Intelligent precise irrigation and nutrient regulation and control integrated system and method for rice

Abstract

The invention discloses an intelligent precise irrigation and nutrient regulation and control integrated system and method for rice, and particularly relates to the technical field of intelligent agriculture, and the system comprises a multi-mode sensing module, an edge computing node, an intelligent decision module, a water and fertilizer cooperative execution module and a credible monitoring module. Rice rhizosphere and leaf environment data are collected through an underground nano sensor and a canopy multispectral imaging device, data fusion and model reasoning are carried out through edge nodes, an irrigation strategy value is output in combination with a mechanism model and an AI model, and a pulse injection device is controlled to achieve precise water and fertilizer integrated operation. The system supports network disconnection operation and night model updating, and the stability and the self-adaptive capability are improved. According to the method, the environment data acquisition precision and the irrigation decision response speed are improved, the interpretability and robustness of the model are enhanced, the water resource utilization efficiency and the intelligent management level are effectively improved, and the method has wide application and popularization values.

Description

Technical Field

[0001] The present invention relates to the field of intelligent agricultural technology, and in particular to an integrated system and method for intelligent precision irrigation and nutrient regulation of rice. Background Art

[0002] As one of the world's major food crops, efficient and precise irrigation and fertilization management are key factors in ensuring rice yield and quality. Traditional rice irrigation methods rely heavily on manual experience, with extensive control of irrigation timing and fertilization ratios, making them difficult to adapt to the actual needs of different soil types, climate changes, and different rice growth stages. In recent years, with the development of sensor technology and artificial intelligence, "smart agriculture" in rice cultivation has gradually emerged, and some systems have achieved automated irrigation operations based on environmental parameters. However, existing technologies generally have the following problems: First, environmental data collection methods are limited to a single modality, such as surface temperature, humidity, and light, and are slow to respond to changes in the rhizosphere environment, making it difficult to accurately identify the true physiological state of rice. Second, existing intelligent irrigation systems rely heavily on cloud computing for data processing, which suffers from high latency and poor stability, making it difficult to operate continuously in the event of a network outage or instability. Furthermore, traditional AI models are mostly "black box structures" that lack explainability and ignore the physiological mechanisms of rice growth, making them difficult to adapt to cross-regional and cross-year promotion and application. Summary of the Invention

[0003] The main purpose of the present invention is to provide an integrated system and method for intelligent precision irrigation and nutrient regulation of rice, which can achieve precise, efficient and sustainable intelligent irrigation operations while improving the scientific nature of the strategy and the robustness of the system.

[0004] To achieve the above object, the technical solution adopted by the present invention is: An integrated system for intelligent precision irrigation and nutrient regulation of rice, comprising: A multimodal sensing module for acquiring information about the rice growing environment, comprising an underground nanosensor array for collecting rhizosphere pH, redox potential, and nitrate nitrogen concentration at different coatings, and a canopy multispectral imager for acquiring spectral information and leaf area index of rice leaves; An edge computing module, connected to the multimodal perception module, is used to perform spatial-temporal alignment and fusion processing on underground and canopy data to generate a standardized feature vector; an intelligent decision-making module, in communication with the edge computing module, configured to output a water and fertilizer irrigation decision based on the standardized vector generated by the edge computing module; A water-fertilizer coordinated execution module, connected to the intelligent decision-making module, for executing water-fertilizer integrated operations according to the generated irrigation decision; The monitoring module is used to ensure stable system operation. It specifically includes a displacement detection unit based on ZigBee signal strength analysis and an anomaly recognition unit based on waveform similarity. The displacement detection unit is used to determine whether the sensor has displaced and trigger an alarm if the displacement exceeds 10 cm. The anomaly recognition unit is used to continuously monitor sensor data and issue an anomaly alert when the similarity is lower than 0.85.

[0005] Preferably, the underground nanosensor array includes a three-layer buried depth structure, which is respectively arranged at depths of 15 cm, 30 cm and 50 cm, for realizing multi-level soil information collection.

[0006] Preferably, the edge computing module has a nighttime data encryption upload function, which is used to encrypt the locally collected data and upload it to the cloud between 00:00 and 04:00 every day. The cloud training generates incremental model parameters and sends them to the edge nodes to realize model fusion update.

[0007] Preferably, the intelligent decision-making module includes a leaf age development mechanism model constructed based on the thermal time equation, a long short-term memory network AI compensation model based on transfer learning optimization, and a model fusion unit for weighted fusion of the mechanism model output and the AI ​​model output to form a final irrigation strategy value.

[0008] Preferably, the water-fertilizer coordinated execution module includes a pulse injection device with pressure-flow decoupling control capability, and the injection frequency of the pulse injection device satisfies:

[0009] in, is the pressure difference at both ends of the nozzle, is the density of fertilizer solution, is the adjustment coefficient, ∈[5,50].

[0010] Preferably, the edge computing module and the multimodal perception module exchange data through a local wireless communication protocol, and the communication delay is less than 50 milliseconds.

[0011] A method for integrating intelligent precision irrigation and nutrient regulation of rice, comprising the following steps: Step 1: Multimodal data acquisition: collecting rice rhizosphere and canopy status information through underground nanosensor arrays and canopy multispectral imagers to form an original data set; Step 2: The collected data is pre-processed and aligned by the edge node, and then input into the mechanism model and AI compensation model to obtain output values. The final irrigation strategy is generated through the following weighted fusion algorithm: ; in, This variable represents the weighted coefficient of model fusion, is the sensitive factor, is the output of the mechanism model, is the output of the AI ​​model, The irrigation strategy is finally generated; Step 3: Send the irrigation strategy y to the water-fertilizer coordinated execution module to control the pulse jet device to complete the precise irrigation operation of the target area.

[0012] Preferably, in step 2, the time interval for collecting the multimodal data does not exceed 10 minutes, and the sensing frequency is automatically adjusted according to the solar cycle to avoid imaging errors.

[0013] Preferably, after the daily perception-execution cycle, the system automatically archives the key indicator change trajectory and uses it to generate the feature template required for nightly incremental model training.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention integrates underground nanosensors and canopy multispectral imaging equipment to build a multimodal data perception system. By combining edge computing with a dynamic model sharding mechanism, it effectively improves the collection accuracy and processing efficiency of rice rhizosphere and leaf environmental information, and significantly enhances the response speed and adaptability of irrigation strategies.

[0015] 2. The fusion design of the mechanism model and the AI ​​model in this invention not only improves the interpretability of decision-making, but also dynamically balances the outputs of the two types of models through an adaptive weighting mechanism, thereby enhancing the robustness of the model to different growth conditions and abnormal situations, thereby ensuring the long-term stable operation of the system in complex farmland environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram of the workflow of the present invention. DETAILED DESCRIPTION

[0017] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0018] like Figure 1 As shown, the present invention discloses an integrated system for intelligent precision irrigation and nutrient regulation of rice, which integrates modules such as multimodal perception, edge computing, mechanism and AI fusion decision-making, and water-fertilizer coordinated execution. The specific implementation methods are as follows: (1) Multimodal perception module In a typical rice-growing area, an underground nanosensor array and a canopy multispectral imager were systematically deployed.

[0019] Underground sensor arrays: Sensors are installed at depths of 15cm, 30cm, and 50cm in the field to monitor the pH, redox potential, and nitrate concentration in the rice rhizosphere in real time. For example, before irrigation, the soil pH may be between 5.8 and 6.2, the nitrate concentration may be 2.5ppm, and the redox potential may be +200mV. This information influences decisions about irrigation and fertilization rates.

[0020] A canopy multispectral imager, installed at a constant height in the field, periodically collects multispectral images of rice leaves. By analyzing the Normalized Difference Vegetation Index (NDVI), PRI (Photosynthetic Reflectance Index), and WBI (Water Status Index), it provides real-time information on rice growth and water status. At a specific moment, an NDVI of 0.85, a PRI of 0.75, and a WBI of 0.65 indicate healthy rice growth in the area. The system then adjusts irrigation strategies based on this data.

[0021] (2) Edge computing module The edge computing module is installed in the field control center and connected to the perception module via a local wireless communication protocol. This node performs data fusion and AI model calculations. Its specific functions are as follows: Model dynamic sharding unit: In this embodiment, the edge computing module divides the AI ​​model into two submodules: a feature extraction submodule (M1) and a decision generation submodule (M2). For example, the feature extraction module first extracts soil data and canopy image data from underground sensors and converts them into standardized feature vectors. This vector is then input into the decision generation submodule, which generates a preliminary irrigation strategy based on the optimized AI model.

[0022] Real-time Memory Scheduler: If the sensor data received by the node primarily comes from canopy images, the real-time memory scheduler loads the image processing module. If the data comes from underground sensors, the real-time memory scheduler loads the soil data processing module. For example, if the soil moisture data in a measurement is low, the system will prioritize loading the underground sensor data processing submodule to optimize irrigation.

[0023] Data Preprocessing and Synchronization Module: This module spatially and temporally aligns underground and canopy data. Specifically, underground data is collected every 10 minutes, while canopy data is collected every 5 minutes. The system adjusts the canopy data collection frequency based on solar radiation intensity, reducing the frequency in strong light conditions to avoid overexposure.

[0024] (3) Intelligent decision-making module In this embodiment, the intelligent decision-making module is used to generate irrigation strategies, combining the mechanism model with the AI ​​model: Mechanism Model: The leaf development mechanism model calculates the growth progress of rice based on the thermo-time equation. If the leaf development model indicates that the rice is entering a period of high irrigation demand, the strategy value will be increased accordingly.

[0025] LSTM compensation model: This model uses historical data to learn the nonlinear patterns of rice growth and makes real-time adjustments. The LSTM model dynamically adjusts its strategy based on current and historical data. For example, if an area was previously over-irrigated, the LSTM model will automatically reduce water and fertilizer supply in that area.

[0026] Model fusion: According to the formula: ; in, This variable represents the weighted coefficient of model fusion, is the sensitive factor, is the output of the mechanism model, is the output of the AI ​​model, The final irrigation strategy is generated.

[0027] For example, when the output difference between the mechanistic model and the AI ​​model is small, ω is close to 0.5, indicating that the two have similar contributions. However, when the output difference between the two is large (for example, the AI ​​model's prediction is more accurate), the value of ω will be smaller, indicating that the AI ​​model has a greater contribution.

[0028] (4) Water and fertilizer coordinated execution module This module implements integrated irrigation and fertilization operations using a pulsed injection device. In practice, the system controls the operating state of the injection device based on the calculated irrigation strategy value y. For example, when the irrigation strategy value is high, the injection frequency and flow rate will be increased to meet the high demand of rice; conversely, the injection frequency and flow rate will be reduced accordingly.

[0029] For example, if the pressure difference of the nozzle is ΔP=0.5bar, the density of the fertilizer liquid is ρ=1.2g / cm 3 , according to the formula:

[0030] The spray device will spray water and fertilizer at a frequency of 7.2 times per second.

[0031] (5) Monitoring module To ensure the stability of the system, the trusted monitoring module implements the following functions: Displacement detection: If a sensor experiences a displacement of more than 10 cm, the ZigBee signal analysis module will immediately issue an alarm and locate the problem area.

[0032] Waveform similarity monitoring: When the system finds that the similarity between the data curve of the rhizosphere sensor and the historical waveform is lower than 0.85, it will automatically issue an abnormal reminder to timely detect and correct equipment failures or data anomalies.

[0033] 2. Irrigation method process description In combination with the above system structure, the intelligent irrigation method of the present invention is mainly implemented through the following steps: Step S201: Multimodal data acquisition In practice, the sensing module collects real-time environmental data from the rice rhizosphere and canopy. Multimodal data is collected every 10 minutes, and the sensing frequency is automatically adjusted based on the daylight cycle to avoid imaging errors. For example, the system collects soil pH, nitrate nitrogen concentration, and redox potential every 10 minutes, while simultaneously capturing spectral images of canopy leaves every 5 minutes. Depending on sunlight intensity, the system may increase the imaging frequency during the day and decrease it at night.

[0034] Step S202: Data fusion and strategy generation After the collected raw data is sent to the edge computing module, the node performs data preprocessing and spatial and temporal alignment. The data is converted into a standardized feature vector and input into the mechanism model and AI model. For example, if the policy value output by the mechanism model is y_p=70 and the policy value output by the AI ​​model is y_a=60, then the formula is:

[0035] Calculate ω and weight it to get the final policy value y. For example, if ω=0.55, the final policy is:

[0036] This value is the irrigation strategy implemented by the system.

[0037] Step S203: Execute the irrigation strategy. The value y is transmitted to the water-fertilizer coordination execution module, which controls the pulse jet device's operating state based on this value. If the value y is high, the system will increase the jet frequency and flow rate to ensure that the rice plants receive sufficient water and fertilizer during the growing season.

[0038] This invention also supports nighttime model updates and incremental learning: Every night from midnight to 4:00 AM, edge nodes compile the day's perception data and execution logs and upload them to the cloud. The cloud then updates the AI ​​model through incremental learning and transmits the latest parameters back to the edge nodes, ensuring the system can adapt to ever-changing environmental conditions.

[0039] After the daily perception-execution cycle, the system automatically archives the change trajectory of key indicators and uses it to generate feature templates required for nightly incremental model training.

[0040] In summary, the present invention effectively improves the accuracy, efficiency and system stability of intelligent rice irrigation by integrating multimodal perception, edge computing, mechanism and AI model fusion decision-making mechanism, as well as water and fertilizer coordinated execution module, and has important promotion significance.

[0041] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An integrated system for intelligent precision irrigation and nutrient regulation of rice, characterized in that: include: A multimodal sensing module for acquiring information about the rice growing environment, comprising an underground nanosensor array for collecting rhizosphere pH, redox potential, and nitrate nitrogen concentration at different coatings, and a canopy multispectral imager for acquiring spectral information and leaf area index of rice leaves; An edge computing module, connected to the multimodal perception module, is used to perform spatial-temporal alignment and fusion processing on underground and canopy data to generate a standardized feature vector; an intelligent decision-making module, in communication with the edge computing module, configured to output a water and fertilizer irrigation decision based on the standardized vector generated by the edge computing module; A water-fertilizer coordinated execution module, connected to the intelligent decision-making module, for executing water-fertilizer integrated operations according to the generated irrigation decision; The monitoring module is used to ensure stable system operation. It specifically includes a displacement detection unit based on ZigBee signal strength analysis and an anomaly recognition unit based on waveform similarity. The displacement detection unit is used to determine whether the sensor has displaced and trigger an alarm if the displacement exceeds 10 cm. The anomaly recognition unit is used to continuously monitor sensor data and issue an anomaly alert when the similarity is lower than 0.

85.

2. The integrated intelligent precision irrigation and nutrient regulation system for rice according to claim 1, characterized in that: The underground nanosensor array includes a three-layer buried depth structure, which is respectively set at depths of 15 cm, 30 cm and 50 cm, and is used to realize multi-level soil information collection.

3. The integrated intelligent precision irrigation and nutrient regulation system for rice according to claim 1, characterized in that: The edge computing module has a nighttime data encryption upload function, which is used to encrypt the locally collected data and upload it to the cloud between 00:00 and 04:00 every day. The cloud-based training generates incremental model parameters and sends them to the edge nodes to achieve model fusion updates.

4. The integrated intelligent precision irrigation and nutrient regulation system for rice according to claim 1, characterized in that: The intelligent decision-making module includes a leaf age development mechanism model constructed based on the thermal time equation, a long short-term memory network AI compensation model based on transfer learning optimization, and a model fusion unit for weighted fusion of the mechanism model output and the AI ​​model output to form a final irrigation strategy value.

5. The integrated intelligent precision irrigation and nutrient regulation system for rice according to claim 1 is characterized by: The water-fertilizer coordinated execution module includes a pulse injection device with pressure-flow decoupling control capability, and the injection frequency of the pulse injection device satisfies: ; in, is the pressure difference at both ends of the nozzle, is the density of fertilizer solution, is the adjustment coefficient, ∈[5,50].

6. The integrated intelligent precision irrigation and nutrient regulation system for rice according to claim 1, characterized in that: The edge computing module and the multimodal perception module exchange data through a local wireless communication protocol, and the communication delay is less than 50 milliseconds.

7. A method for integrating intelligent precision irrigation and nutrient regulation of rice, controlled by the system according to any one of claims 1 to 6, characterized in that , including the following steps: Step 1: Multimodal data acquisition: collecting rice rhizosphere and canopy status information through underground nanosensor arrays and canopy multispectral imagers to form an original data set; Step 2: The collected data is pre-processed and aligned by the edge node, and then input into the mechanism model and AI compensation model to obtain output values. The final irrigation strategy is generated through the following weighted fusion algorithm: ； ; in, This variable represents the weighted coefficient of model fusion, is the sensitive factor, is the output of the mechanism model, is the output of the AI ​​model, The irrigation strategy is finally generated; Step 3: Send the irrigation strategy y to the water-fertilizer coordinated execution module to control the pulse jet device to complete the precise irrigation operation of the target area.

8. The integrated system for intelligent precision irrigation and nutrient regulation of rice according to claim 7, characterized in that: In step 2, the time interval for collecting the multimodal data does not exceed 10 minutes, and the sensing frequency is automatically adjusted according to the solar cycle to avoid imaging errors.

9. The integrated system and method for intelligent precision irrigation and nutrient regulation of rice according to claim 7, characterized in that: After the daily perception-execution cycle, the system automatically archives the change trajectory of key indicators and uses it to generate feature templates required for nightly incremental model training.

Images