Supply system combining siphon driving and fertilizer concentration feedback regulation

Abstract

The invention relates to the technical field of intelligent agriculture and agricultural automation control, and discloses a supply system combining siphon driving and fertilizer concentration feedback regulation, comprising: a cultivation module for bearing crops and accommodating a supply liquid; the active controllable siphon device module is used for discharging the supply liquid in a controllable manner; the sensing network module at least comprises a supply end sensor arranged at the supply end of the supply liquid and a return end sensor arranged at the return end of the supply liquid; the central controller module is used for acquiring supply liquid parameter data, collected by the sensing network module, of the supply end and the return end; determining the net absorption amount of the crops to the supply liquid; adaptively updating a preset system dynamic model; cooperative control instructions are predictively generated. According to the invention, the problems of rigid control and blind supply caused by incapability of accurately sensing real absorption of crops in the prior art are solved, and accurate and foresight management of the crop growth environment is realized.

Description

Technical Field

[0001] This invention relates to the field of smart agriculture and agricultural automation control technology, specifically a supply system that combines siphon drive with fertilizer and pesticide concentration feedback regulation. Background Technology

[0002] Integrated water and fertilizer technology in facility agriculture is key to improving agricultural production efficiency and resource utilization. Among them, the "two-stage cultivation method" and other tidal irrigation technologies, which combine the advantages of substrate cultivation and nutrient solution cultivation, have attracted attention due to their simple structure and ability to provide a good water and air environment for the root system. However, in the process of in-depth research and practice on existing technical solutions, the inventors found that such systems still have several deep-seated technical bottlenecks that urgently need to be addressed in order to achieve truly precise and intelligent management.

[0003] First, in existing tidal or siphon irrigation systems, the design of the siphon drainage device often relies on a fixed physical structure. Key hydraulic parameters such as the trigger liquid level and drainage rate are fixed once manufactured. This design is essentially a passive response mechanism. It cannot proactively and adaptively adjust to the varying root aeration requirements of crops at different growth stages (such as seedling and fruiting stages). Therefore, the water-air ratio in the root zone remains in a suboptimal compromise state for extended periods, limiting the realization of crop potential.

[0004] Secondly, traditional fertilizer solution management methods also have inherent limitations in their sensing capabilities. They typically deploy sensors only in the main storage tank. This single-point measurement method only reflects the initial concentration of the supplied solution, completely failing to detect changes in the composition of the liquid returning after passing through the crop roots. This leaves the system in a black box regarding the crop's true net absorption. The control system can only operate in an open-loop manner based on preset, empirical fertilization procedures, and its decisions lack real-time, accurate physiological data support.

[0005] Furthermore, the control logic of existing technologies is often static. Control strategies are based on a set of fixed empirical parameters or growth models. However, crop physiological metabolism is a highly dynamic and nonlinear process. As crops grow, their water and fertilizer requirements change significantly. The deviation between the fixed model and the actual state of the crop accumulates over time. Ultimately, this leads to the system's control behavior becoming increasingly distant from the crop's true needs, resulting in resource misallocation.

[0006] Furthermore, most existing systems employ a reactive control philosophy. They typically use simple threshold logic for judgment and execution, initiating compensation only when a certain indicator falls below a lower limit. This lagging control approach means it can only passively correct past deviations, unable to anticipate and mitigate potential future risks. It also struggles to handle synergistic optimization problems among multiple objectives, such as achieving a dynamic balance between water conservation and fertilizer retention. This keeps overall management at a relatively rudimentary level. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a supply system that combines siphon drive with fertilizer and pesticide concentration feedback regulation. This solves the problem of traditional siphon irrigation systems being unable to accurately sense the real-time absorption status of crops and adaptively predict and regulate accordingly, resulting in blind water and fertilizer supply and rigid root zone environment control.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a supply system combining siphon drive and fertilizer / pesticide concentration feedback regulation, comprising:

[0009] The cultivation module is used to hold the crops and contain the supply solution;

[0010] An active, controllable siphon device module is connected to the cultivation module and is used to discharge the supply liquid in a controllable manner;

[0011] The sensor network module includes at least a supply-end sensor deployed at the supply end of the supply liquid and a return-end sensor deployed at the supply liquid return end.

[0012] The central controller module is connected to the active controllable siphon device module and the sensor network module. The central controller module is used for:

[0013] Acquire the supply liquid parameter data collected by the sensor network module at the supply end and the return end;

[0014] Based on the parameter data from the supply end and the return end, determine the net absorption amount of the supplied liquid by the crop;

[0015] Based on the net absorption, the preset system dynamic model is adaptively updated; and

[0016] Based on the updated system dynamic model, predictive collaborative control commands are generated to collaboratively control the actuators of the active controllable siphon device module and the supply unit.

[0017] Preferably, the active controllable siphon device module includes a piezoelectric ceramic microfluidic valve or shape memory alloy actuator disposed on the siphon pipeline, and the central controller module controls the siphon drainage rate by outputting a continuously variable electrical signal to adjust the opening of the valve or actuator.

[0018] Preferably, the sensor network module includes ion-selective electrodes for measuring the concentration of specific nutrient ions at both the supply and return ends.

[0019] Preferably, the central controller module determines the net absorption of the supplied liquid by the crop, specifically by calculating the net absorption of one or more ions by the crop within one or more irrigation cycles based on the ion concentration difference between the supply end and the return end and the total amount of circulating liquid.

[0020] Preferably, the system dynamic model is a state-space model, whose state vector includes at least the ion concentration in the storage tank, the liquid level in the cultivation unit, and the estimated ion concentration in the root zone.

[0021] Preferably, the system dynamic model couples the nonlinear fluid dynamics model of the active controllable siphon device module with a parameterized ion absorption model characterizing the crop's absorption capacity.

[0022] Preferably, the central controller module adaptively updates the system dynamic model, specifically through an extended Kalman filter or a moving time-domain estimator, which uses the net absorption as an observation to calibrate the state and parameters in the model online.

[0023] Preferably, the central controller module predictively generates the cooperative control instructions by solving a model predictive control optimization problem to obtain the optimal control sequence in a future prediction time domain.

[0024] Preferably, the cost function of the model predictive control optimization problem includes a state tracking term, a control cost term, and an agronomy-engineering coupled optimization term.

[0025] The agronomy-engineering coupling optimization term includes at least one of the following:

[0026] Root water-air ratio optimization term used to regulate the root wet-dry cycle;

[0027] Ion balance penalty used to maintain the optimal ratio of key ions in the nutrient solution;

[0028] Fertilizer and pesticide risk mitigation measures are used to ensure safe concentration and duration of action when applying pesticides.

[0029] This invention provides a supply system that combines siphon-driven operation with fertilizer and pesticide concentration feedback regulation. It offers the following advantages:

[0030] 1. This invention introduces an active, controllable siphon device module, enabling real-time, software-defined control of the drainage rate of the cultivation unit. This achieves precise shaping of the root zone's wet-dry cycle. Compared to existing passive siphon methods that rely on fixed physical properties, this invention overcomes the fundamental shortcomings of rigid drainage processes and the inability to provide optimal root water and air environments for different growth stages or environmental conditions, significantly improving root health.

[0031] 2. The supply system of this invention can accurately detect the real-time physiological needs of crops. It employs a dual-channel sensor network at both the supply and return ends, and obtains the net absorption data of fertilizers and pesticides by crops through differential calculation. This is fundamentally different from the open-loop supply mode in existing technologies that rely solely on single-point monitoring of the storage tank or pre-programmed procedures. Therefore, this invention overcomes the blindness in the implementation of existing technologies, fundamentally avoiding excessive waste of water and fertilizer and nutrient imbalance, and achieving true on-demand supply.

[0032] 3. This invention constructs a self-evolving system. It uses crop net uptake as a key observation value and employs algorithms such as extended Kalman filtering to calibrate and update core physiological parameters in the system's dynamic model online. This solves the decision-making failure problem caused by the disconnect between the model and the actual crop growth state in existing technologies, endowing the system with adaptive capabilities that learn as the crop grows.

[0033] 4. This invention employs a model predictive control strategy, elevating control from a simple reactive approach to a predictive one. The system can anticipate future disturbances and plan the optimal control path in advance. Its cost function incorporates agronomic-engineering coupled optimization terms, achieving coordinated management of multiple objectives such as water-air ratio and ion balance. Compared to existing technologies based on threshold-triggered hysteresis control, this invention overcomes the limitations of short-sighted control behavior, inability to avoid future risks, and difficulty in balancing multiple agronomic needs, thus achieving forward-looking and refined management. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the system module flow of the present invention;

[0035] Figure 2 This is a schematic diagram of the cultivation module of the present invention;

[0036] Figure 3 This is a schematic diagram of the active controllable siphon device module of the present invention;

[0037] Figure 4 This is a schematic diagram of the sensor network module of the present invention;

[0038] Figure 5 This is a schematic diagram of the central controller module of the present invention. Detailed Implementation

[0039] 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.

[0040] Please see the appendix Figure 1-5 This invention provides a supply system combining siphon-driven and fertilizer / pesticide concentration feedback regulation, comprising:

[0041] The cultivation module is used to hold the crops and contain the supply solution;

[0042] Specifically, the cultivation unit, exemplarily, can adopt a two-layer cultivation structure with an upper substrate and a lower nutrient solution. The upper layer is an inert substrate with stable physical and chemical properties, such as perlite, vermiculite, or rock wool, whose main function is to fix the crop roots and ensure excellent aeration in the root zone.

[0043] The lower part of the cultivation unit is a liquid storage space used to temporarily store the supply liquid during irrigation, allowing the crop roots to be immersed in it to absorb water and nutrients. This structure provides the roots with a growth environment in which gas, liquid, and solid phases coexist.

[0044] Furthermore, the active controllable siphon device is connected to the drain outlet of the cultivation unit. This device is used to actively drain the supply liquid from the cultivation unit in a manner that can be defined in real time by software, rather than relying on a fixed physical siphon phenomenon.

[0045] In this embodiment, the active controllable siphon device includes a siphon pipeline, and, exemplary, a piezoelectric ceramic microfluidic valve is integrated at a key location of the siphon pipeline, at the top of its U-shaped bend.

[0046] The central controller outputs a continuously variable electrical signal, such as a DC voltage signal of 0 to 5 volts, to the piezoelectric ceramic microfluidic valve through its internal digital-to-analog conversion module.

[0047] The piezoelectric ceramic microfluidic valve generates precise and predictable physical deformation based on the received electrical signal, which directly changes the effective flow area of ​​the siphon pipe. This relationship transforms the siphon drainage process from passive to active and controllable.

[0048] Furthermore, the sensor network is physically integrated into the supply liquid circulation path of the system, and it includes at least two sets of sensors respectively located at the supply end and the return end of the supply system.

[0049] A sensor array, exemplarily, is deployed within the main storage tank to monitor the raw chemical parameters of the supply solution to be pumped into the cultivation unit.

[0050] The return-end sensor array, exemplarily, is deployed in the drain outlet or return pipeline of the cultivation unit to monitor the chemical parameters of the supply liquid discharged by the active controllable siphon device after absorption by the crop roots.

[0051] Furthermore, each set of sensors at the supply end and the return end preferably includes an ion-selective electrode for measuring the concentration of specific nutrient ions, so as to achieve accurate quantification of the concentration of key nutrients such as nitrate ions, potassium ions, and calcium ions.

[0052] The central controller, which may be an industrial personal computer or a high-performance embedded system, is electrically connected and communicates with the active controllable siphon device and the sensor network. The central controller is used to execute the core control method of this invention.

[0053] The method first acquires the supply-side parameter data and return-side parameter data collected by the sensor network within the same irrigation cycle.

[0054] Subsequently, the central controller determines the net absorption of the supplied liquid by the crop based on the parameter data from the supply and return ends. Specifically, for the i-th ion, its net absorption mass within this cycle is calculated using the following formula:

[0055]

[0056] In the formula, ΔM obs,i V represents the periodic net absorbance mass observation of the i-th ion. cycle This indicates the total volume of liquid supplied in the current irrigation cycle. This represents the periodic average concentration of the i-th ion measured at the supply side. This represents the periodic average concentration of the i-th ion measured at the return end.

[0057] After determining the net absorption of the crop, the central controller adaptively updates a preset system dynamic model that describes the overall dynamic behavior of the system based on the net absorption.

[0058] In this embodiment, the system dynamic model is a nonlinear state-space model. The advantage of this model lies in its ability to describe the complex coupling relationships between multiple variables in the system. Its state vector exemplarily includes the ion concentration in the storage tank, the liquid level in the cultivation unit, and the estimated root zone ion concentration, which cannot be directly measured but is crucial for control.

[0059] To achieve an accurate description of the system, the system dynamic model is further coupled with several sub-models. One of them is the nonlinear hydrodynamic model of the active controllable siphon device, which is used to characterize the functional relationship between the siphon drainage rate and the control voltage output by the central controller.

[0060] The second is a parameterized crop ion absorption model, which describes the relationship between the crop's absorption rate of a specific ion and the current ion concentration in the root zone, environmental factors, and intrinsic physiological parameters related to the crop's own growth stage.

[0061] The process by which the central controller adaptively updates the system dynamic model can be specifically implemented using an extended Kalman filter algorithm. This algorithm uses the previously calculated net absorption ΔM obs,i As a key external observation.

[0062] The Extended Kalman Filter (EPF) compares model predictions with actual observations through iterative prediction and update steps. Based on the discrepancy, the algorithm can calibrate and correct state variables in the system's dynamic model, such as root zone ion concentration, and model parameters, such as the crop's maximum ion uptake rate, online and synchronously.

[0063] This process enables the system model to learn and adapt, allowing it to continuously track and fit changes in the physiological state of crops as they grow and their environment changes.

[0064] Finally, the central controller predictively generates coordinated control commands based on the real-time updated system dynamic model that more accurately reflects the true state of the system.

[0065] The generation of this instruction is specifically achieved by solving a model predictive control optimization problem. Model predictive control algorithms can calculate the optimal control input sequence based on consideration of the system's behavior in a future prediction time domain.

[0066] The cost function of the model predictive control optimization problem is specially designed, which includes not only the conventional state tracking term and control cost term, but also an agronomic engineering coupled optimization term.

[0067] The agronomic engineering coupling optimization term, for example, includes at least one of the following sub-terms: a root water-air ratio optimization term for regulating the root wet-dry cycle; an ion balance penalty term for maintaining the optimal ratio between key ions in the nutrient solution to prevent nutrient antagonism; and a fertilizer and pesticide risk avoidance term for ensuring absolute safety in concentration and duration of action when applying pesticides.

[0068] The design of this cost function transforms abstract, empirical agronomic knowledge into precise mathematical constraints that can be solved by computers, thereby achieving optimization results far exceeding those of conventional control methods.

[0069] In an exemplary workflow, the system first senses the crop's nutrient uptake over a cycle through a sensor network. The central controller calculates the net uptake and uses this data to update the system model via an extended Kalman filter, making the model parameters more closely reflect the crop's current actual uptake capacity.

[0070] Subsequently, the model prediction control module uses this updated model to predict the crop's needs in the future and, taking into account multiple objectives such as root health and nutrient balance, solves for the optimal control scheme.

[0071] The solution may include: fine-tuning the irrigation time of the main water pump, precise setting of the drainage rate of the active controllable siphon device, and coordinated adjustment of the amount of concentrated fertilizer solution added by peristaltic pumps.

[0072] The central controller sends the first step instruction of the scheme to each actuator for execution. In the next control cycle, the entire process of perception-learning-prediction-decision will be repeated, forming a highly intelligent closed-loop control system that is continuously optimized.

[0073] An active, controllable siphon device module is connected to the cultivation module and is used to discharge the supply liquid in a controllable manner;

[0074] Specifically, the active controllable siphon device includes a siphon pipe with a standard U-shaped bend in its physical structure. This pipe is connected to the drain outlet of the cultivation unit and is used to discharge the supply liquid in the lower liquid storage space of the cultivation unit.

[0075] To achieve active control of the siphon process, an electrically controlled microflow regulating component is integrated at a key hydraulic location in the siphon pipeline. For example, this component could be a piezoelectric ceramic microfluidic valve integrated at the top of the U-bend of the siphon pipeline, i.e., the location with the lowest pressure within the pipeline.

[0076] The piezoelectric ceramic microfluidic valve is connected to a digital-to-analog converter (DAC) output channel of the central controller via an electrical signal cable. This connection enables the central controller to output a precise, continuously variable control voltage signal to the piezoelectric valve.

[0077] In this embodiment, the control voltage signal can be in the range of 0 to 5 volts DC. The piezoelectric ceramic material has an inverse piezoelectric effect; when it receives a voltage signal from the central controller, it will produce a physical deformation at the micrometer level that is proportional to the voltage magnitude.

[0078] This physical deformation directly affects the flow channel within the siphon pipe, altering its effective flow area. Therefore, by outputting a specific voltage value through software, a specific physical flow area can be set, thus transforming the siphon pipe's drainage capacity from a fixed physical property into a variable that can be defined in real-time by the software.

[0079] As another optional implementation, the electrically controlled micro-flow regulating component can also employ a shape memory alloy actuator. By controlling the current flowing through the shape memory alloy actuator to control its temperature, and utilizing the deformation generated by its phase change to adjust the effective flow area of ​​the siphon pipe, active control of the drainage process can also be achieved.

[0080] To enable the central controller to accurately utilize the active controllable siphon device in its predictive control algorithm, a mathematical model needs to be established. This model describes the functional relationship between the siphon drainage rate and the control signal and system state.

[0081] Specifically, the instantaneous siphon drainage rate can be characterized by the following formula:

[0082]

[0083] In the formula, Q siphon (t) represents the instantaneous siphon drainage rate at time t, C d A represents the flow coefficient. eff (V ACS (t) represents the control voltage V at time t. ACS (t) determines the effective flow area of ​​the siphon pipe, g represents the acceleration due to gravity, and h bed (t) represents the real-time liquid level in the cultivation unit at time t.

[0084] Function A eff (V ACS The core characteristic function of this device is denoted by , which characterizes the relationship between control input and physical execution effect. This functional relationship can be obtained through a one-time experimental calibration of a specific device, and the data is stored in the central controller in the form of a lookup table or fitting function.

[0085] The mathematical model of the active, controllable siphon device, as a key sub-model, is embedded into the overall dynamic model of the system. This enables the central controller, particularly when executing model predictive control algorithms, to possess the following capabilities.

[0086] The model predictive control module of the central controller can use the above formula to accurately predict, at each time step within a future prediction time domain, the effect of applying a specific control voltage sequence on the liquid level h of the cultivation unit.bed What kind of impact will it have?

[0087] This predictive capability is fundamental to achieving advanced optimization of the root zone environment. For example, the root water-to-air ratio optimization term included in the cost function of the model predictive control optimization problem is implemented through this mechanism.

[0088] In an exemplary control process, in order to achieve an optimal ratio of root wetting time to drying time, the model predictive control algorithm will proactively plan an optimal control voltage V through optimization. ACS Trajectory.

[0089] This trajectory may include: outputting a lower voltage to obtain the maximum flow area when rapid drainage is needed, outputting a higher voltage to reduce the flow area when it is necessary to prolong root wetting time or slow down drainage, or even outputting the maximum voltage to completely interrupt the siphon process when necessary.

[0090] In this way, the active controllable siphon device described in this invention is no longer a simple switch or drainage channel. It becomes an actuator endowed with a precise mathematical model, which can be directly invoked by advanced control algorithms to finely shape the root zone microenvironment, thereby achieving an unprecedentedly precise control capability over the crop growth environment.

[0091] The sensor network module includes at least a supply-end sensor deployed at the supply end of the supply liquid and a return-end sensor deployed at the supply liquid return end.

[0092] Specifically, the sensor network is physically integrated into the supply liquid circulation path of the system. Its structural feature is that it includes at least two sets of sensor arrays respectively set at the supply end and return end of the supply system, forming a dual-channel differential measurement architecture.

[0093] The first sensor array, namely the supply-side sensor array, is exemplarily deployed within the system's main storage tank. Its function is to monitor, in real-time, the raw reference chemical parameters of the supply solution to be pumped into the cultivation unit by the main water pump.

[0094] The second sensor array, the return-end sensor array, is exemplarily deployed downstream of the drainage outlet of the cultivation unit or in the return pipeline. Its function is to monitor, in real-time, the chemical parameters of the supply solution discharged by the active, controllable siphon device after absorption by the crop roots.

[0095] Furthermore, in order to achieve accurate analysis of crop nutrient absorption behavior, each sensor array at the supply end and the return end preferably includes an ion-selective electrode (ISE) for measuring the concentration of specific nutrient ions.

[0096] These ion-selective electrodes, by way of example, may include nitrate ions (NO3). - Electrode, potassium ion K + Electrode, calcium ions Ca 2+ Electrode, magnesium ions Mg 2+ Electrode and phosphate ions (H2PO4) - Electrodes, etc., to cover the main nutrients required for crop growth.

[0097] In addition, each sensor array may include sensors for measuring general parameters of the supply fluid, such as conductivity (EC) sensors and pH sensors. The configuration and model of both sensor arrays are kept consistent to ensure the consistency and comparability of the measurement data.

[0098] All sensors in the sensor network are electrically connected and communicate with the data acquisition interface of the central controller via signal cables. The central controller can synchronously acquire real-time data from both the supply and return channels at a preset frequency.

[0099] The core technological advantage of the dual-channel sensor network lies in its ability to enable the central controller to accurately quantify the net change in the chemical composition of the supply liquid caused by crop life activities within a single irrigation cycle through differential calculation.

[0100] Specifically, when determining the net amount of fertilizer absorbed by the crop, the central controller first processes and integrates data from the sensor network after a complete irrigation cycle has ended.

[0101] In the control process of this invention, the net absorbed mass observation value ΔM obs,i The observations are used as external inputs to the extended Kalman filter algorithm running in the central controller.

[0102] The extended Kalman filter algorithm uses this real observation to correct its internal theoretical predictions based on the system dynamics model. The discrepancy between the two, i.e., the innovation, is used to drive the filter to update the state and parameters of the system model.

[0103] Therefore, the sensor network is not an isolated measurement module. By providing accurate and quantifiable crop net absorption data, it provides a basis for the adaptive learning of the system model, and is an indispensable technical feature for realizing the transformation from a static, open-loop control system to a dynamic, learning-capable closed-loop intelligent system.

[0104] The central controller module is connected to the active controllable siphon device module and the sensor network module. The central controller module is used for:

[0105] Acquire the supply liquid parameter data collected by the sensor network module at the supply end and the return end;

[0106] Based on the parameter data from the supply end and the return end, determine the net absorption amount of the supplied liquid by the crop;

[0107] Based on the net absorption, the preset system dynamic model is adaptively updated; and

[0108] Based on the updated system dynamic model, predictive collaborative control commands are generated to collaboratively control the actuators of the active controllable siphon device module and the supply unit.

[0109] Specifically, the central controller connects electrically and communicates with other functional modules in the system via its input / output interfaces. Specifically, its data acquisition interface connects to the sensor network to obtain fluid supply parameter data; its control output interface connects to both the actuator assembly and the active controllable siphon device to send coordinated control commands.

[0110] The central controller is used to execute the core control method of the present invention, which includes a series of logical steps such as: acquiring sensor network data, determining the net absorption of crops, adaptively updating the system dynamic model, and predictively generating cooperative control instructions.

[0111] To implement this control method, the central controller first presets and maintains a nonlinear state-space model that describes the overall dynamic behavior of the system. This model mathematically represents the intrinsic relationships between the system's variables, and its general form can be expressed as:

[0112]

[0113] In its formula, Let f(t) represent the derivative of the state variable x(t) with respect to time t, x(t) be the state vector of the system at time t, f(...) be a function describing the relationship between the state variable, control input and other parameters, u(t) be the control input at time t, p(t) be the parameter at time t, and d(t) be the disturbance or disturbance at time t.

[0114] The system's dynamic model is further coupled with the nonlinear fluid dynamics model of the active controllable siphon device and the parameterized ion absorption model characterizing the crop's absorption capacity.

[0115] After determining the net crop uptake, the central controller performs an adaptive model update step. The purpose of this step is to use real measurement data to calibrate and correct the system's dynamic model online, enabling it to continuously and accurately reflect the actual state of the system.

[0116] In this embodiment, the adaptive update process is implemented using an Extended Kalman Filter (EKF) algorithm. To simultaneously estimate the system state and model parameters, an augmented state vector x is first constructed. a It merges the original state vector x and the time-varying parameter vector p.

[0117] The execution process of the extended Kalman filter algorithm includes iterative prediction and update steps.

[0118] In the prediction step, the filter uses the discretized system dynamic model f based on the posterior estimate from the previous time step and the control input. d To predict the prior state estimate at the current moment. and the prior error covariance matrix P k |k-1.

[0119]

[0120]

[0121] In its formula, This represents the state estimate at time k, given information at time k-1, f. d (...) represents a dynamic model function. U represents the state estimate at time k-1. k-1 d represents the control input at time k-1. k P represents the disturbance or interference at time k. k|k-1 Let A represent the covariance matrix of the state estimate at time k, based on information from time k-1. k P represents the state transition matrix, which describes the state transition process from time k-1 to time k. k-1|k-1 Let W represent the covariance matrix of the state estimate at time k-1. k This represents the process noise or uncertainty matrix.

[0122] In the update step, the filter utilizes the crop net uptake ΔM determined by the sensor network. obs,i The actual observed value z at the current moment k .

[0123] By calculating the Kalman gain K k Furthermore, by incorporating the deviation between the observed and predicted values, the prior estimate is corrected to obtain the posterior state estimate at the current time. and the posterior error covariance matrix P k|k .

[0124]

[0125]

[0126] P k|k =(IK k H k )P k|k-1 ;

[0127] In the formula, K k This represents the Kalman gain at time k. H represents the observation matrix k The transpose of V k Let k represent the observation noise covariance matrix at time k. The inverse of the covariance matrix of the observation equation is represented. Let z represent the posterior state estimate at time k. k This represents the actual observed value at time k. Represents the measurement model function. P represents the residual between the actual observed value and the predicted value of the measurement model. k|k Let I denote the posterior state estimate covariance matrix at time k, and let K denote the identity matrix. k H k This represents the product of the Kalman gain and the observation matrix, (IK) k H k ) represents the adjustment factor in the improved state covariance matrix.

[0128] Finally, based on the updated system dynamic model that more accurately reflects the actual system situation, the central controller executes the step of predictively generating cooperative control commands. In this embodiment, this step is achieved by solving a model predictive control optimization problem. At the beginning of each control cycle, the central controller solves the following optimal control problem to determine the future prediction time domain N. p The optimal control sequence within:

[0129]

[0130] In its formula, The objective function J to be minimized is represented by the control input. Represents the range from 0 to N p Summing N to find -1 p It is the length of the predicted time domain. Represents state x k With reference state x ref,k The weighted sum of squared errors between the given values, where Q is the corresponding weighting matrix. Indicates control input u k The weighted sum of squares, R is the corresponding weighting matrix, ΔJ agro,kThis indicates additional costs or dynamic adjustments related to a particular type of agriculture.

[0131] Solving this optimization problem is subject to multiple constraints, including the system's dynamic model, actuator physical limitations, and system state safety boundaries. In the above formula, x... k and u k These represent the predicted future state and the control input, respectively; x ref,k The ideal state reference trajectory is defined based on the agronomic knowledge base; Q and R are weight matrices; ΔJ a gro,k is a specially designed agronomic engineering coupling optimization term.

[0132] The agronomic engineering coupled optimization terms transform advanced agronomic requirements into mathematical forms. For example, they may include: a root water-air ratio optimization term for optimizing the root wet-dry cycle; an ion balance penalty term for actively preventing nutrient antagonism; and a fertilizer and pesticide risk avoidance term for ensuring safe pesticide application. After obtaining the optimal control sequence, the central controller only executes the first control command in the sequence. The data is then sent to each actuator for execution. In the next control cycle, the entire process of perception-learning-prediction-decision-making described above will be repeated based on the new system state, thus forming a closed-loop control system with rolling optimization.

[0133] 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 supply system combining siphon-driven and fertilizer / pesticide concentration feedback regulation, characterized in that, include: The cultivation module is used to hold the crops and contain the supply solution; An active, controllable siphon device module is connected to the cultivation module and is used to discharge the supply liquid in a controllable manner; The sensor network module includes at least a supply-end sensor deployed at the supply end of the supply liquid and a return-end sensor deployed at the supply liquid return end. The central controller module is connected to the active controllable siphon device module and the sensor network module. The central controller module is used for: Acquire the supply liquid parameter data collected by the sensor network module at the supply end and the return end; Based on the parameter data from the supply end and the return end, determine the net absorption amount of the supplied liquid by the crop; Based on the net absorption, the preset system dynamic model is adaptively updated; as well as Based on the updated system dynamic model, predictive collaborative control commands are generated to collaboratively control the actuators of the active controllable siphon device module and the supply unit.

2. The supply system combining siphon drive and fertilizer / pesticide concentration feedback regulation according to claim 1, characterized in that, The active controllable siphon device module includes a piezoelectric ceramic microfluidic valve or shape memory alloy actuator installed on the siphon pipeline. The central controller module controls the siphon drainage rate by outputting a continuously variable electrical signal to adjust the opening of the valve or actuator.

3. The supply system combining siphon drive and fertilizer / pesticide concentration feedback regulation according to claim 1, characterized in that, The sensor network module includes ion-selective electrodes for measuring the concentration of specific nutrient ions at both the supply and return ends.

4. The supply system combining siphon drive and fertilizer / pesticide concentration feedback regulation according to claim 1, characterized in that, The central controller module determines the net absorption of the supplied liquid by the crop, specifically by calculating the net absorption of one or more ions by the crop within one or more irrigation cycles based on the ion concentration difference between the supply end and the return end and the total amount of circulating liquid.

5. The supply system combining siphon drive and fertilizer / pesticide concentration feedback regulation according to claim 1, characterized in that, The system dynamic model is a state-space model, and its state vector includes at least the ion concentration in the storage tank, the liquid level in the cultivation unit, and the estimated ion concentration in the root zone.

6. The supply system combining siphon drive and fertilizer / pesticide concentration feedback regulation according to claim 5, characterized in that, The system dynamic model is coupled with the nonlinear fluid dynamics model of the active controllable siphon device module and the parameterized ion absorption model characterizing the crop's absorption capacity.

7. The supply system combining siphon drive and fertilizer / pesticide concentration feedback regulation according to claim 1, characterized in that, The central controller module adaptively updates the system dynamic model, specifically through an extended Kalman filter or a moving time-domain estimator. This filter or estimator uses the net absorption as an observation to calibrate the state and parameters in the model online.

8. The supply system combining siphon drive and fertilizer / pesticide concentration feedback regulation according to claim 1, characterized in that, The central controller module predictively generates the coordinated control instructions by solving a model predictive control optimization problem to obtain the optimal control sequence in a future prediction time domain.

9. A supply system combining siphon drive and fertilizer / pesticide concentration feedback regulation according to claim 8, characterized in that, The cost function of the model predictive control optimization problem includes a state tracking term, a control cost term, and an agronomy-engineering coupled optimization term.

10. The supply system combining siphon drive and fertilizer / pesticide concentration feedback regulation according to claim 9, characterized in that, The agronomy-engineering coupling optimization term includes at least one of the following: Root water-air ratio optimization term used to regulate the root wet-dry cycle; Ion balance penalty used to maintain the optimal ratio of key ions in the nutrient solution; Fertilizer and pesticide risk mitigation measures are used to ensure safe concentration and duration of action when applying pesticides.

Images