A fertilization control system and method for liquid fertilizers

Abstract

This invention relates to the field of fertilization technology, specifically to a fertilization control system and method for liquid fertilizers, comprising: an interaction module for inputting crop type and loading corresponding crop knowledge graph; an acquisition module; an instruction generation unit for constructing a liquid fertilizer loss dynamics model and dynamically optimizing it using a deep deterministic policy gradient reinforcement learning algorithm, predicting future meteorological parameters using an LSTM time series model, and generating a fertilization plan based on the optimized liquid fertilizer loss dynamics model, crop knowledge graph, and predicted future meteorological parameters, and generating fertilization control instructions according to the fertilization plan; and a fertilization control module for receiving fertilization equipment control instructions, executing fertilization equipment control instructions, and providing real-time feedback. This solves the problems of low accuracy, serious resource waste, and limited data dimensions, poor model adaptability, and imperfect anomaly response mechanisms in existing agricultural fertilization technologies.

Description

Technical Field

[0001] This invention relates to the field of fertilization technology, specifically to a fertilization control system and method for liquid fertilizers. Background Technology

[0002] Traditional agricultural fertilization methods have long relied on manual experience, resulting in low precision and significant resource waste. On the one hand, fertilization plans are mostly based on fixed cycles or empirical values, failing to dynamically adjust according to real-time crop growth, soil fertility changes, and meteorological conditions. This leads to over- or under-application of fertilizer, affecting crop yield and quality, and easily causing environmental problems such as soil compaction and eutrophication. On the other hand, while existing simple fertilization systems incorporate some sensors to monitor soil moisture or fertilizer concentration, the data dimensions are limited, lacking comprehensive consideration of meteorological factors such as sunlight, wind speed, and rainfall. In particular, they ignore the loss patterns of liquid fertilizers under different environments (such as leaching due to rainfall and volatilization and drift caused by high temperatures and strong winds), making it difficult to achieve truly precise fertilization.

[0003] With the development of intelligent agriculture, some fertilization control systems have attempted to introduce data models to assist decision-making, but significant limitations remain. Most systems use static models or simple algorithms, which cannot adapt to complex and ever-changing field environments. Model parameters require frequent manual calibration, limiting their practicality. At the same time, the equipment operation status monitoring and anomaly response mechanisms are imperfect. When pump set failures, pipeline pressure anomalies, or environmental parameter changes occur, it is difficult to provide timely warnings and handle the situation, which not only affects fertilization effectiveness but may also cause equipment damage or fertilizer waste. Summary of the Invention

[0004] This application provides a fertilization control system and method for liquid fertilizers to solve the problems of low precision, serious resource waste, single data dimension, poor model adaptability, and imperfect abnormal response mechanism in agricultural fertilization in the prior art.

[0005] The first aspect of this application provides a fertilization control system for liquid fertilizer, comprising: an interaction module, an acquisition module, an instruction generation unit, and a fertilization control module; wherein, the interaction module is used to input crop type and load corresponding crop knowledge graph; the acquisition module is used to collect air temperature and humidity, wind speed, photosynthetic photon flux density, and rainfall, and deploy sensor modules to collect soil EC value, NPK content, fertilizer solution concentration, fertilizer solution flow rate, pipeline pressure, and crop growth status data; the instruction generation unit is used to construct a liquid fertilizer loss dynamics model, dynamically optimize it using a deep deterministic strategy gradient reinforcement learning algorithm, predict future meteorological parameters using an LSTM time series model, and generate a fertilization plan based on the optimized liquid fertilizer loss dynamics model, the crop knowledge graph, and the predicted future meteorological parameters, and generate fertilization control instructions according to the fertilization plan, wherein the liquid fertilizer loss dynamics model includes a rainfall correction function and a photosynthetic photon flux density-wind speed joint correction function, used to quantify the fertilizer solution loss rate under different meteorological conditions; the fertilization control module is used to receive the fertilization equipment control instructions, execute the fertilization equipment control instructions, and provide real-time feedback.

[0006] Preferably, the interaction module includes an interactive interface unit and a crop knowledge graph loading unit, wherein the interactive interface unit is used to support users in interactively modifying the generated fertilization plan; the crop knowledge graph loading unit is used to load the corresponding crop knowledge graph according to the information, and the crop knowledge graph includes the nutrient requirements of crops at different growth stages and fertilization response curves.

[0007] Preferably, the acquisition module includes: a field weather station unit, a sensor unit, and a data processing unit, wherein the field weather station unit is used to collect air temperature and humidity, wind speed, photosynthetic photon flux density, and rainfall; the sensor unit is used to collect soil EC value, NPK content, fertilizer solution concentration, fertilizer solution flow rate, pipeline pressure, and crop growth status; and the data processing unit is used to perform noise reduction and standardization processing on the data acquired by the field weather station unit and the sensor unit.

[0008] Preferably, the instruction generation unit includes a loss model unit and an instruction generation engine unit. The loss model unit is used to construct and continuously update the liquid fertilizer loss kinetic model based on the actual loss data after fertilization. The instruction generation engine unit is used to generate a fertilization plan and use a deep deterministic policy gradient algorithm to dynamically optimize the parameters of the liquid fertilizer loss kinetic model with the optimization objective of minimizing the deviation between predicted fertilizer loss and actual loss. The instruction generation engine unit supports users to interactively adjust the optimized fertilization plan and generate fertilization control instructions based on the adjusted fertilization plan.

[0009] Preferably, the instruction generation engine unit further includes: during the fertilization process, if the received real-time meteorological data or equipment feedback data deviates from the predicted value by more than a preset threshold, an instruction regeneration process is triggered to re-optimize the liquid fertilizer loss kinetic model based on real-time data and generate new fertilization control instructions.

[0010] Preferably, the formula for the liquid fertilizer loss kinetic model is:

[0011]

[0012] in, This refers to the amount of solute lost per unit volume. For time, This is the value of the rainfall correction function. These are the correction function values ​​for photosynthetic photon flux density and wind speed. This refers to the concentration of the fertilizer solution. For model error, Photosynthetic photon flux density Wind speed;

[0013] The formula for the rainfall correction function value is as follows:

[0014]

[0015] in, The critical rainfall for runoff generation, This represents the actual rainfall. For nonlinear influence intensity, For the degree of linear sloping, Base correction value;

[0016] The expressions for the photosynthetic photon flux density and wind speed correction functions are as follows:

[0017]

[0018] in, For light compensation point density, The light saturation point density, Critical wind speed is required to promote transpiration. The critical wind speed that dominates the drift. Based on the basic absorption efficiency coefficient The evapotranspiration promotion coefficient is for low wind speeds. This is the drift loss coefficient at medium wind speeds. This serves as the baseline coefficient for high wind speed absorption. This is the high wind speed drift attenuation coefficient. is the specular suppression coefficient, and e is the natural constant.

[0019] Preferably, the fertilization control module includes a fertilization unit and an instruction execution unit, wherein the fertilization unit is used to deploy field fertilization equipment, the field fertilization equipment including a fertilizer storage tank, a water storage tank, a pipeline network, adjustable nozzles, and a frequency converter pump; the instruction execution unit is used to perform fertilization according to the control instructions of the fertilization equipment.

[0020] Preferably, the instruction execution unit further includes an anomaly determination unit, wherein the anomaly determination unit is used to collect pump group operating status and valve opening and closing data in real time, receive fertilizer solution flow rate, pipeline pressure, air temperature and humidity, and soil EC value data from the data acquisition module, compare the collected data with preset thresholds, and determine an anomaly when the preset conditions are met, and trigger an alarm.

[0021] Preferably, the preset conditions include: the pump group vibration amplitude, temperature or current value exceeds the normal operating range of the equipment; the valve fails to complete the opening and closing action within the specified time, or an abnormal opening and closing state occurs during non-instructed periods; the fertilizer solution flow rate is continuously lower than the minimum design flow rate or continuously higher than the maximum design flow rate; the pipeline pressure is continuously higher than the safe pressure range or lower than the minimum pressure to maintain the normal operation of the system; the air temperature and humidity or soil EC value continuously deviates from the suitable range for crop growth or the system operation safety threshold.

[0022] Preferably, the fertilization plan includes a region number, fertilizer type, fertilizer amount, and fertilization time; the fertilization control command includes a region number, fertilizer storage tank number, opening time, closing time, frequency converter pump frequency, and pump group pressure valve control value.

[0023] A second aspect of this application provides a fertilization control method for liquid fertilizer, comprising: acquiring a crop knowledge graph, crop growth status, and growth cycle; determining the crop growth stage based on the crop growth status and growth cycle, determining the NPK amount required for the growth stage based on the crop knowledge graph, and calculating the total amount of NPK fertilizer required by the crop and the daily amount of fertilizer based on the soil NPK content and the concentration of NPK fertilizer; constructing and optimizing a liquid fertilizer loss kinetic model, using an LSTM time series model to predict air temperature and humidity, rainfall, photosynthetic photon flux density, and wind speed for the next 24 hours, and inputting the predicted rainfall, photosynthetic photon flux density, and wind speed into the optimization model. The loss kinetics model of the liquid fertilizer after processing yields the expected solute loss per unit volume. Combined with the predicted air temperature and humidity, the optimal fertilization time window is obtained. Based on the optimal fertilization time window, the corresponding fertilization duration and the expected fertilizer loss are calculated. The actual fertilizer application amount for the day is calculated based on the daily fertilizer application amount and the expected fertilizer loss. The fertilizer liquid flow rate is calculated based on the actual fertilizer application amount for the day and the fertilization duration, and mapped to the frequency of the variable frequency pump. Based on the adjustable nozzle model, the optimal spraying pressure is determined and converted into the pump group pressure valve control value to generate a fertilization plan. Based on the fertilization plan, control instructions for the fertilization equipment are generated, and fertilization is carried out according to the control instructions for the fertilization equipment.

[0024] Therefore, this application has the following beneficial effects:

[0025] This application embodiment acquires multi-dimensional data such as air temperature and humidity, wind speed, soil EC value, and crop growth status through an acquisition module. Combined with nutrient requirements at different growth stages from a crop knowledge graph, it avoids the problems of over- or under-fertilization caused by traditional fertilization relying on fixed cycles or empirical values, significantly improving fertilizer utilization and reducing resource waste and negative environmental impact. In addition to incorporating soil and crop data, it also focuses on integrating meteorological factors such as light, rainfall, and wind speed. A liquid fertilizer loss kinetic model characterizes the loss patterns of fertilizer solution under different environments (such as rainfall leaching and strong wind drift), and a deep deterministic gradient reinforcement learning algorithm is used to dynamically optimize model parameters. This overcomes the limitations of existing systems with single data dimensions and poor adaptability of static models, enabling fertilization plans to accurately match the complex and ever-changing actual conditions in the field. The fertilization control module, through a real-time feedback mechanism combined with the anomaly detection function in the instruction execution unit, can promptly monitor the status of equipment such as pumps, valves, and pipeline pressure, as well as abnormal environmental parameters. This solves the problem of imperfect anomaly response mechanisms in existing systems, preventing fertilization effects from being affected or equipment damage caused by malfunctions, and improving the stability and reliability of system operation. This solves the problems of low precision, serious resource waste, single data dimension, poor model adaptability, and imperfect abnormal response mechanism in existing agricultural fertilization technologies.

[0026] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0027] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0028] Figure 1 This is a schematic diagram of a fertilization control system for liquid fertilizer provided according to an embodiment of this application;

[0029] Figure 2 This is a schematic diagram of a fertilization control system for liquid fertilizer according to an embodiment of this application;

[0030] Figure 3 This is a flowchart of a fertilization control method for liquid fertilizer provided according to an embodiment of this application. Detailed Implementation

[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] The following description, with reference to the accompanying drawings, illustrates a fertilization control system and method for liquid fertilizer according to embodiments of this application. Addressing the serious resource waste problem mentioned in the background art, this application provides a fertilization control system for liquid fertilizer. In this system, a multi-dimensional data acquisition module collects data such as air temperature and humidity, wind speed, soil EC value, and crop growth status. Combined with nutrient requirements at different growth stages from a crop knowledge graph, this avoids the problem of excessive or insufficient fertilizer application caused by traditional fertilization relying on fixed cycles or empirical values. This significantly improves fertilizer utilization and reduces resource waste and negative environmental impacts. In addition to incorporating soil and crop data, it also integrates meteorological factors such as sunlight, rainfall, and wind speed. A liquid fertilizer loss kinetics model characterizes the loss patterns of fertilizer solution under different environments (such as leaching from rainfall and drifting in strong winds). Furthermore, a deep deterministic strategy gradient reinforcement learning algorithm dynamically optimizes the model parameters, overcoming the limitations of existing systems with their single data dimension and poor adaptability of static models. This allows fertilization schemes to accurately match the complex and ever-changing actual conditions in the field. The fertilization control module, through a real-time feedback mechanism combined with the anomaly detection function in the instruction execution unit, can promptly monitor the status of equipment such as pumps, valves, and pipeline pressure, as well as abnormal environmental parameters. This solves the problem of imperfect anomaly response mechanisms in existing systems, preventing fertilization effects from being affected or equipment damage caused by malfunctions, and improving the stability and reliability of system operation. Thus, it addresses the problems of low precision, serious resource waste, and limited data dimensions, poor model adaptability, and imperfect anomaly response mechanisms in existing agricultural fertilization technologies.

[0033] Figure 1 This is a schematic diagram of a fertilization control system for liquid fertilizer provided in an embodiment of this application.

[0034] This application provides a fertilization control system for liquid fertilizers. The fertilization control system 10 for liquid fertilizers includes: an interaction module 100, an acquisition module 200, an instruction generation unit 300, and a fertilization control module 400.

[0035] The interactive module 100 is used to input crop type and load corresponding crop knowledge graph; the acquisition module 200 is used to collect air temperature and humidity, wind speed, photosynthetic photon flux density, and rainfall, and deploy sensor modules to collect soil EC value, NPK content, fertilizer solution concentration, fertilizer solution flow rate, pipeline pressure, and crop growth status data; the instruction generation unit 300 is used to construct a liquid fertilizer loss dynamics model, dynamically optimize it using a deep deterministic strategy gradient reinforcement learning algorithm, predict future meteorological parameters using an LSTM time series model, and generate a fertilization plan based on the optimized liquid fertilizer loss dynamics model, crop knowledge graph, and predicted future meteorological parameters, and generate fertilization control instructions according to the fertilization plan; the fertilization control module 400 is used to receive fertilization equipment control instructions, execute fertilization equipment control instructions, and provide real-time feedback.

[0036] The liquid fertilizer loss kinetics model can include a rainfall correction function and a photosynthetic photon flux density-wind speed joint correction function to quantify the fertilizer loss rate under different meteorological conditions.

[0037] It is understood that in this embodiment, the interaction module provides targeted crop growth requirements for fertilization decisions by loading a crop knowledge graph, the acquisition module builds a comprehensive decision-making basis by collecting meteorological, soil, crop and equipment operation data from multiple dimensions, the instruction generation unit achieves precise adaptation of fertilization schemes by using a dynamically optimized liquid fertilizer loss kinetic model and future weather forecasts, and the fertilization control module ensures efficient and reliable operation of the system through instruction execution and real-time feedback, effectively improving the accuracy of fertilization, solving problems such as resource waste, single data dimension, poor model adaptability and insufficient operational reliability in traditional fertilization, and realizing intelligent and precise application of liquid fertilizer.

[0038] In this embodiment, the interaction module 100 includes an interaction interface unit and a crop knowledge graph loading unit.

[0039] The interactive interface unit supports users in interactively modifying the generated fertilization plan; the crop knowledge graph loading unit loads the corresponding crop knowledge graph based on the information. The crop knowledge graph contains the nutrient requirements of crops at different growth stages and the fertilization response curve.

[0040] It is understood that in this embodiment, the interactive module supports users to interactively modify the generated fertilization plan through the interactive interface unit, which not only retains the efficiency of intelligent decision-making, but also improves the flexibility and practical adaptability of the plan through human intervention; the crop knowledge graph loading unit loads a knowledge graph containing nutrient requirements and fertilization response curves at different growth stages based on crop type, providing a scientific basis for fertilization decisions that fits the biological characteristics of the crop, avoiding the blindness of general plans. The combination of the two enables the fertilization plan to accurately match the crop growth pattern and flexibly cope with the complex actual situation in the field, further improving the pertinence and reliability of fertilization control.

[0041] Specifically, when a user selects corn, A03 zone, and jointing stage, a corn knowledge graph can be loaded. Producing 1 ton of corn kernels requires 15.9 kg of nitrogen, 4.1 kg of phosphorus, and 13.8 kg of potassium. During the jointing stage, nitrogen, phosphorus, and potassium are required at 50%, respectively. The calculated amount of nitrogen, phosphorus, and potassium is 7.95 kg, phosphorus, and potassium. Based on the planting area and average yield, the actual NPK required is calculated. The fertilization response curve of corn during the jointing stage is parabolic. Before reaching the nutrient threshold, fertilization has a positive benefit and can promote corn growth. However, after reaching the nutrient threshold, the benefit drops sharply and quickly becomes a negative benefit.

[0042] In this embodiment of the application, the acquisition module 200 includes: a field weather station unit, a sensor unit, and a data processing unit.

[0043] The field weather station unit is used to collect air temperature and humidity, wind speed, photosynthetic photon flux density, and rainfall; the sensor unit is used to collect soil EC value, NPK content, fertilizer solution concentration, fertilizer solution flow rate, pipeline pressure, and crop growth status; and the data processing unit is used to denoise and standardize the data acquired by the field weather station unit and the sensor unit.

[0044] It is understood that in this embodiment, the acquisition module comprehensively collects meteorological, soil, fertilizer solution, and crop growth data through the field weather station unit and sensor unit, constructing a basic data system for the entire scenario, breaking through the limitation of single data in traditional systems; the data processing unit performs noise reduction and standardization processing on the collected data, effectively improving data quality and providing accurate and consistent input for subsequent model calculation and decision generation, ensuring that the fertilization plan is based on real and reliable field information, and guaranteeing the scientific and accurate nature of the decision from the source.

[0045] Specifically, the collected air temperature data is (30℃, 32℃, 34℃, 20℃, 32℃). After data noise reduction, it can be obtained as (30℃, 31℃, 33℃, 32℃, 32℃), eliminating the outlier value of 20℃. The mean is set to 30℃, and after Z-score normalization, it is obtained as (0, 0.5, 1.5, 1, 1).

[0046] In this embodiment, the instruction generation unit 300 includes a loss model unit and an instruction generation engine unit.

[0047] The loss model unit is used to construct and continuously update the liquid fertilizer loss kinetic model based on the actual loss data after fertilization. The instruction generation engine unit is used to generate fertilization plans and use the deep deterministic strategy gradient algorithm to dynamically optimize the parameters of the liquid fertilizer loss kinetic model with the goal of minimizing the deviation between the predicted fertilizer loss and the actual loss. The instruction generation engine unit supports users to interactively adjust the optimized fertilization plan and generate fertilization control instructions based on the adjusted fertilization plan.

[0048] It is understood that in this embodiment, the loss model unit uses actual loss data after fertilization as the core to construct and continuously update the liquid fertilizer loss dynamics model, providing a loss prediction basis that fits the actual scenario for subsequent scheme optimization, and avoiding prediction deviations caused by a fixed model. The instruction generation engine unit first generates a preliminary fertilization scheme based on the model, and then uses a deep deterministic strategy gradient algorithm to dynamically optimize the model parameters with the goal of minimizing the deviation between predicted fertilizer loss and actual loss, ensuring the scientificity and effectiveness of the scheme in loss control. At the same time, it supports users to interactively adjust the optimized scheme, taking into account both algorithm accuracy and practical operation flexibility. Finally, it generates fertilization control instructions based on the adjusted scheme, which can reduce liquid fertilizer waste and reduce environmental burden, and can also adapt to the personalized needs of different users and actual field conditions, improving the intelligence and practicality of fertilization operations.

[0049] It should be noted that the deep deterministic policy gradient algorithm designs a continuous state space (soil moisture, temperature, real-time fertilizer loss rate, etc.) and an action space (adjustments to parameters such as diffusion coefficient and adsorption rate) to minimize the deviation between prediction and actual loss. It constructs a reward function (including deviation penalty and parameter change regularization term) to achieve dynamic optimization of high-dimensional continuous control problems. The Actor network outputs parameter adjustment actions, the Critic network evaluates the Q value, and the target network performs soft updates (τ=0.001) to improve stability. With the continuous input of actual field loss data (such as leaching and volatilization) and environmental parameters (soil moisture, temperature, etc.), the parameters are iterated in real time to adapt to changes in different soil, crop, and climate conditions, avoiding the limitations of static models.

[0050] In this embodiment of the application, the instruction generation engine unit further includes: during the fertilization process, if the received real-time meteorological data or equipment feedback data deviates from the predicted value by more than a preset threshold, an instruction regeneration process is triggered to re-optimize the liquid fertilizer loss kinetic model based on real-time data and generate new fertilization control instructions.

[0051] The preset threshold can be determined according to the specific circumstances.

[0052] It is understood that in this embodiment of the application, when the deviation between real-time meteorological data (such as sudden rainfall or high temperature) or equipment feedback (such as abnormal flow or drip irrigation blockage) and the predicted value exceeds a preset threshold, the process is immediately initiated: the real-time abnormal data is incorporated into the liquid fertilizer loss kinetics model for optimization, the model parameters are quickly corrected to match the actual working conditions, and then a new fertilization control command is generated. This mechanism breaks the static execution closed loop, can resolve the interference of uncontrollable factors in real time, avoid fertilizer waste or nutrient deficiency, enhance the accuracy of fertilization and the fault tolerance of equipment, ensure crop needs while reducing environmental burden, and provide support for the stable implementation of intelligent fertilization.

[0053] Specifically, a smart agriculture project implemented liquid nitrogen fertilizer drip irrigation during the jointing stage of summer maize. Based on the meteorological forecast data 12 hours before fertilization (predicted as "sunny, no precipitation, wind speed 2-3"), the initial liquid fertilizer loss kinetic model calculated that: fertilization should be carried out continuously for 4 hours at a drip irrigation flow rate of 8L / mu·hour to ensure that the nitrogen fertilizer retention rate in the root zone of the soil reaches more than 85% and meet the nutrient requirements of maize during the jointing stage.

[0054] Two hours into the fertilization process, the meteorological monitoring equipment transmitted real-time data: a sudden, short-duration heavy rainfall (25 mm / hour) and a rapid increase in wind speed to level 6-7. At this point, it was necessary to first determine if the deviation exceeded the preset thresholds—the project's preset threshold of "rainfall deviation ≥ 15 mm / hour, wind speed deviation ≥ level 3" triggers a regeneration process. The current real-time data clearly exceeded these thresholds, therefore, a regeneration command was immediately initiated.

[0055] Model optimization: Real-time data such as "short-term heavy rainfall (which will accelerate the loss of liquid fertilizer with surface runoff) and high wind speed (which will increase the evaporation loss of liquid fertilizer during drip irrigation)" were input into the model, and core parameters such as "rainfall loss coefficient" and "evaporation loss coefficient" were recalibrated. The corrected model predicts that if fertilizer is applied at the original flow rate, the nitrogen fertilizer retention rate in the root zone will drop to below 50%, and fertilizer damage may occur due to soil water accumulation.

[0056] New instructions generated: Based on the optimized model, new fertilization control instructions are generated: ① Immediately suspend drip irrigation fertilization to prevent fertilizer loss with rainwater; ② After the rainfall stops (real-time weather monitoring shows that the rain will stop after 1 hour), adjust the drip irrigation flow rate to 5L / acre·hour and shorten the fertilization time to 1.5 hours (to replenish some lost nutrients and prevent the soil from becoming too wet); ③ Simultaneously activate the soil moisture sensor for real-time monitoring. If the soil moisture content exceeds 80% of the field capacity, further delay fertilization.

[0057] In this embodiment of the application, the formula for the liquid fertilizer loss kinetic model is as follows:

[0058]

[0059] in, This refers to the amount of solute lost per unit volume. For time, This is the value of the rainfall correction function. These are the correction function values ​​for photosynthetic photon flux density and wind speed. This refers to the concentration of the fertilizer solution. For model error, Photosynthetic photon flux density Wind speed;

[0060] The formula for the rainfall correction function value is as follows:

[0061]

[0062] in, The critical rainfall for runoff generation, This represents the actual rainfall. For nonlinear influence intensity, For the degree of linear sloping, Base correction value;

[0063] The expressions for the photosynthetic photon flux density and wind speed correction functions are as follows:

[0064]

[0065] in, For light compensation point density, The light saturation point density, Critical wind speed is required to promote transpiration. The critical wind speed that dominates the drift. Based on the basic absorption efficiency coefficient The evapotranspiration promotion coefficient is for low wind speeds. This is the drift loss coefficient at medium wind speeds. This serves as the baseline coefficient for high wind speed absorption. This is the high wind speed drift attenuation coefficient. is the specular suppression coefficient, and e is the natural constant.

[0066] It should be noted that, It is the lighting in the low-light area. It is the beneficial ratio of low wind speed to fertilizer absorption rate. It is the beneficial ratio of medium wind speed to fertilizer absorption rate. It is the rate of deviation of high wind speed from fertilizer absorption rate. It is the lighting in the highlight area.

[0067] In addition, wind speed is negligible during periods of strong sunlight because plants do not perform photosynthesis at this time, and the absorption rate of fertilizer is negligible.

[0068] Specifically, for example, in greenhouse tomato cultivation, with a fertilizer solution concentration of 200 g / L, zero greenhouse rainfall, a photosynthetic photon flux density of 600 μmol / m²·s, and a wind speed of 1 m / s, L(t) ≈ 53.6 g / L, and a loss rate of approximately 26.8%. The initial values ​​of the coefficients in the photosynthetic photon flux density and wind speed correction functions can be obtained from laboratory experiments. Taking greenhouse tomatoes as an example, a = 0.002, b = 0.01, c = 0.05. =100 μmol / m²·s =800 μmol / m²·s =1.5m / s, =3.0m / s, d=0.85, g=0.3, h=1.2, j=0.6, k=1.5, m=0.002. Assuming that 12kg of nitrogen needs to be added after deducting the nitrogen provided by the soil, and the fertilizer solution concentration is 200g / L, then a total of 600L of nitrogen fertilizer is needed. Assuming that this growth stage lasts for 30 days, then 20L of fertilizer needs to be applied per day. According to weather forecasts and liquid fertilizer loss kinetics models, the optimal fertilization time window is from 8:00 to 10:00 AM, requiring approximately 27.32L of fertilizer. The fertilizer solution flow rate is approximately 0.227L / min. A fertilization plan (Zone 1, 27.32L, 8:00-10:00 AM) can be generated. The user modifies it to (Zone 1, 27.32L, 8:00-9:00 AM), generating the following control instructions for the fertilization equipment: (Zone 1, No. 1 fertilizer storage tank (N fertilizer), 7.58-9.02 (open 2 minutes in advance, close 2 minutes later), 35Hz, 0.3MPa). The device was turned on at 7:58 and turned off at 9:02, with no abnormalities.

[0069] Understandably, this application introduces a liquid fertilizer loss kinetic model during the liquid fertilizer application process. The rainfall correction function fully considers the impact of rainfall on fertilizer loss before and after reaching a critical value. Before reaching the critical value, the soil moisture content is not saturated, and effective runoff has not formed, so the contribution of rainfall to fertilizer loss is negligible. When the rainfall reaches the critical value, surface runoff occurs, and rainwater washes away residual fertilizer solution on the surface, even carrying away surface soil particles. As rainfall increases, fertilizer loss increases rapidly with rainfall intensity. The photosynthetic photon flux density and wind speed correction function considers that light and photon flux density I and wind speed V affect the plant's absorption rate of fertilizer, so the function is divided into four stages: suitable light and low wind zone... When I exceeds the compensation point but is less than the saturation point, and the wind speed V is lower than the critical wind speed for transpiration promotion, the plant's absorption rate of fertilizer mainly comes from basic photosynthesis, with no significant drift loss. This is a suitable area with moderate light and light wind. When I exceeds the compensation point but is less than the saturation point, and the wind speed exceeds the critical wind speed for promoting transpiration but is less than the critical wind speed for dominating drift, the plant can perform photosynthesis. By enhancing leaf transpiration, the fertilizer droplets on the leaf surface adhere more evenly around the stomata, while simultaneously promoting the flow of water from the root zone to the root system (nutrients migrate with the water), indirectly improving absorption efficiency. This is suitable for areas with strong light and wind. When I exceeds the compensation point but is below the saturation point, and the wind speed exceeds the drift-dominant critical wind speed, the high wind speed causes a large number of fertilizer droplets on the leaf surface to drift (deviating from the target leaf). The residual droplets rapidly evaporate and concentrate due to strong transpiration, resulting in a sharp drop in absorption efficiency. (High light area) When I exceeds the saturation point, the thylakoid membrane of chloroplasts is damaged, the PSII reaction center is inactivated, the photosynthetic electron transport chain is blocked, ATP production decreases, and the active absorption efficiency declines exponentially, rendering the effect of wind speed on absorption efficiency essentially ineffective. Dynamic optimization of the model using a deep deterministic policy gradient reinforcement learning algorithm ensures the model's adaptability to different regions and environments.

[0070] In this embodiment, the fertilization control module 400 includes a fertilization unit and an instruction execution unit.

[0071] The fertilization unit is used to deploy field fertilization equipment, which includes fertilizer storage tanks, water storage tanks, pipeline networks, adjustable nozzles, and variable frequency pumps; the instruction execution unit is used to fertilize according to the control instructions of the fertilization equipment.

[0072] It is understood that in this embodiment, the fertilization unit constructs a complete field fertilization execution system by deploying equipment such as fertilizer storage tanks, variable frequency pumps, and adjustable nozzles, providing hardware support for precision fertilization. The instruction execution unit drives the equipment to operate according to control instructions, realizing precise control of parameters such as fertilizer amount, spraying pressure, and range. It transforms the decision-making scheme of the instruction generation unit into actual operation, which not only ensures the automation and standardization of the fertilization process, but also adapts to the needs of different crops and different growth stages through adjustable equipment, thereby improving the practicality of the system and the stability of the fertilization effect.

[0073] In this embodiment of the application, the instruction execution unit 400 further includes an exception determination unit.

[0074] The anomaly detection unit is used to collect pump group operating status and valve opening and closing data in real time, receive fertilizer solution flow rate, pipeline pressure, air temperature and humidity, and soil EC value data from the data acquisition module, compare the collected data with preset thresholds, and determine an anomaly when the preset conditions are met, and trigger an alarm.

[0075] It is understood that the anomaly determination unit in the instruction execution unit of this application embodiment collects real-time operating data of equipment such as pump sets and valves, as well as parameters such as fertilizer flow rate, pipeline pressure, and ambient temperature and humidity, and compares them with preset thresholds to achieve anomaly monitoring and alarm. This can not only detect equipment failures (such as pump set abnormalities or valve malfunctions) and environmental parameter abnormalities (such as pressure exceeding limits or unsuitable temperature and humidity) in a timely manner, avoiding equipment damage or fertilization interruption caused by the expansion of the fault, but also ensure the safety and continuity of the fertilization process through early warning, making up for the shortcomings of the traditional system's delayed anomaly response.

[0076] In this embodiment, the preset conditions include: the pump vibration amplitude, temperature or current value exceeding the normal operating range of the equipment; the valve failing to complete the opening and closing action within the specified time, or exhibiting abnormal opening and closing status during non-instructed periods; the fertilizer solution flow rate continuously being lower than the minimum design flow rate or continuously being higher than the maximum design flow rate; the pipeline pressure continuously exceeding the safe pressure range or falling below the minimum pressure required to maintain normal system operation; and the air temperature and humidity or soil EC value continuously deviating from the suitable range for crop growth or the system's safe operating threshold.

[0077] Specifically, the pump unit operation status monitoring involves real-time acquisition of vibration sensor data, temperature sensor data, and current transformer data. When the vibration amplitude exceeds the stable operation threshold set by the equipment manufacturer (such as abnormal vibration caused by impeller wear or bearing failure), the casing temperature exceeds the upper limit of the motor's allowable operating temperature (such as the risk of overload caused by continuous high temperature), or the operating current deviates from the rated current by ±20% (such as a sudden increase in current caused by pipe blockage or a sudden drop in current caused by impeller idling), the pump unit is immediately identified as abnormal.

[0078] Valve status monitoring: The valve opening and closing status is obtained through stroke sensors or position feedback modules. When a control command is received, if the valve fails to complete its action within its designed full stroke time (such as 1-3 seconds for solenoid valves and 5-15 seconds for electric valves depending on the diameter), or if it opens or closes unexpectedly without a control command (such as malfunction caused by valve stem jamming), and this state continues for more than 5 sampling cycles (about 10 seconds), the valve is judged to be abnormal.

[0079] Fertilizer solution flow rate and pipeline pressure monitoring: Based on real-time flow rate data acquisition using an ultrasonic flow meter, if the flow rate remains below the minimum irrigation flow rate designed by the system (such as the lower limit calculated based on the water requirement of a single crop plant) or above the maximum safe flow rate (such as the upper limit to avoid pipeline overload) for 3 consecutive minutes, it is determined to be an abnormal flow rate. At the same time, the pipeline pressure is monitored by a pressure transmitter. If the pressure remains above 1.1 times the rated pressure of the pipeline for 1 minute (such as 0.6 MPa for PVC pipelines, exceeding the limit may cause pipe burst) or below 0.1 MPa (such as a sudden pressure drop due to pipeline leakage), it is determined to be an abnormal pressure.

[0080] Environmental and soil parameter monitoring: For air temperature and humidity, when the temperature is continuously higher than the critical high temperature for crop growth (such as 35℃ for vegetables) or lower than 5℃ for 1 hour (which may cause fertilizer solution crystallization), and the relative humidity is continuously higher than 90% for 2 hours (which can easily lead to equipment condensation and short circuits) or lower than 15% (extreme drought affects fertilizer solution absorption), it is judged as an abnormal temperature and humidity. For soil EC value, when it is continuously higher than the upper limit of crop salt tolerance (such as about 2.5mS / cm for strawberries) or lower than the basic soil fertility threshold (such as below 0.5mS / cm reflecting soil infertility) for 1 hour, it is judged as an abnormal EC value.

[0081] It is understood that the embodiments of this application, through explicit parameter association and time threshold settings, not only avoid false alarms caused by instantaneous fluctuations, but also respond promptly in the early stages of faults or abnormal trends, ensuring that the system forms an effective protection between equipment safety, fertilization effect and crop growth environment, and further enhancing the reliability of the entire fertilization control system.

[0082] In this embodiment, the fertilization plan includes a region number, fertilizer type, fertilizer amount, and fertilization time; the fertilization control command includes a region number, fertilizer storage tank number, opening time, closing time, frequency converter pump frequency, and pump group pressure valve control value.

[0083] Specifically, the fertilizer storage tanks include nitrogen fertilizer storage tanks, phosphate fertilizer storage tanks, potash fertilizer storage tanks, and organic fertilizer storage tanks. Each time fertilizer is applied, the pipeline should be operated independently. Before and after the application, the pipeline should be rinsed with clean water from the storage tank to avoid cross-contamination. The fertilizer storage tanks should be numbered 1, 2, and 3, and the corresponding valves should be opened when applying the corresponding fertilizer.

[0084] This application proposes a fertilization control system for liquid fertilizers. By acquiring multi-dimensional data such as air temperature and humidity, wind speed, soil EC value, and crop growth status through an acquisition module, and combining this data with nutrient requirements at different growth stages from a crop knowledge graph, it avoids the problems of over- or under-fertilization caused by traditional fertilization relying on fixed cycles or empirical values. This significantly improves fertilizer utilization, reduces resource waste, and minimizes negative environmental impacts. In addition to incorporating soil and crop data, it also integrates meteorological factors such as sunlight, rainfall, and wind speed. A liquid fertilizer loss kinetics model characterizes the loss patterns of fertilizer solution under different environments (such as leaching from rainfall and drifting in strong winds). Furthermore, a deep deterministic strategy gradient reinforcement learning algorithm dynamically optimizes the model parameters, overcoming the limitations of existing systems with their single data dimension and poor adaptability of static models. This allows fertilization schemes to accurately match the complex and ever-changing actual conditions in the field. The fertilization control module, through a real-time feedback mechanism combined with the anomaly detection function in the instruction execution unit, can promptly monitor the status of equipment such as pumps, valves, and pipeline pressure, as well as abnormal environmental parameters. This solves the problem of imperfect anomaly response mechanisms in existing systems, preventing fertilization effects from being affected or equipment damage caused by malfunctions, and improving the stability and reliability of system operation. Thus, it addresses the problems of low precision, serious resource waste, and limited data dimensions, poor model adaptability, and imperfect anomaly response mechanisms in existing agricultural fertilization technologies.

[0085] The following will illustrate a fertilization control system for liquid fertilizers through a specific embodiment, taking the flowering and fruit-setting period of greenhouse tomatoes as an example. Figure 2 As shown, it includes:

[0086] In the interactive interface unit of the interactive module, the user inputs the following information via the touchscreen: Crop type: tomato, planting area: E-002, growth cycle: flowering and fruit setting period.

[0087] The crop knowledge graph loading unit loads the wheat seedling stage knowledge graph and obtains key parameters: nutrient requirements: N:P:K=3:1:4. Before the nutrient requirements reach the critical value, the fertilization response curve is nearly parabolic, and after exceeding the critical value, it drops rapidly.

[0088] The data acquisition module uses a field weather station unit to collect data in real time: air temperature: 25℃, relative humidity: 62%, wind speed: 1.8m / s, photosynthetic photon flux density: 800μmol / m²・s, rainfall: 0mm. The sensor unit collects data in real time: soil EC value: 2.5mS / cm, soil NPK content: N=150mg / kg, P=50mg / kg, K=200mg / kg, fertilizer solution concentration: N=150mg / L, P=50mg / L, K=100mg / L, pipeline pressure: 0MPa, and crop growth status: average fruit diameter is about 5cm. The data processing module performs noise reduction and standardization on the data.

[0089] The loss model unit in the instruction generation module constructs a liquid fertilizer loss kinetic model.

[0090]

[0091] Laboratory experimental parameters: critical rainfall for soil runoff generation 15mm, light compensation point density. =100 μmol / m²·s, light saturation point density =800 μmol / m²·s, critical wind speed for transpiration promotion =1.5m / s, the critical wind speed for drift dominance =3.0m / s, fertilizer concentration is 200mg / L, photosynthetic photon flux density I=600μmol / m²・s, wind speed V=1m / s, a=0.003, b=0.02, c=0.06, d=0.89, g=0.2, h=1.1, j=0.7, k=1.5, m=0.002;

[0092] After optimization using a deep deterministic policy gradient reinforcement learning algorithm, a=0.002, b=0.01, c=0.05. =100 μmol / m²·s =800 μmol / m²·s =1.5m / s, =3.0m / s, d=0.85, g=0.3, h=1.2, j=0.6, k=1.5, m=0.002.

[0093] The instruction generation engine unit generates a fertilization plan through calculation: Taking P as an example, based on the crop knowledge graph, during the flowering and fruit-setting period, the region requires a total of 5.8 kg of P, with a fertilizer concentration of 200 mg / L. The flowering and fruit-setting period lasts for 30 days, with a daily fertilizer application of approximately 9.74 L. Using an LSTM time series model, hourly air temperature and humidity, rainfall, light and photon flux density, and wind speed for the next 24 hours are obtained. These data are then substituted into a liquid fertilizer loss kinetics model, and combined with air temperature and humidity, the optimal fertilization time window is determined to be from 8:00 AM to 12:00 PM. The calculated fertilizer loss is 12.48L, and the fertilizer solution flow rate is 0.05L / min. A fertilizer application plan is generated: E-002, P fertilizer, 12.48L, 8:00 to 12:00. The user modifies the plan between 8:00 and 10:00, changing it to: E-002, P fertilizer, 11.19L, 8:00 to 10:00. The fertilizer application equipment control command is generated: E-002, 003 (P fertilizer tank), 11.19L, 7.58 open, 10.02 close, 30Hz, 0.3MPa.

[0094] In the fertilizer equipment control module, the instruction execution unit obtains and executes the fertilizer equipment control instructions. At 7:58, the fertilization is started and the fertilization process is monitored. No abnormalities occur during the process. At 10:02, the process is shut down normally.

[0095] After fertilization was completed, based on real-time weather conditions, the actual fertilizer loss was 0.19 kg, the expected fertilizer loss was 0.29 kg, and the calculated difference was -0.1 kg. The daily fertilizer application rate was reduced from 9.74 L to 9.72 L.

[0096] In summary, this application's embodiments obtain nutrient requirements and proportions for different growth stages of crops by acquiring crop knowledge graphs, providing a basis for calculating the required fertilizer amount; deploy field weather stations and sensors to collect data, and perform noise reduction and standardization on the data. Noise reduction increases data reliability, and standardization facilitates system recognition; construct and optimize a liquid fertilizer loss kinetic model, generate a fertilization plan based on predicted future meteorological parameters and the optimized liquid fertilizer loss kinetic model and crop knowledge graph, transmit the plan to the interactive interface for users to modify parameters, and finally determine the plan and generate control instructions for the fertilization equipment. This can improve the accuracy of fertilization, reduce fertilizer waste and pollution, and avoid yield reduction due to insufficient fertilizer. Thus, it solves the problems of lacking consideration for complex weather conditions in the field and lacking in-depth modeling of fertilizer loss.

[0097] Next, referring to the accompanying drawings, a fertilization control method for liquid fertilizer is described according to an embodiment of this application.

[0098] like Figure 3 As shown, the fertilization control method for liquid fertilizers includes the following steps:

[0099] In step S101, the crop knowledge graph, crop growth status, and generation cycle are obtained.

[0100] It is understood that by acquiring crop knowledge graphs, crop growth status, and growth cycle, the embodiments of this application can accurately anchor the nutrient requirements and growth characteristics of crops at different stages, providing a scientific basis for fertilization decisions based on crop biological characteristics.

[0101] In step S102, the crop growth stage is obtained based on the crop growth status and growth cycle, the NPK amount required for the growth stage is obtained based on the crop knowledge graph, and the total amount of NPK fertilizer required by the crop and the concentration of NPK fertilizer are calculated in combination with the soil NPK content and NPK fertilizer concentration.

[0102] It is understood that the embodiments of this application determine the growth stage by combining the crop growth status and cycle, obtain the required NPK amount for the stage based on the crop knowledge graph, and then calculate the total fertilizer application amount and daily fertilizer application amount by associating the existing NPK content of the soil and fertilizer concentration. This not only accurately makes up for the gap between soil nutrients and crop needs, avoiding nutrient excess or deficiency caused by blind fertilization, but also achieves on-demand supply by refining the daily fertilizer application amount, making the fertilization rhythm highly matched with the crop growth rhythm, and effectively improving fertilizer utilization.

[0103] For example, if tomatoes are detected to be in the fruit expansion stage (judged based on fruit diameter and growth cycle), the crop knowledge graph indicates that this stage requires 1.2 kg / mu of nitrogen, 0.8 kg / mu of phosphorus, and 1.5 kg / mu of potassium. Sensor detection shows that the current soil nitrogen, phosphorus, and potassium contents are 0.5 kg / mu, 0.6 kg / mu, and 0.7 kg / mu, respectively, meaning that an additional 0.7 kg / mu of nitrogen, 0.2 kg / mu of phosphorus, and 0.8 kg / mu of potassium is needed. Combined with... The NPK concentration in the applied liquid fertilizer (e.g., 100g / L nitrogen, 80g / L phosphorus, and 120g / L potassium) can be used to calculate the total fertilizer requirement (7L / mu for nitrogen, 2.5L / mu for phosphorus, and 6.7L / mu for potassium). This requirement is then spread over 20 days during the fruit expansion period to obtain the daily fertilizer application (0.35L / mu for nitrogen, 0.125L / mu for phosphorus, and 0.335L / mu for potassium). This approach precisely meets the nutrient needs of tomatoes during the fruit expansion period while avoiding fertilizer waste.

[0104] In step S103, a liquid fertilizer loss kinetic model is constructed and optimized. An LSTM time series model is used to predict the air temperature and humidity, rainfall, photosynthetic photon flux density, and wind speed for the next 24 hours. The predicted rainfall, photosynthetic photon flux density, and wind speed are input into the optimized liquid fertilizer loss kinetic model to obtain the expected solute loss per unit volume. Combined with the predicted air temperature and humidity, the optimal fertilization time window is obtained, and the corresponding fertilization duration and expected fertilizer loss are calculated based on the optimal fertilization time window.

[0105] It is understood that the embodiments of this application construct and optimize the liquid fertilizer loss kinetic model, combine it with the LSTM time series model to predict meteorological parameters for the next 24 hours, incorporate key meteorological factors into the fertilizer liquid loss calculation, and accurately obtain the expected solute loss per unit volume; at the same time, combine the air temperature and humidity prediction to determine the optimal fertilization time window, and calculate the fertilization duration and expected fertilizer loss accordingly. This not only achieves scientific prediction of fertilizer liquid loss under different meteorological conditions, but also minimizes losses by selecting appropriate time periods for fertilization, effectively making up for the fertilizer waste problem caused by neglecting environmental impact in traditional fertilization, and further improving the accuracy of fertilization plans and fertilizer utilization rate.

[0106] It should be noted that the LSTM time series model predicts meteorological parameters for the next 24 hours by constructing a neural network model containing an input layer, a hidden layer (containing LSTM units), and an output layer. The core formula involved in the prediction process can be summarized as follows: For the input historical meteorological parameter sequence (such as time step data such as air temperature and humidity, and rainfall), information is filtered and updated through the forget gate, input gate, and output gate in the LSTM unit. The forget gate uses the sigmoid function to decide which historical state information to discard. The input gate updates the cell state by combining the sigmoid function and the tanh function. The output gate outputs the current hidden state based on the sigmoid function and the tanh function. Finally, the predicted values ​​of each meteorological parameter for the next 24 hours are obtained by mapping through the output layer. This provides a reliable input of future environmental parameters for the liquid fertilizer loss kinetics model and improves the accuracy of fertilizer loss prediction.

[0107] In step S104, the actual amount of fertilizer to be applied on the day is calculated based on the daily amount of fertilizer applied and the expected amount of fertilizer loss. The fertilizer liquid flow rate is calculated based on the actual amount of fertilizer to be applied on the day and the fertilization time, and mapped to the frequency of the variable frequency pump. The optimal spraying pressure is determined based on the adjustable nozzle model and converted into the pump group pressure valve control value to generate a fertilization plan. The fertilization equipment control command is generated based on the fertilization plan, and fertilization is carried out according to the fertilization equipment control command.

[0108] It is understood that the embodiments of this application combine the daily fertilizer application amount with the expected fertilizer loss to calculate the actual fertilizer application amount for the day, which can ensure that sufficient nutrients are provided to crops while taking into account fertilizer loss, and avoid insufficient fertilizer supply due to loss. Secondly, the fertilizer flow rate is calculated based on the actual fertilizer application amount and fertilization duration and mapped to the frequency of the variable frequency pump, realizing a precise correspondence between the fertilizer application amount and the equipment operating parameters, ensuring accurate execution of the fertilizer application amount. Furthermore, the optimal spraying pressure is determined based on the adjustable nozzle model and converted into the pump group pressure valve control value, taking into account both fertilization effect and equipment safety, and finally generating and executing the fertilization plan and control instructions.

[0109] According to the embodiments of this application, a fertilization control method for liquid fertilizer is proposed. By acquiring multi-dimensional data such as air temperature and humidity, wind speed, soil EC value, and crop growth status through an acquisition module, and combining the nutrient requirements of different growth stages in the crop knowledge graph, it avoids the problem of excessive or insufficient fertilizer caused by traditional fertilization relying on fixed cycles or experience values. This significantly improves fertilizer utilization and reduces resource waste and negative environmental impact. In addition to incorporating soil and crop data, it also focuses on integrating meteorological factors such as light, rainfall, and wind speed. The liquid fertilizer loss kinetics model is used to characterize the loss pattern of fertilizer liquid under different environments (such as leaching by rainfall and drift by strong winds). The deep deterministic strategy gradient reinforcement learning algorithm is used to dynamically optimize the model parameters, which solves the limitations of existing systems with single data dimensions and poor adaptability of static models. This allows the fertilization plan to accurately match the complex and ever-changing actual conditions in the field. The fertilization control module, through a real-time feedback mechanism combined with the anomaly detection function in the instruction execution unit, can promptly monitor the status of equipment such as pumps, valves, and pipeline pressure, as well as abnormal environmental parameters. This solves the problem of imperfect anomaly response mechanisms in existing systems, preventing fertilization effects from being affected or equipment damage caused by malfunctions, and improving the stability and reliability of system operation. Thus, it addresses the problems of low precision, serious resource waste, and limited data dimensions, poor model adaptability, and imperfect anomaly response mechanisms in existing agricultural fertilization technologies.

[0110] The following is a specific embodiment illustrating a fertilization control method for liquid fertilizers, using cucumber cultivation as an example, including the following steps:

[0111] In step one, the crop type is input as "greenhouse cucumber" through the interactive module, and the corresponding crop knowledge graph is loaded. This graph contains the nutrient requirements of cucumbers during the germination, seedling, flowering and fruiting stages (such as the significant increase in nitrogen and potassium requirements during the flowering and fruiting stage) and the fertilization response curve. At the same time, the growth status data such as cucumber leaf color, plant height, and number of fruits are collected by the image sensor of the acquisition module. Combined with the planting records, the growth cycle is determined to be 90 days.

[0112] In step two, based on leaf expansion (average leaf area of ​​25 cm²) and growth cycle (45 days after planting), the cucumber is determined to be in the flowering and fruiting stage. From the knowledge graph, the required nitrogen, phosphorus, and potassium for this stage are extracted to be 2.0 kg / mu, 0.8 kg / mu, and 2.5 kg / mu. Sensor detection shows that the current soil nitrogen, phosphorus, and potassium contents are 1.2 kg / mu, 0.6 kg / mu, and 1.0 kg / mu, respectively, requiring an additional 0.8 kg / mu of nitrogen, 0.2 kg / mu of phosphorus, and 1.5 kg / mu of potassium. Combined with the concentration of the applied liquid fertilizer (150 g / L nitrogen, 100 g / L phosphorus, and 200 g / L potassium), the total fertilizer requirement is calculated to be 5.3 L / mu of nitrogen solution, 2.0 L / mu of phosphorus solution, and 7.5 L / mu of potassium solution. This is then evenly distributed over a 30-day flowering and fruiting period to obtain the daily fertilizer application rate (0.18 L / mu of nitrogen solution, 0.07 L / mu of phosphorus solution, and 0.25 L / mu of potassium solution).

[0113] In step three, a liquid fertilizer loss kinetic model is constructed and the parameters are optimized using the DDPG algorithm. The LSTM model is used to predict the meteorological parameters for the next 24 hours: air temperature 18-28℃, humidity 60-70%, no rainfall, photosynthetic photon flux density 800-1200 μmol / m²・s, and wind speed 0.5-1.2 m / s. The meteorological parameters are input into the model to obtain the expected solute loss per unit volume as 8%. Combined with the temperature and humidity prediction, the optimal fertilization time window is determined to be 9:00-11:00 the next day (when the temperature is suitable and the wind speed is low), corresponding to a fertilization duration of 2 hours, and the expected fertilizer loss is 8% of the daily fertilizer application amount.

[0114] In step four, the actual amount of fertilizer to be applied that day is calculated: nitrogen solution 0.196 L / mu (0.18 L ÷ 92%), phosphorus solution 0.076 L / mu, and potassium solution 0.272 L / mu. Based on the 2-hour fertilization duration, the total flow rate of the fertilizer solution is calculated to be 0.272 L / h (total fertilizer requirement ÷ 2h), which is mapped to a frequency of 35 Hz for the variable frequency pump. Based on the parameters of the adjustable nozzle (model PT-15), the optimal spraying pressure is determined to be 0.25 MPa, which is converted into a pump pressure valve control value of 4.2V. A fertilization plan and control instructions containing "Region A - Cucumber - Flowering and Fruiting Period - Nitrogen Solution 0.196 L / mu - 9:00 Open - 11:00 Close - Pump Frequency 35 Hz - Pressure 0.25 MPa" are generated and executed by the fertilization control module.

[0115] During the process, the anomaly detection unit monitored in real time: the pump current remained stable at 3.2A (within the range of rated current 3.0±15%), the valves completed their opening and closing actions within 2 seconds, the pipeline pressure was maintained at 0.25MPa, and the fertilizer solution flow rate remained stable at 0.136L / h. All parameters met the preset thresholds, and the system operated normally. After this fertilization, the chlorophyll content of cucumber leaves increased by 12%, the fruit enlargement rate accelerated, the fertilizer utilization rate increased by 23% compared to traditional fertilization methods, and no leaf curling caused by nutrient excess was observed.

[0116] In summary, the embodiments of this application, from acquiring crop knowledge graphs, growth status and cycle, to determining growth stages and calculating total and daily fertilizer application amounts, then using models to predict meteorological parameters and fertilizer liquid loss to determine the optimal fertilization window, and finally calculating the actual fertilizer application amount, converting equipment parameters and executing fertilization, achieve precise fertilization by combining multi-dimensional data and intelligent models. Furthermore, anomaly monitoring ensures stable system operation, and in practical applications, it effectively improves fertilizer utilization and crop growth effects.

[0117] In the description of this specification, the references to "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0118] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0119] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0120] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any of the following techniques known in the art, or a combination thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0121] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0122] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A fertilization control system for liquid fertilizers, characterized in that, include: The module comprises an interaction module, an acquisition module, an instruction generation unit, and a fertilization control module; among which, The interactive module is used to input crop types and load the corresponding crop knowledge graph; The acquisition module is used to collect air temperature and humidity, wind speed, photosynthetic photon flux density, and rainfall, and deploy sensor modules to collect soil EC value, NPK content, fertilizer solution concentration, fertilizer solution flow rate, pipeline pressure, and crop growth status data. The instruction generation unit is used to construct a liquid fertilizer loss dynamics model and dynamically optimize it using a deep deterministic policy gradient reinforcement learning algorithm. It employs an LSTM time series model to predict future meteorological parameters and generates a fertilization plan based on the optimized liquid fertilizer loss dynamics model, the crop knowledge graph, and the predicted future meteorological parameters. It then generates control instructions for the fertilization equipment according to the fertilization plan. The liquid fertilizer loss dynamics model includes a rainfall correction function and a photosynthetic photon flux density-wind speed joint correction function to quantify the fertilizer loss rate under different meteorological conditions. The formula for the liquid fertilizer loss dynamics model is: ； in, This refers to the amount of solute lost per unit volume. For time, This is the value of the rainfall correction function. These are the correction function values ​​for photosynthetic photon flux density and wind speed. This refers to the concentration of the fertilizer solution. For model error, Photosynthetic photon flux density Wind speed; The fertilization control module is used to receive control commands from the fertilization equipment, execute the control commands, and provide real-time feedback.

2. The fertilization control system for liquid fertilizer according to claim 1, characterized in that, The interaction module includes an interactive interface unit and a crop knowledge graph loading unit. The interactive interface unit is used to support users in interactively modifying the generated fertilization plan. The crop knowledge graph loading unit is used to load the corresponding crop knowledge graph based on the information. The crop knowledge graph includes the nutrient requirements of crops at different growth stages and fertilization response curves.

3. The fertilization control system for liquid fertilizer according to claim 1, characterized in that, The acquisition module includes: a field weather station unit, a sensor unit, and a data processing unit, wherein... The field weather station unit is used to collect air temperature and humidity, wind speed, photosynthetic photon flux density, and rainfall. The sensor unit is used to collect soil EC value, NPK content, fertilizer solution concentration, fertilizer solution flow rate, pipeline pressure, and crop growth status. The data processing unit is used for noise reduction and standardization of the data acquired by the field weather station unit and the sensor unit.

4. The fertilization control system for liquid fertilizer according to claim 1, characterized in that, The instruction generation unit includes a loss model unit and an instruction generation engine unit. The loss model unit is used to construct and continuously update the liquid fertilizer loss kinetic model based on the actual loss data after fertilization. The instruction generation engine unit is used to generate fertilization plans and dynamically optimize the parameters of the liquid fertilizer loss kinetic model using a deep deterministic policy gradient algorithm with the optimization objective of minimizing the deviation between predicted fertilizer loss and actual loss. The instruction generation engine unit supports users in interactively adjusting the optimized fertilization plan and generates fertilization control instructions based on the adjusted fertilization plan.

5. The fertilization control system for liquid fertilizer according to claim 1, characterized in that, The formula for the rainfall correction function value is: ； in, The critical rainfall for runoff generation, This represents the actual rainfall. For nonlinear influence intensity, For the degree of linear sloping, Base correction value; The expressions for the photosynthetic photon flux density and wind speed correction functions are as follows: ； in, For light compensation point density, The light saturation point density, Critical wind speed is required to promote transpiration. The critical wind speed that dominates the drift. Based on the basic absorption efficiency coefficient The evapotranspiration promotion coefficient is for low wind speeds. This is the drift loss coefficient at medium wind speeds. This serves as the baseline coefficient for high wind speed absorption. This is the high wind speed drift attenuation coefficient. is the specular suppression coefficient, and e is the natural constant.

6. The fertilization control system for liquid fertilizer according to claim 1, characterized in that, The fertilization control module includes: a fertilization unit and an instruction execution unit, wherein... The fertilization unit is used to arrange field fertilization equipment, which includes a fertilizer storage tank, a water storage tank, a pipeline network, adjustable nozzles, and a variable frequency pump. The instruction execution unit is used to apply fertilizer according to the control instructions of the fertilizer application equipment.

7. The fertilization control system for liquid fertilizer according to claim 6, characterized in that, The instruction execution unit further includes an anomaly determination unit, wherein the anomaly determination unit is used to collect pump group operating status and valve opening and closing data in real time, receive fertilizer solution flow rate, pipeline pressure, air temperature and humidity, and soil EC value data from the data acquisition module, compare the collected data with preset thresholds, and determine an anomaly when the preset conditions are met, and trigger an alarm.

8. The fertilization control system for liquid fertilizer according to claim 7, characterized in that, The preset conditions include: the pump group vibration amplitude, temperature or current value exceeding the normal operating range of the equipment; the valve failing to complete the opening and closing action within the specified time, or exhibiting abnormal opening and closing status during non-instructed periods; the fertilizer solution flow rate continuously being lower than the minimum design flow rate or continuously being higher than the maximum design flow rate; the pipeline pressure continuously exceeding the safe pressure range or falling below the minimum pressure required to maintain normal system operation; and the air temperature and humidity or soil EC value continuously deviating from the suitable range for crop growth or the system's safe operating threshold.

9. The fertilization control system for liquid fertilizer according to claim 1, characterized in that, The fertilization plan includes the area number, fertilizer type, fertilizer amount, and fertilization time; the fertilization control instructions include the area number, fertilizer storage tank number, opening time, closing time, frequency converter pump frequency, and pump group pressure valve control value.

10. A method for applying a fertilization control system for liquid fertilizer according to any one of claims 1-9, characterized in that, The method includes: Obtain crop knowledge graphs, crop growth status, and growth cycle; The crop growth stage is obtained based on the crop growth status and growth cycle. The NPK amount required for the growth stage is obtained based on the crop knowledge graph. Combined with the soil NPK content and the concentration of NPK fertilizer, the total amount of NPK fertilizer required by the crop and the daily amount of fertilizer are calculated. A liquid fertilizer loss kinetic model was constructed and optimized. An LSTM time series model was used to predict the air temperature and humidity, rainfall, photosynthetic photon flux density, and wind speed for the next 24 hours. The predicted rainfall, photosynthetic photon flux density, and wind speed were input into the optimized liquid fertilizer loss kinetic model to obtain the expected solute loss per unit volume. Combined with the predicted air temperature and humidity, the optimal fertilization time window was obtained. The corresponding fertilization duration and expected fertilizer loss were calculated based on the optimal fertilization time window. The actual amount of fertilizer to be applied on the day is calculated based on the daily fertilizer application amount and the expected fertilizer loss. The fertilizer solution flow rate is calculated based on the actual amount of fertilizer to be applied on the day and the fertilization duration, and mapped to the frequency of the variable frequency pump. The optimal spraying pressure is determined based on the adjustable nozzle model and converted into the pump group pressure valve control value to generate a fertilization plan. The fertilization equipment control instructions are generated based on the fertilization plan, and fertilization is carried out according to the fertilization equipment control instructions.

Images