A water saving fertilization variable control method suitable for sandy soil

Abstract

The application relates to the technical field of intelligent control, and discloses a water-saving and fertilization variable control method suitable for sandy soil, which comprises the following steps: collecting data based on a sensor; establishing and updating an infiltration-storage-leaching prediction model of the sandy soil according to the collected data; estimating the state of soil moisture and nutrient state, and calculating the root layer moisture deviation, the fertilizer concentration deviation and the deep seepage risk; performing model prediction control in a rolling time domain to obtain a control result; performing water-saving and fertilization operation according to the control result; in the execution, fusing soil volume water content, soil water potential, soil conductivity and shallow seepage observation data, correcting the prediction model, and dynamically adjusting the control result according to the correction result; and when the environmental wind speed exceeds a set threshold, correcting the sprinkling irrigation parameters. The application improves the fine level of control, can simultaneously inhibit deep seepage, guarantee root layer water and fertilizer balance and cope with external disturbance under the condition of sandy soil.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and more specifically, to a water-saving fertilization variable control method suitable for sandy soils. Background Technology

[0002] In modern agriculture, especially in precision agriculture under sandy soil conditions, integrated water and fertilizer management has become an important means to improve water and fertilizer use efficiency and crop yield. Sandy soils are characterized by high infiltration rates, low water retention capacity, and easy fertilizer loss due to water seepage. Therefore, irrigation and fertilization must balance soil moisture retention and nutrient utilization, which places higher demands on the dynamic control of irrigation cycles, fertilizer concentrations, and application methods. Traditional irrigation and fertilization devices mostly focus on equipment structure design or simple valve adjustments, making it difficult to achieve predictive and closed-loop control based on sensor data. This results in limited control accuracy and an inability to fully address the problems of deep seepage and fertilizer waste under sandy soil conditions.

[0003] For example, patent CN113966714A discloses an automatic irrigation fertilization device and method for field fields. It uses meteorological parameters and soil moisture as inputs, outputs three discrete irrigation coefficients via a three-layer neural network, and adjusts the nozzle opening accordingly. While this solution achieves automation to some extent, it still suffers from the following problems: limited sensor data dimensions, lacking real-time perception of soil conductivity, nutrient concentration, and shallow seepage, making it difficult to comprehensively estimate soil water and fertilizer status; control output is limited to three opening levels (maximum, intermediate, and minimum), representing coarse-grained adjustment and failing to achieve continuous variable control through rolling optimization; and the lack of a closed-loop mechanism between prediction and control prevents real-time correction of model biases and environmental disturbances, leading to problems such as over-irrigation, deep seepage, and nutrient loss in sandy soil conditions.

[0004] Therefore, it is necessary to design a water-saving fertilization variable control method suitable for sandy soils to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a water-saving fertilization variable control method suitable for sandy soil, aiming to solve the problems of insufficient perception of process variables, lack of adaptive prediction models, extensive optimization of control variables, and insufficient stability under external disturbances.

[0006] This invention proposes a water-saving fertilization variable control method suitable for sandy soils, comprising:

[0007] Data is acquired based on sensors, including soil volumetric water content and / or soil water potential, soil electrical conductivity, nutrient concentration, crop canopy stress index, microclimate parameters, pipeline flow rate, pressure, and online fertilizer solution concentration data.

[0008] Based on the collected data, an infiltration-storage-leaching prediction model for sandy soil was established and updated. The state of soil moisture and nutrients was estimated, and the root zone moisture deviation, fertilizer concentration deviation, and deep leakage risk were calculated.

[0009] In the rolling time domain, model predictive control is performed based on the prediction model to optimize the irrigation cycle, pulse duty cycle, nozzle pressure, execution unit moving speed and fertilizer solution target concentration, so as to obtain control results that meet the requirements of root soil volumetric water content or soil water potential within a preset range, root fertilizer concentration meeting the stage requirements and deep seepage flux below the threshold.

[0010] The control results are sent to the execution unit to drive the variable ratio fertilizer pump, sprinkler / drip irrigation device and moving mechanism to perform water-saving fertilizer operation according to the cycle, duty cycle, pressure, speed and fertilizer concentration.

[0011] During implementation, soil volumetric moisture content, soil water potential, soil electrical conductivity, and shallow seepage observation data are integrated to correct the prediction model. The control results are dynamically adjusted based on the correction results. When the ambient wind speed exceeds the set threshold, the sprinkler irrigation parameters are corrected.

[0012] Furthermore, when acquiring data based on sensors, this includes:

[0013] Within each control zone, the sensors are deployed in pairs. Soil volumetric moisture content sensors and soil water potential sensors are buried at root zone depths of 10–30 cm and 30–60 cm, respectively, and are coaxially arranged with soil conductivity sensors at the same borehole location. Nutrient concentration sensors are configured as nitrate nitrogen and ammonium nitrogen ion selective electrodes. The crop canopy stress index is obtained using near-infrared leaf temperature and ambient temperature / humidity optical / thermal integrated sensors. Microclimate parameters include wind speed, air temperature, relative humidity, and solar radiation. Pipeline flow and pressure sensors are respectively located downstream of the variable ratio mixer and downstream of the booster pump. The online fertilizer solution concentration is measured in parallel using both conductivity and refractive index methods, and cross-checked. All sensors are time-synchronized and unit-consistent at the edge computing node with a basic sampling cycle of 1–5 minutes.

[0014] Furthermore, when establishing and updating the infiltration-storage-leaching prediction model for sandy soil based on the collected data, the following steps are included:

[0015] The infiltration-storage-leaching prediction model is composed of a Green-Ampt infiltration sub-model, a root zone water balance sub-model, and a one-dimensional convection-diffuse nutrient transport sub-model coupled together. Its parameters are identified online by the measured trajectory of soil volumetric water content and soil water potential based on short-cycle pulsed irrigation data at the edge computing node, and the deep infiltration flux boundary constraint is set by shallow leakage observation.

[0016] Furthermore, when estimating the state of soil moisture and nutrients, the following should be included:

[0017] Extended Kalman filtering is used to fuse soil volumetric water content, soil water potential, soil electrical conductivity, nutrient concentration, and shallow seepage observations to estimate root zone water and root zone fertilizer concentration, and output root zone water deviation and fertilizer concentration deviation. At the same time, a deep seepage risk index is constructed by cross-checking the predicted deep seepage flux and shallow seepage observations, and the deviation and risk index are fed back as constraints and weight update amounts for model predictive control.

[0018] Furthermore, when performing model predictive control based on the predicted model within the rolling time domain to optimize the irrigation cycle, pulse duty cycle, nozzle pressure, execution unit moving speed, and fertilizer solution target concentration, the following is included:

[0019] Apply constraints on the irrigation cycle, pulse duty cycle, nozzle pressure, execution unit moving speed, and upper and lower limits and rate of change of fertilizer solution target concentration;

[0020] Based on the infiltration-storage-leakage prediction model, the instantaneous application intensity is less than or equal to the infiltration capacity constraint.

[0021] Feasible domain constraints of nozzle pressure-flow rate-spray width and upper limit constraints of fertilizer solution target concentration;

[0022] Set control increment penalty and terminal deviation penalty, and only issue control quantities that represent the current sampling period.

[0023] Furthermore, the model predictive control incorporates short-term predictions of microclimate parameters and pipeline pressure fluctuation scenarios for scenario robust optimization, satisfying deep leakage flux constraints at a given confidence level.

[0024] Rolling re-optimization is performed with a sampling period of 1 to 5 minutes and a warm start is adopted. When feasibility degradation is detected, a safety degradation strategy is implemented: successively reducing the nozzle pressure, reducing the pulse duty cycle, lowering the target concentration of fertilizer solution, or pausing. Based on the pressure-flow-spray width mapping and the moving speed-unit application rate mapping obtained by calibration, the conversion of control quantity to execution unit instruction is completed.

[0025] Furthermore, when the control results are sent to the execution unit to drive the variable ratio fertilizer pump, sprinkler / drip irrigation device, and moving mechanism to perform water-saving fertilization operations according to the cycle, duty cycle, pressure, speed, and fertilizer concentration, the following steps are included:

[0026] The control results are sent to the execution unit via fieldbus based on the edge computing node, and the irrigation cycle, pulse duty cycle, nozzle pressure, execution unit moving speed and fertilizer solution target concentration are synchronized in time according to the same sampling period;

[0027] The pulse duty cycle is achieved by the PWM of the partition solenoid valve, the nozzle pressure is tracked by the frequency converter of the booster pump and the closed loop of the pressure regulating valve, the movement speed of the execution unit is tracked by the speed closed loop drive module, and the target concentration of fertilizer solution is given by the variable ratio fertilizer pump. During execution, the inner loop correction is achieved by the feedback of pipeline flow rate, pressure and online concentration of fertilizer solution.

[0028] Furthermore, when driving the variable-ratio fertilizer pump, sprinkler / drip irrigation device, and moving mechanism to perform water-saving fertilization operations according to the aforementioned cycle, duty cycle, pressure, speed, and fertilizer concentration, the driving process includes:

[0029] A variable-ratio fertilizer pump and a Venturi proportioner form a variable-ratio fertilizer channel to track the target concentration of fertilizer solution; the sprinkler / drip irrigation device performs water-saving fertilization according to the irrigation cycle and pulse duty cycle, and automatically shortens the spray width or switches to drip irrigation when the wind speed exceeds the threshold; at the end of the execution of each control zone, a flushing sequence and check / anti-siphon interlock are triggered, and when pressure or concentration exceeds the limit, a safety degradation is performed in sequence: reducing the nozzle pressure, reducing the pulse duty cycle, lowering the target concentration of fertilizer solution, and pausing.

[0030] Furthermore, when calibrating the prediction model by integrating soil volumetric moisture content, soil water potential, soil electrical conductivity, and shallow seepage observation data during implementation, the following steps are included:

[0031] The parameters of the infiltration-storage-leaching prediction model are updated using the innovation of the extended Kalman filter through recursive least squares. These parameters include infiltration parameters, effective porosity, and dispersion coefficient. Based on this, the state estimates of root zone soil volumetric water content and root zone fertilizer concentration, as well as the deep leakage risk index, are recalculated. When the observation variance or the absolute value of innovation exceeds the threshold, the sampling period is shortened to 30–60 seconds, and the weights and constraints in the model's predictive control are adjusted simultaneously. This increases the cost weights of root zone soil volumetric water content deviation, root zone fertilizer concentration deviation, and deep leakage flux as the risk increases.

[0032] Furthermore, the control results are dynamically adjusted based on the correction results. When the ambient wind speed exceeds a set threshold, the irrigation parameters are corrected, including:

[0033] The wind correction mapping proportionally reduces the nozzle pressure and correspondingly reduces the spray width, while simultaneously lowering the pulse duty cycle and the upper limit of the execution unit's movement speed. When the wind speed continuously exceeds the threshold and triggers the secondary wind speed threshold, the drip irrigation mode is automatically switched, and the original sprinkler irrigation settings are restored after the wind speed drops back to the recovery threshold with hysteresis. Throughout the correction process, the constraints such as the instantaneous application intensity being less than or equal to the infiltration capacity and the upper limit of the target fertilizer concentration remain unchanged.

[0034] Compared with existing technologies, the advantages of this invention are as follows: By collecting key variables such as soil volumetric water content, soil water potential, soil electrical conductivity, nutrient concentration, crop canopy stress index, and microclimate parameters through multi-source sensors, and combining pipeline flow rate, pressure, and online fertilizer concentration information, an infiltration-storage-leaching prediction model is established and dynamically updated. This enables real-time estimation of root zone moisture and nutrient status, and calculates deviations in root zone soil volumetric water content, root zone nutrient concentration, and deep seepage risk. The model predicts and controls the continuous optimization of irrigation cycle, pulse duty cycle, sprinkler pressure, execution unit movement speed, and target fertilizer concentration in the rolling time domain, thereby obtaining accurate control results and sending them to the execution unit to drive the variable ratio fertilizer pump, sprinkler / drip irrigation device, and moving mechanism to perform water-saving fertilizer application operations. During execution, sensor data and shallow seepage observations are further integrated to perform closed-loop correction of the prediction model, and the sprinkler parameters are automatically corrected when wind speed disturbances exceed the threshold, thus ensuring stable and reliable control even under uncertain environments. It improves the precision of water conservation and fertilizer control, and can simultaneously inhibit deep seepage, ensure the water and fertilizer balance of the root zone, and cope with external disturbances under sandy soil conditions. Attached Figure Description

[0035] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0036] Figure 1 A flowchart of a water-saving fertilization variable control method suitable for sandy soils provided in an embodiment of the present invention. Detailed Implementation

[0037] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0038] In traditional fertigation control systems for sandy soils, sensor data acquisition is limited to soil moisture and meteorological parameters, lacking real-time monitoring of soil conductivity, nutrient concentration, and shallow seepage flux. This results in an inability to accurately estimate the dynamic distribution of root zone moisture and nutrients. Control strategies rely on discrete speed adjustments, failing to achieve continuous optimization of irrigation cycles, pulse duty cycles, sprinkler pressure, and fertilizer concentration. Furthermore, the lack of a closed-loop feedback mechanism between the predictive model and control execution makes the system prone to exceeding root zone moisture limits, accumulating fertilizer concentration deviations, and exacerbating the risk of deep seepage when soil infiltration rates change abruptly or environmental disturbances occur.

[0039] For example, in traditional three-layer neural network-controlled field sprinkler irrigation systems, soil conductivity and shallow infiltration observation data are not included in the input layer, making it impossible to identify the migration trend of nitrate nitrogen below the root zone. The control output can only adjust the application rate through three nozzle opening levels, failing to dynamically match the combination parameters of pulse duty cycle and movement speed based on soil infiltration capacity. When sudden wind speed changes cause spray amplitude deviation, the system lacks a feasible domain constraint for pressure-flow-spray amplitude, leading to instantaneous application intensity exceeding the infiltration rate in local areas, resulting in excessive deep infiltration flux. Simultaneously, fertilizer concentration adjustment is limited to a fixed ratio channel, preventing real-time closed-loop correction based on root zone nutrient requirements.

[0040] If the above problems are not addressed, the moisture and nutrients in sandy soils will continuously deviate from the target range due to a lack of accurate state estimation, leading to water stress and nutrient imbalance in crop roots. Coarse-grained control strategies cannot suppress random fluctuations in deep seepage fluxes, resulting in wasted water and fertilizer resources and the risk of groundwater pollution. Open-loop control architectures are prone to mismatches between execution parameters and soil infiltration capacity under environmental disturbances, reducing control accuracy and ultimately hindering crop yield increases and the achievement of sustainable production goals.

[0041] For this, please refer to Figure 1 As shown, this application proposes a water-saving fertilization variable control method suitable for sandy soils, including:

[0042] S100: Data is acquired based on sensors, including soil volumetric water content and / or soil water potential, soil electrical conductivity, nutrient concentration, crop canopy stress index, microclimate parameters, pipeline flow rate, pressure, and online fertilizer solution concentration data.

[0043] S200: Based on the collected data, establish and update the infiltration-storage-leaching prediction model for sandy soil, estimate the state of soil moisture and nutrients, and calculate the root zone moisture deviation, fertilizer concentration deviation, and deep leakage risk.

[0044] S300: In the rolling time domain, based on the prediction model, the model predictive control is executed to optimize the irrigation cycle, pulse duty cycle, nozzle pressure, execution unit moving speed and fertilizer solution target concentration, so as to obtain control results that meet the requirements of root soil volumetric water content or soil water potential within the preset range, root fertilizer concentration meeting the stage requirements and deep seepage flux below the threshold.

[0045] S400: Sends control results to the execution unit, driving the variable ratio fertilizer pump, sprinkler / drip irrigation device and moving mechanism to perform water-saving fertilizer operation according to cycle, duty cycle, pressure, speed and fertilizer concentration.

[0046] S500: During execution, it integrates soil volumetric moisture content, soil water potential, soil electrical conductivity and shallow seepage observation data to correct the prediction model, and dynamically adjusts the control results based on the correction results. When the ambient wind speed exceeds the set threshold, it corrects the sprinkler irrigation parameters.

[0047] Specifically, sensor data acquisition refers to the real-time monitoring of various soil and environmental parameters using multiple sensors. These can include soil volumetric moisture sensors, soil water potential sensors, soil conductivity sensors, nitrate and ammonium nitrogen ion selective electrodes, near-infrared leaf temperature sensors, integrated optical / thermal temperature and humidity sensors, pipeline flow sensors, pressure sensors, and online fertilizer solution concentration sensors with dual parallel channels using conductivity and refractive index methods. These sensors, deployed in layers and coaxially, can comprehensively acquire dynamic data on root zone water, nutrients, crop stress, and the environment, providing multi-dimensional input for model prediction. The infiltration-storage-leaching prediction model is a mathematical model used to simulate the dynamic changes in water and nutrients in sandy soils. It is specifically constructed by coupling the Green–Ampt infiltration sub-model, the root zone water balance sub-model, and the one-dimensional convection-diffuse nutrient transport sub-model. This model uses short-cycle pulsed irrigation data for online parameter identification and combines shallow seepage observations with constraints on deep seepage flux boundaries. It can accurately predict root zone moisture deviations, fertilizer concentration deviations, and deep seepage risks, providing a dynamic basis for optimized control. Model predictive control refers to multivariate optimization based on the predictive model within the rolling time domain. Specifically, it sets upper and lower bounds and rate of change constraints for irrigation cycle, pulse duty cycle, sprinkler pressure, execution unit movement speed, and target fertilizer concentration. Combined with instantaneous application intensity, the feasible region of sprinkler pressure-flow-spray width, and upper limit constraints on fertilizer concentration, it generates control results that meet root zone moisture and nutrient requirements while suppressing deep seepage. It achieves smooth output of control quantities through incremental penalties and terminal deviation penalties, ensuring stable operation in dynamic environments. The process of sending control results to the execution unit involves converting optimized control parameters into equipment operation commands. Specifically, this is achieved through adjusting fertilizer concentration using a variable-ratio fertilizer pump, executing irrigation according to cycles and duty cycles using sprinkler / drip irrigation devices, adjusting the movement speed of the moving mechanism, and utilizing PWM control of zoned solenoid valves, frequency conversion of the booster pump, closed-loop tracking of the pressure regulating valve, and a speed closed-loop drive module to ensure precise parameter execution. Inner loop flow, pressure, and concentration feedback corrections ensure consistency between the execution process and the commands. The predictive model correction and dynamic adjustment involve updating model parameters by integrating real-time observation data during execution. Specifically, extended Kalman filtering is used to fuse soil volumetric water content, soil water potential, soil conductivity, and shallow infiltration observation data to recursively correct infiltration parameters, effective porosity, and dispersion coefficient, while simultaneously adjusting control weights and constraints. When the ambient wind speed exceeds a threshold, wind correction mapping is used to reduce sprinkler pressure, narrow the spray width, and adjust the duty cycle and movement speed, or switch to drip irrigation mode to maintain infiltration capacity and concentration constraints.

[0048] This application constructs a dynamic prediction model for water and fertilizer in sandy soil by fusing data from multiple sensor sources. Combined with rolling time-domain optimization and real-time feedback correction, it achieves multi-variable coordinated control of irrigation cycle, pulse duty cycle, sprinkler pressure, moving speed, and fertilizer concentration. This closed-loop control mechanism can dynamically match the root zone's water and nutrient requirements while suppressing deep seepage, effectively solving the problems of easy water and fertilizer loss in sandy soils and the inefficient nature of traditional control methods.

[0049] The working process and principle of this application are as follows: First, data is acquired based on multiple sensors, including soil volumetric water content, soil water potential, soil electrical conductivity, nutrient concentration, crop canopy stress index, microclimate parameters, pipeline flow rate, pressure, and online fertilizer solution concentration data. This multi-dimensional data provides comprehensive dynamic information on soil water and fertilizer for subsequent modeling and control.

[0050] Based on the collected data, an infiltration-storage-leaching prediction model for sandy soils was established and updated. This model is used to estimate the state of soil moisture and nutrients, calculate root zone moisture deviation, fertilizer concentration deviation, and deep seepage risk. The establishment of the prediction model lays the foundation for subsequent optimized control.

[0051] Model predictive control is implemented based on a predictive model within the rolling time domain. Multiple variables, including irrigation cycle, pulse duty cycle, sprinkler pressure, actuator movement speed, and target fertilizer concentration, are optimized. The optimization objective is to ensure that the root zone soil volumetric water content or soil water potential is within a preset range, the root zone fertilizer concentration meets the phased requirements, and the deep percolation flux remains below a threshold. This multivariate optimization control enables precise water and fertilizer management.

[0052] The optimized control results are sent to the execution unit, driving the variable ratio fertilizer pump, sprinkler / drip irrigation device, and moving mechanism to perform water-saving fertilization operations according to the optimized cycle, duty cycle, pressure, speed, and fertilizer concentration. This step translates the optimization results into actual irrigation and fertilization actions.

[0053] During implementation, data on soil volumetric moisture content, soil water potential, soil electrical conductivity, and shallow seepage are continuously integrated to correct the prediction model in real time. The control results are dynamically adjusted based on the correction results, forming a closed-loop feedback mechanism. When the ambient wind speed exceeds a set threshold, the sprinkler irrigation parameters are corrected to adapt to environmental changes.

[0054] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0055] Multiple sensors were deployed in sandy soil farmland. Soil volumetric moisture content sensors and soil water potential sensors were buried at depths of 10 cm and 30 cm, respectively. Soil conductivity sensors and nutrient concentration sensors were installed in the same location. Crop canopy stress index was measured by near-infrared leaf temperature sensors. Microclimate parameters were collected by a weather station. Pipeline flow and pressure sensors were installed on irrigation pipes. Online fertilizer solution concentration was measured by conductivity sensors.

[0056] After data collection, an infiltration-storage-leaching prediction model was established. This model includes a Green-Ampt infiltration sub-model, a root zone water balance sub-model, and a one-dimensional convection-diffuse nutrient transport sub-model. Model parameters were identified online using short-cycle pulsed irrigation data.

[0057] An extended Kalman filter algorithm was used to fuse multi-source observation data to estimate root zone moisture and fertilizer concentration. Based on the estimation results, moisture deviation, fertilizer concentration deviation, and deep seepage risk index were calculated.

[0058] Model predictive control is executed within the rolling time domain. Control variables include irrigation cycle, pulse duty cycle, sprinkler pressure, actuator movement speed, and target fertilizer concentration. The optimization objective is to maintain root zone moisture content at 18%-22%, fertilizer concentration to meet crop growth stage requirements, and deep percolation flux below 5 mm / d. Optimization calculations are performed every 5 minutes.

[0059] The optimization results are transmitted to the execution unit via fieldbus. The variable-ratio fertilizer pump prepares the fertilizer solution according to the target concentration. The sprinkler irrigation system executes irrigation according to the optimized cycle and duty cycle. The moving mechanism moves at the optimized speed. Pipeline flow rate, pressure, and fertilizer solution concentration are monitored in real time, and inner-loop corrections are performed.

[0060] During execution, multi-source observation data is continuously integrated to correct the model. When a wind speed exceeding 3 m / s is detected, the sprinkler pressure is automatically reduced by 20%, and the pulse duty cycle and movement speed are adjusted accordingly to maintain uniform irrigation.

[0061] Through the above-described scheme, this application achieves comprehensive perception and precise control of the water and fertilizer status of sandy soils. Multi-dimensional data acquisition and dynamic prediction models improve the accuracy of state estimation. Rolling optimization and closed-loop correction mechanisms enhance adaptability to environmental changes. Variable control strategies can flexibly adjust irrigation parameters based on soil infiltration characteristics. Wind speed correction ensures the uniformity of sprinkler irrigation. These technical features work together to effectively solve the problems of deep infiltration and nutrient loss under sandy soil conditions, thereby improving water and fertilizer use efficiency.

[0062] In some of the solutions described above in this application, when data is acquired based on sensors, the real-time monitoring accuracy of soil moisture, nutrients and environmental parameters is limited due to unreasonable sensor arrangement or insufficient data acquisition dimensions. This makes it impossible to provide accurate input for subsequent model prediction and control, thereby affecting the regulation effect of irrigation and fertilization.

[0063] This application further proposes deploying sensors in pairs within each control zone. Soil volumetric moisture content sensors and soil water potential sensors are buried at root zone depths of 10–30 cm and 30–60 cm, respectively, and coaxially arranged with a soil conductivity sensor at the same borehole location. Nutrient concentration sensors are configured as nitrate nitrogen and ammonium nitrogen ion selective electrodes. The crop canopy stress index is obtained using near-infrared leaf temperature and ambient temperature and humidity optical / thermal integrated sensors. Microclimate parameters include wind speed, air temperature, relative humidity, and solar radiation. Pipeline flow and pressure sensors are respectively located downstream of the variable ratio mixer and the booster pump. Online fertilizer solution concentration is measured using a dual-channel parallel method combining conductivity and refractive index, and cross-calibrated. All sensors are time-synchronized and unit-consistent at the edge computing node with a basic sampling cycle of 1–5 minutes.

[0064] The soil volumetric moisture content sensor and soil water potential sensor are layered and buried at different depths in the root zone to monitor the moisture status of shallow and deep soil layers, respectively. A soil conductivity sensor arranged coaxially in the same borehole can simultaneously acquire soil salinity information from the same vertical profile, avoiding errors caused by spatial heterogeneity. Nutrient concentration sensors use nitrate and ammonium nitrogen ion-selective electrodes to directly detect the concentration of the main nitrogen forms in the soil. The crop canopy stress index quantifies the degree of crop water stress by fusing near-infrared spectroscopy and thermal imaging data with environmental temperature and humidity parameters. Microclimate parameters, including wind speed, air temperature, relative humidity, and solar radiation, provide environmental disturbance inputs for irrigation decisions. Pipeline flow and pressure sensors are installed downstream of the mixer and the booster pump, respectively, to monitor the flow stability after fertilizer solution mixing and the pressure fluctuation after pressurization. Online fertilizer solution concentration is measured using a dual-channel method combining conductivity and refractive index, utilizing the complementary nature of these two physical principles to cross-validate the reliability of the concentration data. All sensor data are timestamped and unit-consistent at the edge computing node, eliminating the impact of temporal deviations and dimensional differences on multi-source data fusion.

[0065] Specifically, when sensors are deployed in pairs within the control zone, soil volumetric moisture sensors are buried in the shallow root zone (10–30 cm) to monitor real-time moisture in the crop's main water-absorbing layer. Soil water potential sensors are buried in the deep root zone (30–60 cm) to reflect the downward infiltration trend of water. Soil conductivity sensors and moisture sensors, arranged coaxially at the same borehole location, form a vertical measurement chain, ensuring spatial consistency of moisture and salinity data at the same location. Nitrate and ammonium nitrogen ion-selective electrodes are directly embedded in the soil, achieving in-situ detection of nitrogen concentration through potential difference measurement, avoiding the lag of traditional sampling methods. Near-infrared leaf temperature sensors capture canopy thermal radiation, and combined with environmental temperature and humidity data, calculate the crop water stress index to characterize the crop's physiological state. Microclimate parameters such as wind speed and air temperature are collected in real-time by meteorological stations, providing environmental input for adjusting sprinkler irrigation parameters. Pipeline flow sensors are located downstream of the variable ratio mixer to monitor the actual flow rate of the mixed fertilizer solution. Pressure sensors are located downstream of the booster pump to provide feedback on the pressure stability of the sprinkler irrigation system. The fertilizer solution concentration is measured in parallel using both conductivity and refractive index methods. Measurement accuracy is improved through joint inversion of conductivity and refractive index, and cross-validation is used to eliminate errors from single methods. All sensors synchronously acquire data at 1-5 minute intervals, and time alignment and unit standardization are performed at edge computing nodes to ensure consistency of multi-source data in both time and space, providing high-precision input for model prediction and control.

[0066] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0067] Sensors were deployed in pairs within each control zone. Soil volumetric moisture content and soil water potential sensors were buried at root zone depths of 20cm and 45cm, respectively, and coaxially arranged with soil conductivity sensors at the same borehole locations. Nutrient concentration sensors were configured with nitrate nitrogen and ammonium nitrogen ion selective electrodes. The crop canopy stress index was obtained using near-infrared leaf temperature and ambient temperature and humidity optical / thermal integrated sensors. Microclimate parameters included wind speed, air temperature, relative humidity, and solar radiation. Pipeline flow and pressure sensors were installed downstream of the variable ratio mixer and the booster pump, respectively. Online fertilizer solution concentration was measured using a dual-channel parallel method combining conductivity and refractive index measurements, with cross-verification performed. All sensors were time-synchronized and unit-consistent at the edge computing node with a 3-minute basic sampling cycle.

[0068] Through the above technical solution, this application achieves comprehensive perception of the water and fertilizer status of sandy soil. By deploying sensors at multiple levels and dimensions, comprehensive information including soil moisture, nutrients, crop physiological status, and environmental parameters is acquired. The use of paired and coaxial sensor deployment improves data reliability and representativeness. By setting reasonable sampling periods and edge computing nodes, the real-time performance and consistency of the data are ensured. This lays a solid data foundation for subsequent model prediction and precise control, effectively solving the problems of insufficient perception dimensions and low data accuracy in traditional methods.

[0069] In some of the schemes mentioned above in this application, when establishing a prediction model for infiltration-storage-leaching of sandy soil, traditional methods lack multi-sub-model coupling and dynamic parameter update mechanisms, making it difficult to accurately simulate the dynamic transport process of water and nutrients in sandy soil. This results in insufficient model prediction accuracy and an inability to effectively constrain the risk of deep seepage.

[0070] This application further proposes an infiltration-storage-leaching prediction model consisting of a Green–Ampt infiltration sub-model, a root zone water balance sub-model, and a one-dimensional convection-diffuse nutrient transport sub-model coupled together. Its parameters are identified online by using short-cycle pulsed irrigation data at the edge computing nodes to identify the measured trajectories of soil volumetric water content and soil water potential, and shallow seepage observations are used to set deep infiltration flux boundary constraints.

[0071] The Green–Ampt infiltration sub-model is used to simulate the infiltration rate of water and the propagation process of the wetting front in sandy soils. The root zone water balance sub-model dynamically tracks the storage and consumption of root zone soil water by calculating input and output terms. The one-dimensional convection-dispersion nutrient transport sub-model combines solute transport equations to predict the concentration distribution and leaching trend of nutrients in the soil. Online parameter identification is driven by short-cycle pulsed irrigation data, and the hydraulic conductivity and matrix potential parameters in the model are dynamically calibrated using the measured trajectories of soil volumetric water content and soil water potential. The deep infiltration flux boundary constraint is set based on shallow seepage observation data, limiting the deep seepage flux predicted by the model to not exceed a preset threshold.

[0072] Specifically, in the edge computing nodes, short-cycle pulsed irrigation data triggers iterative optimization of model parameters. By comparing the measured soil volumetric water content and soil water potential change trajectories with the model's predicted values, the least squares method is used to adjust the infiltration parameters of the Green-Ampt model and the porosity parameters of the water balance model online. The dispersion coefficient of the one-dimensional convection-diffusion model is updated through inversion using measured soil conductivity and nutrient concentration data. Shallow seepage observation data is converted into deep seepage flux boundary conditions, forcing the deep seepage flux output by the model to not exceed this boundary value, thereby suppressing seepage risk in advance during the prediction stage. The three sub-models achieve linked prediction of water infiltration, storage, and nutrient transport through coupled solution. The output of the infiltration sub-model serves as the input of the water balance sub-model, and the water distribution of the water balance sub-model further drives the concentration calculation of the nutrient transport sub-model. The synergistic effect of online parameter identification and boundary constraints enables the model to adapt to the dynamic characteristics of sandy soils while ensuring that the prediction results conform to the physical limitations of actual seepage observations.

[0073] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0074] The infiltration-storage-leaching prediction model is composed of a coupled Green-Ampt infiltration sub-model, a root zone water balance sub-model, and a one-dimensional convection-diffuse nutrient transport sub-model. The Green-Ampt infiltration sub-model describes the vertical infiltration process of water in the soil surface, the root zone water balance sub-model calculates water changes in the root zone, and the one-dimensional convection-diffuse nutrient transport model simulates nutrient migration in the soil profile.

[0075] The model parameters are identified online by edge computing nodes based on the measured trajectories of soil volumetric water content and soil water potential using short-cycle pulsed irrigation data. Specifically, before irrigation begins, a short-duration pulsed irrigation is conducted to collect dynamic change data on soil volumetric water content and soil water potential. Then, the edge computing nodes use this data, employing least squares or other optimization algorithms, to identify and correct key parameters in the model, such as saturated hydraulic conductivity and soil suction.

[0076] Furthermore, shallow seepage observations are used to set boundary constraints for deep seepage flux. Seepage observation devices are installed at shallow locations within the soil profile to monitor infiltration volume in real time. This observational data is used as an upper limit constraint for deep seepage flux in the model to avoid unreasonably large amounts of deep seepage predicted by the model.

[0077] Therefore, by combining short-cycle pulsed irrigation and shallow seepage observation, dynamic correction of model parameters and setting of boundary constraints were achieved, thereby improving the accuracy of the prediction model in describing the water transport characteristics of sandy soils.

[0078] Through the above technical solutions, this application achieves accurate modeling of the infiltration-storage-leaching process in sandy soils. The coupling of the Green-Ampt infiltration sub-model, the root zone water balance sub-model, and the one-dimensional convection-diffuse nutrient transport sub-model comprehensively describes the transport patterns of water and nutrients in sandy soils. Online parameter identification based on short-cycle pulsed irrigation data enables the model to adapt to dynamic changes in soil properties. The introduction of shallow seepage observations provides experimental constraints for deep seepage prediction. This method, which combines multi-model coupling, dynamic parameter identification, and experimental constraints, improves the accuracy of the prediction model in describing the water and fertilizer transport characteristics of sandy soils.

[0079] In some of the solutions mentioned above in this application, a method for establishing a prediction model based on sensor data and executing model predictive control was proposed. However, when fusing multi-source sensor data, due to the spatiotemporal heterogeneity of soil volumetric water content, soil water potential, soil electrical conductivity, nutrient concentration and shallow seepage observation data, directly using single sensor data may lead to deviations in the state estimation of root zone water and fertilizer concentration, which in turn affects the constraints and weight settings of model predictive control and cannot effectively suppress the risk of deep seepage.

[0080] This application further proposes to use extended Kalman filtering to fuse soil volumetric water content, soil water potential, soil electrical conductivity, nutrient concentration, and shallow seepage observations to estimate root zone water and root zone nutrient concentration, and output root zone water deviation and fertilizer concentration deviation. At the same time, a deep seepage risk index is constructed by cross-checking the predicted deep seepage flux and shallow seepage observations, and the deviation and risk index are fed back as constraints and weight update amounts for model prediction control.

[0081] The extended Kalman filter (EPF) establishes state and observation equations, using soil volumetric water content, soil water potential, soil electrical conductivity, and shallow seepage observation data as inputs to estimate root zone water and nutrient concentration in real time. The predicted deep seepage flux is calculated by the infiltration-storage-leaching prediction model and compared with shallow seepage observation data to generate a deep seepage risk index. The feedback loop transforms root zone water deviation, fertilizer concentration deviation, and the risk index into constraints and optimization weight parameters for model prediction and control, enabling dynamic adjustment.

[0082] Specifically, soil volumetric moisture sensors are buried at a depth of 10–30 cm, soil water potential sensors at a depth of 30–60 cm, and soil conductivity sensors are arranged coaxially with these sensors to form stratified observation data. The state equation of the extended Kalman filter is constructed based on the root zone water balance sub-model and the nutrient transport sub-model, and the observation equation is mapped from the sensor measurements to the state space. The predicted deep infiltration flux is calculated using the infiltration-storage-leaching prediction model, and the difference is calculated with the measured values ​​of shallow infiltration. When the difference exceeds a threshold, the risk index is updated. During the feedback process, the root zone moisture deviation is used to adjust the preset range of soil volumetric moisture content or soil water potential predicted and controlled by the model, the fertilizer concentration deviation is used to correct the phased demand target, and the deep infiltration risk index dynamically constrains the infiltration flux threshold. For example, when the risk index increases, the coefficient of the infiltration flux penalty term in the optimization weights increases, forcing the control results to reduce irrigation intensity or fertilizer concentration.

[0083] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0084] An extended Kalman filter (EPF) was employed to fuse observational data on soil volumetric water content, soil water potential, soil electrical conductivity, nutrient concentration, and shallow seepage. The EPF algorithm comprises a prediction step and an update step. In the prediction step, the state estimate and error covariance matrix for the current time step are predicted based on the state equation. In the update step, the Kalman gain is calculated using the observation equation and actual observations, and then the state estimate and error covariance matrix are updated.

[0085] Specifically, state variables include root zone moisture and root zone nutrient concentration. Observational variables include soil volumetric water content, soil water potential, soil electrical conductivity, nutrient concentration, and shallow seepage. The state equation is based on an infiltration-storage-leaching prediction model, while the observation equation describes the relationship between the observational variables and the state variables.

[0086] The optimal estimates of root zone water and nutrient concentrations are obtained using the extended Kalman filter algorithm. Further, the root zone water deviation, i.e., the difference between the estimated value and the target value, is calculated. Similarly, the fertilizer concentration deviation is calculated. These deviation values ​​serve as constraints and weight updates for the model's predictive control, guiding subsequent irrigation and fertilization decisions.

[0087] Therefore, a deep leakage risk index is constructed. Specifically, the predicted deep leakage flux is cross-checked with shallow leakage observation data; a large deviation between the two indicates a higher risk of deep leakage. The deep leakage risk index is also fed back to the model's predictive control module to adjust the control strategy.

[0088] For example, if a large deviation in root zone moisture is detected, the irrigation amount can be increased accordingly. If a large deviation in fertilizer concentration is detected, the fertilizer application rate should be adjusted. When the risk index of deep seepage is high, the amount of irrigation per session should be reduced and the frequency of irrigation increased to reduce the risk of deep seepage.

[0089] Through the above technical solution, this application achieves accurate estimation of the moisture and nutrient status of sandy soils. The fusion of multi-source sensor data improves the accuracy and reliability of the state estimation. By calculating root zone moisture deviation, fertilizer concentration deviation, and deep seepage risk index, it provides important basis for subsequent irrigation and fertilization decisions. This data-driven closed-loop control method can effectively address the special characteristics of sandy soils, achieve precision irrigation and fertilization, improve water and fertilizer use efficiency, and reduce resource waste caused by deep seepage.

[0090] In some of the solutions mentioned above in this application, there are problems such as sudden changes in control quantity or exceeding the physical limits of equipment during the model predictive control process, which may lead to delayed response of the execution unit or mechanical damage. At the same time, the instantaneous application intensity exceeding the soil infiltration capacity may cause the risk of deep leakage. Furthermore, the lack of a penalty mechanism for control increment and terminal deviation can easily lead to the optimization results deviating from the actual needs.

[0091] This application further proposes upper and lower bounds and rate of change constraints for the irrigation cycle, pulse duty cycle, sprinkler pressure, execution unit movement speed, and target fertilizer concentration. It also includes a constraint that the instantaneous application intensity is less than or equal to the infiltration capacity, based on an infiltration-storage-leaching prediction model. Furthermore, it establishes constraints on the feasible region of sprinkler pressure-flow rate-spray width and the upper limit of the target fertilizer concentration. Control increment penalties and terminal deviation penalties are set, and only control quantities representing the current sampling period are issued.

[0092] The system employs several key constraints: upper and lower bounds, which set minimum and maximum values ​​for irrigation cycle, pulse duty cycle, sprinkler pressure, moving speed, and fertilizer concentration to ensure the execution unit operates within physical limits; a rate of change constraint, which limits the maximum variation of each control variable within adjacent cycles to avoid mechanical shock; an instantaneous application intensity constraint, based on the infiltration capacity calculation results of the prediction model, which limits the application rate per unit time to within the soil's absorbable range; and a sprinkler pressure-flow rate-spray width feasible region constraint, which establishes the correspondence between pressure, flow rate, and spray width using calibrated experimental data to ensure spray uniformity. A control increment penalty adds a quadratic term to the rate of change of the control variable in the objective function to suppress frequent fluctuations. A terminal deviation penalty adds a deviation term between the state variable and the setpoint at the end of the objective function, forcing the optimization result to converge towards the desired terminal state.

[0093] Specifically, during the rolling optimization process, the feasible domain of the control quantity is limited by upper and lower bounds and the rate of change constraints to prevent the nozzle pressure from exceeding the rated range or the moving speed from exceeding the driving capacity. Based on the dynamic matching of instantaneous application intensity and infiltration capacity, excessive application is avoided to prevent water accumulation or nutrient loss. The feasible domain constraints of nozzle pressure-flow rate-spray width, combined with calibration mapping data, ensure that the flow rate and spray width simultaneously meet the spray coverage requirements when adjusting the pressure. The control increment penalty uses weighted coefficients to adjust the smoothness of the control quantity changes. For example, when the target fertilizer concentration needs to be increased from 1.5 g / L to 2.0 g / L, the optimization process will be implemented in multiple gradual steps rather than a single abrupt change. The terminal deviation penalty sets terminal weights for root layer moisture content and fertilizer concentration to ensure that the predicted state at the end of the rolling time domain is close to the set value. The optimization result only outputs the control command for the current sampling period. For example, within a 5-minute period, the nozzle pressure is gradually increased from 0.25 MPa to 0.28 MPa, while the fertilizer concentration increases in increments of 0.1 g / L, thus achieving precise and gradual adjustment.

[0094] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0095] When performing model predictive control based on a predictive model within the rolling time domain, the irrigation cycle, pulse duty cycle, sprinkler pressure, actuator movement speed, and target fertilizer concentration are optimized. Specifically, upper and lower bounds and rate of change constraints are first applied to the irrigation cycle, pulse duty cycle, sprinkler pressure, actuator movement speed, and target fertilizer concentration. For example, the irrigation cycle is set to 1-7 days, the pulse duty cycle to 20%-80%, the sprinkler pressure to 200-500 kPa, the actuator movement speed to 0.1-1 m / s, and the target fertilizer concentration to 0.1%-0.5%. The rate of change constraint can be set to ensure that the change between two consecutive controls does not exceed 20%.

[0096] Furthermore, based on the infiltration-storage-leaching prediction model, a constraint is set that the instantaneous application intensity is less than or equal to the infiltration capacity. This can avoid surface water accumulation and runoff loss. Specifically, the instantaneous infiltration capacity under the current soil conditions can be calculated according to the Green-Ampt infiltration model and used as the upper limit constraint for the application intensity.

[0097] In addition, the feasible domain constraints of nozzle pressure-flow rate-spray width and the upper limit constraints of fertilizer solution target concentration must also be considered. For example, the relationship between flow rate and spray width within a given pressure range can be determined based on the nozzle performance curve and used as a constraint. The upper limit of fertilizer solution concentration can be set according to crop tolerance, such as not exceeding 0.5%.

[0098] In the objective function, control increment penalties and terminal deviation penalties are set, and only the control quantity representing the current sampling period is issued. Specifically, the quadratic forms of the control increment and the terminal state deviation can be added to the objective function. The control increment penalty is used to suppress drastic changes in the control quantity, and the terminal deviation penalty is used to ensure that the state at the predicted terminal time meets the requirements. Finally, only the optimized control quantity for the current sampling period is issued and executed to achieve rolling optimization.

[0099] Through the above technical solution, this application enables refined control of the irrigation and fertilization process. By setting multiple constraints, the feasibility and safety of the control results are ensured. The rolling optimization strategy can promptly respond to environmental changes and model errors, improving control accuracy. Simultaneously, by issuing control quantities only for the current sampling period, the computational burden is reduced, and the response speed is improved. This model-prediction-based optimization control method, compared to traditional fixed-strategy control, is better suited to the characteristics of sandy soils, reducing deep seepage and nutrient loss, and improving water and fertilizer use efficiency.

[0100] In some of the schemes mentioned above in this application, the model prediction control in the water-saving fertilization variable control of sandy soil may cause the optimization results to deviate from the actual working conditions due to environmental disturbances and pipeline pressure fluctuations. The deep seepage flux constraint cannot meet the confidence level, and there is no safe degradation mechanism when feasibility degrades, which may lead to over-irrigation or fertilizer waste.

[0101] This application further proposes a model predictive control approach that incorporates short-term predictions of microclimate parameters and a scenario set of pipeline pressure fluctuations for scenario-based robust optimization, satisfying deep seepage flux constraints at a given confidence level. Rolling re-optimization is performed with a sampling period of 1–5 minutes and a warm start is adopted. When feasibility degradation is detected, a safety degradation strategy is implemented: sequentially reducing nozzle pressure, reducing pulse duty cycle, lowering fertilizer solution target concentration, or pausing the process. Furthermore, based on the calibrated pressure-flow-spray width mapping and moving speed-unit application rate mapping, the conversion of control variables into execution unit commands is completed.

[0102] The scenario-based robust optimization process generates short-term predictions of microclimate parameters and a set of pipeline pressure fluctuation scenarios, covering uncertainties such as wind speed, temperature, and pressure fluctuations. It combines these with an infiltration-storage-leakage prediction model to perform parallel simulations across multiple scenarios, selecting optimized solutions that satisfy deep leakage flux constraints. Finally, it selects the optimal combination of control variables at a given confidence level. Rolling re-optimization updates the control sequence within the prediction time domain at 1-5 minute intervals, employing a warm-start mechanism to inherit the optimization results from the previous cycle to accelerate convergence. The safety degradation strategy, based on the real-time detection of constraint violations, executes pressure reduction, duty cycle reduction, concentration reduction, or pause operations in priority order to avoid deep leakage and over-fertilization. Pressure-flow-spray width mapping and moving speed-unit application rate mapping are pre-established through calibration experiments, converting control variables into executable command parameters for the execution unit.

[0103] Specifically, within each sampling period, edge computing nodes generate short-term prediction data of microclimate parameters and a set of pipeline pressure fluctuation scenarios. These are then combined with an infiltration-storage-leaching prediction model for parallel optimization across multiple scenarios, selecting the optimal combination of control variables that meets the deep seepage flux threshold at a 95% confidence level. Rolling optimization is triggered in 1-5 minute cycles, using a warm-start mechanism to use the previous cycle's optimization results as initial guesses, thus shortening the solution time. When soil volumetric moisture content or deep seepage flux exceeds limits, the sprinkler pressure is sequentially reduced to a preset safety value, the pulse duty cycle is reduced to the lower limit, and the fertilizer solution target concentration is lowered to the minimum threshold. If feasibility cannot be restored, irrigation is suspended. Control variables are converted from pressure-flow-spray width mapping obtained through calibration into sprinkler opening commands, and movement speed-unit application rate mapping into drive motor speed commands, ensuring that the execution unit accurately tracks the optimization results.

[0104] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0105] Model predictive control incorporates short-term predictions of microclimate parameters and a scenario set of pipeline pressure fluctuations for scenario-based robust optimization, satisfying deep seepage flux constraints at a given confidence level. Rolling re-optimization is performed with a 3-minute sampling period and a warm start is adopted. When feasibility degradation is detected, a safety degradation strategy is implemented: sequentially reducing nozzle pressure, decreasing pulse duty cycle, lowering fertilizer solution target concentration, or pausing. The conversion from control variables to execution unit commands is completed based on the calibrated pressure-flow-spray width mapping and moving speed-unit application rate mapping.

[0106] Specifically, the system first acquires predicted data on parameters such as temperature, humidity, and wind speed for the next two hours using microclimate sensors. Then, based on historical data, it generates multiple scenarios for pipeline pressure fluctuations, such as normal fluctuations and sudden fluctuations. These predicted data and scenarios are then input into the model predictive controller for robust optimization calculations. The optimization objective is to ensure that deep seepage flux does not exceed a set threshold at a 95% confidence level.

[0107] The optimization calculations are performed in 3-minute cycles. Each calculation uses the optimization result from the previous cycle as the initial value, i.e., a warm-start approach, to accelerate convergence. If an optimization problem is detected as infeasible, a preset safety degradation strategy is implemented: first, the nozzle pressure is reduced by 10%; if that is still infeasible, the pulse duty cycle is reduced by 20%; if that still doesn't work, the target fertilizer concentration is lowered by 30%; finally, the operation is paused.

[0108] Finally, the optimized control quantities such as pressure and flow rate are converted into command parameters that can be directly executed by the actuator through a pre-calibrated pressure-flow-spray width and moving speed-unit application rate mapping relationship. For example, a pressure setpoint of 4 bar is converted into a frequency command for the frequency converter.

[0109] Through the above technical solutions, this application can effectively address environmental uncertainties and equipment fluctuations under sandy soil conditions, ensuring that deep seepage flux can be controlled within a safe range under various circumstances. Simultaneously, the adoption of rolling optimization and warm-start strategies improves computational efficiency. A safety degradation mechanism further enhances reliability. Furthermore, precise conversion from control variables to execution instructions is achieved through mapping relationships, avoiding execution errors. Therefore, this solution enables precise and efficient water-saving fertilization control under sandy soil conditions.

[0110] In some of the solutions described above in this application, when the control results are sent to the execution unit, different control parameters may be out of sync due to transmission delays or differences in the response of the execution unit. This can lead to deviations in the execution of irrigation cycle, pulse duty cycle, nozzle pressure, moving speed, and fertilizer concentration, affecting the accuracy of root zone water and fertilizer status control and increasing the risk of deep leakage.

[0111] This application further proposes a method based on edge computing nodes transmitting control results to the execution unit via fieldbus, and synchronizing the irrigation cycle, pulse duty cycle, sprinkler pressure, execution unit movement speed, and fertilizer solution target concentration using the same sampling period. The pulse duty cycle is implemented using PWM of the zoned solenoid valves; the sprinkler pressure is tracked by the frequency converter of the booster pump and the closed-loop tracking of the pressure regulating valve; the execution unit movement speed is tracked by the speed closed-loop drive module; and the fertilizer solution target concentration is given by the variable ratio fertilizer pump. During execution, inner-loop correction is achieved through feedback from pipeline flow rate, pressure, and online fertilizer solution concentration.

[0112] In this system, edge computing nodes issue control commands via fieldbus protocol, ensuring the real-time and consistent reception of commands by each execution unit. A time synchronization mechanism for the same sampling period is triggered by the clock signal of the edge nodes, aligning irrigation cycle, duty cycle, pressure, speed, and fertilizer concentration parameters on the time axis. The zoned solenoid valves use PWM signals to control the duty cycle, adjusting the pulse width to match the preset duty cycle target. A pressure closed loop is formed between the booster pump frequency converter and the pressure regulating valve, dynamically adjusting the motor speed and valve opening to track the target pressure. The speed closed-loop drive module corrects movement speed deviations in real time through encoder feedback. The variable ratio fertilizer pump adjusts the mixing ratio of the concentrate and water according to the target concentration. Inner loop correction dynamically compensates the output of the execution units using real-time feedback data from flow meters, pressure sensors, and fertilizer concentration sensors.

[0113] Specifically, after completing model predictive control optimization, the edge computing nodes package irrigation cycle, pulse duty cycle, nozzle pressure, moving speed, and target fertilizer concentration parameters into a unified timestamp instruction frame, which is then broadcast to each execution unit via the fieldbus. Upon receiving the instruction, the execution unit synchronously initiates parameter adjustments at the start of the next sampling cycle based on the clock signal. The zoned solenoid valve receives the PWM signal and achieves precise start and stop of pulse irrigation through duty cycle adjustment. The booster pump frequency converter adjusts the motor speed based on the deviation between the pressure setpoint and the feedback value, while the pressure regulating valve makes fine adjustments to eliminate pressure fluctuations. The speed closed-loop drive module adjusts the output torque of the drive motor based on the difference between the encoder pulse count and the target speed. The variable ratio fertilizer pump controls the injection volume of the concentrate according to the concentration setpoint and achieves mixing ratio adjustment through a Venturi proportionalizer. During execution, the flow meter monitors the pipeline flow in real time, the pressure sensor detects the actual nozzle pressure, and the fertilizer concentration sensor measures the mixed solution concentration online. The feedback data from these three sources is sent to the inner loop controller to dynamically compensate the output of the execution unit, thereby eliminating execution errors. For example, when the fertilizer solution concentration feedback value is lower than the target value, the inner loop controller increases the speed of the concentrate pump to improve the mixing ratio. When the nozzle pressure drops due to increased pipeline resistance, the frequency converter increases the speed of the booster pump to maintain the target pressure. Through the above synchronization mechanism and closed-loop control, the output parameters of each execution unit remain coordinated in time and space, ensuring that the intensity of water and fertilizer application matches the soil infiltration capacity and avoiding local over-irrigation or nutrient concentration deviations caused by asynchronous execution.

[0114] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0115] The control results are transmitted from the edge computing node to the execution unit via fieldbus, and the irrigation cycle, pulse duty cycle, nozzle pressure, execution unit movement speed, and fertilizer solution target concentration are synchronized in time according to the same sampling period. The edge computing node can be an industrial-grade embedded computer, communicating with the execution unit via CAN bus or Modbus RTU protocol. The sampling period can be set to 2 minutes to balance control accuracy and load.

[0116] The pulse duty cycle is achieved using the PWM of the zoned solenoid valve. For example, a 24V DC solenoid valve can be used, with the PWM frequency set to 10Hz and the duty cycle range of 20%-80%. The nozzle pressure is tracked by a frequency converter in the booster pump and a closed-loop pressure regulating valve. Specifically, a frequency converter can be used to control the pump speed, while a proportional pressure regulating valve is used to control the pressure within the range of 100-400kPa. The movement speed of the actuator is tracked by a speed closed-loop drive module, which can be driven by a servo motor with a speed range of 0.1-2m / min. The target concentration of the fertilizer solution is given by a variable ratio fertilizer pump, which can be continuously adjusted within the range of 0-2000mg / L using a peristaltic pump.

[0117] During execution, inner-loop correction is achieved through feedback from pipeline flow rate, pressure, and online fertilizer solution concentration. Flow rate is measured using an electromagnetic flowmeter, pressure is detected using a pressure transmitter, and fertilizer solution concentration is monitored online using a conductivity sensor. When the deviation between the actual value and the target value exceeds 5%, the corresponding actuator parameters will be automatically adjusted to ensure control accuracy.

[0118] Through the above technical solutions, this application achieves precise control and real-time correction of the execution unit. The distributed architecture based on edge computing nodes improves response speed and reduces communication latency. Using PWM to control the solenoid valve enables continuous adjustment of irrigation water volume, which is more precise than traditional on / off control. The closed-loop control strategy effectively suppresses external disturbances, ensuring consistency between the actual output and the set value. The inner-loop correction mechanism allows the control results to adapt to complex and changing field environments. This refined control method improves water and fertilizer use efficiency and reduces deep seepage and nutrient loss.

[0119] In some of the solutions described above in this application, when driving the variable ratio fertilizer pump, sprinkler / drip irrigation device and moving mechanism to perform water-saving fertilizer application operations, there are risks of sudden wind speed changes causing spray width deviation, insufficient fertilizer concentration tracking accuracy, and cross-contamination caused by pipeline residues in sprinkler irrigation mode.

[0120] This application further proposes a variable-ratio fertilizer application channel formed by a variable-ratio fertilizer pump and a Venturi proportioner to track the target concentration of the fertilizer solution. The sprinkler / drip irrigation system performs water-saving fertilization according to the irrigation cycle and pulse duty cycle, and automatically shortens the spray width or switches to drip irrigation when the wind speed exceeds a threshold. At the end of execution in each control zone, a flushing sequence and check / anti-siphon interlock are triggered, and a safety degradation process is sequentially executed upon detecting pressure or concentration exceeding limits: reducing sprinkler pressure—reducing the pulse duty cycle—lowering the target fertilizer solution concentration—pausing.

[0121] The variable-ratio fertilizer pump and Venturi proportioner achieve dynamic mixing of fertilizer stock solution and irrigation water through parallel channels. The mixing ratio is jointly adjusted by the opening of the solenoid valve and the pump speed, for example, using a concentration adjustment range of 0.5% to 5.0%. The wind speed response mechanism of the sprinkler irrigation system is realized through a pressure-spray width mapping table. When the wind speed reaches 5 m / s, the sprinkler head pressure is reduced by a proportional coefficient of 0.8 to 0.9. Drip irrigation switching is completed within 10 seconds via a solenoid three-way valve. The flushing sequence includes flushing with clean water of 3 to 5 times the pipeline volume, and a check valve and anti-siphon device are linked to prevent fertilizer backflow.

[0122] Specifically, the variable-ratio fertilizer channel draws in the mother liquor through the Venturi negative pressure effect. The mixing ratio is controlled in a closed loop by the stroke frequency of the variable-ratio fertilizer pump and the Venturi inlet pressure, achieving a concentration tracking accuracy of ±5%. When the sprinkler irrigation system detects excessive wind speed, it reduces the nozzle working pressure to 85% of its original value according to the pre-calibrated pressure-spray width curve, simultaneously adjusting the upper limit of the moving speed to 0.8 m / s to ensure a constant application rate per unit area. When the wind speed continuously exceeds 8 m / s for 30 seconds, the electromagnetic three-way valve switches to the drip irrigation branch, and the dripper array operates at a pressure of 0.1 MPa. After each control zone's operation is completed, the flushing solenoid valve opens for 120 seconds to flush the pipeline of residual fertilizer solution at a flow rate of 4 L / s. The check valve closes immediately after flushing to prevent siphoning. When the pressure sensor detects that the pipeline pressure exceeds 1.2 MPa or the fertilizer concentration deviation exceeds 15%, the control unit will sequentially reduce the nozzle pressure by 0.1 MPa, reduce the duty cycle by 20%, and lower the target concentration to the safe threshold. If the abnormality continues, the unit will enter a shutdown state.

[0123] As a preferred embodiment, the solution of this application is implemented as follows: In an irrigation system deployed in a sandy soil area, a variable ratio fertilizer pump and a Venturi proportioner form an independent channel through parallel pipelines. The variable ratio fertilizer pump adjusts the mixing ratio of mother liquor and water in real time according to the target concentration of fertilizer solution. The sprinkler irrigation device uses a rotary sprinkler head, which performs intermittent sprinkler irrigation according to a preset irrigation cycle and pulse duty cycle. When the wind speed sensor detects an ambient wind speed exceeding 5 m / s, the sprinkler head pressure is automatically reduced from 0.3 MPa to 0.25 MPa, the spray width is reduced from 12 meters to 8 meters, and the pulse duty cycle is adjusted from 50% to 40%. When the wind speed continuously exceeds 8 m / s, the system switches to drip irrigation mode, and the drip irrigation tape operates at a constant pressure. After irrigation of each control zone is completed, the execution unit triggers a 0.5-minute clean water flushing sequence, while simultaneously closing the zone solenoid valve and activating the anti-siphon valve. When the pressure sensor detects that the pipeline pressure exceeds the safety threshold or the fertilizer concentration sensor detects that the concentration deviation exceeds ±10%, the following actions are executed in sequence: reducing the nozzle pressure to the lower limit, reducing the pulse duty cycle to 30%, and adjusting the fertilizer target concentration to the preset lower limit. If the exceedance still cannot be eliminated, irrigation will be suspended and an alarm will be issued.

[0124] Through the above technical solutions, this application achieves adaptive adjustment of sprinkler irrigation parameters under conditions of sudden wind speed changes, effectively avoiding water drift loss caused by excessive spray width. A safety degradation mechanism gradually reduces operational intensity, preventing equipment damage or uneven fertilization caused by abnormal pressure or concentration. Flushing and anti-siphon interlocking operations reduce the risk of residual fertilizer crystallization in pipelines, ensuring the accuracy and operational reliability of water and fertilizer application in sandy soils.

[0125] In some of the solutions described above in this application, when driving the variable ratio fertilizer pump, sprinkler / drip irrigation device and mobile mechanism to perform water-saving fertilizer operation, there are problems such as insufficient fertilizer concentration tracking accuracy, uneven distribution of sprinkler irrigation due to wind speed interference, and the risk of blockage or siphoning caused by residual fertilizer solution in the execution unit.

[0126] This application further proposes a variable-ratio fertilizer application channel formed by a variable-ratio fertilizer pump and a Venturi proportioner to track the target concentration of the fertilizer solution. The sprinkler / drip irrigation system performs water-saving fertilization according to the irrigation cycle and pulse duty cycle, and automatically shortens the spray width or switches to drip irrigation when the wind speed exceeds a threshold. At the end of execution in each control zone, a flushing sequence and check / anti-siphon interlock are triggered, and a safety degradation process is sequentially executed upon detecting pressure or concentration exceeding limits: reducing sprinkler pressure—reducing the pulse duty cycle—lowering the target fertilizer solution concentration—pausing.

[0127] The system incorporates a variable-ratio fertilizer pump and a Venturi proportioner, forming a two-stage adjustment structure through parallel channels. The Venturi proportioner draws in fertilizer mother liquor under negative pressure, while the variable-ratio fertilizer pump compensates for the nonlinear error of the Venturi proportioner by adjusting the plunger stroke. When the wind speed exceeds 5 m / s, the sprinkler system uses a servo motor to shrink the nozzle orifice, reducing the spray width to 60%–80% of its original length. When the wind speed consistently exceeds 8 m / s for more than 30 seconds, it switches to drip irrigation mode. The flushing sequence includes 3–5 pulse flushes, each lasting 10–15 seconds with a pulse interval of 5–8 seconds. A check valve and anti-siphon device are interlocked based on pressure difference to prevent backflow of residual fertilizer solution in the pipeline. The safety degradation strategy features a pressure reduction gradient of 0.1 MPa per flush, a duty cycle reduction step of 10%, and a fertilizer concentration reduction of 20% of the set value.

[0128] Specifically, the two-stage adjustment structure of the variable-ratio fertilizer pump and the Venturi proportioner achieves fertilizer concentration tracking through feedforward-feedback composite control. The Venturi proportioner generates the initial ratio based on the target concentration, while the variable-ratio fertilizer pump fine-tunes the concentration based on the actual concentration deviation measured by conductivity, keeping the concentration error within ±5%. When the wind speed exceeds the limit, the sprinkler irrigation system adjusts the nozzle orifice diameter to change the droplet size distribution, increasing the proportion of large-diameter droplets to over 70% to reduce drift loss. Simultaneously, the nozzle pressure is dynamically reduced according to the wind speed proportionality coefficient of 0.85~0.95. After the execution unit completes zoned irrigation, it triggers a high-pressure water pulse to flush the pipeline and nozzle flow channel. The flushing water flow rate is 1.2~1.5 times the working flow rate, and the flushing pressure is increased to 0.35~0.4 MPa. When the pressure sensor detects that the pipeline pressure exceeds the safety threshold of 0.5 MPa or the fertilizer concentration deviates from the target value by 15%, the control unit sequentially reduces the nozzle pressure to 0.3 MPa, gradually reduces the pulse duty cycle from 80% to 50%, and adjusts the fertilizer concentration to 80% of the original value. If the situation does not improve after three consecutive adjustments, a pause command is triggered.

[0129] As a preferred embodiment, the specific implementation of this application is as follows: During irrigation and fertilization, soil volumetric moisture content sensors, soil water potential sensors, soil electrical conductivity sensors, and shallow seepage observation devices continuously collect data and transmit the data to edge computing nodes. An extended Kalman filter algorithm is used to fuse the collected soil volumetric moisture content, soil water potential, soil electrical conductivity, and shallow seepage observation data to generate an innovation quantity sequence. The infiltration parameters, effective porosity, and dispersion coefficient of the infiltration-storage-leaching prediction model are updated in real time using the recursive least squares method, with the residual sum of squares of the innovation quantity serving as the objective function for parameter updates. The updated model parameters are input into the prediction model to recalculate the state estimates of root zone soil volumetric moisture content and root zone fertilizer concentration, and to generate a deep seepage risk index. When the variance of the observed data exceeds the preset threshold or the absolute value of the innovation output of the extended Kalman filter exceeds the allowable range for three consecutive sampling periods, the edge computing node shortens the sampling period from 5 minutes to 30 seconds. At the same time, it increases the cost weight coefficients of root soil volumetric moisture content deviation, root fertilizer concentration deviation, and deep seepage flux in the optimization objective function of model predictive control, so that the control results prioritize suppressing state deviations with higher risks.

[0130] Through the above technical solution, this application achieves dynamic correction of the water and fertilizer transport model in sandy soils, solving the problem of model prediction bias caused by time-varying soil parameters and environmental disturbances. By updating recursive parameters and adjusting risk weights based on real-time observation data, the risk of deep seepage is effectively reduced, ensuring that irrigation and fertilization control amounts always match the current soil conditions, thus improving the accuracy of water and fertilizer application and seepage prevention capabilities.

[0131] In some of the solutions mentioned above in this application, model predictive control is proposed to optimize and adjust irrigation parameters. However, in actual implementation, fluctuations in ambient wind speed may cause the spray width and application rate of the sprinkler irrigation device to deviate from the preset range, resulting in uneven irrigation or an increased risk of deep seepage. At the same time, the existing solutions do not clearly define the parameter correction mechanism when the wind speed exceeds the limit, making it difficult to ensure the control stability under sandy soil conditions.

[0132] This application further proposes to dynamically adjust the control results based on the calibration results, and to correct the sprinkler irrigation parameters when the ambient wind speed exceeds the set threshold.

[0133] The wind correction mapping mechanism experimentally calibrates the proportional relationship between sprinkler pressure and spray width under different wind speeds, creating a table corresponding to the pressure reduction coefficient and the spray width reduction ratio. After the sprinkler pressure is reduced proportionally, the spray width is simultaneously reduced to maintain a match between the application rate per unit area and the infiltration capacity. Lowering the pulse duty cycle and the upper limit of the execution unit's movement speed reduces the duration and speed of a single pulse irrigation, preventing localized over-irrigation caused by the reduced spray width. When the wind speed continuously exceeds the threshold and triggers the secondary wind speed threshold, the drip irrigation mode is automatically switched to eliminate wind disturbance, and the original sprinkler irrigation settings are restored after the wind speed drops back to the recovery threshold with hysteresis. During the correction process, maintaining the instantaneous application intensity less than or equal to the infiltration capacity and the upper limit of the fertilizer solution target concentration ensures the dynamic balance of water and fertilizer in sandy soils.

[0134] Specifically, when the wind speed sensor detects that the ambient wind speed exceeds the first threshold, a pre-stored wind correction mapping table is invoked. The pressure reduction coefficient corresponding to the current wind speed is applied to the sprinkler pressure closed-loop control module, causing the sprinkler pressure to decrease proportionally. As the pressure decreases, the sprinkler flow rate also decreases. At this point, based on the calibrated pressure-flow-spray width mapping relationship, the sprinkler width is synchronously reduced to ensure the application rate per unit area matches the infiltration capacity of sandy soil. To adapt to the change in coverage caused by the reduced spray width, the pulse duty cycle reduces the proportion of solenoid valve opening time via the PWM module. Simultaneously, the upper limit of the execution unit's movement speed is lowered by the speed closed-loop drive module to avoid repeated irrigation of the same area. If the wind speed continues to rise and exceeds the second threshold, a mode switching command is triggered, shutting down the sprinkler system and starting the drip irrigation pipeline to apply water directly through the drippers to eliminate the influence of wind deviation. When the wind speed drops, a comparator with hysteresis is used to determine the recovery conditions, avoiding frequent switching. Throughout the correction process, the infiltration capacity constraints and fertilizer concentration upper limit constraints of the model prediction control are still enforced to ensure that the root zone moisture content, fertilizer concentration and deep leakage risk are within a controllable range.

[0135] As a preferred embodiment, the solution of this application is implemented as follows: When the wind speed sensor detects that the ambient wind speed reaches the first-level threshold of 5.4 m / s, the control system automatically calls the pre-stored wind correction mapping table, adjusts the nozzle pressure from the initial set value of 0.25 MPa to 0.21 MPa by a proportional coefficient of 0.85, and reduces the spray width from 8 m to 6.8 m. The pulse duty cycle is reduced from 70% to 55%, and the upper limit of the moving speed is reduced from 2.5 m / min to 2.1 m / min. If the wind speed exceeds the second-level threshold of 7.9 m / s for 6 consecutive minutes, the system switches to drip irrigation mode, closes the nozzle, and starts the solenoid valve of the drip irrigation branch. During drip irrigation, the fertilizer concentration is kept constant, and the moving speed is reduced to 1.8 m / min. When the wind speed drops back to 5.0 m / s and stabilizes for 10 minutes, the original irrigation parameters are restored, and the nozzle pressure is gradually increased back to the initial value at a rate gradient of 0.05 MPa / 30 s. During the correction process, the instantaneous application intensity is checked in real time to see if it exceeds the current soil infiltration capacity, and the concentration is ensured to be within the preset upper limit by the fertilizer concentration sensor.

[0136] Through the above technical solution, this application effectively solves the problems of sprinkler irrigation water drift and deep seepage caused by sudden changes in wind speed. By dynamically adjusting the coordinated matching of pressure, spray width, and movement speed, the uniformity of water and fertilizer spatial distribution is ensured while maintaining soil infiltration capacity constraints. The automatic switching mechanism of drip irrigation mode avoids ineffective irrigation under extreme wind speeds, while the hysteresis recovery strategy prevents equipment damage caused by frequent mode switching. This solution achieves stable control of water and fertilizer application in sandy soils under complex wind field environments, reducing the risk of nutrient loss with runoff.

[0137] In the above embodiments, key variables such as soil volumetric moisture content, soil water potential, soil electrical conductivity, nutrient concentration, crop canopy stress index, and microclimate parameters are collected by multi-source sensors. Combined with pipeline flow rate, pressure, and online fertilizer concentration information, an infiltration-storage-leaching prediction model is established and dynamically updated to achieve real-time estimation of root zone moisture and nutrient status. The model also calculates deviations in root zone soil volumetric moisture content, root zone fertilizer concentration, and deep seepage risk. Using model predictive control, irrigation cycle, pulse duty cycle, sprinkler pressure, actuator movement speed, and target fertilizer concentration are continuously optimized in the rolling time domain. This yields accurate control results, which are then sent to the actuator to drive the variable ratio fertilizer pump, sprinkler / drip irrigation device, and moving mechanism to perform water-saving fertilization operations. During execution, sensor data and shallow seepage observations are further integrated to perform closed-loop correction of the prediction model. Sprinkler parameters are automatically corrected when wind speed disturbances exceed the threshold, ensuring stable and reliable control even under uncertain environments. It improves the precision of water conservation and fertilizer control, and can simultaneously inhibit deep seepage, ensure the water and fertilizer balance of the root zone, and cope with external disturbances under sandy soil conditions.

[0138] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A water-saving fertilization variable control method suitable for sandy soils, characterized in that, include: Data is acquired based on sensors, including soil volumetric water content and / or soil water potential, soil electrical conductivity, nutrient concentration, crop canopy stress index, microclimate parameters, pipeline flow rate, pressure, and online fertilizer solution concentration data. Based on the collected data, an infiltration-storage-leaching prediction model for sandy soil was established and updated. The state of soil moisture and nutrients was estimated, and the root zone moisture deviation, fertilizer concentration deviation, and deep leakage risk were calculated. In the rolling time domain, model predictive control is performed based on the prediction model to optimize the irrigation cycle, pulse duty cycle, nozzle pressure, execution unit moving speed and fertilizer solution target concentration, so as to obtain control results that meet the requirements of root soil volumetric water content or soil water potential within a preset range, root nutrient concentration meeting the stage requirements and deep seepage flux below the threshold. The control results are sent to the execution unit to drive the variable ratio fertilizer pump, sprinkler / drip irrigation device and moving mechanism to perform water-saving fertilization operation according to the irrigation cycle, pulse duty cycle, nozzle pressure, execution unit moving speed and fertilizer solution target concentration. During implementation, soil volumetric moisture content, soil water potential, soil electrical conductivity, and shallow seepage observation data are integrated to correct the prediction model. The control results are dynamically adjusted based on the correction results. When the ambient wind speed exceeds the set threshold, the sprinkler irrigation parameters are corrected. When establishing and updating the infiltration-storage-leaching prediction model for sandy soil based on the collected data, the following steps are included: The infiltration-storage-leaching prediction model is composed of a Green-Ampt infiltration sub-model, a root zone water balance sub-model, and a one-dimensional convection-diffuse nutrient transport sub-model coupled together. Its parameters are identified online by the measured trajectory of soil volumetric water content and soil water potential based on short-cycle pulsed irrigation data at the edge computing node, and the deep infiltration flux boundary constraint is set by shallow infiltration observation. When integrating soil volumetric moisture content, soil water potential, soil electrical conductivity, and shallow seepage observation data to correct the prediction model during implementation, the following steps are included: The parameters of the infiltration-storage-leaching prediction model are updated using the innovation of the extended Kalman filter through recursive least squares. These parameters include infiltration parameters, effective porosity, and dispersion coefficient. Based on this, the state estimates of root zone soil volumetric water content and root zone nutrient concentration, as well as the deep leakage risk index, are recalculated. When the observation variance or the absolute value of innovation exceeds the threshold, the sampling period is shortened to 30–60 seconds, and the weights and constraints in the model's predictive control are adjusted simultaneously. This increases the cost weights of root zone soil volumetric water content deviation, root zone nutrient concentration deviation, and deep leakage flux as the risk increases. The control results are dynamically adjusted based on the calibration results. When the ambient wind speed exceeds the set threshold, the sprinkler irrigation parameters are corrected, including: The wind correction mapping proportionally reduces the nozzle pressure and correspondingly reduces the spray width, while simultaneously lowering the pulse duty cycle and the upper limit of the execution unit's movement speed. When the wind speed continuously exceeds the threshold and triggers the secondary wind speed threshold, the drip irrigation mode is automatically switched, and the original sprinkler irrigation settings are restored after the wind speed drops back to the recovery threshold with hysteresis. Throughout the correction process, the constraints such as the instantaneous application intensity being less than or equal to the infiltration capacity and the upper limit of the target fertilizer concentration remain unchanged.

2. The water-saving fertilization variable control method suitable for sandy soils according to claim 1, characterized in that, When acquiring data based on sensors, the following are included: Within each control zone, the sensors are deployed in pairs. Soil volumetric moisture content sensors and soil water potential sensors are buried at root zone depths of 10–30 cm and 30–60 cm, respectively, and are coaxially arranged with soil conductivity sensors at the same borehole location. Nutrient concentration sensors are configured as nitrate nitrogen and ammonium nitrogen ion selective electrodes. The crop canopy stress index is obtained using near-infrared leaf temperature and ambient temperature / humidity optical / thermal integrated sensors. Microclimate parameters include wind speed, air temperature, relative humidity, and solar radiation. Pipeline flow and pressure sensors are respectively located downstream of the variable ratio mixer and downstream of the booster pump. The online fertilizer solution concentration is measured in parallel using both conductivity and refractive index methods, and cross-checked. All sensors are time-synchronized and unit-consistent at the edge computing node with a basic sampling cycle of 1–5 minutes.

3. The method for controlling water-saving fertilization variables suitable for sandy soils according to claim 1, characterized in that, When estimating the state of soil moisture and nutrients, the following should be included: Extended Kalman filtering is used to fuse soil volumetric water content, soil water potential, soil electrical conductivity, nutrient concentration, and shallow seepage observations to estimate root zone water and root zone nutrient concentration, and output root zone water deviation and nutrient concentration deviation. At the same time, a deep seepage risk index is constructed by cross-checking the predicted deep seepage flux and shallow seepage observations, and the deviation and risk index are fed back as constraints and weight update amounts for model predictive control.

4. The water-saving fertilization variable control method suitable for sandy soils according to claim 1, characterized in that, When performing model predictive control based on the predicted model within the rolling time domain to optimize the irrigation cycle, pulse duty cycle, sprinkler pressure, execution unit moving speed, and fertilizer solution target concentration, the following are included: Apply constraints on the irrigation cycle, pulse duty cycle, nozzle pressure, execution unit moving speed, and upper and lower limits and rate of change of fertilizer solution target concentration; Based on the infiltration-storage-leakage prediction model, the instantaneous application intensity is less than or equal to the infiltration capacity constraint. Feasible domain constraints of nozzle pressure-flow rate-spray width and upper limit constraints of fertilizer solution target concentration; Set control increment penalty and terminal deviation penalty, and only issue control quantities that represent the current sampling period.

5. The water-saving fertilization variable control method suitable for sandy soils according to claim 4, characterized in that, The model predictive control incorporates short-term predictions of microclimate parameters and pipeline pressure fluctuation scenarios for scenario robust optimization, satisfying deep leakage flux constraints at a given confidence level. Rolling re-optimization is performed with a sampling period of 1 to 5 minutes and a warm start is adopted. When feasibility degradation is detected, a safety degradation strategy is implemented: successively reducing the nozzle pressure, reducing the pulse duty cycle, lowering the target concentration of fertilizer solution, or pausing. Based on the pressure-flow-spray width mapping and the moving speed-unit application rate mapping obtained by calibration, the conversion of control quantity to execution unit instruction is completed.

6. The water-saving fertilization variable control method suitable for sandy soils according to claim 1, characterized in that, When the control results are sent to the execution unit to drive the variable ratio fertilizer pump, sprinkler / drip irrigation device, and moving mechanism to perform water-saving fertilization operations according to the irrigation cycle, pulse duty cycle, nozzle pressure, execution unit moving speed, and target fertilizer concentration, the following are included: The control results are sent to the execution unit via fieldbus based on the edge computing node, and the irrigation cycle, pulse duty cycle, nozzle pressure, execution unit moving speed and fertilizer solution target concentration are synchronized in time according to the same sampling period; The pulse duty cycle is achieved by the PWM of the partition solenoid valve, the nozzle pressure is tracked by the frequency converter of the booster pump and the closed loop of the pressure regulating valve, the movement speed of the execution unit is tracked by the speed closed loop drive module, and the target concentration of fertilizer solution is given by the variable ratio fertilizer pump. During execution, the inner loop correction is achieved by the feedback of pipeline flow rate, pressure and online concentration of fertilizer solution.

7. The method for controlling water-saving fertilization variables suitable for sandy soils according to claim 6, characterized in that, When the variable-ratio fertilizer pump, sprinkler / drip irrigation device, and moving mechanism are driven to perform water-saving fertilization operations according to the irrigation cycle, pulse duty cycle, nozzle pressure, actuator moving speed, and target fertilizer concentration, the driving process includes: A variable-ratio fertilizer pump and a Venturi proportioner form a variable-ratio fertilizer channel to track the target concentration of fertilizer solution; the sprinkler / drip irrigation device performs water-saving fertilization according to the irrigation cycle and pulse duty cycle, and automatically shortens the spray width or switches to drip irrigation when the wind speed exceeds the threshold; at the end of the execution of each control zone, a flushing sequence and check / anti-siphon interlock are triggered, and when pressure or concentration exceeds the limit, a safety degradation is performed in sequence: reducing the nozzle pressure, reducing the pulse duty cycle, lowering the target concentration of fertilizer solution, and pausing.

Images