Intelligent water and fertilizer integrated optimization method and device based on Internet of Things technology

Abstract

The invention relates to the technical field of water and fertilizer integration, and provides an intelligent water and fertilizer integration optimization method and device based on the Internet of Things technology, and the method comprises the following three core stages and interaction mechanisms: 1, a data perception and edge processing stage; 2, an intelligent decision-making and dynamic optimization stage; and step 3, an accurate execution and closed-loop control stage. Soil parameters, meteorological parameters and crop physiological parameters are comprehensively collected through a multi-modal sensor network, the data transmission efficiency is improved by using an LoRa / NB-IoT hybrid transmission protocol, data cleaning and standardization processing are completed in combination with edge computing nodes, and the data fusion depth and quality are improved; a multi-objective optimization function is constructed based on a crop growth model, a water and fertilizer regulation and control strategy is dynamically generated through a machine learning algorithm, and collaborative optimization of water saving, yield increasing and environmental protection is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of integrated water and fertilizer management, and particularly to an intelligent integrated water and fertilizer optimization method and device based on Internet of Things technology. Background Art

[0002] With the transformation of modern agriculture towards intelligence and precision, the integrated water and fertilizer technology, as the core means to improve resource utilization efficiency, has become a research hotspot in the field of agricultural Internet of Things. Traditional water and fertilizer management mostly relies on manual experience judgment or single environmental parameter control, resulting in problems such as extensive irrigation and fertilization and serious resource waste. In recent years, through the combination of sensor networks and remote control systems, the Internet of Things technology has initially realized farmland environmental monitoring and basic irrigation control. However, the existing systems still have obvious shortcomings in terms of data fusion depth, decision model adaptability, and execution accuracy, and it is difficult to meet the multi-objective optimization requirements in complex farmland environments.

[0003] The current technical system faces multiple bottlenecks: First, sensor networks are mostly limited to single parameter collection (such as only monitoring soil moisture), lacking the collaborative perception of multi-dimensional data of soil - meteorology - crops, resulting in one-sided decision-making basis; Second, traditional optimization models mostly adopt static thresholds or single-objective optimization, and cannot dynamically balance the contradictory relationships among water conservation, yield increase, quality improvement, and environmental protection; Third, the execution system relies on fixed program control, with a lag in response to sudden environmental changes (such as heavy rain, drought), and lacks a closed-loop feedback mechanism, which is prone to cause over-irrigation or fertilization pollution. In addition, existing devices generally have problems such as low hardware redundancy and poor coordination between algorithms and devices, resulting in insufficient system reliability and difficulty in adapting to the long-term stable operation of large-scale farmland.

[0004] The purpose of the present invention is to solve the problems of the existing integrated water and fertilizer system in terms of data fusion depth, decision model adaptability, and execution accuracy. Summary of the Invention

[0005] The purpose of the present invention is to solve the problems of the existing integrated water and fertilizer system in terms of data fusion depth, decision model adaptability, and execution accuracy. The present invention adopts the following technical solutions:

[0006] An intelligent integrated water and fertilizer optimization method based on Internet of Things technology, including the following three core stages and their interaction mechanisms:

[0007] Step 1: Data perception and edge processing stage:

[0008] Real-time collection of multi-dimensional farmland data through a multi-modal sensor network, including soil parameters (humidity, nutrient content, pH value), meteorological parameters (temperature and humidity, wind speed, rainfall), and crop physiological parameters (leaf area index, stem diameter);

[0009] Data transmission is carried out using the LoRa / NB-IoT protocol, and data cleaning (filtering, denoising, outlier removal) and standardization processing (unit unification, missing value imputation) are completed at the edge computing node;

[0010] Step 2: Intelligent decision-making and dynamic optimization stage:

[0011] Based on the crop growth model, a multi-objective optimization function is constructed to optimize water and fertilizer consumption, crop yield, quality indicators, and environmental risks simultaneously;

[0012] Use machine learning algorithms to generate dynamic regulation strategies and perform adaptive correction in combination with real-time environmental changes;

[0013] Step 3: Precise execution and closed-loop control stage:

[0014] Implement precise water and fertilizer application through IoT actuators, monitor the execution status in real time, and trigger safety threshold control;

[0015] Establish a digital twin-driven feedback mechanism to compare actual and predicted data to continuously optimize system parameters.

[0016] An intelligent water and fertilizer integration optimization method based on IoT technology as described above, and the first step further includes the following steps:

[0017] S1.1. The multi-modal sensor network includes, but is not limited to, soil moisture sensors, soil nutrient sensors, pH value sensors, temperature and humidity sensors, anemometers, rain gauges, leaf area index measuring instruments, and stem diameter measuring instruments. Among them: the soil moisture sensor uses the frequency domain reflectometry (FDR) measurement principle, and the measurement accuracy error ≤ ±2% FS; the temperature sensor uses a PT100 platinum resistance, and the measurement accuracy is ±0.5°C; the leaf area index measuring instrument uses multi-spectral imaging technology, and the measurement error ≤ ±0.3;

[0018] S1.2. Data transmission uses a LoRa / NB-IoT hybrid networking mode. Among them: the LoRa module uses the SX1276 chip, and the transmission distance ≥ 3 km (line of sight); the NB-IoT module supports B5 / B8 frequency bands and is built-in with the AES-128 encryption algorithm; the data transmission protocol uses an improved MQTT-SN protocol, and the data packet size ≤ 512 bytes;

[0019] S1.3. The edge computing node is configured as follows: the processor uses an ARM Cortex-A53 quad-core architecture with a main frequency ≥ 1.2 GHz; the data cleaning algorithm combination: Kalman filter + wavelet denoising + 3σ anomaly detection; the standardization processing uses the Z-Score normalization algorithm, and the missing value filling uses the cubic spline interpolation method.

[0020] An intelligent integrated water and fertilizer optimization method based on Internet of Things technology as described above, the step S1.1 further includes the following steps:

[0021] S1.1.1. The spatial layout adopts a hierarchical deployment strategy: Basic layer: Deploy 1 set of basic sensors (temperature and humidity, soil humidity) per 10 mu; Enhancement layer: Deploy 1 set of enhanced sensors (nutrients, pH value) per 5 mu; Key area: Deploy 3 sets of redundant sensors within a range of 50 m from the irrigation hub.

[0022] S1.1.2. The sensor power supply system adopts: Main power supply: Monocrystalline silicon solar panel (20W) + Lithium iron phosphate battery (12V / 24Ah); Backup power supply: Replaceable lithium thionyl chloride battery pack (ER34615), with a battery life of ≥ 30 days; The power management system supports dynamic voltage regulation, and the power consumption in sleep mode is ≤ 10 μA.

[0023] S1.1.3. The data quality assurance measures include: Execute a self-check program at 3:00 am every day to generate a device health status report; Set up a three-level early warning mechanism: Level 1 early warning (data deviation > 15%): Trigger automatic calibration, Level 2 early warning (3 consecutive anomalies): Start the backup sensor, Level 3 early warning (device offline > 2h): Notify the maintenance personnel.

[0024] An intelligent integrated water and fertilizer optimization method based on Internet of Things technology as described above, the step two further includes the following steps:

[0025] S2.1. Build a crop growth model: The crop growth model is built using a hybrid modeling method; Mechanistic model: A water demand model based on the FAO-56 Penman formula; Data-driven model: An LSTM neural network to predict the crop growth curve; Fusion method: Use Bayesian inference to update the model parameters.

[0026] S2.2. Design a multi-objective optimization function: The multi-objective optimization function expression is min F(x) = [f1(x), -f2(x), f3(x), f4(x)];

[0027] Where:

[0028] f1(x): Water consumption per unit area (m 3 / ha);

[0029] f2(x): Predicted yield (kg / ha);

[0030] f3(x): Nitrate leaching amount (kg / ha);

[0031] f4(x): Quality deviation degree (%).

[0032] S2.3. Generate a dynamic regulation strategy: The generation of the dynamic regulation strategy adopts a reinforcement learning framework; State space: includes 12-dimensional environmental parameters and 6-dimensional crop parameters; Action space: irrigation amount (0 - 50 mm), fertilization amount (0 - 300 kg / ha); Reward function: R = α·yield + β·quality - γ·water consumption - δ·pollution.

[0033] An intelligent integrated water and fertilizer optimization method based on Internet of Things technology as described above, the S2.1 further includes the following steps:

[0034] S2.1.1. Field experiment design: Set 3 control areas: traditional irrigation, intelligent irrigation benchmark scheme, and the scheme of the present invention; Monitoring indicators include: soil moisture content in the root layer (measured every 6 hours); leaf surface temperature (infrared thermal imaging twice a day); dry matter accumulation (sampled weekly).

[0035] S2.1.2. Model verification criteria: Relative error of water use efficiency ≤ 8%; Yield prediction error ≤ 5%; Correlation of quality indicators R 2 ≥ 0.85.

[0036] S2.1.3. Parameter update mechanism: Perform model retraining once a week; Trigger immediate update when the environmental mutation index ΔE > 0.5; The updated data retains a sliding window of the most recent 30 days.

[0037] An intelligent integrated water and fertilizer optimization method based on Internet of Things technology as described above, the S2.2 further includes the following steps:

[0038] S2.2.1. The multi-objective solution uses an improved NSGA-III algorithm. Reference point generation: Divide the objective space based on the Das-Dennis method; Selection strategy: Adopt a dynamic reference point adaptive adjustment mechanism; Mutation operator: The polynomial mutation probability is set to 0.1;

[0039] S2.2.2. Solution acceleration strategy, pre-screening: Use K-means clustering to reduce the solution set scale; Parallel computing: Divide the population into 4 sub-populations for synchronous evolution; Early stopping mechanism: Terminate when the improvement in 10 consecutive generations < 1%;

[0040] S2.2.3. Decision support system, provide a Pareto front visualization interface; Support manual preference setting (yield priority / environmental protection priority); Generate multiple sets of alternative solutions for decision-makers to choose.

[0041] An intelligent integrated water and fertilizer optimization method based on Internet of Things technology as described above, the step three further includes the following steps:

[0042] S3.1. Actuator system configuration; Irrigation system: Pressure-compensated drip irrigation tape (operating pressure 0.1 - 0.3 MPa); Fertilizer applicator: Venturi injection proportional fertilizer pump (accuracy ±2%); Control valve: Electric ball valve (response time ≤ 3 s);

[0043] S3.2. Irrigation safety threshold: Single maximum irrigation volume ≤ 80% of field capacity; Daily cumulative irrigation volume ≤ 120% of crop water requirement; Fertilization safety threshold: Nitrogen fertilizer concentration ≤ 200 mg / L; Phosphorus fertilizer application rate ≤ 90% of soil adsorption capacity;

[0044] S3.3. Digital twin system architecture, Physical entity layer: Sensor + actuator network; Virtual model layer: PlantSim crop growth simulation engine; Data interaction layer: Bidirectional communication implemented by OPC UA protocol.

[0045] An intelligent water and fertilizer integration optimization method based on Internet of Things technology as described above, said S3.1 further includes the following steps:

[0046] S3.1.1. Precise control of irrigation volume, using PID + feedforward composite control algorithm; Flowmeter accuracy class: 0.5 level; End pressure fluctuation compensation algorithm: ΔP = Kp·e(t) + Ki·∫e(t)dt + Kd·de(t) / dt where e(t) = set pressure - measured pressure;

[0047] S3.1.2. Control of fertilizer solution concentration, Online EC / pH monitor (measurement period ≤ 10 s); Fertilizer solution ratio using fuzzy control algorithm; Establish a fertilizer solubility-temperature compensation database;

[0048] S3.1.3. Fault diagnosis system, Feature extraction: Wavelet packet decomposition of vibration signals; Fault classification: SVM multi-classifier (accuracy ≥ 92%); Self-recovery mechanism: Automatic switching time of standby pipeline ≤ 15 s.

[0049] An intelligent water and fertilizer integration optimization method based on Internet of Things technology as described above, said S3.2 further includes the following steps:

[0050] S3.2.1. Dynamic safety threshold adjustment mechanism: Dynamically correct the irrigation and fertilization safety thresholds based on real-time environmental data (soil humidity, weather forecast, crop growth stage), specifically including:

[0051] Dynamic threshold of irrigation volume: Single maximum irrigation volume = initial threshold × (1 + 0.05 × ΔT) - 0.1 × R_7d (where ΔT is the change in daily temperature compared to the reference value, and R_7d is the cumulative rainfall in the previous 7 days);

[0052] The daily cumulative irrigation volume threshold is adjusted in real time according to the crop evapotranspiration (ETc), and the formula is: Q_max = min(1.2×ETc, field capacity×0.8 - W_current) (where W_current is the current soil moisture content in the root layer);

[0053] The dynamic threshold of fertilization concentration: the upper limit of nitrogen fertilizer concentration = basic threshold×(1 - 0.02×S_moisture) (where S_moisture is the reading of the soil moisture sensor, unit: %).

[0054] The threshold of phosphate fertilizer application rate = soil adsorption capacity×(0.9 - 0.01×pH_dev) (where pH_dev is the absolute deviation between the current pH value and the optimal value of the crop).

[0055] An intelligent integrated water and fertilizer optimization method based on Internet of Things technology as described above. The dedicated device for implementing this method includes: a multi-modal sensor network (integrating high-precision sensors for soil, meteorological, and crop physiological parameters, using LoRa / NB-IoT hybrid communication protocol), an edge computing node (equipped with data cleaning and standardization processing algorithms), a cloud decision-making engine (a multi-objective optimization module based on crop growth models and reinforcement learning), a precise water and fertilizer execution mechanism (pressure-compensated drip irrigation system, Venturi proportional fertilizer pump, and safety redundancy control loop), a digital twin feedback system (virtual-real data mapping and dynamic parameter optimization module), and a hybrid energy supply and communication unit (solar power supply and encrypted data transmission module). Each module is interconnected through a unified interface protocol to form a closed-loop control system from data acquisition, intelligent decision-making to precise execution.

[0056] Implementing the embodiments of the present invention has the following beneficial effects:

[0057] 1. The present invention comprehensively collects soil parameters, meteorological parameters, and crop physiological parameters through a multi-modal sensor network, improves data transmission efficiency using the LoRa / NB-IoT hybrid transmission protocol, and combines the edge computing node to complete data cleaning and standardization processing, significantly improving the depth and quality of data fusion; constructs a multi-objective optimization function based on crop growth models, dynamically generates water and fertilizer regulation strategies through machine learning algorithms, realizes the coordinated optimization of water conservation, yield increase, and environmental protection, and adapts to environmental changes in real time; precisely controls water and fertilizer application with the help of Internet of Things actuators, and continuously optimizes system parameters in combination with the digital twin feedback mechanism, ultimately effectively improving the utilization efficiency of water and fertilizer resources.

[0058] In summary, the present invention solves the problems of insufficient data fusion depth, adaptability of decision-making models, and execution accuracy in existing integrated water and fertilizer systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0060] Figure 1 It is a block diagram of the steps of an intelligent integrated water and fertilizer optimization method based on Internet of Things technology of the present invention. Specific embodiments

[0061] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0062] As <{ Figure 1 shown, the present invention proposes an intelligent integrated water and fertilizer optimization method based on Internet of Things technology.

[0063] It includes the following three core stages and their interaction mechanisms:

[0064] Step 1: Data perception and edge processing stage:

[0065] Real-time collection of multi-dimensional data of farmland through a multi-modal sensor network, including soil parameters (humidity, nutrient content, pH value), meteorological parameters (temperature and humidity, wind speed, rainfall), and crop physiological parameters (leaf area index, stem diameter);

[0066] Use the LoRa / NB-IoT protocol for data transmission, and complete data cleaning (filtering, denoising, outlier removal) and standardization processing (unit unification, missing value imputation) at the edge computing node;

[0067] S1.1. The multi-modal sensor network includes, but is not limited to, a soil humidity sensor, a soil nutrient sensor, a pH value sensor, a temperature and humidity sensor, an anemometer, a rain gauge, a leaf area index measuring instrument, and a stem diameter measuring instrument. Among them: the soil humidity sensor uses the frequency domain reflectometry (FDR) measurement principle, and the measurement accuracy error ≤ ±2% FS; the temperature sensor uses a PT100 platinum resistance, and the measurement accuracy is ±0.5 °C; the leaf area index measuring instrument uses multi-spectral imaging technology, and the measurement error ≤ ±0.3;

[0068] S1.1.1. The spatial layout adopts a hierarchical deployment strategy: Basic layer: Deploy 1 set of basic sensors (temperature and humidity, soil humidity) per 10 mu; Enhancement layer: Deploy 1 set of enhanced sensors (nutrients, pH value) per 5 mu; Critical area: Deploy 3 sets of redundant sensors within a range of 50 m from the irrigation hub radius;

[0069] S1.1.2. The sensor power supply system adopts: Main power supply: Monocrystalline silicon solar panel (20W) + Lithium iron phosphate battery (12V / 24Ah); Backup power supply: Replaceable lithium thionyl chloride battery pack (ER34615), with a battery life of ≥ 30 days; The power management system supports dynamic voltage regulation, and the power consumption in sleep mode ≤ 10 μA;

[0070] S1.1.3. Data quality assurance measures include: Execute a self-check program at 3:00 am every day to generate a device health status report; Set up a three-level early warning mechanism: Level 1 early warning (data deviation > 15%): Trigger automatic calibration, Level 2 early warning (abnormal three times in a row): Start the backup sensor, Level 3 early warning (device offline > 2h): Notify the maintenance personnel

[0071] S1.2. Data transmission adopts a LoRa / NB-IoT hybrid networking mode, where: The LoRa module uses the SX1276 chip, and the transmission distance ≥ 3 km (line of sight); The NB-IoT module supports B5 / B8 frequency bands and is built-in with the AES-128 encryption algorithm; The data transmission protocol adopts an improved MQTT-SN protocol, and the packet size ≤ 512 bytes;

[0072] S1.3. The edge computing node is configured as follows: The processor adopts an ARM Cortex-A53 quad-core architecture with a main frequency ≥ 1.2 GHz; Data cleaning algorithm combination: Kalman filter + Wavelet denoising + 3σ anomaly detection; Standardization processing adopts the Z-Score normalization algorithm, and missing value filling uses the cubic spline interpolation method

[0073] Step 2: Intelligent decision-making and dynamic optimization stage:

[0074] Build a multi-objective optimization function based on the crop growth model, and simultaneously optimize water and fertilizer consumption, crop yield, quality indicators, and environmental risks;

[0075] Use machine learning algorithms to generate dynamic regulation strategies and perform adaptive correction in combination with real-time environmental changes;

[0076] S2.1. Build a crop growth model: The crop growth model is built using a hybrid modeling method; Mechanistic model: A water demand model based on the FAO-56 Penman formula; Data-driven model: An LSTM neural network to predict the crop growth curve; Fusion method: Use Bayesian inference to update the model parameters.

[0077] S2.1.1. Field experiment design: Set up 3 control areas: traditional irrigation, intelligent irrigation baseline scheme, and the scheme of the present invention; Monitoring indicators include: soil moisture content in the root layer (measured every 6 hours); leaf surface temperature (twice a day by infrared thermal imaging); dry matter accumulation (sampled weekly).

[0078] S2.1.2. Model verification criteria: Relative error of water use efficiency ≤ 8%; Yield prediction error ≤ 5%; Correlation of quality indicators R 2 ≥ 0.85.

[0079] S2.1.3. Parameter update mechanism: Perform model retraining once a week; Trigger immediate update when the environmental mutation index ΔE > 0.5; Retain the sliding window of the most recent 30 days for updated data.

[0080] S2.2. Design a multi-objective optimization function: The expression of the multi-objective optimization function is min F(x) = [f1(x), -f2(x), f3(x), f4(x)].

[0081] Where:

[0082] f1(x): Water consumption per unit area (m 3 / ha);

[0083] f2(x): Predicted yield (kg / ha);

[0084] f3(x): Nitrate leaching amount (kg / ha);

[0085] f4(x): Quality deviation degree (%).

[0086] S2.2.1. The multi-objective solution adopts the improved NSGA-III algorithm. Reference point generation: Divide the objective space based on the Das-Dennis method; Selection strategy: Adopt the dynamic reference point adaptive adjustment mechanism; Mutation operator: Set the polynomial mutation probability to 0.1;

[0087] S2.2.2. Solution acceleration strategy, pre-screening: Use K-means clustering to reduce the solution set scale; Parallel computing: Divide the population into 4 sub-populations for synchronous evolution; Early stopping mechanism: Terminate when the improvement is < 1% for 10 consecutive generations;

[0088] S2.2.3. Decision support system, providing a Pareto front visualization interface; Supporting manual preference setting (yield priority / environmental protection priority); Generating multiple sets of alternative solutions for decision-makers to choose from.

[0089] S2.3. Generate dynamic regulation strategies: The generation of dynamic regulation strategies adopts a reinforcement learning framework; State space: includes 12-dimensional environmental parameters and 6-dimensional crop parameters; Action space: irrigation amount (0 - 50 mm), fertilization amount (0 - 300 kg / ha); Reward function: R = α · yield + β · quality - γ · water consumption - δ · pollution.

[0090] Step 3: Precise execution and closed-loop control stage:

[0091] Implement precise water and fertilizer application through IoT actuators, monitor the execution status in real time and trigger safety threshold control;

[0092] Establish a digital twin-driven feedback mechanism to compare actual and predicted data to continuously optimize system parameters.

[0093] S3.1. Actuator system configuration; Irrigation system: pressure-compensated drip irrigation tape (working pressure 0.1 - 0.3 MPa); Fertilizer applicator: Venturi injection proportional fertilizer pump (accuracy ±2%); Control valve: electric ball valve (response time ≤3 s).

[0094] S3.1.1. Precise control of irrigation amount, using a PID + feedforward composite control algorithm; Flowmeter accuracy class: 0.5 level; End pressure fluctuation compensation algorithm: ΔP = Kp · e(t) + Ki · ∫e(t)dt + Kd · de(t) / dt where e(t) = set pressure - measured pressure;

[0095] S3.1.2. Control of fertilizer solution concentration, online EC / pH monitor (measurement period ≤10 s); Fertilizer solution ratio uses a fuzzy control algorithm; Establish a fertilizer solubility-temperature compensation database;

[0096] S3.1.3. Fault diagnosis system, Feature extraction: Wavelet packet decomposition of vibration signals; Fault classification: SVM multi-classifier (accuracy ≥92%); Self-recovery mechanism: Standby pipeline automatic switching time ≤15 s

[0097] S3.2. Irrigation safety threshold: Single maximum irrigation amount ≤80% of field capacity; Daily cumulative irrigation amount ≤120% of crop water requirement; Fertilization safety threshold: Nitrogen fertilizer concentration ≤200 mg / L; Phosphorus fertilizer application amount ≤90% of soil adsorption capacity.

[0098] S3.2.1. Dynamic safety threshold adjustment mechanism: Dynamically correct the safety thresholds of irrigation and fertilization based on real-time environmental data (soil humidity, weather forecast, crop growth stage), specifically including:

[0099] Dynamic threshold of irrigation volume: Maximum single irrigation volume = Initial threshold × (1 + 0.05 × ΔT) - 0.1 × R_7d (where ΔT is the change in the daily temperature compared to the reference value, and R_7d is the cumulative rainfall in the previous 7 days);

[0100] The daily cumulative irrigation volume threshold is adjusted in real time according to the crop evapotranspiration (ETc). Formula: Q_max = min(1.2 × ETc, Field capacity × 0.8 - W_current) (W_current is the current soil moisture content in the root layer);

[0101] Dynamic threshold of fertilization concentration: Upper limit of nitrogen fertilizer concentration = Basic threshold × (1 - 0.02 × S_moisture) (S_moisture is the reading of the soil moisture sensor, unit: %).

[0102] Threshold of phosphate fertilizer application amount = Soil adsorption capacity × (0.9 - 0.01 × pH_dev) (pH_dev is the absolute deviation between the current pH value and the optimal value for the crop).

[0103] S3.3, Digital twin system architecture, Physical entity layer: Sensor + actuator network; Virtual model layer: PlantSim crop growth simulation engine; Data interaction layer: Bidirectional communication implemented by OPC UA protocol.

[0104] S3.3.1, Definition of state and action:

[0105] The state space contains 18-dimensional parameters (12-dimensional environmental parameters: Soil moisture, Soil nitrate nitrogen content, Soil pH value, Air temperature, Air humidity, Wind speed, Light intensity, Rainfall, Leaf area index, Stem diameter, Crop growth stage, Root layer depth; 6-dimensional crop parameters: Leaf nitrogen content, Transpiration rate, Photosynthetically active radiation absorption rate, Dry matter accumulation, Fruit swelling rate, Water stress index), and Min-Max normalization is used.

[0106] The action space defines the irrigation volume (0 - 50 mm, continuously adjustable) and the fertilization amount (0 - 300 kg / ha, stepwise adjustable), which are mapped through one-hot encoding.

[0107] S3.3.2, Algorithm selection and training:

[0108] Proximal Policy Optimization (PPO) algorithm is adopted. The Actor network outputs the action probability, and the Critic network evaluates the state value.

[0109] Reward function design:

[0110] The training data is based on historical farmland data (≥10,000 records), and the training period ≤ 24 hours.

[0111] S3.3.2, Dynamic Policy Execution and Revision

[0112] Real-time Regulation and Feedback:

[0113] Update the policy every 30 minutes and generate an irrigation and fertilization plan based on real-time sensor data.

[0114] Response to Environmental Sudden Changes: When the temperature suddenly changes by ΔT > 5°C or the rainfall suddenly increases by > 10 mm, automatically switch to the safe mode (halve the irrigation amount and suspend fertilization).

[0115] Digital Twin Verification: Input the policy into the PlantSim simulation model. If the predicted yield deviation > 8%, re-optimize the policy.

[0116] The dedicated device for implementing this method includes: a multi-modal sensor network (integrating high-precision sensors for soil, meteorological, and crop physiological parameters, using the LoRa / NB-IoT hybrid communication protocol), an edge computing node (equipped with data cleaning and standardization processing algorithms), a cloud decision-making engine (a multi-objective optimization module based on crop growth models and reinforcement learning), a precise water and fertilizer execution mechanism (a pressure-compensated drip irrigation system, a Venturi proportional fertilizer pump, and a safety redundancy control loop), a digital twin feedback system (a virtual-real data mapping and dynamic parameter optimization module), and a hybrid power supply and communication unit (a solar power supply and encrypted data transmission module). Each module is interconnected through a unified interface protocol to form a closed-loop control system from data acquisition, intelligent decision-making to precise execution.

[0117] Example 1:

[0118] Application Scenario: A 200-acre terraced millet planting area in Yuci District, Jinzhong City, located on the eastern edge of the Loess Plateau, with an average annual rainfall of 480 mm, yellow loess soil (organic matter content 0.8% - 1.2%), a spring drought occurrence frequency of 70%, and serious soil erosion. The traditional irrigation water use coefficient is only 0.6, and the fertilizer utilization rate is less than 35%.

[0119] Technical Challenges: The fragmented terrain makes it difficult to deploy the sensor network; the contradiction between seasonal drought and soil erosion is prominent; traditional water and fertilizer management is extensive and lacks precise regulation means.

[0120] Solutions:

[0121] (I) Deployment of the Multi-modal Sensing Network

[0122] 1. Sensor Selection and Layout

[0123] Soil Parameter Monitoring:

[0124] Adopt a TEROS12 three-parameter sensor (humidity, temperature, conductivity), with a measurement range of 0 - 100% VWC and an accuracy of ±3%.

[0125] Layered burial: tillage layer (20cm), active root layer (40cm), deep moisture monitoring (60cm).

[0126] Deployment density: 0.5m from the edge of each terrace, horizontal spacing of 15m, and one group added for every 5m of vertical height difference.

[0127] Meteorological monitoring:

[0128] A six-element weather station is set up on the top of the slope to monitor wind speed (0-60m / s), light (0-1400W / m 2 ), rainfall (resolution 0.2mm).

[0129] The data is integrated with precipitation data from the CMA-CMORPH satellite of the China Meteorological Administration, with a temporal resolution of 30 minutes.

[0130] Crop physiological monitoring:

[0131] The drone is equipped with a multispectral camera to obtain NDVI (Normalized Difference Vegetation Index) and NDWI (Water Index) every week.

[0132] Stem flow meters monitor transpiration, with one set deployed for every 50 mu.

[0133] 2. Communication network optimization

[0134] Hybrid communication architecture:

[0135]

[0136] Anti-interference design:

[0137] Adopt frequency hopping spread spectrum technology (FHSS), channel switching cycle is 10 minutes

[0138] 3. Edge computing node configuration

[0139] Hardware selection:

[0140] Processor: Rockchip RK3399 (dual-core A72 + quad-core A53), main frequency 2.0GHz.

[0141] Storage: 128GB eMMC + 1TB SSD.

[0142] Protection level: IP67 waterproof and dustproof, operating temperature -30℃~70℃.

[0143] Data processing flow:

[0144] Kalman filter (process noise Q = 0.1, observation noise R = 0.5).

[0145] Wavelet denoising (db8 wavelet basis, 5 - layer decomposition).

[0146] 3σ anomaly detection (sliding window of 7 days).

[0147] Missing value processing:

[0148] Combination of cubic spline interpolation and spatio - temporal Kriging interpolation.

[0149] (II) Intelligent decision - making and dynamic optimization

[0150] 1. Construction of dry - farming crop models

[0151] Mechanistic model:

[0152] Water production function:

[0153] (Y_m = potential yield 350 kg / mu, k_y = 1.25)

[0154] Root water uptake model:

[0155] (z_r = maximum root depth 80 cm, p = 2.5)

[0156] Data - driven model:

[0157] LSTM network structure: 12 - node input layer, 2×64 GRU hidden layers, 3 - node output layer. Training data: 2018 - 2022 field experiment dataset (n = 15,000 records)

[0158] 2. Multi - objective optimization implementation

[0159] Objective function:

[0160] minF(x)=[f1(x), - f2(x), f3(x), f4(x)]

[0161] f1: Irrigation water consumption (m 3 / ha)

[0162] f2: Predicted yield (kg / ha)

[0163] f3: Soil erosion amount (t / ha)

[0164] f4: Nitrate nitrogen leaching amount (kg / ha)

[0165] Improved NSGA - III algorithm:

[0166] parameter set value reference point generation Hierarchical Sampling Method (HSS) population size 256 crossover probability 0.9

[0167] (III) Precise execution control

[0168] 1. Design of Terrace Irrigation System

[0169] Three - stage Pressure - regulating System:

[0170] level equipment model technical parameter primary level centrifugal pump <![CDATA[Head 120m, flow rate 50m 3 / h]]> secondary level variable frequency pump The pressure can be adjusted from 0.6 to 1.2 MPa tertiary level pressure-compensated drip irrigation tape operating pressure 0.1 - 0.6 MPa

[0171] Anti - clogging Design:

[0172] Prefilter: Automatic back - flushing laminated type (precision 13μm).

[0173] 2. Dynamic Safety Threshold Control

[0174] Dynamic Adjustment of Irrigation Volume:

[0175] Q max = min(1.2×ET c ,θ FC ×0.8×D r -θ current ×Dr)

[0176] Fertilization Safety Rule:

[0177] When the soil conductivity EC > 2.5 mS / cm, automatically switch to the clear - water flushing mode.

[0178] (IV) Digital Twin Feedback Optimization

[0179] 1. 3D Modeling and Simulation

[0180] Terrain Modeling:

[0181] Generate a DEM (Digital Elevation Model) with a resolution of 5 cm by UAV aerial survey.[[ID=5)]

[0182] Crop Growth Simulation:

[0183] Use the PlantSim engine to simulate the growth process of millet under different water - fertilizer strategies.

[0184] 2. Parameter Optimization Mechanism

[0185] Model Calibration Trigger Condition:

[0186] index threshold calibration action >15％ transpiration prediction error

[0187] III. Verification of Implementation Effects

[0188] 1. Water Resource Utilization Efficiency

[0189] LSTM model incremental learning index traditional system system of the present invention improvement range 0.62 0.91 +46.8％

[0190] 2. Agricultural Production Benefits

[0191] irrigation water use coefficient index traditional system system of the present invention improvement range 230 315 +37.0％

[0192] 3. Ecological and environmental benefits

[0193]

[0194] 4. System reliability

[0195] millet yield per mu (kg) index test result standard requirement ≤380 ≤500

[0196] Example 2:

[0197] Application scenario: A 300 - mu mountain jujube forest in Linxian County, Lvliang City. The soil is sandy cinnamon soil (pH 7.8 - 8.3), with an average annual rainfall of 420 mm. Drought occurs frequently during the fruit expansion period (July - September), and the cracking rate of fruits caused by traditional irrigation is as high as 25%.

[0198] Technical difficulties: The water and fertilizer requirements of jujubes follow complex rules and need to be precisely regulated according to the phenological period; the irrigation pressure distribution on the mountain is uneven.

[0199] (I) Deployment of multi - modal perception network

[0200] 1. Special configuration of sensors

[0201] Trunk physiological monitoring:

[0202] Deploy trunk micro - change sensors for every 20 plants, with a resolution of 0.0 / mm

[0203] Root layer monitoring:

[0204] Capacitive soil sensors are buried at depths of 40 / 60 / 80 cm

[0205] 2. Optimization of communication networking

[0206] Mountain - adapted networking:

[0207] control instruction delay (ms) communication mode deployment location coverage radius LoRa ridge line relay tower 5.2 km ZigBee 80m

[0208] (II) Intelligent decision - making and dynamic optimization

[0209] 1. Special model for jujubes

[0210] Phenological period recognition rules:

[0211] inter-forest cluster network phenological period temperature accumulation (℃) moisture threshold (%) ≥80 18-22 germination period ≥350 20-25

[0212] 2. Multi - objective optimization

[0213] Quality - oriented function:

[0214] Qscore = 0.4×SSC + 0.3×single - fruit weight+0.3×fruit shape index

[0215] (III) Precise execution control

[0216] 1. Pulse irrigation system

[0217] Working mode: Irrigation for 5 minutes → Stop irrigation for 15 minutes in a cycle

[0218] Pressure compensation:

[0219] (4) Digital twin feedback optimization

[0220] 1. Virtual test system:

[0221] Simulate continuous high-temperature scenarios and automatically generate emergency irrigation strategies.

[0222] III. Implementation effect verification

[0223] 1. Yield and quality

[0224] flowering period index traditional system system of the present invention improvement range 1200 1650 +37.5％

[0225] 2. Resource efficiency

[0226]

[0227] 3. System reliability

[0228] fresh jujube yield per mu (kg) index test result standard requirement fault self-diagnosis accuracy rate 92％ ≥85％

[0229] In summary, the present invention solves the problems of the deficiencies in the depth of data fusion, the adaptability of the decision-making model, and the execution accuracy of the existing water and fertilizer integration system.

[0230] It should be understood that in the present invention, terms such as "first" and "second" are used to describe various information, but these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present invention, "first" information can also be called "second" information, and similarly, "second" information can also be called "first" information. In addition, the orientation or positional relationship indicated by terms such as "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention.

[0231] The above is the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and deformations can be made, and these improvements and deformations are also regarded as the protection scope of the present invention.

Claims

1. An intelligent integrated water and fertilizer optimization method based on Internet of Things technology, characterized in that, It includes the following three core stages and their interaction mechanisms: Step 1: Data perception and edge processing stage: Multidimensional data of farmland are collected in real time through a multi-modal sensor network, including soil parameters (humidity, nutrient content, pH value), meteorological parameters (temperature and humidity, wind speed, rainfall), and crop physiological parameters (leaf area index, stem diameter); The LoRa / NB-IoT protocol is used for data transmission, and data cleaning (filtering, denoising, outlier removal) and standardization processing (unit unification, missing value imputation) are completed at the edge computing node; Step 2: Intelligent decision-making and dynamic optimization stage: Based on the crop growth model, a multi-objective optimization function is constructed to optimize water and fertilizer consumption, crop yield, quality indicators, and environmental risks simultaneously; Machine learning algorithms are used to generate dynamic regulation strategies, which are adaptively corrected in combination with real-time environmental changes; Step 3: Precise execution and closed-loop control stage: Precise water and fertilizer application is implemented through IoT actuators, and the execution status is monitored in real time and safety threshold control is triggered; A digital twin-driven feedback mechanism is established to compare actual and predicted data to continuously optimize system parameters.

2. The intelligent integrated water and fertilizer optimization method based on Internet of Things technology according to claim 1, characterized in that The said Step 1 further includes the following steps: S1.

1. The multi-modal sensor network includes, but is not limited to, soil humidity sensors, soil nutrient sensors, pH value sensors, temperature and humidity sensors, anemometers, rain gauges, leaf area index measuring instruments, and stem diameter measuring instruments. Among them: The soil humidity sensor adopts the frequency domain reflectometry (FDR) measurement principle, and the measurement accuracy error ≤ ±2% FS; The temperature sensor adopts a PT100 platinum resistance, and the measurement accuracy is ±0.5°C; The leaf area index measuring instrument adopts multi-spectral imaging technology, and the measurement error ≤ ±0.3; S1.

2. The data transmission adopts a LoRa / NB-IoT hybrid networking mode. Among them: The LoRa module uses an SX1276 chip, and the transmission distance ≥ 3 km (line of sight); The NB-IoT module supports B5 / B8 frequency bands and is built-in with the AES-128 encryption algorithm; The data transmission protocol adopts an improved MQTT-SN protocol, and the data packet size ≤ 512 bytes; S1.

3. The edge computing node is configured as follows: The processor adopts an ARM Cortex-A53 quad-core architecture, and the main frequency ≥ 1.2 GHz; The data cleaning algorithm combination: Kalman filter + wavelet denoising + 3σ anomaly detection; The standardization processing adopts the Z-Score normalization algorithm, and the missing value filling uses the cubic spline interpolation method.

3. The intelligent integrated water and fertilizer optimization method based on Internet of Things technology according to claim 2, wherein, The said S1.1 further includes the following steps: S1.1.

1. The spatial layout adopts a hierarchical deployment strategy: Basic layer: 1 group of basic sensors (temperature and humidity, soil humidity) is deployed per 10 mu; Enhancement layer: 1 group of enhanced sensors (nutrients, pH value) is deployed per 5 mu; Key area: 3 groups of redundant sensors are deployed within a range of 50 m from the irrigation hub radius; S1.1.

2. The sensor power supply system adopts: Main power supply: Monocrystalline silicon solar panel (20W) + lithium iron phosphate battery (12V / 24Ah); Backup power supply: Replaceable lithium thionyl chloride battery pack (ER34615), with a battery life ≥ 30 days; The power management system supports dynamic voltage regulation, and the power consumption in the sleep mode ≤ 10 μA; S1.1.

3. Data quality assurance measures include: executing a self-check program at 3:00 am every day to generate a device health status report; setting up a three-level early warning mechanism: Level 1 early warning (data deviation > 15%): triggering automatic calibration, Level 2 early warning (three consecutive anomalies): starting a backup sensor, Level 3 early warning (device offline > 2h): notifying the maintenance personnel.

4. An intelligent integrated water and fertilizer optimization method based on Internet of Things technology according to claim 1, characterized in that, The second step further includes the following steps: S2.

1. Build a crop growth model: The crop growth model is built using a hybrid modeling method; Mechanistic model: A water demand model based on the FAO-56 Penman formula; Data-driven model: An LSTM neural network to predict the crop growth curve; Fusion method: Using Bayesian inference to update model parameters; S2.

2. Design a multi-objective optimization function: The multi-objective optimization function expression is min F(x) = [f1(x), -f2(x), f3(x), f4(x)]; Where: f1(x): Water consumption per unit area (m 3 / ha); f2(x): Predicted yield (kg / ha); f3(x): Nitrate leaching amount (kg / ha); f4(x): Quality deviation percentage (%). S2.

3. Generate a dynamic regulation strategy: The dynamic regulation strategy is generated using a reinforcement learning framework; State space: Includes 12-dimensional environmental parameters and 6-dimensional crop parameters; Action space: Irrigation amount (0 - 50mm), fertilization amount (0 - 300kg / ha); Reward function: R = α·yield + β·quality - γ·water consumption - δ·pollution.

5. The intelligent integrated water and fertilizer optimization method based on Internet of Things technology according to claim 4, characterized in that, The S2.1 further includes the following steps: S2.1.

1. Field experiment design: Set up 3 control areas: traditional irrigation, intelligent irrigation benchmark scheme, and the scheme of the present invention; Monitoring indicators include: Root zone soil moisture content (measured every 6 hours); Leaf surface temperature (infrared thermal imaging twice a day); Dry matter accumulation amount (sampled weekly). S2.1.

2. Model verification criteria: relative error of water use efficiency ≤ 8%; yield prediction error ≤ 5%; correlation of quality indicators R 2 ≥ 0.

85. S2.1.

3. Parameter update mechanism: Perform model retraining once a week; Trigger immediate update when the environmental mutation index ΔE > 0.5; The updated data retains a sliding window of the most recent 30 days.

6. The intelligent water-fertilizer integration optimization method based on Internet of Things technology according to claim 4 is characterized in that: The S2.2 further includes the following steps: S2.2.

1. The multi-objective solution uses an improved NSGA-III algorithm. Reference point generation: Divide the objective space based on the Das-Dennis method; Selection strategy: Use a dynamic reference point adaptive adjustment mechanism; Mutation operator: The polynomial mutation probability is set to 0.1; S2.2.

2. Solution acceleration strategy, pre-screening: Use K-means clustering to reduce the solution set scale; Parallel computing: Divide the population into 4 sub-populations for synchronous evolution; Early stopping mechanism: Terminate when the improvement in 10 consecutive generations < 1%; S2.2.

3. Decision support system, providing a Pareto front visualization interface; Supporting artificial preference settings (yield priority / environmental protection priority); Generating multiple alternative solutions for decision-makers to choose from.

7. An intelligent integrated water and fertilizer optimization method based on Internet of Things technology according to claim 1, characterized in that, The third step further includes the following steps: S3.

1. Actuator system configuration; Irrigation system: Pressure-compensated drip irrigation tape (operating pressure 0.1 - 0.3MPa); Fertilizer applicator: Venturi injection proportional fertilizer pump (accuracy ±2%); Control valve: Electric ball valve (response time ≤ 3s); S3.

2. Irrigation safety threshold: The maximum single irrigation volume ≤ 80% of the field capacity; The daily cumulative irrigation volume ≤ 120% of the crop water requirement; Fertilization safety threshold: Nitrogen fertilizer concentration ≤ 200 mg / L; Phosphorus fertilizer application rate ≤ 90% of the soil adsorption capacity; S3.

3. Digital twin system architecture, Physical entity layer: Sensor + actuator network; Virtual model layer: PlantSim crop growth simulation engine; Data interaction layer: OPC UA protocol for two-way communication.

8. An intelligent integrated water and fertilizer optimization method based on Internet of Things technology according to claim 7, characterized in that S3.1 also includes the following steps: S3.1.

1. Precise control of irrigation volume, using PID + feedforward composite control algorithm; Flowmeter accuracy class: 0.5 level; End pressure fluctuation compensation algorithm: ΔP = Kp·e(t)+Ki·∫e(t)dt+Kd·de(t) / dt where e(t) = set pressure - measured pressure; S3.1.

2. Control of fertilizer solution concentration, Online EC / pH monitor (measurement period ≤ 10 s); Fertilizer solution ratio using fuzzy control algorithm; Establish a fertilizer solubility-temperature compensation database; S3.1.

3. Fault diagnosis system, Feature extraction: Wavelet packet decomposition of vibration signals; Fault classification: SVM multi-classifier (accuracy rate ≥ 92%); Self-recovery mechanism: Automatic switching time of standby pipeline ≤ 15 s.

9. An intelligent integrated water and fertilizer optimization method based on Internet of Things technology according to claim 7, characterized in that S3.2 also includes the following steps: S3.2.

1. Dynamic safety threshold adjustment mechanism: Dynamically correct the irrigation and fertilization safety thresholds based on real-time environmental data (soil humidity, weather forecast, crop growth stage), specifically including: Dynamic threshold of irrigation volume: The maximum single irrigation volume = initial threshold × (1 + 0.05 × ΔT)-0.1 × R_7d (where ΔT is the change in daily temperature compared to the reference value, and R_7d is the cumulative rainfall in the previous 7 days); The daily cumulative irrigation volume threshold is adjusted in real time according to the crop evapotranspiration (ETc), formula: Q_max = min(1.2 × ETc, field capacity × 0.8 - W_current) (W_current is the current soil water content in the root layer); Dynamic threshold of fertilization concentration: Upper limit of nitrogen fertilizer concentration = basic threshold × (1 - 0.02 × S_moisture) (S_moisture is the reading of the soil humidity sensor, unit: %); Phosphorus fertilizer application rate threshold = soil adsorption capacity × (0.9 - 0.01 × pH_dev) (pH_dev is the absolute deviation between the current pH value and the optimal value of the crop).

10. The intelligent integrated water and fertilizer optimization method based on the Internet of Things technology according to any one of claims 1-9, characterized in that, The dedicated device for implementing this method includes: a multi-modal sensor network (integrating high-precision sensors for soil, meteorological, and crop physiological parameters, using a LoRa / NB-IoT hybrid communication protocol), an edge computing node (equipped with data cleaning and standardization processing algorithms), a cloud decision-making engine (a multi-objective optimization module based on crop growth models and reinforcement learning), a precise water and fertilizer execution mechanism (a pressure-compensated drip irrigation system, a Venturi proportional fertilizer pump, and a safety redundancy control loop), a digital twin feedback system (a virtual-real data mapping and dynamic parameter optimization module), and a hybrid energy supply and communication unit (a solar power supply and encrypted data transmission module). Each module is interconnected through a unified interface protocol to form a closed-loop control system from data collection, intelligent decision-making to precise execution.

Images