A wheat-cotton interplanting differentiated water and fertilizer integrated intelligent management system

Abstract

This invention discloses a differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping, relating to the field of agricultural intelligent irrigation and water and fertilizer management technology. It includes a field deployment module, a data acquisition module, a differentiated irrigation and fertilization decision-making module, and an intelligent branch control module. The intelligent branch control module is connected to the field deployment module, the data acquisition module, and the differentiated irrigation and fertilization decision-making module respectively. The field deployment module is used to lay shallow-buried drip irrigation tape in the wheat strip and mulched drip irrigation tape in the cotton strip. The shallow-buried drip irrigation tape is buried below the soil surface between the wheat rows, and the mulched drip irrigation tape is laid under the mulch film between the cotton rows. The shallow-buried drip irrigation tape and the mulched drip irrigation tape are each independently connected to corresponding branch water supply pipelines. This invention achieves independent water and fertilizer supply for the wheat strip and the cotton strip by laying shallow-buried drip irrigation tape in the wheat strip and mulched drip irrigation tape in the cotton strip, and connecting them to independent branch water supply pipelines.

Description

Technical Field

[0001] This invention relates to the field of agricultural intelligent irrigation and water and fertilizer management technology, and in particular to a differentiated integrated water and fertilizer management system for wheat-cotton intercropping. Background Technology

[0002] Wheat-cotton intercropping is a composite planting model based on the growth habits and developmental characteristics of wheat and cotton. It typically involves placing wheat and cotton strips adjacent to each other within the same field, allowing for complementary use of food and cash crops in both time and space. During the wheat-cotton co-growth period, wheat is mostly in the grain-filling stage, while cotton is mostly in the seedling stage. After wheat harvest, cotton gradually enters the budding and boll-forming stages. Different crops and different growth stages have different water and nutrient requirements and root water absorption characteristics. Integrated water and fertilizer technology usually achieves simultaneous water and fertilizer supply through drip irrigation tape, water supply pipelines, water and fertilizer mixing devices, solenoid valves, soil temperature and humidity sensors, and control units. Irrigation and fertilization management is based on information such as soil moisture, crop growth, meteorological data, reference evapotranspiration, and crop coefficients.

[0003] Existing irrigation and fertilization systems are mostly designed for single crops or general zoned planting scenarios. They typically use soil moisture thresholds for uniform irrigation control, lacking independent slurry supply structures for wheat and cotton intercropping scenarios. They also struggle to dynamically calculate target irrigation amounts, target fertilizer amounts, and fertilizer formulas based on the individual growth stages of wheat and cotton, root water content, crop growth, reference evapotranspiration, and crop coefficients. Furthermore, existing decision-making methods do not adequately consider the wheat-cotton co-existence period, the transition between cotton budding and boll-forming stages after wheat harvest, predictions of future soil moisture changes, and the comprehensive optimization of yield, water use efficiency, fertilizer use efficiency, and environmental risks. This makes it difficult to achieve differentiated, dynamic, and forward-looking precision water and fertilizer management in wheat-cotton intercropping fields. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping to solve the problem that the water and fertilizer requirements of wheat and cotton strips are inconsistent in the wheat-cotton intercropping mode, and that it is difficult to achieve independent supply, dynamic decision-making and forward-looking regulation for each strip with a unified irrigation and fertilization method.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping, comprising a field deployment module, a data acquisition module, a differentiated irrigation and fertilization decision module, and an intelligent branch control module. The intelligent branch control module is connected to the field deployment module, the data acquisition module, and the differentiated irrigation and fertilization decision module, respectively. The field deployment module is used to lay shallow-buried drip irrigation tape in the wheat strip and mulched drip irrigation tape in the cotton strip. The shallow-buried drip irrigation tape is buried below the soil surface between the wheat rows, and the mulched drip irrigation tape is laid under the mulch film between the cotton rows. The shallow-buried drip irrigation tape and the mulched drip irrigation tape are independently connected to corresponding branch water supply pipelines. The data acquisition module is used to collect soil volumetric water content and soil temperature at different depths in the wheat strip and cotton strip through soil temperature and humidity sensors deployed in the wheat strip and the cotton strip, and to measure these data through crop growth... The monitoring device collects crop growth data for the wheat and cotton strips; the differentiated irrigation and fertilization decision module is used to calculate the root zone weighted average water content, target irrigation amount, target fertilization amount, and fertilization formula for the wheat and cotton strips respectively, based on the differentiated water and fertilizer requirements parameters of the wheat and cotton at their respective growth stages, combined with soil volumetric water content, soil temperature, crop growth data, reference evapotranspiration, and crop coefficient collected by the data acquisition module, and generates irrigation and fertilization control parameters for the wheat and cotton strips respectively; the intelligent branch control module is used to receive the irrigation and fertilization control parameters for the wheat and cotton strips through the control unit, and control the solenoid valves and water and fertilizer mixing devices set on each branch water supply pipeline respectively, so that the shallow buried drip irrigation tape and the film-covered drip irrigation tape can independently supply water and fertilizer to the wheat and cotton strips according to the corresponding target irrigation amount, target fertilization amount, and fertilization formula.

[0007] As a preferred embodiment of the differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping described in this invention, the branch water supply pipeline includes a wheat strip water supply pipeline connected to a shallow buried drip irrigation tape and a cotton strip water supply pipeline connected to a film-covered drip irrigation tape. The wheat strip water supply pipeline and the cotton strip water supply pipeline are each equipped with independently controlled solenoid valves.

[0008] As a preferred embodiment of the differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping described in this invention, the soil temperature and humidity sensors are arranged in layers along the root depth direction of the wheat and cotton strips to collect soil volumetric water content and soil temperature at different soil layers. The data acquisition module collects soil volumetric water content and soil temperature at a preset acquisition frequency, and generates soil temperature and humidity data for the corresponding wheat and cotton belts during the decision-making period.

[0009] As a preferred embodiment of the differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping described in this invention, the crop growth monitoring device includes a multispectral camera. The multispectral camera is used to acquire multispectral images of wheat and cotton belts; The crop growth data includes canopy spectral information and normalized vegetation index; The normalized vegetation index is calculated based on the near-infrared and red light reflectance in the canopy spectral information.

[0010] As a preferred embodiment of the differentiated water and fertilizer integration intelligent management system for wheat-cotton intercropping described in this invention, the root layer weighted average water content is calculated based on the soil volumetric water content at different depths and the crop root distribution weight. The crop root distribution weights are configured according to the root water absorption areas corresponding to the growth stages of wheat and cotton, respectively. The differentiated irrigation and fertilization decision module calculates the soil moisture deficit based on the root zone weighted average moisture content, the target soil moisture content at the current growth stage, and the wilting point.

[0011] As a preferred embodiment of the differentiated water and fertilizer integration intelligent management system for wheat-cotton intercropping described in this invention, the target irrigation amount is calculated based on the target soil moisture content, root zone weighted average moisture content, planned wetting layer depth, planned wetting ratio, and unit conversion factor at the current growth stage. The differentiated irrigation and fertilization decision module calculates the actual crop evapotranspiration based on the reference evapotranspiration and crop coefficient, and uses the actual crop evapotranspiration to correct the irrigation interval and target irrigation amount.

[0012] As a preferred embodiment of the differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping described in this invention, the target fertilizer application rate and fertilizer formula are calculated based on the target yield, growth stage, available soil nutrients, and fertilizer utilization rate. The amount of nitrogen fertilizer applied is determined based on the target yield, growth stage, available nitrogen in the soil, and fertilizer utilization rate. The amount of phosphorus fertilizer applied and the amount of potassium fertilizer applied are adjusted according to the corresponding nutrient requirements and the remaining growth stage.

[0013] As a preferred embodiment of the differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping described in this invention, the differentiated irrigation and fertilization decision module determines the growth stage of wheat and cotton based on the current date, accumulated temperature data, and wheat harvest confirmation signal. During the wheat-cotton co-existence period, the wheat belt uses the grain-filling stage sub-model, while the cotton belt uses the seedling stage model. After the wheat harvest, the wheat strip model switches to a dormant state, while the cotton strip model switches to one of the budding stage sub-model or the flowering and boll-forming stage sub-model, which corresponds to the current growth stage of cotton.

[0014] As a preferred embodiment of the differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping described in this invention, the differentiated irrigation and fertilization decision module further includes an LSTM time series prediction module and a multi-objective optimization module. The LSTM time series prediction module is a long short-term memory network time series prediction module. Based on historical soil moisture content, meteorological forecast data, growth stage and historical irrigation amount, it predicts the baseline change of root soil moisture content in wheat belt and cotton belt within a preset time period. The multi-objective optimization module, based on the predicted root soil moisture content baseline change, superimposes the moisture content change formed by the candidate irrigation amount under each group of candidate irrigation amount, nitrogen application amount, phosphorus application amount, potassium application amount and irrigation interval. Taking the irrigation amount, nitrogen application amount, phosphorus application amount, potassium application amount and irrigation interval of wheat belt and cotton belt as optimization variables, and taking the maximization of total yield, water use efficiency and nitrogen fertilizer partial productivity and the minimization of nitrogen leaching risk as optimization objectives, it generates a Pareto frontier scheme set. When the multi-objective optimization module does not generate a Pareto front scheme set, the differentiated irrigation and fertilization decision module generates the target irrigation amount based on the root zone weighted average water content, and generates the target fertilization amount and fertilization formula based on the target yield, growth stage, available soil nutrients and fertilizer utilization rate.

[0015] As a preferred embodiment of the differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping described in this invention, the intelligent branch control module includes a control unit, a water and fertilizer mixing device, a water supply pipeline for wheat, a water supply pipeline for cotton, and a solenoid valve. The solenoid valves are respectively installed on the water supply pipelines for wheat and cotton. The wheat belt water supply pipeline is connected to the shallow buried drip irrigation belt; The cotton strip water supply pipeline is connected to the mulched drip irrigation tape; The control unit receives irrigation and fertilization control parameters for wheat strips and cotton strips, and controls the solenoid valves and water-fertilizer mixing devices on the corresponding water supply pipelines according to the irrigation and fertilization control parameters for wheat strips and cotton strips, so that the shallow-buried drip irrigation tape and the film-covered drip irrigation tape can independently supply water and fertilizer to the wheat strips and cotton strips according to the corresponding target irrigation amount, target fertilization amount and fertilization formula.

[0016] The beneficial effects of this invention are as follows: By laying shallow-buried drip irrigation tape in the wheat belt and mulched drip irrigation tape in the cotton belt, and connecting them to independent branch water supply pipelines, independent water and fertilizer supply is achieved for the wheat and cotton belts, avoiding water and fertilizer supply mismatch caused by unified irrigation and fertilization; by collecting soil volumetric water content, soil temperature, and crop growth data for the two crop belts respectively, and combining the growth stage, reference evapotranspiration, and crop coefficient to calculate the target irrigation amount, target fertilizer amount, and fertilizer formula, the adaptability of irrigation and fertilization decisions during the wheat-cotton co-op period and at different stages after wheat harvest is improved; by using the LSTM time series prediction module and multi-objective optimization module to predict future root zone soil water content, and generating optimized schemes around yield, water use efficiency, nitrogen fertilizer partial productivity, and nitrogen leaching risk, the foresight and comprehensive coordination ability of water and fertilizer management is improved; by using the intelligent branch control module to convert irrigation and fertilization control parameters into electromagnetic valve opening and closing control and water and fertilizer mixing control, precise branch irrigation and fertilization management of wheat-cotton intercropping fields is achieved. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. 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.

[0018] Figure 1 A schematic diagram of a differentiated integrated water and fertilizer management system for wheat-cotton intercropping.

[0019] Figure 2 A schematic diagram of the overall structure of a differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping.

[0020] Figure 3 This is a schematic diagram of the plan layout of the field deployment system.

[0021] Figure 4 This is a structural block diagram of the intelligent branch control module.

[0022] Figure 5 A flowchart illustrating the workflow of a differentiated irrigation and fertilization decision-making model. Detailed Implementation

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0026] Reference Figures 1-5 This is one embodiment of the present invention, which provides a differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping, comprising the following steps: The field deployment module is used to lay shallow-buried drip irrigation tape in wheat strips and mulched drip irrigation tape in cotton strips. The shallow-buried drip irrigation tape is buried below the soil surface between wheat rows, and the mulched drip irrigation tape is laid under the mulch film between cotton rows. The shallow-buried drip irrigation tape and the mulched drip irrigation tape are independently connected to the corresponding branch water supply pipelines.

[0027] It should be noted that the wheat-cotton strip intercropping method is adopted in the field, with wheat strips and cotton strips arranged alternately. In Xinjiang, the row spacing for wheat and cotton strips includes 35cm+65cm, 40cm+60cm, 70cm+80cm, 45cm+55cm, 66cm+76cm, 80cm+70cm, 60cm+40cm, and 66cm+120cm for the wheat+cotton intercropping pattern. Outside of Xinjiang, the width and row spacing of the wheat and cotton strips can be determined according to local wheat-cotton intercropping agronomic requirements. Shallow-buried planting materials are laid within the wheat strips. The drip irrigation tape is buried at a depth defined as the vertical distance from the ground surface to the top surface of the shallow-buried drip irrigation tape. For example, the burial depth is 3-8 cm, preferably 5 cm. The shallow-buried drip irrigation tape uses a low-flow drip irrigation tape with a dripper spacing of 30 cm, which can be determined according to the soil infiltration capacity and the water requirement of wheat. The mulched drip irrigation tape is laid inside the cotton strip, with the mulched drip irrigation tape located under the plastic film and between the cotton rows. The rough side of the drip irrigation tape faces upward and is tightly attached to the inside of the film. The dripper flow rate is not greater than 2.0 liters / hour. For example, the dripper flow rate is selected as 2.0 liters / hour.

[0028] The inlet end of the shallow-buried drip irrigation tape is independently connected to the branch water supply pipeline corresponding to the wheat belt, and the inlet end of the mulched drip irrigation tape is independently connected to the branch water supply pipeline corresponding to the cotton belt. Each branch water supply pipeline is equipped with an independently controlled solenoid valve. During the land preparation and sowing stage, when sowing wheat in autumn, a seeding and pipe-laying integrated machine is used to simultaneously complete sowing and laying of shallow-buried drip irrigation tape. When sowing cotton in spring, a mulched seeding and pipe-laying integrated machine is used to simultaneously complete mulching, sowing, and laying of mulched drip irrigation tape. For fields where drip irrigation tape has been laid, the shallow-buried drip irrigation tape and mulched drip irrigation tape are directly connected to the corresponding branch water supply pipeline.

[0029] The above-mentioned field deployment modules enable the wheat strip to form a shallow-buried drip irrigation water supply channel, the cotton strip to form a mulched drip irrigation water supply channel, and provide a pipeline foundation for subsequent differentiated irrigation and fertilization of different strips.

[0030] The data acquisition module is used to collect soil volumetric water content and soil temperature at different depths in the wheat and cotton belts through soil temperature and humidity sensors deployed in the wheat belt and the cotton belt, and to collect crop growth data in the wheat and cotton belts through a crop growth monitoring device.

[0031] It should be noted that the data acquisition module includes at least three soil temperature and humidity sensors deployed in the wheat belt and at least three soil temperature and humidity sensors deployed in the cotton belt. In this embodiment, three soil temperature and humidity sensors are buried in the middle of the wheat belt and in the main root distribution area between cotton rows in the cotton belt, respectively, at measurement depths of 10 cm, 20 cm, and 30 cm, to collect soil volumetric water content and soil temperature at different depths of the root layer of the corresponding crop belt in real time. The data acquisition module collects soil volumetric water content and soil temperature once per hour and stores them in a local data cache according to time series. During daily decision-making, the soil volumetric water content is taken as the arithmetic mean of 24 collected values ​​for the day, and the soil temperature is taken as the average value for the day, which is used for accumulated temperature accumulation and input features of the Long Short-Term Memory Network (LSTM) time series prediction module. The LSTM time series prediction module calls hourly raw data to retain the diurnal fluctuation information of soil moisture content.

[0032] The crop growth monitoring device includes a multispectral camera, which collects multispectral images of wheat and cotton strips during the daytime around noon when sunlight is abundant, according to a preset collection cycle. For example, the preset collection cycle is once a day. When there are consecutive cloudy days or the current light intensity is lower than the effective light intensity threshold for image collection, the current collection is skipped and the most recent effective image data is used. Before each collection, the multispectral camera takes a standard whiteboard reference image and performs radiometric correction on the reflectance of each band based on the standard whiteboard reference image. After the collection is completed, the data acquisition module performs cloud shadow detection, shadow area removal, and high reflectance area masking processing on the image, retaining only the reflectance data of the crop canopy area.

[0033] When the proportion of shadowed areas in an image exceeds the image shadowed area proportion threshold, the current acquisition is deemed invalid, and the previous valid acquisition data is reused; the Normalized Difference Vegetation Index (NDVI) is calculated based on the multispectral image, using the following formula: ; in, Indicates the normalized vegetation index; Indicates near-infrared reflectivity; This indicates the reflectivity in the red light band.

[0034] Canopy spectral information is used to generate near-infrared and red light reflectance and participates in the calculation of normalized vegetation index; the differentiated irrigation and fertilization decision module determines the crop growth correction coefficient based on the normalized vegetation index.

[0035] The data acquisition module compares the current NDVI with the previous valid acquisition value. When the magnitude of a single change exceeds the NDVI single jump anomaly threshold, the current NDVI is marked as an anomaly, the anomaly is removed, and the previous and subsequent valid acquisition values ​​are used for linear smoothing to complete the result. When the subsequent valid acquisition value has not yet been formed, the previous valid acquisition value is temporarily used, and the smoothing result is updated after the subsequent valid acquisition value is formed.

[0036] Furthermore, the effective image acquisition light intensity threshold, image shadow area ratio threshold, and NDVI single jump anomaly threshold are determined through initial calibration and operational correction: In the initial deployment phase, multispectral images of wheat and cotton belts under sunny, cloudy, partially shaded, and mulch reflective conditions are continuously acquired, and the light intensity, shadow area ratio, and NDVI variation amplitude of adjacent effective acquisition cycles are recorded simultaneously; the acquired images are manually labeled for validity, the data distribution of valid and invalid images is statistically analyzed, and the boundary value that can retain valid images and remove invalid images is selected as a candidate threshold; the candidate threshold is verified using historical image data not included in the statistics, and after verification, it is written into the data acquisition module; if the valid image rejection rate continuously increases or the number of NDVI abnormal jumps continuously increases during operation, the above statistical and verification process is re-executed; for example, the effective image acquisition light intensity threshold is set to 30,000 lux, the image shadow area ratio threshold is set to 15%, and the NDVI single jump anomaly threshold is set to 0.15; through the above data acquisition module, soil temperature and humidity data and crop growth data are generated for differentiated irrigation and fertilization decisions.

[0037] The differentiated irrigation and fertilization decision module is used to calculate the root zone weighted average water and fertilizer requirements, target irrigation amount, target fertilization amount, and fertilization formula for wheat and cotton strips based on the differentiated water and fertilizer requirements parameters of wheat and cotton at their respective growth stages, combined with soil volumetric water content, soil temperature, crop growth data, reference evapotranspiration, and crop coefficient collected by the data acquisition module. It also generates irrigation and fertilization control parameters for wheat strips and cotton strips.

[0038] It should be noted that the differentiated irrigation and fertilization decision module is built into the control unit and includes a basic decision sub-model and an optimization enhancement module. Further, the control unit receives soil temperature and humidity data, crop growth data, and meteorological data generated by the data acquisition module, and performs calculations of irrigation amount, fertilization amount, and fertilization formula from the basic decision sub-model. The cloud computing server performs LSTM time-series prediction and NSGA-II multi-objective optimization calculations when the activation conditions of the optimization enhancement module are met, and returns the prediction results and Pareto front scheme set to the control unit. The remote monitoring terminal displays candidate schemes, receives user-set target yield, weight preferences, and parameter update information, but does not directly issue execution commands to the solenoid valves and water-fertilizer mixing devices. Finally, the irrigation and fertilization control parameters are generated by the control unit, which then sends control commands to the solenoid valves, water-fertilizer mixing devices, and fertilizer pumps.

[0039] The model database pre-stores suitable soil moisture ranges, crop coefficients, recommended nitrogen, phosphorus, and potassium application rates, nutrient ratios, and crop root distribution weights for wheat and cotton at different growth stages. Among them, the growth stages of wheat include jointing stage, flag-raising stage, grain-filling stage, and maturity stage, while the growth stages of cotton include seedling stage, budding stage, flowering and boll-forming stage, and boll-opening stage. The parameters in the model database are derived from agronomic specifications, irrigation systems, statistical results of historical field trial data, and quantitative results of expert experience, and support regional updates through remote monitoring terminals.

[0040] The operation flow of the differentiated irrigation and fertilization decision-making module is as follows: The first step is to read the current date and accumulated temperature data to determine the growth stage of wheat and cotton.

[0041] The second step involves reading soil temperature and humidity data and crop growth data generated by the data acquisition module, calculating the current root zone weighted average water content, and determining crop growth correction information.

[0042] The third step is to call the basic water decision sub-model, and combine it with the appropriate water content threshold for the corresponding growth stage to calculate the water deficit and basic irrigation amount.

[0043] The fourth step is to invoke the basic fertilization decision sub-model to calculate the basic fertilization amount based on the target yield and available soil nutrients.

[0044] The fifth step involves the differentiated irrigation and fertilization decision module determining the activation conditions for the optimization and enhancement module. These conditions include: network connectivity between the control unit and the cloud computing server; available cloud computing resources meeting the optimization computing resource availability threshold; optimization calculations being completed within the optimization calculation timeout period; and the control unit currently having no irrigation or fertilization tasks. The optimization and enhancement module is activated when all four conditions are met simultaneously. If any condition is not met, the output of the basic decision sub-model is directly used as the final irrigation and fertilization control parameters. For example, the optimization calculation timeout period is set to 10 minutes; if the calculation times out, optimization is terminated and the system reverts to the basic decision sub-model output. The optimization and enhancement module runs the LSTM time series prediction module, which, based on the soil moisture content of the past 10 days, weather forecasts, growth stage, and historical irrigation amounts, outputs the daily root soil moisture content baseline for the next 7 days. The system generates a Pareto frontier scheme set. In each candidate scheme of the NSGA-II multi-objective optimization, the differentiated irrigation and fertilization decision module overlays the root zone soil moisture content change curve after water supply onto the baseline change curve of root zone soil moisture content according to the irrigation amount and irrigation interval in that candidate scheme, thus obtaining the root zone soil moisture content change curve for the corresponding candidate scheme. The multi-objective optimization module calculates the total yield, water use efficiency, nitrogen fertilizer partial productivity, and nitrogen leaching risk based on the root zone soil moisture content change curve of the corresponding candidate scheme, combined with nitrogen application rate, phosphorus application rate, potassium application rate, and irrigation interval, generating a Pareto frontier scheme set. According to the current preset selection mode, a unique execution scheme is determined from the Pareto frontier scheme set. The unique execution scheme includes irrigation amount, fertilizer application rate, ratio, and irrigation interval. When the optimization enhancement module does not generate a Pareto frontier scheme set, the output of the basic decision sub-model is used.

[0045] The optimized computing resource availability threshold refers to the minimum available computing resources required for the cloud computing server to complete LSTM time series prediction and NSGA-II multi-objective optimization within a preset timeout period. For example, the optimized computing resource availability threshold is set as follows: the current cloud processor utilization rate is not higher than 70%, the available memory ratio is not lower than 30%, and the current queued task waiting time is not more than 2 minutes. The value is based on the historical runtime statistics of optimization computing tasks. When the cloud computing resources meet the above conditions, the optimization calculation can usually be completed within a preset timeout period of 10 minutes. When the above conditions are not met, the differentiated irrigation and fertilization decision module directly adopts the output of the basic decision sub-model to avoid the delayed generation of irrigation and fertilization control parameters.

[0046] The sixth step involves generating irrigation and fertilization control parameters for wheat and cotton strips based on the output of the basic decision sub-model or the unique execution scheme determined by the optimization and enhancement module, and sending them to the intelligent branch control module for execution.

[0047] Furthermore, the root layer weighted average moisture content is calculated based on the soil volumetric moisture content at depths of 10 cm, 20 cm, and 30 cm and the corresponding crop root distribution weights; the formula for calculating the root layer weighted average moisture content is: ; in, This represents the weighted average moisture content of the root zone. This indicates the volumetric water content of the soil in a 10-centimeter soil layer. This indicates the volumetric water content of the soil in a 20-centimeter soil layer. This indicates the volumetric water content of the soil in a 30 cm soil layer. This represents the weight of crop root distribution corresponding to a 10 cm soil layer. This represents the weight of crop root distribution corresponding to a 20 cm soil layer. This represents the weight of crop root distribution corresponding to a 30 cm soil layer.

[0048] The default weights are determined based on the root water absorption distribution corresponding to the crop type and growth stage. , and These represent the crop root distribution weights corresponding to soil layers of 10 cm, 20 cm, and 30 cm, respectively; the crop root distribution weights corresponding to each growth stage of wheat are as follows: jointing stage 0.4 0.4 and The value is 0.2, during the flag-raising period. 0.3 0.4 and The value is 0.3, during the grouting period. 0.2 0.5 and It is 0.3, at maturity. 0.3 0.4 and The weight of crop root distribution at the cotton seedling stage is 0.3; 0.4 0.4 and The weights for crop root distribution during the cotton budding and boll-forming stages are 0.2; 0.2 0.4 and It is 0.4.

[0049] Furthermore, the differentiated irrigation and fertilization decision module determines the target soil moisture content and the lower limit threshold based on the suitable soil moisture content range at the current growth stage. The target soil moisture content is taken as the median of the suitable soil moisture content range and is used for water deficit calculation, basic irrigation amount calculation, and priority calculation. The lower limit threshold is taken as the lower limit of the suitable soil moisture content range and is used for maximum irrigation interval calculation and irrigation trigger judgment. For example, the suitable soil moisture content range for wheat during the grain-filling stage is 70%-80% of field capacity, the target soil moisture content corresponds to 75% of field capacity, and the lower limit threshold corresponds to 70% of field capacity. The suitable soil moisture content range for cotton during the seedling stage is 55%-65% of field capacity, and the target soil moisture content corresponds to 60% of field capacity.

[0050] Furthermore, the differentiated irrigation and fertilization decision module calculates the soil moisture deficit based on the root zone weighted average moisture content, using the following formula: ; in, Indicates the degree of soil moisture deficit; This indicates the target soil moisture content for the current growth stage. The target soil moisture content is the volumetric water content converted from the percentage of field capacity. This represents the weighted average moisture content of the root zone. The wilting point is the volumetric water content, which is usually 40%-50% of the field capacity.

[0051] The basic water decision sub-model calculates the basic irrigation amount using a threshold combined with water balance, and the calculation formula is as follows: ; in, This indicates the basic irrigation volume, expressed in cubic meters per mu (unit of land area). It represents the target soil moisture content at the current growth stage, which is volumetric water content, with the unit being cm³ / cm³. It is calculated from the percentage of the median of the suitable moisture content range to the field capacity. This represents the actual weighted average moisture content of the root layer, expressed as volumetric moisture content in cm³ / cm³. This indicates the planned depth of the wetting layer, in centimeters. This represents the planned wetting ratio, which is dimensionless and ranges from 0.6 to 1.0. It is used to characterize the proportion of the actual wetted area to the planted area under drip irrigation conditions. For example, it is 0.8 for shallow buried drip irrigation and 0.9 for mulched drip irrigation. This represents the unit conversion factor, with the unit being cubic meters per (acre·cm). An example value of 6.667 is used to convert a water depth of 1 cm into the water supply volume per acre.

[0052] When the basic irrigation volume is used for actual water supply, the control unit determines the target water supply volume based on the basic irrigation volume and the effective irrigation area of ​​the corresponding crop belt, and calculates the opening time of the solenoid valve based on the actual flow rate of the pipeline.

[0053] Furthermore, the differentiated irrigation and fertilization decision-making module calculates the growth correction coefficient based on the normalized vegetation index generated by the data acquisition module. :when hour, A value of 0.8 indicates excessive growth; when... hour, A value of 1.2 indicates poor growth; in other cases, Take 1.0; where, This represents the theoretical reference value for the corresponding growth stage; for example, the grain-filling stage of wheat. Approximately 0.75, during the cotton seedling stage. Approximately 0.55; the corrected irrigation amount is obtained after adjusting the base irrigation amount for growth. The calculation formula is: ; in, This indicates the amount of irrigation after growth correction; This represents the basic irrigation amount calculated by the basic water decision sub-model.

[0054] The calculation of subsequent target irrigation volume, maximum irrigation interval, and actual irrigation volume are all based on The derivation is based on this.

[0055] Furthermore, the target yield is set by the user through a remote monitoring terminal. If no target yield is set, the differentiated irrigation and fertilization decision module determines the default target yield based on the historical average yield of the same variety in the same region in recent years. The daily nitrogen, phosphorus, and potassium requirements and the remaining growing days at the current growth stage are read from the model database according to crop type, variety type, and current growth stage. The available nitrogen, phosphorus, and potassium in the soil are determined by pre-sowing soil testing data, periodic soil testing data, or soil background testing data from multiple sampling points in the field, and the average available nutrient content of the root zone of the corresponding crop is obtained through spatial interpolation. The fertilizer utilization rate is preset by the model database and allows users to make regional adjustments based on local soil type, fertilization method, and historical fertilization effects.

[0056] Furthermore, the basic fertilization decision sub-model uses the improved Stanford formula to calculate the fertilization amount; the improved Stanford formula has the following four improvements on the conventional nutrient balance formula: First, the target nitrogen requirement The calculation has been changed from the total nitrogen requirement for the entire reproductive period to a segmented calculation based on reproductive stages; the conventional Stanford formula uses the total nitrogen requirement for the target yield over the entire reproductive period as... This plan allocates the total nitrogen requirement according to the daily nitrogen intensity of the current reproductive stage and the remaining reproductive days, so that... Reflecting the actual nutrient requirements at the current stage; the formula for calculating nitrogen fertilizer application rate is as follows: ; in, This indicates the amount of nitrogen fertilizer applied during the current growth stage; This represents the target nitrogen requirement for the current stage, determined based on the target yield, the daily nitrogen demand intensity at the current growth stage, and the remaining growth days. Indicates the amount of nitrogen available in the soil; This indicates fertilizer utilization rate.

[0057] Second, the amount of nitrogen available in the soil The method was changed from single-point detection values ​​to spatial interpolation estimation; the conventional Stanford formula directly uses single soil test values ​​as... This scheme uses soil background monitoring data from multiple field sampling points as a basis and employs spatial interpolation methods (exemplarily, inverse distance weighting or Kriging interpolation) to estimate the average available nitrogen in the current crop root zone, thereby improving... Spatial representativeness.

[0058] Third, fertilizer utilization rate The fixed constant has been changed to a configurable parameter; the conventional Stanford formula typically uses a fixed fertilizer utilization rate, while this scheme... The value ranges from 0.4 to 0.6, and users can adjust the configuration according to local soil type and fertilization method through a remote monitoring terminal.

[0059] Fourth, the application rates of phosphate and potash fertilizers are dynamically adjusted according to the remaining growth stages. While the conventional Stanford formula uses a fixed nutrient requirement ratio for phosphate and potash fertilizers, this scheme determines the weights of phosphate and potash fertilizers based on the nutrient ratio parameters corresponding to the current and remaining growth stages in the model database. These nutrient ratio parameters are derived from agronomic standards, irrigation systems, statistical results of historical field trial data, and quantitative results from expert experience, and are pre-set in the model database according to crop type, variety, and growth stage. The differentiated irrigation and fertilization decision module reads the nutrient ratio parameters corresponding to the current crop zone, determines the phosphate fertilizer requirement as the phosphate fertilizer weight based on the proportion of phosphate and potash fertilizer requirements to the total phosphate and potash fertilizer requirements, and adjusts the application rates of phosphate and potash fertilizers accordingly. The calculation formulas for phosphate and potash fertilizer application rates have the same structure as the nitrogen fertilizer application formula, calculated based on the target nutrient requirement, soil availability, and utilization rate, respectively.

[0060] Furthermore, when the calculated When the value is negative, it indicates that the current available nitrogen in the soil has met the stage's requirements, and the differentiated irrigation and fertilization decision module will... Set to zero, meaning no additional nitrogen fertilizer will be applied during this stage; the application rates of phosphorus and potassium fertilizers will be treated similarly.

[0061] Furthermore, the differentiated irrigation and fertilization decision module includes a growth stage switching logic, which embeds a growth period calendar and is corrected based on accumulated temperature data. During the wheat-cotton co-existence period, the wheat belt uses the grain-filling stage sub-model, with the target soil moisture content set at 75% of field capacity; the cotton belt uses the seedling stage sub-model, with the target soil moisture content set at 60% of field capacity. The corresponding irrigation amount, fertilization amount, and fertilizer formula are all calculated according to the aforementioned basic water decision sub-model and basic fertilization decision sub-model. After wheat harvest, the wheat belt sub-model is switched to a dormant state based on the current date and wheat harvest confirmation signal, and the cotton belt model is switched to one of the budding stage sub-model or the flowering and boll-forming stage sub-model corresponding to the current growth stage of cotton. The irrigation amount, fertilization amount, and fertilizer formula of the cotton belt are recalculated according to the target soil moisture content, root zone weighted average moisture content, and nitrogen, phosphorus, and potassium nutrient requirements of the current growth stage of cotton.

[0062] The grain-filling stage sub-model, seedling stage sub-model, budding stage sub-model, and flowering and boll-forming stage sub-model are all parameter tables and calculation rule sets for the corresponding growth stages. Each sub-model includes at least the target soil moisture content, lower limit threshold of moisture content, crop coefficient, root distribution weight, daily nutrient intensity, remaining growth days, and nutrient ratio parameters for the corresponding growth stage. After determining the current growth stage, the differentiated irrigation and fertilization decision module calls the parameters in the corresponding sub-model and generates irrigation and fertilization control parameters for the corresponding crop zone according to the basic irrigation amount calculation formula, growth correction formula, and fertilization amount calculation formula.

[0063] Furthermore, the differentiated irrigation and fertilization decision module incorporates actual crop evapotranspiration into the basic water decision sub-model. Reference evapotranspiration is calculated in real time using meteorological station data, which includes temperature, humidity, wind speed, and radiation data. Crop coefficients are obtained by looking up tables based on crop growth stages, and the actual crop evapotranspiration is calculated based on the reference evapotranspiration and the crop coefficients.

[0064] Reference evaporation The calculation is performed using the FAO-56 Penman-Monteith formula, as follows: ; in, This represents the reference evapotranspiration, expressed in millimeters per day. This indicates net radiation, expressed in megajoules per square meter per day. This represents soil heat flux, expressed in megajoules per square meter per day. This indicates the average daily temperature at a height of 2 meters, expressed in degrees Celsius. This indicates the wind speed at a height of 2 meters, expressed in meters per second. This represents saturated vapor pressure, expressed in kilopascals (kPa). This indicates the actual water vapor pressure, expressed in kilopascals (kPa). This represents the slope of the saturated vapor pressure curve, expressed in kilopascals per degree Celsius. This represents the wet / dry constant, with units of kPa / degree Celsius.

[0065] Crop coefficients are read from a pre-set table in the model database based on crop type and growth stage. value; The data is obtained from a table based on the reproductive stage and is not dynamically adjusted with NDVI to avoid conflict with the growth correction factor. Functional overlap; Responsible for correcting actual growth based on NDVI. This reflects the basic evapotranspiration characteristics of crops at each growth stage, and the two work together through multiplicative combination in subsequent calculations; for example, the evapotranspiration characteristics of wheat during the grain-filling stage... The value is 1.05-1.15, during the cotton seedling stage. The value is 0.35-0.50 during the cotton bud stage. The value is 0.70-0.85 during the cotton flowering and boll-forming stage. It ranges from 1.05 to 1.20.

[0066] Furthermore, when the topsoil moisture content is high after irrigation or rainfall, the soil evaporation component of the actual crop evapotranspiration increases. The differentiated irrigation and fertilization decision module incorporates the soil evaporation coefficient. right The formula for calculating actual crop evapotranspiration has been supplemented and corrected; the corrected formula is as follows: ; in, This indicates the actual evapotranspiration of the crop, expressed in millimeters per day. This represents the soil evaporation coefficient, a dimensionless value, calculated based on the topsoil moisture content and the proportion of exposed area after irrigation or rainfall. It is measured when the topsoil moisture content drops to near the wilting point or when the crop canopy coverage reaches or exceeds the canopy coverage evaporation suppression threshold. Approaching 0, meaning the soil evaporation component is negligible; among them, the canopy cover evaporation suppression threshold is determined based on the statistical results of historical canopy images and the monitoring results of surface soil evaporation. For example, the canopy cover evaporation suppression threshold is set to 80%.

[0067] Furthermore, the differentiated irrigation and fertilization decision-making module utilizes the actual evapotranspiration of the crop. The irrigation decision is revised; before calculating the maximum allowable irrigation interval, the differentiated irrigation and fertilization decision module first determines the actual crop evapotranspiration. Is it greater than zero; when When the maximum allowable irrigation interval formula is not executed, and it is determined that irrigation will not be triggered by evapotranspiration on that day, only the immediate irrigation judgment when the root zone weighted average water content is lower than the lower limit threshold is retained; when The maximum allowable irrigation interval is calculated using the following formula: ; in, Indicates the maximum permissible irrigation interval, in days; This represents the current weighted average moisture content of the root zone; This indicates the lower limit threshold for moisture content; This indicates the planned depth of the wetting layer, in centimeters. This represents the conversion factor from centimeter-level water layer to millimeter-level water layer, with a value of 10 mm / cm.

[0068] By setting ,make and Maintain consistency across water layer units.

[0069] when At that time, root layer water content Close to or below When irrigation is triggered, it is determined that irrigation must be carried out today; when If necessary, irrigation should be postponed, and a decision should be made based on crop growth and weather forecasts.

[0070] When the root layer weighted average water content is monitored in real time Below the lower limit threshold of moisture content At that time, the differentiated irrigation and fertilization decision module triggers an immediate irrigation command without waiting for the next decision cycle.

[0071] Furthermore, the actual irrigation amount can be adjusted as follows: ; in, Indicates the actual amount of irrigation; Indicates the planned irrigation cycle, usually taken as ; Indicates irrigation efficiency; This represents the conversion factor between area per mu (unit of land area) and water layer in millimeters to water volume in cubic meters.

[0072] Furthermore, the optimization and enhancement module includes an LSTM time series prediction module and a multi-objective optimization module. The LSTM time series prediction module is a long short-term memory network time series prediction module, which establishes long short-term memory neural network models for wheat belt and cotton belt respectively. The inputs are soil moisture content, weather forecast, growth stage and historical irrigation amount for the past 10 days, and the output is the daily average soil moisture content baseline change in the root layer for the next 7 days. Among them, the weather forecast includes temperature, rainfall, radiation and wind speed.

[0073] The model structure of the LSTM time series prediction module is as follows: input layer, LSTM layer, Dropout layer, LSTM layer, fully connected layer, and output layer. For example, the model structure is as follows: the input layer has 10 days × 12 features, the first LSTM layer has 64 units, the parameter of the Dropout layer is 0.2, the second LSTM layer has 32 units, the fully connected layer has 16 neurons and uses the ReLU activation function, and the output layer outputs 7 predicted soil moisture values.

[0074] Furthermore, the LSTM time-series prediction module underwent initial training before system deployment. The initial training data came from historical field monitoring data, historical irrigation records, and historical meteorological records of the same region. The input to the initial training samples consisted of a 10-day feature sequence, labeled with the measured average root zone soil moisture content for the next 7 days following the 10-day sequence. The 12 input features in the 10-day × 12 feature set were: soil volumetric water content at 10 cm depth, soil volumetric water content at 20 cm depth, soil volumetric water content at 30 cm depth, root zone weighted average water content, daily average soil temperature, and daily average air temperature. Rainfall, solar radiation, wind speed, reference evapotranspiration, current growth stage code, and historical irrigation amount are included. The current growth stage code is generated by the growth stage switching logic, and the historical irrigation amount is obtained from the irrigation execution results recorded by the control unit. Each continuous numerical feature is standardized according to the mean and standard deviation of the initial training samples. Missing values ​​are filled in with the most recent valid value, and a missing value identifier is recorded in the local data cache. The missing value identifier is used to distinguish between the actual collected values ​​and the filled values, and is synchronously called during model evaluation, anomaly tracing, and remote monitoring display. It is not involved in the calculation of the soil moisture content value itself.

[0075] The LSTM time series prediction module performs online fine-tuning every 7 days using the most recent 30 days of measured data to maintain prediction accuracy; the specific method of online fine-tuning is as follows: (1) Training sample construction: Based on the measured data of the most recent 30 days, training samples are constructed using a sliding window method; the input of each training sample is a feature sequence of 10 consecutive days, and the label is the measured root soil moisture content value of the 7 consecutive days after 10 days; for example, 30 days of data can construct about 14 sets of training samples; the online fine-tuning stage uses 12 input features consistent with the initial training stage, and the 12 input features are, in order, the soil volume moisture content of 10 cm soil layer, soil volume moisture content of 20 cm soil layer, soil volume moisture content of 30 cm soil layer, root layer weighted average moisture content, daily average soil temperature, daily average air temperature, rainfall, solar radiation, wind speed, reference evapotranspiration, current growth stage code and historical irrigation amount; among them, the historical irrigation amount is obtained from the irrigation execution results recorded by the control unit, which can reflect the impact of irrigation events on subsequent changes in root soil moisture content.

[0076] (2) Fine-tuning strategy: Online fine-tuning adopts the transfer learning method, using the currently running model parameters as the initial weights, and using the above training samples to perform gradient updates in a small number of rounds (for example, the fine-tuning rounds are 5 to 10 rounds, and the learning rate is 1 / 10 of the initial training learning rate) to avoid overfitting and retain the generalization ability of the model.

[0077] (3) Fault tolerance and rollback mechanism: After online fine-tuning is completed, the model is evaluated using the validation set (the last 7 days of measured values ​​reserved in the last 30 days of data) and the root mean square error of prediction is calculated. When the prediction error of the model after fine-tuning is less than or equal to the prediction error before fine-tuning, the fine-tuning is confirmed to be effective and the original model parameters are replaced. When the prediction error of the model after fine-tuning is greater than the prediction error before fine-tuning, or when the training does not converge during the online fine-tuning process, the fine-tuning is judged to have failed and the model is automatically rolled back to the original model parameters before fine-tuning. The original model continues to run until the next fine-tuning cycle.

[0078] In the multi-objective optimization module, yield, water use efficiency, nitrogen fertilizer partial productivity, and nitrogen leaching risk are all evaluated using a rolling optimization window of the next 7 days. Irrigation amount, effective rainfall, actual crop evapotranspiration water consumption, and changes in root zone soil water storage are all uniformly converted to cubic meters per acre, while nitrogen application amount, nitrogen brought in by rainfall, and soil mineralized nitrogen are all uniformly converted to kilograms per acre. The index values ​​of each scheme in the Pareto frontier scheme set are calculated according to the same evaluation window and the same unit before being normalized and weighted.

[0079] Furthermore, the multi-objective optimization module employs the NSGA-II algorithm, using a rolling optimization window of the next 7 days to independently optimize the following four objectives for each crop zone: Total yield is maximized by estimating using a lightweight crop growth surrogate model.

[0080] Furthermore, the lightweight crop growth surrogate model is a pre-trained yield estimation model used to quickly evaluate the expected yields corresponding to different irrigation and fertilization schemes during the NSGA-II optimization iteration process, avoiding the need to call the computationally intensive full-growth-cycle crop growth simulation model in the optimization loop.

[0081] The basic form of the lightweight crop growth surrogate model is a multiple regression response surface model. The input variables include five decision variables in the current optimization scheme: irrigation amount, nitrogen application amount, phosphorus application amount, potassium application amount and irrigation interval, as well as state variables such as current growth stage, meteorological forecast data (temperature, radiation and rainfall) and current root zone soil moisture content. The output variable is the expected yield per unit area corresponding to the optimization scheme, in kilograms per acre.

[0082] The training method for the lightweight crop growth surrogate model is as follows: Before system deployment, yield datasets under different irrigation and fertilization combinations are generated using local historical field trial data or full-growth-cycle crop growth simulation models (for example, DSSAT or WOFOST models). The surrogate model is trained using multivariate multinomial regression or lightweight gradient boosting tree methods, with irrigation amount, fertilization amount, irrigation interval, meteorological data, and soil parameters as input features and corresponding yield as labels. After the lightweight crop growth surrogate model is trained, it is stored in the control unit or cloud server. When running online, it is directly called for yield estimation without real-time training updates. When the planting area or crop variety changes, users can upload new training data through a remote monitoring terminal to retrain and update the parameters of the lightweight crop growth surrogate model.

[0083] To maximize water use efficiency, the formula for calculating water use efficiency is as follows: ; in, Water use efficiency is expressed in kilograms per cubic meter. This indicates yield, expressed in kilograms per mu (unit of land area). This indicates the irrigation water volume within the current 7-day rolling optimization window, expressed in cubic meters per mu (unit of land area). This indicates the effective rainfall within the current 7-day rolling optimization window, expressed in cubic meters per acre.

[0084] Nitrogen fertilizer partial productivity maximization, the calculation formula for nitrogen fertilizer partial productivity is as follows: ; in, This indicates that nitrogen fertilizer has a positive impact on productivity; This indicates the amount of nitrogen applied within the current 7-day rolling optimization window, expressed in kilograms per acre.

[0085] To minimize nitrogen leaching risk, a tiered strategy is employed for its calculation. The HYDRUS-1D surrogate model or empirical formula is selected for evaluation based on available computing resources. When cloud computing resources meet the optimized computing resource availability threshold, a pre-trained HYDRUS-1D surrogate model is invoked for nitrogen leaching risk assessment. This model uses current root zone soil moisture content, irrigation amount, nitrogen application rate, soil texture parameters, and weather forecast data as inputs to simulate nitrogen transport in the soil profile over the next 7 days, outputting the cumulative nitrogen leaching at a preset depth below the root zone as a nitrogen leaching risk indicator. When computing resources do not meet the threshold, an empirical formula is used for rapid estimation. This formula is based on irrigation water volume, effective rainfall, nitrogen application rate, nitrogen brought in by rainfall, soil mineralized nitrogen, actual crop evapotranspiration, and changes in root zone soil water storage within the current 7-day rolling optimization window. The formula is as follows: when When this occurs, it indicates that there is no irrigation water and no effective rainfall within the current 7-day rolling optimization window, and the differentiated irrigation and fertilization decision module directly orders... ;when At that time, the nitrogen leaching risk index is calculated according to the following formula: ; in, Indicators representing nitrogen leaching risk, expressed in kilograms per acre; This represents the estimated amount of nitrogen brought in by rainfall within the current 7-day rolling optimization window, in kilograms per acre; This indicates the amount of nitrogen released by soil mineralization within the current 7-day rolling optimization window, expressed in kilograms per acre. The amount is obtained by referring to a table based on soil organic matter content and temperature. This indicates the actual evapotranspiration water consumption of crops within the current 7-day rolling optimization window, in cubic meters per acre. This indicates the change in root zone soil water storage within the current 7-day rolling optimization window, expressed in cubic meters per acre.

[0086] In the above formula, This represents the average nitrogen concentration corresponding to irrigation water and effective rainfall. This represents the amount of deep seepage water exceeding the crop's evapotranspiration and root zone water storage capacity; when the deep seepage water volume is not greater than zero... Take zero.

[0087] The HYDRUS-1D proxy model is trained before system deployment. The training samples are derived from the simulation results of the HYDRUS-1D standard model under different soil textures, irrigation amounts, nitrogen application rates, initial root zone moisture content, and meteorological conditions. It can be corrected by combining field leachate or soil nitrate nitrogen monitoring data below the root zone. The training samples use root zone soil moisture content, irrigation amount, nitrogen application rate, soil texture parameters, and meteorological forecast data as input features, and use the cumulative nitrogen leaching at a preset depth below the root zone output by the HYDRUS-1D standard model or obtained from field monitoring as labels. After the HYDRUS-1D proxy model is trained, it is stored on a cloud computing server. During online optimization, it is only called to quickly assess the risk of nitrogen leaching.

[0088] Furthermore, after the evaluation using the HYDRUS-1D surrogate model is completed, the simulation results are compared with the results calculated using the empirical formula. If the deviation between the HYDRUS-1D surrogate model results and the results calculated using the empirical formula exceeds the nitrogen leaching assessment deviation threshold, the HYDRUS-1D surrogate model results are taken as the standard, and the deviation log is recorded for subsequent model verification. If the deviation does not exceed the nitrogen leaching assessment deviation threshold, the average of the two is taken as the final nitrogen leaching risk index. For example, the nitrogen leaching assessment deviation threshold is set to 30%, and the value is determined based on the statistical deviation between historical assessment results and field monitoring data.

[0089] Furthermore, the optimization variables of the multi-objective optimization module include irrigation amount, nitrogen application rate, phosphorus application rate, potassium application rate, and irrigation interval; for example, the irrigation amount is 0-35 cubic meters / mu, the nitrogen application rate is 0-8 kg / mu, the phosphorus application rate is 0-4 kg / mu, the potassium application rate is 0-6 kg / mu, and the irrigation interval is 2-8 days; the constraints include soil moisture content not exceeding field capacity, not falling below the wilting point, and total water consumption not exceeding the regional quota; the algorithm parameters include population size 100, number of generations 200, crossover probability 0.9, mutation probability 1 / number of variables, and output Pareto front; for example, the output is approximately 50 non- The solution is dominated; where the number of variables is the total number of optimization variables. In this embodiment, the optimization variables include irrigation amount, nitrogen application amount, phosphorus application amount, potassium application amount, and irrigation interval, totaling 5. Each variable is optimized based on the total amount or mean within a 7-day rolling optimization window, without daily splitting. The number of variables is 5, and the mutation probability is 1 / 5, i.e., 0.2. Furthermore, the differentiated irrigation and fertilization decision module sets up a scheme selection mechanism, which is used to select a unique execution scheme from the Pareto front. The scheme selection mechanism includes three modes: human-machine collaboration mode, weight preference mode, and dynamic rule mode. The three modes are executed in the following priority order: The highest priority is the human-machine collaboration mode: when there are online users on the remote monitoring terminal and the users manually select an execution plan within the preset response time after the Pareto front push (for example, the response time is 30 minutes), the plan selected by the user will be the final execution plan, and subsequent modes will not be triggered.

[0090] The next priority is the weighted preference mode: when the human-machine collaboration mode is not triggered, the system checks whether the user has preset the weights corresponding to yield, water use efficiency, nitrogen fertilizer partial productivity and nitrogen leaching risk through the remote terminal; if preset, the weighted score of each scheme is calculated according to the weight set by the user, and the solution with the highest weighted score is selected as the final execution scheme.

[0091] The lowest priority is the dynamic rule mode: when neither the human-machine collaboration mode nor the weight preference mode is triggered, the system reads the weights corresponding to yield, water use efficiency, nitrogen fertilizer partial productivity, and nitrogen leaching risk from the weight mapping table preset in the model database according to the current growth stage, and selects a scheme based on the read weights; the weight mapping table is determined by agronomic specifications, irrigation system, statistical results of historical field test data, and quantitative results of expert experience, and satisfies that the sum of the four weights under the same growth stage is 1; after the remote monitoring terminal updates the model database, it synchronously updates the weight mapping table in the dynamic rule mode.

[0092] Furthermore, the three modes are mutually exclusive in their triggering; only the highest-priority triggered mode is executed within the same decision-making cycle, without mode stacking or weight mixing. In the weight preference mode, users set the weights corresponding to yield, water use efficiency, nitrogen fertilizer partial productivity, and nitrogen leaching risk via remote terminal. The system automatically calculates the weighted scores of each option in the Pareto front and selects the solution with the highest weighted score as the final execution option. The weighted score is calculated as follows: ; in, Indicates the weighted score; , , and These represent the weights of yield, water use efficiency, nitrogen fertilizer partial productivity, and nitrogen leaching risk, respectively, with the sum of the four weights being 1. , , and These represent the normalized values ​​of yield, water use efficiency, nitrogen fertilizer partial productivity, and nitrogen leaching risk within the Pareto frontier scheme set, respectively.

[0093] Normalization is performed using the range standardization method within the Pareto front. For the maximization objective, it is calculated according to the following formula: ; in, Indicates output Water use efficiency Nitrogen fertilizer has a relatively low productivity Any of the original indicators; This represents the normalized value of the corresponding indicator; and These represent the maximum and minimum values ​​of the corresponding indicators in the Pareto frontier scheme set, respectively.

[0094] For the minimization target, the risk of nitrogen leaching is calculated using the following formula: ; in, This represents the normalized nitrogen leaching risk index; and These represent the maximum and minimum values ​​of the concentrated nitrogen leaching risk in the Pareto frontier scheme, respectively; the nitrogen leaching risk is determined by... Reverse scoring is performed so that a higher normalized score indicates a lower risk of nitrogen leaching.

[0095] For example, user settings 0.5 For 0.2 and The value is 0.3, and the system will automatically adjust it. Set to 0, which is the default value when the user does not explicitly set the PFP weights; users can also explicitly set the four weight values ​​through a remote terminal; when multiple schemes have the same weighted score in the Pareto front, the scheme with the smallest irrigation amount is selected as the final execution scheme.

[0096] In the human-machine collaboration mode, the Pareto frontier is sent to a remote monitoring terminal in a visual form, where agricultural experts and farmers can manually select it.

[0097] When the optimization enhancement module fails to generate a Pareto front, the differentiated irrigation and fertilization decision module uses the output of the basic decision sub-model as the final irrigation and fertilization control parameters to generate irrigation and fertilization control parameters for the wheat belt and the cotton belt. The triggering scenarios for the optimization enhancement module failing to generate a Pareto front include the following five: (1) Network interruption: The network communication between the control unit and the cloud computing server is disconnected during the optimization calculation process, resulting in the loss of the calculation task or the inability to obtain the return result.

[0098] (2) Calculation timeout: The optimization and enhancement module failed to complete the NSGA-II iteration and return the result within the preset maximum allowed calculation time (for example, the maximum allowed calculation time is 10 minutes).

[0099] (3) Optimization non-convergence: After the NSGA-II algorithm runs to the preset maximum number of generations (200 generations), it still fails to form a non-dominated solution that satisfies the constraints that the soil moisture content does not exceed the field capacity, does not fall below the wilting point, and the total water consumption does not exceed the regional quota. It is judged as optimization non-convergence.

[0100] (4) No feasible solution due to constraints: Under the current constraints of soil moisture content, meteorological conditions and regional water quota, there is no feasible solution in the optimization variable space that simultaneously satisfies all constraints.

[0101] (5) LSTM prediction failure: The LSTM time series prediction module failed to generate an effective baseline curve of root soil moisture content change over the next 7 days, resulting in the lack of initial state input for the multi-objective optimization module.

[0102] When any of the above situations occur, the optimization and enhancement module immediately terminates the current calculation, the differentiated irrigation and fertilization decision module records the corresponding trigger reason identifier, and automatically switches to the output of the basic decision sub-model to ensure that the irrigation and fertilization decision is not interrupted. The trigger cause identifier is simultaneously sent to the remote monitoring terminal for notification and written into the historical operation record for subsequent operation and maintenance analysis and parameter adjustment; the trigger cause identifier does not change the execution of the output of the basic decision sub-model in this operation.

[0103] Furthermore, when the wheat belt and cotton belt simultaneously request irrigation, but water resources and equipment capacity are limited, the differentiated irrigation and fertilization decision module determines the execution order according to priority; among them, the determination condition for the crop belt to issue an irrigation request is that any one of the following three conditions is met: (1) the basic irrigation amount calculated by the basic decision sub-model. (2) Maximum allowable irrigation interval > 0; (3) The root zone weighted average water content monitored in real time. Below the lower limit threshold of moisture content .

[0104] Priority is calculated using an urgency index, as shown in the following formula: ; in, Indicates priority; Indicates the importance coefficient of the reproductive stage; Indicates the degree of soil moisture deficit; for example, during the grain-filling stage of wheat. The value is 1.5, during the cotton seedling stage. It is 0.5, during the cotton flowering and boll-forming period. It is 1.2; when Higher than hour, Take zero.

[0105] The execution rule is: execute first. High crop belt; when When the difference is less than 0.1, priority should be given to crop belts with higher water demand, and rotational irrigation should be adopted when continuous water supply is limited.

[0106] The aforementioned differentiated irrigation and fertilization decision-making module can combine the growth stages of wheat and cotton, real-time soil temperature and humidity data, crop growth data, reference evapotranspiration, and crop coefficient to form irrigation and fertilization control parameters for wheat and cotton respectively, and provide control basis for the subsequent intelligent branch control module to execute independent water and fertilizer supply for each branch.

[0107] The intelligent branch control module is used to receive irrigation and fertilization control parameters for wheat strips and cotton strips through the control unit, and to control the solenoid valves and water-fertilizer mixing devices installed on each branch water supply pipeline, so that the shallow buried drip irrigation tape and the film-covered drip irrigation tape can independently supply water and fertilizer to the wheat strip and cotton strip according to the corresponding target irrigation amount, target fertilization amount and fertilization formula.

[0108] It should be noted that the intelligent branch control module includes a control unit, solenoid valves installed on each branch water supply pipeline, and a water and fertilizer mixing device. The control unit receives irrigation and fertilization control parameters for the wheat belt and the cotton belt generated by the differentiated irrigation and fertilization decision module, and controls the solenoid valves and water and fertilizer mixing devices on the corresponding branch water supply pipelines to achieve independent water and fertilizer supply for the wheat belt and the cotton belt.

[0109] Furthermore, when the fertilizer formulas for wheat and cotton strips are different, since a single water and fertilizer mixing device can only output one ratio of water and fertilizer solution at any given time, the control unit adopts a time-sharing irrigation method to achieve the differentiated supply of fertilizer formulas to different strips.

[0110] Specifically, the control unit controls the opening of the solenoid valves on the corresponding branch water supply pipelines according to the irrigation and fertilization control parameters of the wheat belt and the cotton belt, following the execution order determined by the differentiated irrigation and fertilization decision module. Before each switch, the control unit adjusts the mother liquor intake ratio of the water and fertilizer mixing device so that each crop belt receives the water and fertilizer mixture according to the corresponding fertilizer formula. When the irrigation and fertilization of one crop belt is completed, the control unit closes the corresponding solenoid valve and performs a short-term clean water flushing of the pipeline. Then, it adjusts the water and fertilizer mixing ratio and opens the solenoid valve corresponding to another crop belt, thus completing the independent water and fertilizer supply for each belt in sequence.

[0111] Optionally, when there are high requirements for the timeliness of fertilizer formula switching or the field scale is large, the water and fertilizer mixing device can be equipped with multiple independent mother liquor tanks. Each mother liquor tank stores fertilizer mother liquor with different ratios. Rapid formula switching can be achieved through the corresponding mother liquor suction pipeline and proportional control valve, reducing the waiting time.

[0112] In this embodiment, the control unit adopts an embedded industrial controller or a PLC controller, and the control unit has a built-in data acquisition module, decision calculation module and instruction output module.

[0113] The control unit communicates with field sensors and actuators via wired or wireless means; the sensors connected to the control unit include soil temperature and humidity sensors and pipeline flow sensors; the actuators connected to the control unit include solenoid valves on each branch water supply pipeline and fertilizer pumps and proportional control valves in the water and fertilizer mixing device.

[0114] Furthermore, the Venturi fertilizer applicator in the water-fertilizer mixing device is a mechanical hydraulic component without a communication interface. Its working status is indirectly monitored through pipeline flow sensors and pressure sensors. The liquid level in the mother liquor tank is monitored by a liquid level sensor and uploaded to the control unit. The control unit adjusts the mother liquor intake by sending start / stop commands to the fertilizer pump and opening commands to the proportional control valve, thereby controlling the water-fertilizer mixing ratio. It controls the irrigation on / off of each crop zone by sending opening / closing commands to the solenoid valves on each branch water supply pipeline.

[0115] The water-fertilizer mixing device includes at least one mother liquor tank, one clean water inlet, at least one Venturi fertilizer applicator, and one mixed liquor outlet.

[0116] In this embodiment, the water-fertilizer mixing device is equipped with multiple mother liquor tanks, each storing concentrated fertilizer mother liquor with different nutrients. For example, three mother liquor tanks are set up to store nitrogen fertilizer mother liquor, phosphorus fertilizer mother liquor, and potassium fertilizer mother liquor respectively. Each mother liquor tank is connected to the corresponding Venturi fertilizer applicator or proportional injection pump through an independent fertilizer suction branch. The control unit adjusts the opening of the proportional control valve on each fertilizer suction branch according to the fertilizer formula of the current crop zone, injects the different nutrient mother liquors into the clear water channel according to the target ratio, and outputs the water-fertilizer mixture that meets the target nitrogen, phosphorus, and potassium ratio through the mixed liquid outlet. When it is necessary to output different fertilizer formulas for different crop zones, the control unit adjusts the opening of the proportional control valve of each fertilizer suction branch at the same time as switching the solenoid valve of the branch water supply pipeline to achieve rapid formula switching.

[0117] Optionally, when the field scale is small or the fertilizer formulas are not significantly different, the water-fertilizer mixing device can be simplified to a single mother liquor tank structure, with compound fertilizer mother liquor pre-prepared in the tank. In this case, the differences in fertilizer formulas for different crop zones are achieved through time-sharing irrigation: the control unit executes irrigation and fertilization according to the irrigation and fertilization control parameters for one crop zone, and after flushing the pipeline with clean water, adjusts the irrigation volume and irrigation time, and then executes the irrigation and fertilization operation for another crop zone. Under the single mother liquor tank scheme, each crop zone uses the same proportion of fertilizer mother liquor, and the difference in fertilizer application is adjusted by irrigation time and mother liquor absorption ratio. Furthermore, the outlet of the mixed liquid of the water-fertilizer mixing device is connected to the water supply pipelines for the wheat zone and the cotton zone respectively via a three-way splitter; the water supply pipeline for the wheat zone is connected to the shallow buried drip irrigation tape, and the water supply pipeline for the cotton zone is connected to the mulched drip irrigation tape; independently controlled solenoid valves are installed on the water supply pipelines for the wheat zone and the cotton zone respectively.

[0118] Furthermore, when the fertilizer formulas for the wheat belt and the cotton belt are different, since a single mixed liquid outlet can only output one ratio of water-fertilizer mixture at a time, the control unit adopts a time-sharing liquid supply method to implement differentiated fertilizer formulas. The specific execution process is as follows: The control unit adjusts the mother liquor intake ratio of the water-fertilizer mixing device according to the irrigation and fertilization control parameters of the wheat belt, completes the preparation of water-fertilizer mixture according to the target fertilizer formula of the wheat belt, and simultaneously opens the solenoid valve on the water supply pipeline of the wheat belt and closes the solenoid valve on the water supply pipeline of the cotton belt to supply water-fertilizer mixture with the corresponding formula to the wheat belt; after the irrigation and fertilization of the wheat belt is completed, the control unit closes the solenoid valve of the wheat belt and opens the clean water flushing pipeline, adjusts the mother liquor intake ratio to the target fertilizer formula of the cotton belt, opens the solenoid valve on the water supply pipeline of the cotton belt, and supplies water-fertilizer mixture with the corresponding formula to the cotton belt; the two belts complete irrigation and fertilization in sequence without interfering with each other.

[0119] When the fertilizer formulas for the wheat and cotton strips are the same, the control unit can simultaneously open the solenoid valves on the water supply pipelines of both strips to achieve synchronous irrigation and fertilization, thereby shortening the total execution time. Furthermore, the control unit controls the opening of the solenoid valves on the wheat strip's water supply pipeline according to the irrigation and fertilization control parameters, and controls the water-fertilizer mixing device according to the corresponding target irrigation volume, target fertilization volume, and fertilizer formula, so that the water-fertilizer mixture enters the shallow-buried drip irrigation belt through the wheat strip's water supply pipeline to irrigate and fertilize the wheat strip. The specific method by which the control unit converts decision parameters into execution actions is as follows: (1) Target irrigation volume converted into solenoid valve opening time: The control unit calculates the required total water supply volume based on the target irrigation volume and the effective irrigation area of ​​the crop strip, and then calculates the solenoid valve opening time based on the pipeline flow rate; the calculation formula is as follows: ; in, This indicates the duration the solenoid valve remains open, in minutes. This indicates the target irrigation volume, expressed in cubic meters per mu (unit of land area). This indicates the effective irrigated area of ​​the current crop belt, expressed in mu (approximately 0.16 acres). This indicates the actual measured flow rate or the preset rated flow rate of the pipeline, in liters per minute. This represents the conversion factor for cubic meter uplift.

[0120] When a flow sensor is installed on the pipeline, the control unit adopts a closed-loop control method: after the solenoid valve is opened, the cumulative flow value of the flow sensor is read in real time, and the solenoid valve is automatically closed when the cumulative flow reaches the target water supply volume, so as to achieve accurate metering; when no flow sensor is installed on the pipeline, the control unit adopts an open-loop control method: the opening time of the solenoid valve is calculated according to the above formula, and the solenoid valve is automatically closed after the preset time is reached.

[0121] (2) The target fertilizer application rate and fertilizer formula are converted into a water-fertilizer mixing ratio: The control unit calculates the intake amount of each nutrient mother liquor based on the target fertilizer application rate, target irrigation amount and mother liquor concentration, and then achieves the target mixing ratio by adjusting the fertilizer pump speed or the opening of the proportional control valve; the calculation formula is as follows: ; in, Indicates the first The target inhalation flow rate of the seed nutrient mother liquor, in liters per minute; This indicates the target nutrient application rate, expressed in kilograms per acre. This indicates the concentration of the nutrient stock solution, expressed in kilograms per liter. This indicates the duration the solenoid valve remains open, in minutes.

[0122] When an EC sensor (conductivity sensor) is installed on the pipeline, the control unit adopts a closed-loop control method: it monitors the EC value of the mixture in real time and dynamically adjusts the fertilizer pump speed or the proportional control valve opening through a PID algorithm to keep the EC value of the mixture within the target range; when no EC sensor is installed, the control unit adopts an open-loop control method: it calculates the target suction flow rate according to the above formula, sets the pump speed according to the flow-speed characteristic curve of the fertilizer pump, or sets the valve opening according to the flow-opening characteristic curve of the proportional control valve, and performs irrigation and fertilization according to preset parameters.

[0123] (3) Monitoring and protection during the execution process: The control unit monitors the pipeline pressure sensor data in real time during the irrigation and fertilization process. When the pipeline pressure exceeds the upper limit of pipeline pressure safety or falls below the lower limit of pipeline pressure safety, the solenoid valve and fertilizer pump are automatically shut off to prevent pipeline rupture or dry running. When the liquid level sensor of the mother liquor tank detects that the liquid level is lower than the lower limit of the mother liquor tank liquid level, the control unit issues a liquid shortage alarm and suspends fertilization, performing only clean water irrigation. The upper limit of pipeline pressure safety, the lower limit of pipeline pressure safety, and the lower limit of the mother liquor tank liquid level are determined by the pipeline pressure resistance test, the statistical results of normal operating pressure, and the effective fertilizer absorption liquid level of the mother liquor tank during the system installation and commissioning phase, and are written into the control unit. Furthermore, the control unit controls the opening and closing status of the solenoid valves on each branch water supply pipeline and the water-fertilizer mixing ratio of the water-fertilizer mixing device to achieve wheat The system provides independent outputs for different irrigation volumes, fertilization amounts, and fertilizer formulas for the wheat and cotton strips. The opening and closing states of the solenoid valves control the irrigation on / off status and water supply duration for each crop strip. The adjustment of the water-fertilizer mixing ratio is based on the conversion relationship between the target fertilization amount, mother liquor concentration, pipeline flow rate, and valve opening time. The control unit calculates the target intake flow rate of each nutrient mother liquor based on the target fertilization amount and mother liquor concentration, and achieves the target mixing ratio by adjusting the fertilizer pump speed or the proportional control valve opening. The specific conversion method and closed-loop / open-loop control methods are described in the aforementioned execution control section. When the fertilizer formulas for the wheat and cotton strips are different, the control unit simultaneously adjusts the fertilizer pump speed or the proportional control valve opening of the water-fertilizer mixing device while switching the solenoid valves of the branch water supply pipelines, so that the nitrogen, phosphorus, and potassium ratio of the mixed solution is switched to the target fertilizer formula for the corresponding crop strip. When the irrigation and fertilization control parameters for the wheat belt and the cotton belt are different, the control unit controls the corresponding branch water supply pipelines according to the irrigation and fertilization control parameters of the corresponding crop belt, so that the water and fertilizer supply of the wheat belt and the cotton belt do not interfere with each other.

[0124] Furthermore, the control unit also integrates a wireless communication module, which communicates with a remote monitoring terminal, which is a mobile app or PC-based management platform. The remote monitoring terminal is used for remote setting, real-time monitoring, and data recording of irrigation and fertilization parameters, as well as the display and interactive selection of Pareto frontier schemes. The Pareto frontier schemes are generated by the differentiated irrigation and fertilization decision module and pushed to the remote monitoring terminal for visualization. The remote monitoring terminal receives the user's manual selection results and sends them back to the differentiated irrigation and fertilization decision module. After confirming the final execution plan, the differentiated irrigation and fertilization decision module generates irrigation and fertilization control parameters for the wheat strip and the cotton strip, and then sends them to the intelligent branch control module for execution. The remote monitoring terminal does not directly send execution commands to the intelligent branch control module to ensure unified control of the final execution plan by the decision module.

[0125] Managers can view soil moisture, irrigation records, fertilization records, future moisture prediction curves, and Pareto front plots for wheat and cotton belts in real time through remote monitoring terminals. They can also manually select different optimization schemes or modify irrigation and fertilization plans. When managers manually modify irrigation and fertilization plans through remote monitoring terminals, the modified schemes must be verified by the security constraints of the differentiated irrigation and fertilization decision module before they can be issued for execution.

[0126] The safety constraint verification includes the following four items: (1) the irrigation amount does not exceed the regional water quota limit; (2) the amount of nitrogen, phosphorus and potassium applied does not exceed the single application limit of the corresponding nutrients; (3) the estimated root layer water content after irrigation does not exceed the field water holding capacity; (4) the estimated root layer water content after irrigation is not lower than the wilting point; when the modified plan fails the safety constraint verification, the differentiated irrigation and fertilization decision module returns a verification failure prompt and specific over-limit items to the remote monitoring terminal, does not issue an execution instruction, and the manager adjusts and resubmits.

[0127] Furthermore, the intelligent branch control module executes irrigation and fertilization operations based on the irrigation and fertilization control parameters generated by the differentiated irrigation and fertilization decision module. After the differentiated irrigation and fertilization decision module generates irrigation and fertilization control parameters for the wheat strip and the cotton strip, the intelligent branch control module controls the corresponding solenoid valves and water and fertilizer mixing devices for the wheat strip and the cotton strip respectively, supplying water and fertilizer independently to the wheat strip and the cotton strip with different irrigation amounts, fertilizer amounts, and fertilizer distribution directions.

[0128] Furthermore, during a single irrigation and fertilization process in the wheat-cotton co-existing stage, when both the wheat and cotton strips require irrigation and fertilization simultaneously, the control unit executes the applications sequentially according to the priority order output by the differentiated irrigation and fertilization decision module. For example, when the differentiated irrigation and fertilization decision module determines that the wheat strip needs irrigation and fertilization while the cotton strip needs light water replenishment, the priority formula is used to calculate the priority order for the wheat strip. If the priority is higher than that of the cotton strip, the control unit will first execute the irrigation and fertilization control parameters of the wheat strip, and then execute the water replenishment control parameters of the cotton strip; if the priority calculation results are opposite, the control unit will first execute the irrigation and fertilization control parameters of the cotton strip, and then execute the irrigation and fertilization control parameters of the wheat strip.

[0129] When the differentiated irrigation and fertilization decision module fails to generate an optimization scheme due to network interruption, insufficient computing resources, computation timeout, optimization non-convergence, or constraints with no feasible solution, the control unit executes the irrigation and fertilization operation of the corresponding crop strip according to the irrigation and fertilization control parameters generated by the basic decision sub-model to ensure basic irrigation and fertilization needs. The above rollback scenario is consistent with the trigger scenario where the optimization enhancement module in the differentiated irrigation and fertilization decision module fails to generate a Pareto front, ensuring that the trigger conditions for decision rollback and execution rollback are unified. Through the above-mentioned intelligent branch control module, the irrigation and fertilization control parameters for wheat strip and cotton strip generated by the differentiated irrigation and fertilization decision module can be converted into corresponding solenoid valve opening and closing control and water and fertilizer mixing control.

[0130] The irrigation and fertilization control parameters are structured instruction data, including at least the crop strip identifier, target irrigation amount, target fertilization amount, nitrogen, phosphorus, and potassium ratio, corresponding valve number, opening duration, water-fertilizer mixing ratio, execution sequence, and execution time. The crop strip identifier indicates either the wheat strip or the cotton strip corresponding to the current structured instruction data. The target irrigation amount represents the water required for the current irrigation, in cubic meters per acre. The target fertilization amount represents the application rate of nitrogen, phosphorus, and potassium during the current irrigation and fertilization process, in kilograms per acre. The nitrogen, phosphorus, and potassium ratio represents the absorption ratio of each nutrient solution. The corresponding valve number indicates the solenoid valve corresponding to the target crop strip. The opening duration represents the holding time of the target solenoid valve. Continued opening time; water-fertilizer mixing ratio is used to characterize the target concentration after mixing mother liquor and clean water; execution order is used to characterize the priority order when both wheat and cotton strips have irrigation and fertilization needs; execution time is used to characterize the scheduled execution time; for example, the crop strip is identified as wheat strip, the target irrigation amount is 2.5 cubic meters / acre, the target fertilizer amount is 3 kg / acre of nitrogen fertilizer, 1.5 kg / acre of phosphorus fertilizer, and 2 kg / acre of potassium fertilizer, the nitrogen-phosphorus-potassium ratio is 3:1.5:2, the corresponding valve number is V1, the opening time is 15 minutes, the target conductivity value corresponding to the water-fertilizer mixing ratio is 1.8 mS / cm, the execution order is Priority=0.62 and ranked 1st, and the execution time is 06:00 on the same day.

[0131] After receiving the structured instruction data, the control unit determines the target solenoid valve based on the crop strip identifier and the corresponding valve number, determines the valve opening and closing sequence based on the target irrigation volume and opening duration, adjusts the fertilizer pump speed and proportional control valve opening based on the nitrogen, phosphorus and potassium ratio and water-fertilizer mixing ratio, and sends control instructions to the corresponding solenoid valve, water-fertilizer mixing device and fertilizer pump when the execution time arrives according to the execution sequence, so that the shallow buried drip irrigation tape and the mulched drip irrigation tape supply water and fertilizer to the wheat strip and cotton strip respectively according to their corresponding target irrigation volume, target fertilizer volume and fertilizer formula.

[0132] In summary, this invention achieves independent water and fertilizer supply for wheat and cotton intercropping by laying shallow-buried drip irrigation tape in wheat strips and mulched drip irrigation tape in cotton strips, and connecting them to independent branch water supply pipelines, thus avoiding water and fertilizer supply mismatch caused by unified irrigation and fertilization. It also improves the adaptability of irrigation and fertilization decisions during the wheat-cotton co-cropping period and at different stages after wheat harvest by collecting soil volumetric water content, soil temperature, and crop growth data for both crop strips, and calculating target irrigation amounts, target fertilization amounts, and fertilization formulas based on growth stages, reference evapotranspiration, and crop coefficients. Furthermore, it predicts future root zone soil water content using an LSTM time-series prediction module and a multi-objective optimization module, and generates optimized solutions based on yield, water use efficiency, nitrogen fertilizer partial productivity, and nitrogen leaching risk, improving the foresight and comprehensive coordination of water and fertilizer management. Finally, it converts irrigation and fertilization control parameters into electromagnetic valve opening and closing control and water and fertilizer mixing control through an intelligent branch control module, enabling precise branch irrigation and fertilization management in wheat-cotton intercropping fields.

[0133] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A differentiated integrated water and fertilizer management system for wheat-cotton intercropping, characterized in that: It includes a field deployment module, a data acquisition module, a differentiated irrigation and fertilization decision module, and an intelligent branch control module, wherein the intelligent branch control module is connected to the field deployment module, the data acquisition module, and the differentiated irrigation and fertilization decision module respectively; A field deployment module is used to lay shallow-buried drip irrigation tape in wheat strips and mulched drip irrigation tape in cotton strips. The shallow-buried drip irrigation tape is buried below the soil surface between wheat rows, and the mulched drip irrigation tape is laid under the mulch film between cotton rows. The shallow-buried drip irrigation tape and the mulched drip irrigation tape are independently connected to the corresponding branch water supply pipelines. The data acquisition module is used to collect soil volumetric water content and soil temperature at different depths in the wheat and cotton belts through soil temperature and humidity sensors deployed in the wheat belt and the cotton belt, and to collect crop growth data in the wheat and cotton belts through a crop growth monitoring device. The differentiated irrigation and fertilization decision module is used to calculate the root zone weighted average water content, target irrigation amount, target fertilization amount and fertilization formula for wheat strip and cotton strip respectively, based on the differentiated water and fertilizer requirements parameters of wheat and cotton at their respective growth stages, combined with soil volumetric water content, soil temperature, crop growth data, reference evapotranspiration and crop coefficient collected by the data acquisition module, and to generate irrigation and fertilization control parameters for wheat strip and cotton strip respectively. The intelligent branch control module is used to receive irrigation and fertilization control parameters for wheat strips and cotton strips through the control unit, and to control the solenoid valves and water-fertilizer mixing devices installed on each branch water supply pipeline, so that the shallow buried drip irrigation tape and the film-covered drip irrigation tape can independently supply water and fertilizer to the wheat strip and cotton strip according to the corresponding target irrigation amount, target fertilization amount and fertilization formula.

2. The differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping as described in claim 1, characterized in that: The branch water supply pipelines include a wheat strip water supply pipeline connected to a shallow-buried drip irrigation tape and a cotton strip water supply pipeline connected to a mulched drip irrigation tape. The wheat strip water supply pipeline and the cotton strip water supply pipeline are each equipped with independently controlled solenoid valves.

3. The differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping as described in claim 2, characterized in that: The soil temperature and humidity sensors are arranged in layers along the root depth of the wheat and cotton strips to collect soil volumetric water content and soil temperature at different soil layers. The data acquisition module collects soil volumetric water content and soil temperature at a preset acquisition frequency, and generates soil temperature and humidity data for the corresponding wheat and cotton belts during the decision-making period.

4. The differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping as described in claim 3, characterized in that: The crop growth monitoring device includes a multispectral camera; The multispectral camera is used to acquire multispectral images of wheat and cotton belts; The crop growth data includes canopy spectral information and normalized vegetation index; The normalized vegetation index is calculated based on the near-infrared and red light reflectance in the canopy spectral information.

5. The differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping as described in claim 4, characterized in that: The root layer weighted average water content is calculated based on the soil volumetric water content at different depths and the crop root distribution weights. The crop root distribution weights are configured according to the root water absorption areas corresponding to the growth stages of wheat and cotton, respectively. The differentiated irrigation and fertilization decision module calculates the soil moisture deficit based on the root zone weighted average moisture content, the target soil moisture content at the current growth stage, and the wilting point.

6. The differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping as described in claim 5, characterized in that: The target irrigation amount is calculated based on the target soil moisture content, root zone weighted average moisture content, planned wetting layer depth, planned wetting ratio, and unit conversion factor at the current growth stage. The differentiated irrigation and fertilization decision module calculates the actual crop evapotranspiration based on the reference evapotranspiration and crop coefficient, and uses the actual crop evapotranspiration to correct the irrigation interval and target irrigation amount.

7. The differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping as described in claim 6, characterized in that: The target fertilization amount and fertilization formula are calculated based on the target yield, growth stage, available nutrients in the soil, and fertilizer utilization rate. The amount of nitrogen fertilizer applied is determined based on the target yield, growth stage, available nitrogen in the soil, and fertilizer utilization rate. The amount of phosphorus fertilizer applied and the amount of potassium fertilizer applied are adjusted according to the corresponding nutrient requirements and the remaining growth stage.

8. The differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping as described in claim 1, characterized in that: The differentiated irrigation and fertilization decision module determines the growth stage of wheat and cotton based on the current date, accumulated temperature data, and wheat harvest confirmation signal. During the wheat-cotton co-existence period, the wheat belt uses the grain-filling stage sub-model, while the cotton belt uses the seedling stage model. After the wheat harvest, the wheat strip model switches to a dormant state, while the cotton strip model switches to one of the budding stage sub-model or the flowering and boll-forming stage sub-model, which corresponds to the current growth stage of cotton.

9. The differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping as described in claim 8, characterized in that: The differentiated irrigation and fertilization decision module also includes an LSTM time series prediction module and a multi-objective optimization module; The LSTM time series prediction module is a long short-term memory network time series prediction module. Based on historical soil moisture content, meteorological forecast data, growth stage and historical irrigation amount, it predicts the baseline change of root soil moisture content in wheat belt and cotton belt within a preset time period. The multi-objective optimization module, based on the predicted root soil moisture content baseline change, superimposes the moisture content change formed by the candidate irrigation amount under each group of candidate irrigation amount, nitrogen application amount, phosphorus application amount, potassium application amount and irrigation interval. Taking the irrigation amount, nitrogen application amount, phosphorus application amount, potassium application amount and irrigation interval of wheat belt and cotton belt as optimization variables, and taking the maximization of total yield, water use efficiency and nitrogen fertilizer partial productivity and the minimization of nitrogen leaching risk as optimization objectives, it generates a Pareto frontier scheme set. When the multi-objective optimization module does not generate a Pareto front scheme set, the differentiated irrigation and fertilization decision module generates the target irrigation amount based on the root zone weighted average water content, and generates the target fertilization amount and fertilization formula based on the target yield, growth stage, available soil nutrients and fertilizer utilization rate.

10. The differentiated water and fertilizer integrated intelligent management system for wheat-cotton intercropping as described in claim 1, characterized in that: The intelligent branch control module includes a control unit, a water and fertilizer mixing device, a water supply pipeline for wheat, a water supply pipeline for cotton, and a solenoid valve. The solenoid valves are respectively installed on the water supply pipelines for wheat and cotton. The wheat belt water supply pipeline is connected to the shallow buried drip irrigation belt; The cotton strip water supply pipeline is connected to the mulched drip irrigation tape; The control unit receives irrigation and fertilization control parameters for wheat strips and cotton strips, and controls the solenoid valves and water-fertilizer mixing devices on the corresponding water supply pipelines according to the irrigation and fertilization control parameters for wheat strips and cotton strips, so that the shallow-buried drip irrigation tape and the film-covered drip irrigation tape can independently supply water and fertilizer to the wheat strips and cotton strips according to the corresponding target irrigation amount, target fertilization amount and fertilization formula.

Images