Water and fertilizer integrated irrigation device and method based on Internet of Things and intelligent strategy

Abstract

The invention relates to the technical field of agricultural automation equipment, and discloses a water and fertilizer integrated irrigation device and method based on the Internet of Things and an intelligent strategy, and the device comprises a crop and environment sensing unit which obtains crop image information and environment parameters; the intelligent decision-making unit is used for generating a target irrigation strategy and fertilization concentration; the source water management unit is used for analyzing a water source before irrigation and providing water quality background information; the water-fertilizer synthesis and control unit is used for dynamically synthesizing and outputting a nutrient solution in real time by combining the target concentration and the water quality background; and the cloud management and optimization unit is used for collecting data of each unit and is used for storage, remote monitoring and management. According to the invention, an on-line spectrum analysis technology and a high-frequency pulse microfluidic technology are coupled, a real-time closed-loop control system for directly measuring chemical components is constructed, and high consistency of a digital decision instruction and the chemical components of a physically output nutrient solution is achieved.

Description

Technical Field

[0001] This invention relates to the field of agricultural automation equipment technology, specifically to an integrated water and fertilizer irrigation device and method based on the Internet of Things and intelligent strategies. Background Technology

[0002] In modern intensive agricultural production, especially in the greenhouse cultivation of high-value-added crops, the supply of water and nutrients is the core factor determining crop yield and quality. How to precisely control irrigation and fertilization processes to meet the dynamic nutritional needs of crops while maximizing the utilization rate of water and fertilizer resources is a key technological focus in this field.

[0003] Various automated water and fertilizer irrigation devices already exist in the current technology. These devices automate the execution of irrigation and fertilization, initiating irrigation at designated times according to user-preset programs and injecting fertilizer solution into the water flow by controlling the operation of fertilizer pumps. Some devices integrate conductivity (EC) and pH sensors, enabling basic monitoring and adjustment of the total salt concentration and pH of the mixed nutrient solution. This technology liberates agricultural producers from heavy, repetitive labor, achieving timed and standardized water and fertilizer management. Compared to traditional manual irrigation and fertilization methods, it has significant advantages in improving labor efficiency and achieving basic integrated water and fertilizer management.

[0004] However, existing technologies still have several fundamental limitations in pursuing higher levels of precision and intelligence. First, existing solutions generally ignore the dynamic changes in the source water background. The ionic composition of the source water is not constant, and its fluctuations directly introduce an unknown, systematic error in fertilizer application, causing the final delivered nutrient solution concentration to deviate from the target value. Second, the basis for fertilization decisions is slow to be obtained. The judgment of crop nutrient status either relies on highly subjective manual observation or time-consuming laboratory analysis. This information lag makes it impossible for fertilization adjustments to respond promptly to the actual needs of the crop. Third, the control precision of the fertilizer application process is insufficient. Most systems adopt open-loop control, only executing injection commands without verifying the final result. They cannot compensate for disturbances such as pump performance fluctuations or pipeline pressure changes, thus making it difficult to guarantee the accuracy of the concentration of each nutrient element delivered to the crop root zone. Finally, the system's operational health status lacks active monitoring methods. Minor blockages in drip irrigation pipelines and drippers are difficult to detect in the early stages, and the system cannot provide early warnings or self-maintenance. Problems are usually only discovered after the crop shows poor growth due to uneven water and fertilizer supply, resulting in irreversible losses. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an integrated water and fertilizer irrigation device and method based on the Internet of Things and intelligent strategies, in order to solve problems such as inaccurate fertilizer application due to fluctuations in the source water background, delayed diagnosis of crop nutrient deficiencies, lack of high-precision closed-loop control in the fertilizer application process, and inability to actively perceive the health status of the system.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an integrated water and fertilizer irrigation device and method based on the Internet of Things and intelligent strategies.

[0007] The first aspect of this invention provides an integrated water and fertilizer irrigation device based on the Internet of Things and intelligent strategies, comprising:

[0008] The crop and environment sensing unit is used to acquire crop image information and environmental parameters.

[0009] The intelligent decision-making unit is used to generate a target irrigation strategy and a target fertilizer concentration based on the crop image information output by the crop and environment sensing unit and the environmental parameters.

[0010] The source water management unit is used to perform water quality analysis on the irrigation water source before the target irrigation strategy is implemented, so as to obtain and provide water quality background information;

[0011] The water and fertilizer synthesis and control unit is used to respond to the target irrigation strategy and, based on the target fertilizer concentration output by the intelligent decision unit and the water quality background information provided by the source water management unit, dynamically synthesize and output the nutrient solution of the target concentration in real time, while generating operation data.

[0012] The cloud management and optimization unit is used to receive and store crop image information and environmental parameters from the crop and environment sensing unit, target irrigation strategies and target fertilizer concentrations from the intelligent decision-making unit, and operating data from the water and fertilizer synthesis and control unit, and to realize remote monitoring and management.

[0013] In one specific embodiment, the crop and environment sensing unit includes:

[0014] Environmental sensors are used to collect various environmental parameters in real time, including soil temperature and humidity, soil electrical conductivity, air temperature and humidity, and light intensity.

[0015] Image acquisition equipment, used to periodically acquire crop image information;

[0016] The image processing unit is equipped with a crop nutrient deficiency detection model. It receives crop image information as input, processes the input using the crop nutrient deficiency detection model, identifies visual features that characterize crop nutrient deficiency, and outputs structured crop nutrient diagnosis results.

[0017] In one specific embodiment, the intelligent decision-making unit includes:

[0018] The irrigation strategy decision module is equipped with a random forest decision model. Its inputs are the environmental parameters, the crop nutrition diagnosis results, the preset crop growth stage information and weather forecast data, and its outputs are the target irrigation start and stop times and the total irrigation amount.

[0019] The fertilization strategy decision-making module is equipped with a fertilizer application control model based on an improved genetic algorithm. This model calculates and outputs a target concentration vector c containing multiple nutrient elements based on the crop nutrient diagnosis results and preset crop nutrient requirements. target Its form is c target =[c N,t ,c P,t ,c K,t ,…,c i,t ] T , where c i,t This represents the target concentration of the i-th nutrient element.

[0020] In one specific embodiment, the source water management unit includes:

[0021] The source water spectral acquisition module is used to perform full-band spectral scanning of pure source water to obtain its source water spectral fingerprint S. src (λ), where λ represents the light wavelength; the dynamic baseline establishment module is used to establish the source water spectral fingerprint S. src (λ) is set as the dynamic compensation baseline for a single irrigation task;

[0022] The water quality early warning module is used to calculate the spectral fingerprint S of the source water. src (λ) and a preset historical normal source water spectral fingerprint The Euclidean distance D between them src If the Euclidean distance D src Exceeding a preset threshold τ alert If so, it will output an early warning message about abnormal water quality.

[0023] In one specific embodiment, the water and fertilizer synthesis and control unit includes:

[0024] A high-frequency pulsed microfluidic injection array consisting of multiple independent piezoelectric or peristaltic micropumps is used to receive controller commands and inject various fertilizer mother liquors into the main water flow in the form of high-frequency pulses.

[0025] An online multi-channel spectral analysis module is used to measure the total absorption spectrum of the nutrient solution formed after the main water flow and the fertilizer mother liquor are mixed in real time.

[0026] A data processing module is used to subtract the dynamically compensated baseline from the total absorption spectrum to obtain the net absorption spectrum, and to invert the net absorption spectrum into a real-time measured concentration vector c using a pre-trained partial least squares (PLS) model. meas Its form is c meas =[c N,m ,c P,m ,…,c i,m ] T , where c i,m This represents the real-time measured concentration of the i-th element;

[0027] A high-fidelity predictive feedforward-feedback coupled controller is used to compare the target fertilizer concentration with the real-time measured concentration vector, and dynamically adjust the pulse duty cycle of each micropump in the high-frequency pulse microfluidic injection array according to the comparison result, so as to achieve closed-loop control of nutrient solution concentration.

[0028] Preferably, the high-fidelity predictive feedforward-feedback coupled controller continuously calculates the error vector e(k) between the target concentration and the measured concentration through a multi-channel PID controller, and calculates and applies a pulse duty cycle adjustment Δp for each mother liquor pump based on the error vector e(k). i (k). The error vector e(k) is obtained through e(k) = c target -c meas The relational calculation of (k) and the adjustment amount Δp of the i-th element. i (k) Calculated using the following relation:

[0029]

[0030] Where k represents the current time step, e i (k) represents the concentration error of the i-th element at time step k, K p,i K i,i K d,i Let represent the proportional, integral, and derivative gain coefficients for the i-th element, respectively, and Δt represent the control period.

[0031] Preferably, the water and fertilizer synthesis and control unit further includes:

[0032] The system health self-diagnosis module is used to instruct the high-frequency pulse microfluidic injection array to inject a diagnostic pressure pulse sequence into the pipeline during non-irrigation periods, and at the same time instruct the online multi-channel spectral analysis module to monitor the instantaneous change of absorbance at a specific scattering wavelength to assess the adhesion or scaling on the inner wall of the pipeline and calculate the health index H.

[0033] The pulse physical maintenance module is used when the health index H is lower than a preset health threshold τ.health At that time, the high-frequency pulsed microfluidic injection array is instructed to generate high-frequency, high-energy physical pulses. These pulses induce microscopic cavitation and water hammer effects in the liquid, physically impacting and peeling off deposits on the tube wall and sensor surface. The health index H and the instantaneous change ΔA of absorbance at the specific scattering wavelength are also considered. turb There is an inverse relationship.

[0034] In one specific embodiment, the cloud management and optimization unit includes:

[0035] The data storage module is used to centrally store crop image information and environmental parameters of the crop and environment sensing unit, target irrigation strategy and target fertilizer concentration of the intelligent decision-making unit, and operation data of the water and fertilizer synthesis and control unit.

[0036] The remote monitoring module is used to display equipment status, historical records, and real-time irrigation data through a computer or mobile application interface.

[0037] The model iteration module is used to call the data in the data storage module to periodically retrain and update the parameters of the crop nutrient deficiency detection model, irrigation strategy decision module, and fertilization strategy decision module.

[0038] A second aspect of this invention provides a water and fertilizer integration irrigation method based on the Internet of Things and intelligent strategies, comprising the following steps:

[0039] Acquire crop image information and obtain environmental parameters;

[0040] Based on the crop image information and environmental parameters obtained in the crop and environment status perception step, a target irrigation strategy and target fertilizer concentration are generated.

[0041] Before the target irrigation strategy is implemented, the irrigation water source is analyzed for water quality and background water quality information is provided.

[0042] Based on the target fertilizer concentration generated by the intelligent decision-making step and the water quality background information provided by the source water management step, the nutrient solution of the target concentration is synthesized dynamically in real time and output, while generating operation data.

[0043] The system receives and stores the crop image information, the environmental parameters, the target irrigation strategy, the target fertilizer concentration, and the operational data, and enables remote monitoring and management.

[0044] This invention provides an integrated water and fertilizer irrigation device and method based on the Internet of Things and intelligent strategies, which has the following beneficial effects:

[0045] 1. This invention couples online spectral analysis technology with high-frequency pulse microfluidics technology to construct a real-time closed-loop control system for directly measuring chemical components. This achieves a high degree of consistency between digital decision-making commands and physical output of nutrient solution chemical components. Compared with existing technologies that rely on open-loop control methods such as Venturi pumps or proportional pumps, or indirect feedback methods such as EC conductivity, this invention solves the fundamental defect that it cannot directly measure and verify the true concentration of each nutrient element, resulting in a serious disconnect between commands and execution.

[0046] 2. This invention, through its unique source water spectral fingerprint acquisition and dynamic background compensation mechanism, enables the system to adapt to fluctuations in irrigation water quality. It can actively identify and deduct source water background interference before each irrigation. Existing technologies typically treat irrigation water sources as a constant background for calibration and control. The design of this invention avoids the significant measurement errors introduced by existing technologies due to changes in inherent minerals, organic matter, or turbidity in the source water, ensuring measurement accuracy in complex water source environments.

[0047] 3. This invention creatively reuses its core execution and sensing components. Through a set of preset pulse detection and maintenance protocols, it realizes the system's self-diagnosis and self-maintenance functions, giving the device the ability to monitor and clean itself online. Unlike traditional maintenance solutions that rely on periodic manual disassembly and inspection, chemical cleaning, or the configuration of additional cleaning devices, this invention solves the technical problem of performance degradation and inconvenient maintenance caused by pipeline and sensor contamination in traditional equipment, and improves the long-term operational stability of the system. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the device structure according to an embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram of the internal structure of the crop and environment sensing unit according to an embodiment of the present invention;

[0050] Figure 3 This is a schematic diagram of the internal structure of the intelligent decision-making unit according to an embodiment of the present invention;

[0051] Figure 4 This is a schematic diagram of the internal structure of the source water management unit according to an embodiment of the present invention;

[0052] Figure 5 This is a schematic diagram of the internal structure of the water and fertilizer synthesis and control unit according to an embodiment of the present invention;

[0053] Figure 6 This is a schematic diagram of the internal structure of the cloud management and optimization unit according to an embodiment of the present invention;

[0054] Figure 7 This is a flowchart of the integrated water and fertilizer irrigation method according to an embodiment of the present invention.

[0055] Explanation of icon numbers:

[0056] 10. Crop and Environment Sensing Unit; 20. Intelligent Decision-Making Unit; 30. Source Water Management Unit; 40. Water and Fertilizer Synthesis and Control Unit; 50. Cloud Management and Optimization Unit. Detailed Implementation

[0057] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] Figure 1 This is a schematic diagram of a water and fertilizer integrated irrigation device based on the Internet of Things and intelligent strategies according to an embodiment of the present invention. Figure 1 As shown, this embodiment of the invention provides an integrated water and fertilizer irrigation device based on the Internet of Things and intelligent strategies. The device may include: a crop and environment sensing unit 10, an intelligent decision-making unit 20, a source water management unit 30, a water and fertilizer synthesis and control unit 40, and a cloud management and optimization unit 50.

[0059] The crop and environment sensing unit 10 is used to acquire crop image information and environmental parameters within the deployment area. This crop and environment sensing unit 10 outputs the acquired sensing data to the intelligent decision-making unit 20 and the cloud management and optimization unit 50 via wired or wireless communication.

[0060] The intelligent decision-making unit 20, electrically connected to the crop and environment sensing unit 10, receives and processes crop image information and environmental parameters, and generates target irrigation strategies and target fertilizer concentrations accordingly. The intelligent decision-making unit 20 outputs the generated decision command data to the water and fertilizer synthesis and control unit 40, and uploads the decision process data to the cloud management and optimization unit 50.

[0061] The source water management unit 30 is physically located between the irrigation water source and the water-fertilizer mixing point. It is used to analyze the water quality of the irrigation water source before the irrigation task is performed and to provide the water quality background information obtained from the analysis to the water-fertilizer synthesis and control unit 40.

[0062] The water and fertilizer synthesis and control unit 40 is electrically connected to the intelligent decision-making unit 20 and the source water management unit 30, respectively. This water and fertilizer synthesis and control unit 40 is used to dynamically synthesize a nutrient solution meeting the target concentration in real time, based on the target fertilizer concentration output by the intelligent decision-making unit 20 and combined with the water quality background information provided by the source water management unit 30, and then output it to the crop irrigation area. The operational status data of the water and fertilizer synthesis and control unit 40 is transmitted in real time to the cloud management and optimization unit 50.

[0063] The cloud management and optimization unit 50 is connected via a network to the crop and environment sensing unit 10, the intelligent decision-making unit 20, and the water and fertilizer synthesis and control unit 40. This cloud management and optimization unit 50 receives and structurally stores the sensing data, decision-making process data, and operational status data from these units, and provides a remote monitoring and management interface for users. Furthermore, the cloud management and optimization unit 50 analyzes the stored data and uses the analysis results to update the model parameters within the intelligent decision-making unit 20.

[0064] When the device of this embodiment performs an irrigation task, its overall workflow is as follows: First, the crop and environment sensing unit 10 collects real-time crop and environment data; the data is transmitted to the intelligent decision-making unit 20 to generate precise irrigation and fertilization instructions; before the irrigation and fertilization instructions are executed, the source water management unit 30 performs a spectral scan on the irrigation water source to be used to establish a background baseline; the water and fertilizer synthesis and control unit 40 receives the decision instructions and the background baseline, and synthesizes nutrient solution with high fidelity through a closed-loop control process and delivers it to the crop root zone; the data of the entire process of the irrigation device operation is finally collected in the cloud management and optimization unit 50 for data archiving, remote presentation and subsequent model iteration.

[0065] The following provides a detailed description of the specific composition and operation of the crop and environment sensing unit 10, intelligent decision-making unit 20, source water management unit 30, water and fertilizer synthesis and control unit 40, and cloud management and optimization unit 50 in the embodiments of the present invention.

[0066] See Figure 2 , Figure 2 This is a schematic diagram of the internal structure of a crop and environment sensing unit 10 according to an embodiment of the present invention.

[0067] The crop and environment sensing unit 10 can be specifically implemented as follows: an environmental parameter acquisition module, a crop image information acquisition module, and an image processing and nutrient deficiency diagnosis module. These three modules work together to acquire comprehensive data for subsequent decision-making.

[0068] The environmental parameter acquisition module includes multiple sensors deployed at the agricultural production site. Specifically, these include a soil temperature and humidity sensor for measuring soil moisture content and temperature, a soil electrical conductivity (EC) sensor for measuring the concentration of soluble salts in the soil, an air temperature and humidity sensor for measuring air temperature and humidity, and a light intensity sensor for measuring photosynthetically active radiation intensity. These sensors are distributed in a preset grid layout in the target area and periodically transmit the collected real-time environmental parameter data to a local edge gateway via wireless communication protocols such as Zigbee or LoRa.

[0069] The crop image acquisition module includes one or more high-resolution image acquisition devices mounted on fixed poles or on a drone platform. These devices acquire high-resolution crop images from directly above the crop canopy or from a specific lateral angle at preset time intervals (e.g., at fixed times daily) or according to task instructions. The image information is then transmitted to the image processing and nutrient deficiency diagnosis module.

[0070] The core of the image processing and nutrient deficiency diagnosis module is a pre-trained crop nutrient deficiency detection model based on YOLOv7. This model is trained using supervised learning on a dataset of tens of thousands of crop images manually labeled by agricultural experts with different nutrient deficiency regions and categories (e.g., yellowing leaves, necrotic spots). The model receives images acquired by the crop image acquisition module as input. The internal processing flow of the model is as follows: First, its backbone network extracts features from the input image, obtaining a deep feature map containing multiple dimensions such as color, texture, and shape. Then, its detection head (the component specifically responsible for generating specific target locations and category predictions from the deep feature map) identifies and locates crop leaf regions in the image based on the deep feature map. Finally, for the located leaf regions, the model's classification part compares them with a pre-learned visual feature library representing symptoms of deficiency or excess of different nutrient elements (nitrogen, phosphorus, potassium, etc.), outputting a structured crop nutrient diagnosis result.

[0071] Crop nutrient diagnostic results can be formatted as a crop nutrient diagnostic vector D. diag Its form is:

[0072] D diag =[d N ,d P ,d K ,…,d i ] T ;

[0073] Where, d N ,d P ,dK Each element d in the vector representing the diagnostic results of crops for nitrogen, phosphorus, and potassium elements represents a different crop. i A diagnostic status value corresponding to a specific nutrient element i is generated. This status value is used to characterize the classification result or confidence level of the corresponding element's deficiency, normality, or excess. Finally, the environmental parameters collected by the environmental parameter acquisition module and the crop nutrient diagnosis results output by the image processing and nutrient deficiency diagnosis module are packaged together and sent to the intelligent decision-making unit 20 as the output data of the crop and environment sensing unit 10.

[0074] like Figure 3 As shown, Figure 3 This is a schematic diagram of the internal structure of an intelligent decision-making unit 20 according to an embodiment of the present invention.

[0075] The intelligent decision-making unit 20 receives environmental parameters and crop nutrient diagnosis results output by the crop and environment sensing unit 10, and based on these input data, generates instruction data to guide irrigation and fertilization operations through an internally integrated decision model. Specifically, the intelligent decision-making unit 20 may include an irrigation strategy decision-making module and a fertilization strategy decision-making module.

[0076] The irrigation strategy decision-making module is internally configured with a pre-trained random forest decision-making model. The input data for this random forest decision-making model includes: environmental parameters collected by the crop and environment sensing unit 10 (specifically, soil moisture, soil electrical conductivity, air humidity, and light intensity), and the output crop nutrient diagnostic vector D. diag The system incorporates current crop growth stage information and short-term weather forecast data obtained from external meteorological data services. The random forest decision model, through its integrated multiple decision trees, comprehensively analyzes and votes on the above multi-dimensional input data, ultimately outputting an optimal target irrigation strategy. This strategy includes the target irrigation start and stop times t. start and irrigation duration t dur .

[0077] The fertilization strategy decision-making module is internally configured with an improved genetic algorithm-based fertilization control model. This fertilization control model uses the crop nutrient diagnostic vector D output by the crop and environment sensing unit 10. diag As the core input, and combined with a standard nutrient requirement model for a specific crop variety and growth stage stored in a local database, the goal of this fertilizer control model is to generate a targeted fertilizer formula that can compensate for the current nutrient deficiencies of the crop through optimization.

[0078] In practice, this improved genetic algorithm uses a set of vectors containing the concentrations of various nutrient elements as individuals in the population. The improvement lies in the fact that its objective function is designed as a multi-objective optimization function, capable of simultaneously minimizing crop-diagnosed nutrient deviations, total fertilizer costs, and chemical antagonistic effects between different fertilizers, rather than the single-objective problem typically handled by traditional genetic algorithms.

[0079] The improved genetic algorithm iteratively performs selection, crossover, and mutation operations, and evolves the population using a single objective function: minimizing the difference between the diagnosed nutrient deviation and the applied nutrient amount. Specifically, the objective function can be designed as a multi-objective optimization function that comprehensively considers minimizing nutrient element deviation, minimizing total fertilizer cost, and minimizing chemical antagonistic effects between different fertilizers. After a predetermined number of generations of optimization, the improved genetic algorithm converges and outputs an optimal target concentration vector c. target The vector has the following form:

[0080]

[0081] Among them, c i,t c represents the target concentration of the i-th nutrient element, in ppm or mg / L; N,t ,c P,t ,c K,t These specifically refer to the target concentrations of nitrogen, phosphorus, and potassium, respectively. Ultimately, the target irrigation strategy output by the irrigation strategy decision module and the target concentration vector c output by the fertilization strategy decision module are combined. target Together with the intelligent decision-making unit 20, they are sent to the water and fertilizer synthesis and control unit 40 as output instructions.

[0082] See Figure 4 , Figure 4 This is a schematic diagram of the internal structure of a source water management unit 30 according to an embodiment of the present invention.

[0083] The source water management unit 30 is physically connected in series in the main irrigation pipeline, located between the irrigation water source outlet and the fertilizer mother liquor injection point. This source water management unit 30 is used to perform spectral analysis on the pure source water that is about to flow into the mixing system before each irrigation and fertilization task is officially started, in order to obtain and provide water quality background information for subsequent accurate concentration measurements. Specifically, the source water management unit 30 may include a source water spectral acquisition module, a dynamic baseline establishment module, and a water quality early warning module.

[0084] The source water spectral acquisition module consists of an online multi-channel spectrometer and a corresponding flow cell. During the initial mission startup phase, an upstream valve group is controlled to ensure that only pure source water (free of any fertilizer mother liquor) flows through the flow cell. At this time, the online multi-channel spectrometer performs a full-spectrum scan of the source water flowing through the flow cell, measuring its absorbance at different wavelengths to obtain an absorption spectrum curve. This curve is defined as the source water spectral fingerprint S for this mission. src (λ), where λ represents the wavelength of light.

[0085] The dynamic baseline establishment module is used to process and apply the source water spectral fingerprint S acquired by the source water spectral acquisition module. src (λ). This dynamic baseline establishment module will collect S... src (λ) The data is temporarily stored in the local controller's memory and designated as the dynamic compensation baseline for this irrigation task. This baseline data is then transmitted to the water and fertilizer synthesis and control unit 40 for background subtraction calculations during subsequent measurements of the mixed nutrient solution spectrum.

[0086] The water quality early warning module is used to quickly assess the safety of the water source immediately after collecting the source water spectral fingerprint. This module will display the real-time collected source water spectral fingerprint S... src (λ) and a historical normal source water spectral fingerprint pre-stored in a local database, representing the normal water quality status of the region. The comparison is performed by calculating the Euclidean distance D between the two spectral curves. src This is achieved through a calculation formula, which is:

[0087]

[0088] Among them, D src S represents the calculated Euclidean distance value. src (λ) represents the source water spectral fingerprint collected in real time; This indicates the spectral fingerprint of historical normal source water.

[0089] After the calculation is completed, the module will output the obtained Euclidean distance value D. src With a preset security threshold τ alert Compare them. If D src >τ alert If the system detects an anomaly in the current source water quality, the module immediately generates a water quality anomaly warning. This warning is sent to the user management interface via the cloud management and optimization unit 50, and can also directly trigger the water and fertilizer synthesis and control unit 40 to stop the current irrigation task.

[0090] like Figure 5 As shown, Figure 5 This is a schematic diagram of the internal structure of a water and fertilizer synthesis and control unit 40 according to an embodiment of the present invention.

[0091] The water and fertilizer synthesis and control unit 40, as the actuator of the device of the present invention, is used to receive and accurately execute the fertilization instructions generated by the intelligent decision-making unit 20. This water and fertilizer synthesis and control unit 40 converts the digitized target concentration vector into a physical nutrient solution with specific chemical components. Its specific implementation may include: a high-frequency pulsed microfluidic injection array, an online multi-channel spectral analysis module, a data processing module, and a high-fidelity predictive feedforward-feedback coupled controller.

[0092] The high-frequency pulsed microfluidic injection array is positioned on the main irrigation pipeline, physically located after the source water management unit 30. This array consists of multiple independent micropumps, such as piezoelectric pumps or peristaltic pumps, each connected to a different fertilizer mother liquor tank. Each micropump is independently driven by a high-fidelity predictive feedforward-feedback coupled controller, using pulse width modulation (PWM) signals to precisely inject micro-levels of mother liquor into the clean water flowing through the main pipeline in the form of high-frequency pulses, achieving instantaneous mixing.

[0093] The online multi-channel spectral analysis module has the same structure as the spectral acquisition component in the source water management unit 30. Its physical location is set at a sufficient distance downstream of the high-frequency pulsed microfluidic injection array to ensure thorough mixing of the mother liquor and purified water. This online multi-channel spectral analysis module is used to perform real-time, uninterrupted full-band spectral scanning of the mixed nutrient solution flowing through its internal flow tank, acquiring the total absorption spectrum S. mix (λ).

[0094] The data processing module is integrated with a high-fidelity predictive feedforward-feedback coupled controller to perform real-time concentration inversion calculations. First, the data processing module receives the dynamic compensation baseline S for this task from the source water management unit 30. src (λ), and from the total absorption spectrum S measured in real time. mix Subtracting this from (λ) yields the net absorption spectrum s produced solely by the fertilizer solute. sol (λ), whose calculation formula is:

[0095] s sol (λ)=S mix (λ)-S src (λ);

[0096] Subsequently, the data processing module invokes a partial least squares (PLS) chemometrics model pre-calibrated using extensive experimental data. The calibration process for this PLS model includes: preparing hundreds of standard solutions of single and compound nutrients containing different known concentration gradients (covering the working concentration range); measuring the net absorption spectrum of each standard solution using the same spectral analysis module as the device of this invention; and finally, establishing and validating the PLS regression model and its regression coefficient matrix B using the known concentration data and corresponding spectral data of the standard solutions, and converting the net absorption spectral vector s... sol (λ) is used as input to invert the real-time measured concentration vector c of each nutrient element in the mixed nutrient solution. meas Its inversion formula is:

[0097] c meas =B T s sol ;

[0098] Where B is the regression coefficient matrix of the PLS model, c meas =[c N,m ,c P,m ,…,c i,m ] T , where c N,m ,c P,m c represents the concentration values ​​of nitrogen and phosphorus obtained from real-time measurements, respectively. i,m This represents the real-time measured concentration of the i-th element.

[0099] The high-fidelity predictive feedforward-feedback coupled controller is the core of the entire closed-loop control. This high-fidelity predictive feedforward-feedback coupled controller first determines the target concentration vector c output by the intelligent decision unit 20. target The initial pulse duty cycle vector p required to drive each mother liquor pump is calculated using a pre-calibrated model. init Feedforward control is implemented. During system operation, the controller adjusts the target concentration vector c within each control cycle k. target The real-time measured concentration vector c output by the data processing module meas Compare (k) and calculate the error vector e(k) between the two:

[0100] e(k) = c target -c meas (k);

[0101] A parallel multichannel PID controller calculates an adjustment Δp for the pulse duty cycle of each mother liquor injection pump based on this error vector. i (k). For the i-th element, the formula for calculating its adjustment amount is:

[0102]

[0103] Where k represents the current time step; e i (k) represents the concentration error of the i-th element at time step k; K p,i K i,i K d,i Let represent the proportional, integral, and derivative gain coefficients for the i-th element, respectively; Δt represents the control period. Finally, the controller adds this adjustment to the duty cycle of the previous moment, updating the PWM duty cycle vector p(k) applied to each pump, thus forming a high-frequency closed loop to ensure that the output nutrient solution concentration accurately tracks the target command.

[0104] In a preferred embodiment, the water and fertilizer synthesis and control unit 40 further includes a system health self-diagnosis module and a pulse physical maintenance module.

[0105] The system health self-diagnosis module is activated during non-irrigation periods. Upon activation, this module instructs a high-frequency pulsed microfluidic injection array to inject a specially designed sequence of diagnostic pressure pulses with specific frequencies and amplitudes into the pipeline. Simultaneously, it instructs the online multi-channel spectral analysis module to switch to high-frequency sampling mode and focus on monitoring a specific scattering wavelength λ that is sensitive to changes in turbidity. turb Instantaneous change in absorbance ΔA turb The system's self-diagnostic health module calculates a quantified health index H based on the relationship between pressure pulse excitation and absorbance response. When the inner wall of the pipeline or the surface of the flow tank is clean, fluid disturbance is small, and ΔA... turb A smaller value results in a higher calculated H value; conversely, if there is adhesion or scaling, the fluid turbulence is greater, and ΔA will be lower. turb The larger the value, the lower the calculated H value. The specific calculation formula can be expressed by an exponential decay model:

[0106]

[0107] Where H is the final calculated health index; H max The ideal health index of the system (e.g., set to 100) represents a state of complete cleanliness; ΔA turb is the absolute value of the instantaneous change in absorbance at a specific scattering wavelength; k is a positive system sensitivity coefficient, which is experimentally calibrated to adjust the severity of the effect of absorbance change on the health index H.

[0108] The pulse physical maintenance module is linked to the system health self-diagnosis module. When the health index H falls below a preset health threshold τ... healthWhen activated, the module instructs one or more micropumps to generate continuous high-frequency, high-energy physical pulses. The frequency of these pulses is set between 1 kHz and 10 kHz, and the instantaneous pressure gradient at the pulse leading edge is controlled above a preset threshold sufficient to generate a microscopic cavitation effect, thus achieving effective physical removal of deposits. As these pulses propagate through the liquid, they utilize the generated microscopic cavitation and water hammer effect to physically impact and remove deposits from the pipe walls and sensor optical window surfaces, thereby achieving online, chemical-free self-cleaning maintenance.

[0109] See Figure 6 , Figure 6 This is a schematic diagram of the internal structure of a cloud management and optimization unit 50 according to an embodiment of the present invention.

[0110] The cloud management and optimization unit 50 communicates with the crop and environment sensing unit 10, the intelligent decision-making unit 20, and the water and fertilizer synthesis and control unit 40 via network, serving as a central platform for data aggregation, remote interaction, and model optimization. Specifically, the cloud management and optimization unit 50 may include a data storage module, a remote monitoring module, and a model iteration module.

[0111] The data storage module is used to receive and persistently store the entire process operation data from other units of the device. Specifically, the stored data includes: time series of environmental parameters, crop image information, and structured crop nutrient diagnosis results uploaded by the crop and environment sensing unit 10; and target irrigation strategies and target concentration vector c uploaded by the intelligent decision-making unit 20. target ; and the source water spectral fingerprint S uploaded by the source water management unit 30 src (λ) and the real-time measured concentration vector c uploaded by the water and fertilizer synthesis and control unit 40. meas The system health index H. All data is timestamped and includes the source device identifier upon storage, and is organized in a time-series database for easy retrieval and analysis.

[0112] The remote monitoring module provides users with a human-machine interface, accessible via a web browser on a personal computer or a mobile application. Through this interface, users can view real-time and historical data provided by the data storage module, such as soil moisture change curves displayed in chart form, crop image history, and detailed logs of historical irrigation tasks. Users can also perform management operations through this interface, including manually starting or stopping an irrigation task, and adjusting internal system operating parameters, such as the water quality anomaly warning threshold τ in the source water management unit 30. alert The health threshold τ in the water and fertilizer synthesis and control unit 40 health .

[0113] The model iteration module is used to perform periodic retraining and updates of the decision-making models deployed within the intelligent decision-making unit 20. Based on preset triggering conditions, such as the end of a complete crop growing season, this module automatically extracts all relevant data within a given time period from the data storage module, forming a new training dataset. This dataset includes environmental parameters, crop images, system decision commands, and actual concentration data. The module uses this dataset to retrain the crop nutrient deficiency detection model, irrigation strategy decision-making model, and fertilizer application control model. After training, the performance of the new model is compared and verified with the existing models deployed in the intelligent decision-making unit 20. If the new model's predetermined performance indicators, such as the accuracy of nutrient deficiency identification, are higher than the existing model, the module pushes the newly trained model parameter file to the intelligent decision-making unit 20 to replace the original model parameters, thus completing a closed-loop optimization of the model.

[0114] To further illustrate the collaborative working process of the technical solution of this invention, a specific working scenario example will be used below.

[0115] See Figure 7 , Figure 7 This is a flowchart of an integrated water and fertilizer irrigation method according to an embodiment of the present invention. The application scenario of this embodiment is set as: supplementary integrated water and fertilizer irrigation for greenhouse tomatoes in the fruiting stage.

[0116] S1: Task Initiation and Source Water Calibration. The system initiates the irrigation task according to the preset irrigation plan or user instructions. First, the source water management unit 30 is activated. The upstream valve opens, allowing only pure source water to flow through its internal circulation pool. The source water spectral acquisition module scans the pure source water to obtain the source water spectral fingerprint S for this task. src (λ). The dynamic baseline establishment module will use this S src (λ) is set as the dynamic compensation baseline for this task and sent to the water and fertilizer synthesis and control unit 40. Simultaneously, the water quality early warning module calculates S. src (λ) and historical normal source water spectral fingerprint The Euclidean distance D between them src The calculated result is less than the warning threshold τ alert The water quality was determined to be normal, and the mission continued.

[0117] S2: State Perception and Nutrient Deficiency Diagnosis. The crop and environment perception unit 10 performs data acquisition. Data collected by its environmental parameter acquisition module shows that the current soil moisture is below a preset lower threshold. Its crop image information acquisition module acquires images of the tomato plant canopy. The images are sent to the image processing and nutrient deficiency diagnosis module, where the internal YOLOv7 model analyzes the images, identifying yellowing characteristics at the edges of older leaves in the lower and middle parts of some plants, and outputting a crop nutrient diagnosis vector D. diag The potassium element in this vector corresponds to the state value d. K This indicates that the crop has mild potassium deficiency symptoms.

[0118] S3: Intelligent Decision Generation. The intelligent decision-making unit 20 receives environmental parameters and crop nutrient diagnostic vector D from the crop and environment sensing unit 10. diag Its irrigation strategy decision-making module generates an irrigation duration t based on weather forecast data indicating low soil moisture and no rainfall expected in the next 24 hours. dur The target irrigation strategy is 25 minutes. Its fertilization strategy decision module is based on the prior knowledge that the crop has a high potassium requirement during the fruiting stage, and combines this with D... diag A clear diagnosis of potassium deficiency is made, and a target concentration vector c is generated by solving the problem using an improved genetic algorithm. target The target concentration of potassium in this vector is c. K,t The concentration was set at 180 ppm, while the concentrations of other elements such as nitrogen and phosphorus were maintained at the standard recommended values ​​for this growth stage.

[0119] S4: High-fidelity water and fertilizer synthesis and delivery. The water and fertilizer synthesis and control unit 40 receives the target irrigation strategy and target concentration vector c. target The controller starts the main pipeline water pump and, according to c target The initial pulse duty cycle of each fertilizer mother liquor pump was calculated for feedforward injection. An online multi-channel spectral analysis module was used to measure the total absorption spectrum S of the mixed nutrient solution in real time. mix (λ). The data processing module utilizes the dynamic compensation baseline S obtained in S1. src (λ) Calculate the net absorption spectrum s sol (λ), and then the real-time measured concentration vector c is derived by inverting the PLS model. meas The high-fidelity predictive feedforward-feedback coupled controller uses a PID algorithm to continuously compare c. target With c meas The system detects differences in nutrient solution concentration and adjusts the pulse duty cycle of each micro-pump at high frequency until the error vector e(k) approaches zero. This closed-loop control process continues throughout the 25-minute irrigation period, ensuring that the concentration of nutrient solution delivered to the tomato root zone precisely matches the decision command.

[0120] S5: System Self-Maintenance and Data Archiving. After the irrigation task is completed, the main water pump stops. The system health self-diagnosis module of the water and fertilizer synthesis and control unit 40 is activated, injecting a diagnostic pressure pulse sequence into the pipeline and calculating the health index H based on the change in the absorbance of scattered light. The calculation result is higher than the health threshold τ. health The pipeline was determined to be in good condition and required no physical maintenance. Finally, all data from this task, including environmental parameters, crop images, diagnostic results, decision commands, source water spectra, real-time concentration monitoring records, and the health index H, was packaged and uploaded to the data storage module of the cloud management and optimization unit 50 for archiving, so that users can access it at any time or use it for subsequent model iterations.

[0121] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A water and fertilizer integrated irrigation device based on the Internet of Things and intelligent strategies, characterized in that, include: The crop and environment sensing unit is used to acquire crop image information and environmental parameters. The intelligent decision-making unit is used to generate a target irrigation strategy and a target fertilizer concentration based on the crop image information and the environmental parameters. The source water management unit is used to perform water quality analysis on the irrigation water source before the target irrigation strategy is implemented, so as to obtain and provide water quality background information; The water and fertilizer synthesis and control unit is used to respond to the target irrigation strategy and, based on the target fertilizer concentration output by the intelligent decision unit and the water quality background information provided by the source water management unit, dynamically synthesize and output the nutrient solution of the target concentration in real time, while generating operation data. The cloud management and optimization unit is used to receive and store crop image information and environmental parameters from the crop and environment sensing unit, target irrigation strategies and target fertilizer concentrations from the intelligent decision-making unit, and operating data from the water and fertilizer synthesis and control unit, and to realize remote monitoring and management.

2. The integrated water and fertilizer irrigation device based on the Internet of Things and intelligent strategies according to claim 1, characterized in that, The crop and environment sensing unit includes: Environmental sensors are used to collect various environmental parameters in real time, including soil temperature and humidity, soil electrical conductivity, air temperature and humidity, and light intensity. Image acquisition equipment, used to periodically acquire crop image information; The image processing unit is equipped with a crop nutrient deficiency detection model. It receives the crop image information as input, processes the input using the crop nutrient deficiency detection model to identify crop nutrient deficiency symptoms, and outputs crop nutrient diagnosis results.

3. The integrated water and fertilizer irrigation device based on the Internet of Things and intelligent strategies according to claim 2, characterized in that, The intelligent decision-making unit includes: The irrigation strategy decision module is equipped with a random forest decision model, which is used to generate target irrigation start and stop times and total irrigation amount based on the environmental parameters, crop nutrition diagnosis results, crop growth stage and weather forecast. The fertilization strategy decision-making module is equipped with a fertilizer application control model based on an improved genetic algorithm. This model generates a target concentration vector c containing multiple nutrient elements based on the crop nutrient diagnosis results and crop nutrient requirement patterns. target ; The target concentration vector c target =[c N,t ,c P,t ,c K,t ,…,c i,t ] T , where c N,t ,c P,t ,c K,t Specifically referring to the target concentrations of nitrogen, phosphorus, and potassium, c i,t This represents the target concentration of the i-th nutrient element.

4. The integrated water and fertilizer irrigation device based on the Internet of Things and intelligent strategies according to claim 1, characterized in that, The source water management unit includes: The source water spectral acquisition module is used to perform full-band spectral scanning of pure source water to obtain the source water spectral fingerprint S. src (λ), where λ represents the wavelength of light; The dynamic baseline establishment module is used to establish the source water spectral fingerprint S. src (λ) is set as a dynamic compensation baseline to eliminate the interference of source water background on subsequent concentration measurements; The water quality early warning module is used to generate the source water spectral fingerprint S src (λ) and historical normal source water spectral fingerprint The source water spectral fingerprint S was compared. src (λ) and the historical normal source water spectral fingerprint The Euclidean distance D between them src Exceeding the preset threshold τ alert If so, a water quality anomaly warning message will be sent.

5. The integrated water and fertilizer irrigation device based on the Internet of Things and intelligent strategies according to claim 1, characterized in that, The water and fertilizer synthesis and control unit includes: The high-frequency pulse microfluidic injection array consists of multiple independent piezoelectric or peristaltic micropumps, which are used to inject various fertilizer mother liquors into the main water flow in the form of high-frequency pulses according to controller instructions, and mix them to form a mixed nutrient solution. An online multi-channel spectral analysis module is used to measure the total absorption spectrum of the mixed nutrient solution in real time; The data processing module is used to compensate for the source water background from the total absorption spectrum based on the water quality background information provided by the source water management unit, and to invert the compensated net absorption spectrum into a real-time measured concentration vector c using a pre-trained partial least squares model. meas ; The real-time measured concentration vector c meas =[c N,m ,c P,m ,…,c i,m ] T , where c N,m ,c P,m c represents the concentration values ​​of nitrogen and phosphorus obtained from real-time measurements, respectively. i,m This represents the real-time measured concentration of the i-th element; A high-fidelity predictive feedforward-feedback coupled controller is used to dynamically adjust the pulse duty cycle of the high-frequency pulse microfluidic injection array based on the target fertilizer concentration output by the intelligent decision-making unit and the real-time measured concentration vector output by the data processing module, thereby achieving closed-loop control of the nutrient solution concentration.

6. The integrated water and fertilizer irrigation device based on the Internet of Things and intelligent strategies according to claim 1, characterized in that, The cloud management and optimization unit includes: The data storage module is used to centrally store crop image information and environmental parameters of the crop and environment sensing unit, target irrigation strategy and target fertilizer concentration of the intelligent decision-making unit, and operation data of the water and fertilizer synthesis and control unit. The remote monitoring module is used to remotely view the device status, historical records, and real-time irrigation data via a computer or mobile application. The model iteration module is used to periodically retrain and optimize the crop nutrient deficiency detection model, irrigation strategy decision module, and fertilization strategy decision module using the data in the data storage module.

7. The integrated water and fertilizer irrigation device based on the Internet of Things and intelligent strategies according to claim 5, characterized in that, The high-fidelity predictive feedforward-feedback coupled controller is specifically used to continuously calculate the error vector e(k) between the target concentration and the measured concentration, and based on the error vector e(k), calculates and applies an adjustment amount Δp of the pulse duty cycle for each mother liquor pump through a multi-channel PID controller. i (k), where the formula for calculating the error vector e(k) is e(k) = c target -c meas (k), and the adjustment amount Δp of the i-th element i The formula for calculating (k) is: Where k represents the current time step, e i (k) represents the concentration error of the i-th element at time step k, K p,i K i,i K d,i Let represent the proportional, integral, and differential gain coefficients of the i-th element, respectively, and Δt represent the control period.

8. The integrated water and fertilizer irrigation device based on the Internet of Things and intelligent strategies according to claim 5, characterized in that, The water and fertilizer synthesis and control unit also includes: The system health self-diagnosis module is used to instruct the high-frequency pulse microfluidic injection array to inject a diagnostic pressure pulse sequence during non-irrigation periods, and at the same time instruct the online multi-channel spectral analysis module to monitor the instantaneous change of absorbance at a specific scattering wavelength to assess the adhesion or scaling on the inner wall of the pipeline and calculate the health index H. The pulse physical maintenance module is used when the health index H is lower than a preset health threshold τ. health At that time, the high-frequency pulse microfluidic injection array is instructed to generate high-frequency, high-energy physical pulses to physically impact and peel off the deposits on the tube wall and sensor surface, thereby achieving online self-cleaning.

9. The integrated water and fertilizer irrigation device based on the Internet of Things and intelligent strategies according to claim 8, characterized in that, The health index H calculated by the system's self-diagnosis module is related to the instantaneous change ΔA of absorbance at a specific scattering wavelength. turb There is an inverse proportional relationship, when ΔA turb The larger the value, the smaller the H value.

10. A water and fertilizer integration irrigation method based on the Internet of Things and intelligent strategies, applied to the water and fertilizer integration irrigation device based on the Internet of Things and intelligent strategies as described in any one of claims 1-9, characterized in that, Includes the following steps: Acquire crop image information and obtain environmental parameters; Based on the crop image information and environmental parameters obtained in the crop and environment status perception step, a target irrigation strategy and a target fertilizer concentration are generated. Before the target irrigation strategy is implemented, the irrigation water source is analyzed for water quality and background water quality information is provided. Based on the target fertilizer concentration generated by the intelligent decision-making step and the water quality background information provided by the source water management step, the nutrient solution of the target concentration is synthesized dynamically in real time and output, while generating operation data. The system receives and stores the crop image information, the environmental parameters, the target irrigation strategy, the target fertilizer concentration, and the operational data, and enables remote monitoring and management.

Images