Dynamic prediction of grain crop water requirement based on multi-source sensing and intelligent irrigation system

Abstract

This invention discloses a multi-source sensing-based dynamic water requirement prediction and intelligent irrigation system for grain and cash crops, belonging to the field of agricultural intelligent irrigation technology. It includes a multi-source sensing data acquisition layer, an edge computing processing layer, a dynamic water requirement prediction layer, an intelligent irrigation control layer, and a data storage and interaction layer. The acquisition layer uses multiple types of sensors, such as quantum dot fluorescent probe arrays, to collect multi-dimensional data on soil, crops, and the environment. The edge layer performs noise reduction, fusion, and real-time processing on the data. The prediction layer uses a transfer learning multimodal fusion model, combined with a quantum tunneling effect correction factor, to correct traditional formulas and achieve hourly-scale water requirement prediction. The control layer drives a magnetorheological actuator to complete precise irrigation. The storage and interaction layer enables secure data storage, traceability, and visualization. This invention improves the accuracy and real-time performance of water requirement prediction, enables precise irrigation on demand, improves water resource utilization, reduces labor costs, and ensures high-quality and high-yield grain and cash crops, making it suitable for large-scale planting scenarios.

Description

Technical Field

[0001] This invention relates to the field of intelligent agricultural irrigation technology, specifically to a dynamic prediction and intelligent irrigation system for water demand of grain and cash crops based on multi-source sensing. Background Technology

[0002] Grain and cash crops are the core components of my country's agricultural production. Their water requirements during growth are significantly dynamic, influenced by a combination of factors such as soil moisture, weather conditions, crop growth period, and variety characteristics. Accurately understanding the water requirements of grain and cash crops and implementing irrigation on demand is key to ensuring high-quality and high-yield crops, conserving agricultural water resources, and promoting green and sustainable agricultural development.

[0003] Existing intelligent irrigation systems are mostly designed for single crops and cannot adapt to the different water requirements of various grain and cash crops, such as wheat, corn, and cotton. Furthermore, the water requirement prediction models have poor generalization ability, with prediction accuracy decreasing significantly under different regions and soil conditions, making large-scale application difficult. Therefore, developing an intelligent irrigation system that can accurately sense the dynamic water requirements of grain and cash crops, achieve precise irrigation on demand, and is highly adaptable and reliable has become an urgent need for the current development of intelligent agriculture. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic prediction and intelligent irrigation system for water demand of grain and cash crops based on multi-source sensing, so as to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: The system for dynamic prediction of water demand for grain and cash crops and intelligent irrigation based on multi-source sensing includes a multi-source sensing data acquisition layer, an edge computing processing layer, a dynamic water demand prediction layer, an intelligent irrigation control layer, and a data storage and interaction layer. Each layer is electrically connected in sequence to work together to realize dynamic prediction of water demand for grain and cash crops and intelligent irrigation. The multi-source sensor data acquisition layer includes a soil multi-dimensional sensing module, a crop physiological sensing module, an environmental parameter sensing module, and a drone inspection module, which are used to collect multi-dimensional monitoring data of grain and cash crops in all aspects during their growth process. The edge computing processing layer adopts a heterogeneous computing architecture, integrating FPGA and RISC-V processor, and is used to perform noise reduction, calibration and fusion processing on the raw data collected by the multi-source sensor data acquisition layer, and output standardized and highly reliable feature data. The water demand dynamic prediction layer incorporates a multimodal fusion computing model based on transfer learning. The model introduces a quantum tunneling effect correction factor to modify the traditional Penman-Monteith formula. It inputs the feature data output by the edge computing processing layer and outputs the dynamic water demand value of grain and cash crops at the hourly scale. The intelligent irrigation control layer includes an irrigation control center and a magnetorheological irrigation actuator. The irrigation control center receives the dynamic water demand value output by the dynamic water demand prediction layer, and combines it with the preset water demand thresholds for each growth stage of grain and cash crops to generate a pulse width modulation signal, which drives the magnetorheological irrigation actuator to adjust the irrigation flow and pressure to complete precise irrigation. The data storage and interaction layer adopts a combination of consortium blockchain and cloud storage to store multi-source monitoring data, water demand prediction data, irrigation control data and crop growth parameters. It also provides a visual interactive interface to enable data query, parameter setting and irrigation status monitoring.

[0006] A further improvement of the technical solution of the present invention is that the soil multidimensional sensing module adopts a quantum dot fluorescent probe array, which is composed of CdSe / ZnS core-shell structure quantum dots and mesoporous silica carrier, with an excitation wavelength of 365-405nm, an emission wavelength of 520-680nm, a spatial resolution of not less than 0.1mm, and the probe surface is modified with soil colloid-specific recognition groups.

[0007] The soil sensing unit uses a combination of a quantum dot fluorescent probe array and a soil moisture sensor. The quantum dot fluorescent probe array is composed of CdSe / ZnS core-shell quantum dots and mesoporous silica carrier. The probe surface is modified with soil colloid-specific recognition groups. The soil moisture sensor is embedded in the crop root zone at depths of 10cm, 20cm, and 30cm. The quantum dot fluorescent probe array is embedded at different depths in the root zone of grain and cash crops. Soil moisture data is obtained through the nonlinear mapping relationship between fluorescence intensity decay rate and soil water activity. Simultaneously, real-time temperature data collected by a temperature sensor is used to calibrate the soil moisture data using a temperature compensation formula, which is: ,in To compensate for the reduced moisture content, These are the original measured values. For temperature coefficient, For real-time temperature, For temperature calibration.

[0008] A further improvement of the technical solution of the present invention is that: the crop physiological sensing module includes a terahertz time-domain spectroscopy system, the time resolution of the terahertz time-domain spectroscopy system is ≤5fs, which is used to scan the canopy leaves of grain and cash crops to obtain the intensity of characteristic absorption peaks in the 0.3-3THz frequency band.

[0009] A further improvement of the technical solution of the present invention is that the environmental parameter sensing module adopts an environmental field monitoring module based on millimeter-wave radar, which operates in the 77GHz frequency band and is used to collect real-time wind speed, atmospheric pressure and solar radiation flux density of the growing environment of grain and cash crops.

[0010] A further improvement of the technical solution of the present invention is that: the UAV inspection module is equipped with a terahertz imager to perform a large-scale scan of the canopy of grain and cash crops, and the scan results are fused with the measurement results of the ground terahertz time-domain spectral system to optimize the spatial distribution accuracy of crop physiological water demand signals.

[0011] A further improvement of the technical solution of the present invention is that the frequency of the drone inspection module can be preset, ranging from 1 to 7 days / time. During the inspection, the crop canopy coverage and plant height data are collected in real time and uploaded to the edge computing processing layer to assist in correcting the parameters of the water demand prediction model.

[0012] A further improvement of the technical solution of the present invention is that: the multimodal fusion computing model based on transfer learning uses source domain data from a dataset of water requirements of various grain and cash crops under a controlled laboratory environment, and the target domain is data from the actual field growth environment. Distribution alignment is achieved through an adversarial domain adaptation network.

[0013] A further improvement of the technical solution of the present invention is that the magnetorheological irrigation actuator includes a ring electromagnetic coil and a deformable valve core, the coil current adjustment range is 0-2A, the corresponding irrigation flow adjustment range is 0-50L / h, and the flow control accuracy is ≤±2%FS.

[0014] A further improvement of the technical solution of the present invention is that: the consortium blockchain architecture of the data storage and interaction layer contains more than 3 consensus nodes, which are deployed in the meteorological station, soil monitoring terminal and crop growth monitoring terminal respectively. The consensus mechanism adopts the practical Byzantine fault-tolerant algorithm, and the block generation interval is ≤30s. It is used to perform real-time hash verification on the original data and intermediate results in the calculation process to ensure that the data cannot be tampered with.

[0015] A further improvement of the technical solution of the present invention is that it also includes a fault diagnosis module, which is electrically connected to the multi-source sensor data acquisition layer and the intelligent irrigation control layer. It is used to monitor the operating status of the sensors and irrigation actuators in real time. When a fault is detected, it immediately issues an alarm signal and uploads the fault information to the data storage and interaction layer. At the same time, it automatically switches to the backup irrigation mode to ensure that the irrigation work continues.

[0016] Due to the adoption of the above technical solution, the technical progress achieved by this invention compared to the prior art is as follows: 1. This invention provides a dynamic prediction and intelligent irrigation system for water demand of grain and cash crops based on multi-source sensing. It adopts multi-source sensing technology to collect multi-dimensional data of soil, crops and environment in all aspects. Combined with a multi-modal fusion model based on transfer learning and a quantum tunneling effect correction factor, it can realize the dynamic prediction of water demand of grain and cash crops at the hour scale with a prediction error of ≤3% and the dynamic response speed is improved to the hour level. It solves the problems of staticity and slow response of traditional models and can accurately capture changes in crop water demand.

[0017] 2. This invention provides a dynamic prediction and intelligent irrigation system for water demand of grain and cash crops based on multi-source sensing. It achieves precise regulation of irrigation flow through a magnetorheological irrigation actuator with a flow control accuracy of ≤±2%FS. Combined with the dynamic prediction results of water demand, it realizes "on-demand irrigation and precise irrigation", avoiding over-irrigation and under-irrigation. Experimental verification shows that the water saving rate exceeds 30%, significantly improving the water resource utilization rate.

[0018] 3. This invention provides a dynamic prediction and intelligent irrigation system for water demand of grain and cash crops based on multi-source sensing. Based on a multimodal fusion model of transfer learning, it can adapt to the different water demand differences of different grain and cash crops, such as wheat, corn and cotton. At the same time, it achieves the distribution alignment of source domain and target domain data through an adversarial domain adaptation network, adapting to planting scenarios in different regions and under different soil conditions, and can be promoted and applied on a large scale.

[0019] 4. This invention provides a dynamic prediction and intelligent irrigation system for water demand of grain and cash crops based on multi-source sensing. The system realizes full automation from data acquisition and water demand prediction to irrigation control, eliminating the need for frequent manual inspections and operations. It also supports remote monitoring and management, which can reduce the amount of manual work by more than 80% and reduce labor costs. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the overall system architecture of the present invention; Figure 2 This is a flowchart of the multi-source sensor data acquisition and processing of the present invention; Figure 3 This is a flowchart of the dynamic prediction process for water requirements of grain and cash crops according to the present invention; Figure 4 This is a flowchart of the intelligent irrigation control process of the present invention. Detailed Implementation

[0021] The present invention will be further described in detail below with reference to embodiments:

[0022] Example 1 A multi-source sensing-based dynamic water demand prediction and intelligent irrigation system for grain and cash crops includes a multi-source sensing data acquisition layer, an edge computing processing layer, a dynamic water demand prediction layer, an intelligent irrigation control layer, and a data storage and interaction layer. Each layer is electrically connected in sequence to collaboratively realize dynamic water demand prediction and intelligent irrigation for grain and cash crops.

[0023] Multi-source sensor data acquisition layer The multi-source sensor data acquisition layer is the data foundation of the system. It is used to collect multi-dimensional monitoring data on soil, crops, and environment during the growth process of grain and cash crops in an all-round and high-precision manner, providing data support for water demand prediction and irrigation control. Specifically, it includes a multi-dimensional soil sensing module, a crop physiological sensing module, an environmental parameter sensing module, and a drone inspection module. Edge computing processing layer The edge computing processing layer adopts a heterogeneous computing architecture, integrating FPGA and RISC-V processors. It is used to perform noise reduction, calibration and fusion processing on the raw data collected by the multi-source sensor data acquisition layer, and output standardized and highly reliable feature data to avoid data redundancy and interference. This provides high-quality data input for dynamic water demand prediction, while reducing data transmission pressure and improving system response speed. The multimodal fusion computing model based on transfer learning uses source domain data from water requirement datasets of various grain and cash crops (wheat, corn, cotton, rapeseed, etc.) under controlled laboratory conditions. This dataset includes water requirement data for crops under different soil types, environmental conditions, and growth stages. The target domain is actual field growth environment data. An adversarial domain adaptation network is used to align the distribution of source domain data with target domain data, thereby improving the model's generalization ability and enabling it to adapt to the water requirement prediction needs of grain and cash crops under different regions and soil conditions. To address the issue that the traditional Penman-Monteith formula does not consider soil porosity and crop variety specificity, the model introduces a quantum tunneling effect correction factor to correct for it. The formula for calculating the quantum tunneling effect correction factor is as follows: ,in This is a correction factor for the quantum tunneling effect. This is the soil porosity correction coefficient, determined according to soil type: 0.8-0.9 for sandy soil, 0.9-1.0 for loam, and 1.0-1.1 for clay. For grain and cash crop variety-specific parameters, wheat was set at 0.95-1.05, corn at 1.0-1.1, and cotton at 1.05-1.15. E represents the real-time monitored soil water potential. This is the baseline water potential value.

[0024] Water Demand Dynamic Prediction Layer The water demand dynamic prediction layer is the core of the system. It has a built-in multimodal fusion computing model based on transfer learning to realize the hourly dynamic prediction of water demand for grain and cash crops, and solves the problems of staticity, slow response and poor generalization ability of traditional models. Intelligent irrigation control layer The intelligent irrigation control layer is used to drive the irrigation actuator to complete precise irrigation based on the dynamic prediction results of water demand and the crop growth requirements, so as to achieve "on-demand irrigation and precise irrigation". Specifically, it includes the irrigation control center and the magnetorheological irrigation actuator. Data storage and interaction layer The data storage and interaction layer is used to realize secure data storage, traceability and visualization interaction, and provides users with a convenient operation and monitoring interface. Specifically, it adopts a combination of consortium blockchain and cloud storage. Fault diagnosis module The system also includes a fault diagnosis module, which is electrically connected to the multi-source sensor data acquisition layer and the intelligent irrigation control layer. This module is used to monitor the operating status of sensors and irrigation actuators in real time. When a sensor or irrigation actuator malfunction is detected, an alarm signal is immediately issued (both audible and SMS alerts). The fault information, including fault type, location, and time, is uploaded to the data storage and interaction layer. Simultaneously, the system automatically switches to a backup irrigation mode, such as activating backup sensors or manual irrigation mode, to ensure continuous irrigation and prevent crop water shortages or over-irrigation due to malfunctions.

[0025] Example 2 In this embodiment, wheat is selected as the grain and cash crop, and the planting area is a wheat planting base in the North China Plain. The soil type is loam, and the planting area is 100 mu. The specific system configuration is as follows: 1. Multi-source sensor data acquisition layer: (1) Soil multidimensional sensing module: A quantum dot fluorescent probe array is used, with an excitation wavelength of 380nm, an emission wavelength of 600nm, and a spatial resolution of 0.1mm. The probe surface is modified with soil colloid-specific recognition groups and embedded in the wheat root zone at depths of 20cm, 30cm, and 40cm. One set is arranged for every 5 acres, for a total of 20 sets. The temperature sensor uses a PT100 temperature sensor with a measurement range of -20℃ to 60℃ and an accuracy of ±0.1℃. It is arranged synchronously with the quantum dot fluorescent probe array for temperature compensation of soil moisture data.

[0026] (2) Crop physiological sensing module: A terahertz time-domain spectroscopy system with a time resolution of 5 fs was used to scan wheat canopy leaves at a scanning frequency of 30 minutes / time to obtain the characteristic absorption peak intensity in the 0.3-3 THz frequency band. A quantitative relationship model between the terahertz absorption coefficient and the osmotic pressure of leaf cell sap was established. The input features include 128 terahertz absorption peak characteristic values. The intercellular water transport rate was calculated by the terahertz wave propagation delay difference.

[0027] (3) Environmental parameter sensing module: A 77GHz millimeter-wave radar environmental field monitoring module is used to collect real-time wind speed, atmospheric pressure and solar radiation flux density at a frequency of 30 minutes / time; an integrated temperature and humidity sensor (measurement range: 0~100%RH, -20℃~60℃, accuracy ±0.5℃, ±2%RH), light sensor (measurement range: 0~200000lux, accuracy ±5%), and rainfall sensor (measurement range: 0~4mm / min, accuracy ±0.1mm) are installed at a frequency of 30 minutes / time, with one set placed at the center of the planting base.

[0028] (4) Unmanned aerial vehicle (UAV) inspection module: A multi-rotor UAV is selected, equipped with a terahertz imager. The inspection frequency is 3 days / time. The inspection range covers the entire planting base. Data on wheat canopy coverage and plant height are collected and uploaded to the edge computing processing layer simultaneously.

[0029] 2. Edge computing processing layer: Adopting a heterogeneous computing architecture, integrating an FPGA (Xilinx Zynq-7000) and a RISC-V processor (RV32IMAC). The FPGA is used for real-time Fourier transform of terahertz spectral data, and the RISC-V processor runs data fusion algorithms and preprocessing programs. The data processing latency is ≤10ms. Data preprocessing adopts wavelet threshold denoising algorithm, linear interpolation method and normalization processing. The weights of the weighted fusion algorithm are set as follows: soil moisture data 0.4, crop physiological data 0.3, and environmental parameter data 0.3.

[0030] 3. Water Demand Dynamic Prediction Layer: Based on a multimodal fusion computing model of transfer learning, the source domain data is the water demand dataset of wheat at different growth stages under controlled laboratory conditions, and the target domain is the actual field data of the planting base. Distribution alignment is achieved through an adversarial domain adaptation network. In the calculation of the quantum tunneling effect correction factor, α is 0.95 (soil), β is 1.0 (wheat), and E0 is -10kPa.

[0031] 4. Intelligent Irrigation Control Layer: The irrigation control center adopts an STM32H743 microcontroller, which has built-in water requirement thresholds for each growth stage of wheat: 60%-70% relative soil moisture content during the jointing stage and 70%-80% relative soil moisture content during the grain-filling stage. The magnetorheological irrigation actuator includes a ring electromagnetic coil and a deformable valve core, with an irrigation flow rate adjustment range of 0-50L / h and a flow control accuracy of ±2%FS. One set of irrigation actuators is arranged for every 2 acres, for a total of 50 sets. The built-in nanometer-level flow sensor provides real-time feedback of irrigation flow data, forming a closed-loop control.

[0032] 5. Data Storage and Interaction Layer: The consortium blockchain architecture includes three consensus nodes, deployed at a weather station, a soil monitoring terminal, and a crop growth monitoring terminal, respectively. The consensus mechanism adopts a practical Byzantine fault-tolerant algorithm with a block generation interval of 30 seconds. The cloud storage module uses Alibaba Cloud distributed storage with a storage period of 5 years. The visual interactive interface supports access from PC and mobile APP, displaying real-time monitoring data, water demand prediction curves, irrigation status, and other information.

[0033] 6. Fault Diagnosis Module: Employs an STM32F103 microcontroller to monitor the operating status of sensors and irrigation actuators in real time. When a fault is detected, it issues an audible alarm and notifies the administrator via SMS, while automatically switching to standby irrigation mode.

[0034] The working process of this embodiment is as follows: (1) Data acquisition: The sensors of the multi-source sensor data acquisition layer and the UAV synchronously collect soil moisture, temperature, crop physiological signals, environmental parameters and canopy data in the wheat root zone. The acquisition frequency is 30 minutes / time, and the UAV inspection is carried out once every 3 days.

[0035] (2) Data processing: The edge computing processing layer performs noise reduction, calibration and fusion processing on the collected raw data, removes noise interference, supplements missing data, and outputs standardized feature data after normalization processing. The processing delay is ≤10ms.

[0036] (3) Water demand prediction: The water demand dynamic prediction layer inputs standardized feature data into the multimodal fusion computing model based on transfer learning, and outputs the dynamic water demand value of wheat on an hourly scale through the modified Penman-Monteith formula, with a prediction error of ≤3%.

[0037] (4) Intelligent irrigation: The irrigation control center compares the predicted dynamic water demand value with the water demand threshold of the current growth stage of wheat. When the water demand is higher than the upper limit of the threshold, a pulse width modulation signal is generated to drive the magnetorheological irrigation actuator to start irrigation, adjust the irrigation flow and duration, and ensure that the irrigation amount matches the water demand. When the water demand is lower than the lower limit of the threshold, irrigation is suspended. During the irrigation process, the nanometer-level flow sensor feeds back the irrigation flow data in real time to form a closed-loop control.

[0038] (5) Data storage and interaction: All collected data, forecast data, and irrigation control data are uploaded to the data storage and interaction layer and hashed through the consortium blockchain to ensure that the data is tamper-proof. Users can query data, monitor irrigation status, and set parameters through PC or mobile APP.

[0039] (6) Fault diagnosis: The fault diagnosis module monitors the system operation status in real time. When a fault is detected, it immediately issues an alarm signal and switches to the standby irrigation mode to ensure that irrigation work continues.

[0040] Experimental verification shows that after the system of this embodiment is applied in wheat planting base, the water demand prediction error is ≤3%, the water saving rate reaches 32%, the wheat yield increases by 8%, and the labor cost is reduced by 85%, which significantly improves the intelligent level of wheat planting and water resource utilization, and realizes high-quality and high-yield wheat.

[0041] The present invention has been described in detail above. However, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, any modifications or improvements that do not depart from the spirit of the present invention are within the scope of protection of the present invention.

Claims

1. A dynamic prediction and intelligent irrigation system for water demand of grain and cash crops based on multi-source sensing, characterized in that: It includes a multi-source sensor data acquisition layer, an edge computing processing layer, a water demand dynamic prediction layer, an intelligent irrigation control layer, and a data storage and interaction layer. Each layer is electrically connected in sequence to work together to realize dynamic prediction of water demand for grain and cash crops and intelligent irrigation. The multi-source sensor data acquisition layer includes a soil multi-dimensional sensing module, a crop physiological sensing module, an environmental parameter sensing module, and a drone inspection module, which are used to collect multi-dimensional monitoring data of grain and cash crops in all aspects during their growth process. The edge computing processing layer adopts a heterogeneous computing architecture, integrating FPGA and RISC-V processor, and is used to perform noise reduction, calibration and fusion processing on the raw data collected by the multi-source sensor data acquisition layer, and output standardized and highly reliable feature data. The water demand dynamic prediction layer incorporates a multimodal fusion computing model based on transfer learning. The model introduces a quantum tunneling effect correction factor to modify the traditional Penman-Monteith formula. It inputs the feature data output by the edge computing processing layer and outputs the dynamic water demand value of grain and cash crops at the hourly scale. The intelligent irrigation control layer includes an irrigation control center and a magnetorheological irrigation actuator. The irrigation control center receives the dynamic water demand value output by the dynamic water demand prediction layer, and combines it with the preset water demand thresholds for each growth stage of grain and cash crops to generate a pulse width modulation signal, which drives the magnetorheological irrigation actuator to adjust the irrigation flow and pressure to complete precise irrigation. The data storage and interaction layer adopts a combination of consortium blockchain and cloud storage to store multi-source monitoring data, water demand prediction data, irrigation control data and crop growth parameters. It also provides a visual interactive interface to enable data query, parameter setting and irrigation status monitoring.

2. The intelligent irrigation system for dynamic water demand prediction of grain and cash crops based on multi-source sensing according to claim 1, characterized in that: The soil multidimensional sensing module uses a quantum dot fluorescent probe array, which is composed of CdSe / ZnS core-shell quantum dots and mesoporous silica carrier. The excitation wavelength is 365-405nm, the emission wavelength is 520-680nm, the spatial resolution is not less than 0.1mm, and the probe surface is modified with soil colloid-specific recognition groups. The soil sensing unit uses a combination of a quantum dot fluorescent probe array and a soil moisture sensor. The quantum dot fluorescent probe array is composed of CdSe / ZnS core-shell quantum dots and mesoporous silica carrier. The probe surface is modified with soil colloid-specific recognition groups. The soil moisture sensor is embedded in the crop root zone at depths of 10cm, 20cm, and 30cm. The quantum dot fluorescent probe array is embedded at different depths in the root zone of grain and cash crops. Soil moisture data is obtained through the nonlinear mapping relationship between fluorescence intensity decay rate and soil water activity. Simultaneously, real-time temperature data collected by a temperature sensor is used to calibrate the soil moisture data using a temperature compensation formula, which is: ,in To compensate for the reduced moisture content, These are the original measured values. For temperature coefficient, For real-time temperature, For temperature calibration.

3. The water demand dynamic prediction and intelligent irrigation system for grain and cash crops based on multi-source sensing according to claim 1, characterized in that: The crop physiological sensing module includes a terahertz time-domain spectroscopy system with a time resolution of ≤5 fs, which is used to scan the canopy leaves of grain and cash crops to obtain the intensity of characteristic absorption peaks in the 0.3-3 THz frequency band.

4. The intelligent irrigation system for dynamic water requirement prediction of grain and cash crops based on multi-source sensing according to claim 1, characterized in that: The environmental parameter sensing module adopts an environmental field monitoring module based on millimeter-wave radar, which operates in the 77GHz frequency band and is used to collect real-time wind speed, atmospheric pressure and solar radiation flux density of the growing environment of grain and cash crops.

5. The intelligent irrigation system for dynamic water demand prediction of grain and cash crops based on multi-source sensing according to claim 1, characterized in that: The drone inspection module is equipped with a terahertz imager to perform a large-scale scan of the canopy of grain and cash crops. The scan results are then fused with the measurement results of the ground-based terahertz time-domain spectral system to optimize the spatial distribution accuracy of crop physiological water demand signals.

6. The intelligent irrigation system for dynamic water demand prediction of grain and cash crops based on multi-source sensing according to claim 1, characterized in that: The frequency of the drone inspection module can be preset, ranging from 1 to 7 days / time. During the inspection, it collects crop canopy coverage and plant height data in real time and uploads them to the edge computing processing layer to help correct the parameters of the water demand prediction model.

7. The intelligent irrigation system for dynamic water demand prediction of grain and cash crops based on multi-source sensing according to claim 1, characterized in that: The multimodal fusion computing model based on transfer learning uses source domain data from water requirement datasets of various grain and cash crops under controlled laboratory conditions, and target domain data from actual field growth environments. Distribution alignment is achieved through an adversarial domain adaptation network.

8. The intelligent irrigation system for dynamic water demand prediction of grain and cash crops based on multi-source sensing according to claim 1, characterized in that: The magnetorheological irrigation actuator includes a ring-shaped electromagnetic coil and a deformable valve core. The coil current adjustment range is 0-2A, corresponding to an irrigation flow rate adjustment range of 0-50L / h, and the flow control accuracy is ≤±2%FS.

9. The intelligent irrigation system for dynamic prediction of water demand of grain and cash crops based on multi-source sensing according to claim 1, characterized in that: The consortium blockchain architecture of the data storage and interaction layer contains more than three consensus nodes, which are deployed at meteorological stations, soil monitoring terminals and crop growth monitoring terminals respectively. The consensus mechanism adopts a practical Byzantine fault-tolerant algorithm with a block generation interval of ≤30s, which is used to perform real-time hash verification on the original data and intermediate results in the calculation process to ensure that the data cannot be tampered with.

10. The intelligent irrigation system for dynamic water demand prediction of grain and cash crops based on multi-source sensing according to any one of claims 1-9, characterized in that: The system also includes a fault diagnosis module, which is electrically connected to the multi-source sensor data acquisition layer and the intelligent irrigation control layer. It is used to monitor the operating status of sensors and irrigation actuators in real time. When a fault is detected, an alarm signal is immediately issued and the fault information is uploaded to the data storage and interaction layer. At the same time, it automatically switches to the backup irrigation mode to ensure that irrigation work continues.

Images