response intensity determination module 240, and a soybean predicted yield determination module 250 connected in sequence, wherein: Data acquisition module 210 is used to acquire fertilization management data and yield response data during the soybean planting cycle; The current fertilizer supply trajectory deviation determination module 220 is used to obtain the fertilizer supply trajectory of the sample to be analyzed based on fertilizer management data, and determine the deviation of the current fertilizer supply trajectory from the fertilizer supply benchmark trajectory. The fertilizer deviation yield response sensitivity coefficient determination module 230 is used to determine the fertilizer deviation yield response sensitivity coefficient corresponding to the current fertilizer supply trajectory deviation based on the historical fertilizer supply trajectory deviation and its historical yield response data in the historical samples. The fertilizer deviation critical response intensity determination module 240 is used to determine the fertilizer deviation critical response intensity based on the fertilizer deviation yield response sensitivity coefficient and the critical position of the current fertilizer supply trajectory deviation within its respective deviation response interval. The soybean yield prediction module 250 is used to determine the soybean yield based on the intensity of the fertilization deviation from the critical response, the deviation of the current fertilization supply trajectory, and the historical average yield of the current deviation response interval.

[0077] It should be noted that the above-mentioned soybean yield adaptive prediction system 200 based on big data can realize any one of the functions of the soybean yield adaptive prediction method based on big data provided in the embodiments of the present invention. For the method embodiment section, please refer to the method embodiment section for the way each function is realized and the working principle. The system embodiment section will not repeat it here.

[0078] This invention is now complete.

[0079] In summary, in this embodiment of the invention, fertilization management data and yield response data during the soybean planting cycle are obtained; based on the fertilization management data, the fertilization supply trajectory of the sample to be analyzed is obtained, and the deviation of the current fertilization supply trajectory from the fertilization supply baseline trajectory is determined; based on the historical fertilization supply trajectory deviation and its historical yield response data in historical samples, the fertilization deviation yield response sensitivity coefficient corresponding to the current fertilization supply trajectory deviation is determined; based on the fertilization deviation yield response sensitivity coefficient and the critical position of the current fertilization supply trajectory deviation within its respective deviation response interval, the critical response intensity of the fertilization deviation is determined; based on the critical response intensity of the fertilization deviation, the current fertilization supply trajectory deviation, and the historical average yield of the current deviation response interval, the predicted soybean yield is determined.

[0080] This invention establishes a mapping relationship between differences in fertilization management processes and yield prediction results by constructing fertilization supply trajectory deviation, fertilization deviation yield response sensitivity coefficient, and fertilization deviation critical response intensity. This allows the prediction results to reflect the effect of fertilization regulation, achieving adaptive soybean yield prediction based on fertilization management processes. On the other hand, by accumulating fertilization time, fertilizer application rate per unit area, and fertilizer nutrient content according to growth stage windows to form a fertilization supply trajectory, and comparing its deviation with the baseline trajectory of historical high-yield samples, the discrete fertilization management process is transformed into a quantifiable nutrient supply rhythm deviation index, thereby distinguishing yield response differences in soybean plots with similar environmental conditions but different fertilization management processes. Furthermore, by analyzing the correspondence between fertilization supply trajectory deviation and yield response in historical samples, deviation response intervals are divided, and the yield change rate between adjacent intervals is calculated to obtain the yield response to fertilization deviation. Sensitivity coefficients are used to identify the degree to which yields at different deviation locations respond to changes in fertilization deviation, avoiding a simple linear correlation between deviation and yield decline. Furthermore, by combining the yield response sensitivity coefficient of fertilization deviation with the critical position of the current fertilization supply trajectory deviation within its corresponding response interval, the critical response intensity of fertilization deviation is calculated. This allows for the determination of whether the plot to be predicted is approaching a critical state of a higher deviation interval, enabling early response to yield risks that are about to enter a high deviation interval. In addition, by establishing a mapping relationship between the critical response intensity of fertilization deviation and fertilization parameters under high fertilization deviation conditions, and by adaptively setting the fertilization amount per unit area based on current monitoring data, fertilization decisions can be adjusted specifically based on the fertilization experience of historical high-yield samples. This corrects nutrient supply trajectory deviation through fertilization intervention, and the fertilization amount setting results are automatically corrected with the periodic updates of the historical sample database.

[0081] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A soybean yield adaptive forecasting method based on big data, characterized in that, The method includes: Acquire fertilization management data and yield response data during the soybean growing cycle; Based on the fertilization management data, the fertilization supply trajectory of the sample to be analyzed is obtained, and the deviation of the current fertilization supply trajectory from the fertilization supply baseline trajectory is determined. Based on the historical fertilization supply trajectory deviation and its historical output response data in the historical sample, determine the fertilization deviation output response sensitivity coefficient corresponding to the current fertilization supply trajectory deviation. Based on the fertilizer deviation response sensitivity coefficient and the critical position of the current fertilizer supply trajectory deviation within its respective deviation response interval, the critical response intensity of fertilizer deviation is determined. The predicted soybean yield is determined based on the fertilization deviation from the critical response intensity, the deviation of the current fertilization supply trajectory, and the historical average yield of the current deviation response interval.

2. The adaptive soybean yield prediction method based on big data according to claim 1, characterized in that, The determination of the fertilizer deviation yield response sensitivity coefficient corresponding to the current fertilizer supply trajectory deviation based on historical fertilizer supply trajectory deviation and its historical yield response data from historical samples includes: Based on the historical fertilization management data and historical yield response data of each historical sample, the deviation of the historical fertilization supply trajectory for each historical sample is determined, and the correspondence between the deviation of the historical fertilization supply trajectory and the historical yield response data is established. The historical samples are sorted from smallest to largest according to the deviation of the historical fertilizer supply trajectory, and divided into several consecutive deviation response intervals; Determine the deviation response interval corresponding to the current deviation amount of fertilizer supply trajectory, select the previous deviation response interval and the next deviation response interval adjacent to the deviation response interval, and determine the average fertilizer supply trajectory deviation amount and average historical yield response data of historical samples in the previous deviation response interval and the next deviation response interval respectively; Based on the difference in average historical yield response data between the previous deviation response interval and the subsequent deviation response interval, as well as the difference in average fertilizer supply trajectory deviation between the previous deviation response interval and the subsequent deviation response interval, the fertilizer deviation yield response sensitivity coefficient is determined.

3. The adaptive soybean yield prediction method based on big data according to claim 2, characterized in that, The determination of the fertilizer deviation yield response sensitivity coefficient based on the difference in average historical yield response data between the previous and subsequent deviation response intervals, and the difference in average fertilizer supply trajectory deviation between the previous and subsequent deviation response intervals, includes: Calculate the difference between the average historical production response data of the previous deviation response interval and the average historical production response data of the next deviation response interval to obtain the production response difference; Calculate the difference between the average fertilizer supply trajectory deviation in the subsequent deviation response interval and the average fertilizer supply trajectory deviation in the previous deviation response interval to obtain the deviation difference; Calculate the ratio of the difference in production response to the difference in deviation to obtain the initial sensitivity coefficient; Based on the initial sensitivity coefficient, a fertilization deviation-yield response sensitivity coefficient is determined; wherein, the fertilization deviation-yield response sensitivity coefficient is the larger of the initial sensitivity coefficient and zero.

4. The adaptive soybean yield prediction method based on big data according to claim 1, characterized in that, The determination of the critical response intensity of fertilization deviation based on the sensitivity coefficient of the fertilization deviation from yield and the critical position of the current fertilization supply trajectory deviation within its respective deviation response interval includes: Determine the lower and upper boundaries of the deviation response interval to which the current fertilizer supply trajectory deviation belongs; Based on the degree of approximation between the current fertilizer supply trajectory deviation and the lower and upper boundaries, the critical position of the interval is determined; Based on the sensitivity coefficient of the fertilization deviation from yield and the critical position of the interval, the critical response intensity of the fertilization deviation is determined.

5. The adaptive soybean yield prediction method based on big data according to claim 1, characterized in that, The process of determining the predicted soybean yield based on the fertilization deviation from the critical response intensity, the deviation of the current fertilization supply trajectory, and the historical average yield of the current deviation response interval includes: Obtain the historical average output of the current deviation from the response interval; Based on the intensity of the fertilization deviation from the critical response and the degree of deviation of the current fertilization supply trajectory from the lower boundary of its respective deviation response interval, the critical response reduction amount is determined. The predicted soybean yield is determined based on the historical average yield and the critical response reduction.

6. The adaptive soybean yield forecasting method based on big data according to any one of claims 1-5, characterized in that, The step of obtaining the fertilizer supply trajectory of the sample to be analyzed based on the fertilizer management data, and determining the deviation of the current fertilizer supply trajectory from the fertilizer supply baseline trajectory, includes: The soybean planting cycle is divided into several continuous analysis windows according to preset fixed time intervals or soybean growth stages; wherein, the soybean growth stages include seedling stage, branching stage, flowering stage, grain filling stage and maturity stage; For each analysis window, the cumulative nutrient supply for that analysis window is determined based on the fertilization management data corresponding to that analysis window. Based on the cumulative nutrient supply of all the analysis windows, the fertilizer supply trajectory is obtained; Based on historical samples where the actual yield per unit area is within a preset high-yield range, the baseline nutrient supply and supply fluctuation range corresponding to each analysis window are determined. Based on the baseline nutrient supply of all analysis windows, obtain the fertilizer supply baseline trajectory; Based on the difference between the cumulative nutrient supply and the baseline nutrient supply in each analysis window, as well as the range of supply fluctuations, the deviation of the current fertilization supply trajectory is determined.

7. The adaptive soybean yield prediction method based on big data according to claim 6, characterized in that, The method of determining the deviation of the current fertilization supply trajectory based on the difference between the cumulative nutrient supply and the baseline nutrient supply in each analysis window, and the range of supply fluctuation, includes: Calculate the absolute value of the difference between the cumulative nutrient supply and the baseline nutrient supply for each analysis window to obtain the nutrient supply deviation for each analysis window; Calculate the ratio of the nutrient supply deviation to the supply fluctuation range for each analysis window, and obtain the standardized deviation value for each analysis window; Based on the standardized deviation value of each analysis window and the preset window weight, the deviation of the current fertilizer supply trajectory is determined; wherein, the preset window weight is determined based on the proportion of the duration of the analysis window within the soybean planting cycle.

8. The adaptive soybean yield forecasting method based on big data according to any one of claims 1-5, characterized in that, After determining the predicted soybean yield based on the fertilization deviation critical response intensity, the deviation amount of the current fertilization supply trajectory, and the historical average yield of the current deviation response interval, the method further includes: When the deviation of the current fertilizer supply trajectory is greater than a preset deviation threshold, or when the critical response intensity of the fertilizer deviation is greater than a preset response intensity threshold, the plot to be predicted is determined to be in a high deviation fertilizer supply state. Historical samples are extracted from the historical sample database that are consistent with or within a preset range of the fertilization deviation critical response intensity of the plot to be predicted, and whose actual yield per unit area is within a preset high-yield range. A mapping relationship between the fertilization deviation critical response intensity and the fertilization parameters is established. The fertilization parameters include fertilization time, fertilization frequency, fertilizer type, nutrient type, and fertilization amount per unit area. Based on the critical response intensity of the fertilization deviation of the plot to be predicted, the corresponding historical fertilization parameters are matched in the mapping relationship, the appropriate fertilizer type and nutrient type are determined from the preset candidate fertilizer schemes, and the fertilization amount per unit area is adaptively set in combination with the current monitoring data; wherein, the current monitoring data includes at least one of soil nutrient status, soil moisture, plant growth, growth stage and environmental conditions. After the current planting cycle ends and the actual yield per unit area is obtained, the fertilization management data, fertilization parameters, and actual yield per unit area of ​​the plot to be predicted will be added to the historical sample database.

9. The adaptive soybean yield forecasting method based on big data according to any one of claims 1-5, characterized in that, The acquisition of fertilization management data and yield response data during the soybean planting cycle includes: Fertilization management data and yield response data are collected during the soybean planting cycle; wherein, the fertilization management data includes the fertilization time, fertilizer application rate per unit area, nutrient type, and fertilizer nutrient content for each fertilization; the yield response data includes the actual yield per unit area. The fertilizer management data and the yield response data are preprocessed; wherein the preprocessing includes: Based on plot identifiers and planting cycle identifiers, the fertilization management data and yield response data of the same plot are associated and matched. Convert the fertilization time into a relative time after sowing; The fertilizer management data and the yield response data shall be standardized to the same measurement format; Organize the fertilization records of the same plot in chronological order of fertilization time.

10. A soybean yield adaptive forecasting system based on big data, characterized in that, This includes, in sequence, a data acquisition module, a module for determining the deviation of the current fertilizer supply trajectory, a module for determining the sensitivity coefficient of the yield response to fertilizer deviation, a module for determining the critical response intensity of fertilizer deviation, and a module for determining soybean predicted yield, wherein: The data acquisition module is used to acquire fertilization management data and yield response data during the soybean planting cycle; The current fertilizer supply trajectory deviation determination module is used to obtain the fertilizer supply trajectory of the sample to be analyzed based on the fertilizer management data, and determine the deviation of the current fertilizer supply trajectory from the fertilizer supply benchmark trajectory. The fertilizer deviation yield response sensitivity coefficient determination module is used to determine the fertilizer deviation yield response sensitivity coefficient corresponding to the current fertilizer supply trajectory deviation based on the historical fertilizer supply trajectory deviation and its historical yield response data in the historical samples. The fertilizer deviation critical response intensity determination module is used to determine the fertilizer deviation critical response intensity based on the fertilizer deviation yield response sensitivity coefficient and the critical position of the current fertilizer supply trajectory deviation within its respective deviation response interval. The soybean yield prediction module is used to determine the soybean yield based on the fertilization deviation critical response intensity, the current fertilization supply trajectory deviation, and the historical average yield of the current deviation response interval.

Images