Intelligent irrigation system and method for crop root system

Abstract

The invention relates to the field of intelligent irrigation, and discloses an intelligent irrigation system and method for a crop root system, and the system comprises a root zone sensing module which is used for collecting soil environment parameters through multi-mode soil sensor arrays disposed at soil layers of different depths, generating a sensing matrix, and transmitting the sensing matrix to a root zone sensing module; the sensing matrix is used for representing water potential, pH value, nutrient concentration and temperature information in each soil layer; and the root system density analysis module is connected with the root zone sensing module, and is used for constructing a root system density weight matrix based on root system space distribution information obtained by underground computed tomography, and multiplying the sensing matrix by the root system density weight matrix element by element to generate a root zone state vector. Accurate generation of regulation and control signals is achieved, the micro-fluidic execution module can adjust the irrigation flow and the fertilization proportion in real time according to the actual requirements of the root zone, and therefore it is guaranteed that the rhizosphere environment is always close to the optimal physiological interval.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent irrigation, and in particular to an intelligent irrigation system and method for crop root systems. Background Art

[0002] With the development of facility agriculture and smart agriculture, data-driven water and fertilizer control systems based on sensor networks have been gradually applied in greenhouses, fruit and vegetable production bases, and cash crop cultivation scenarios. These systems aim to improve crop yield, quality, and water and fertilizer use efficiency by precisely controlling irrigation volume and fertilization ratios. However, existing technologies still have many technical limitations in terms of control response mechanisms, control model adaptability, and data feedback utilization, making it difficult to support dynamic and precise management of the crop rhizosphere environment.

[0003] Most existing water and fertilizer control systems still rely on preset thresholds or timed programs, lacking real-time sensing and dynamic feedback mechanisms for the crop's actual root zone status. Although some systems have incorporated sensors based on soil moisture or conductivity, these systems often fail to achieve effective closed-loop control. Even when sensor data is available, they lack targeted modeling and strategy iteration mechanisms. This results in control actions often lagging behind environmental changes, making stable operation difficult under changing weather and soil conditions.

[0004] In addition, most existing feedback mechanisms only make regulatory decisions based on current instantaneous data, lack the accumulation and memory function of historical information, and are unable to identify the differences between short-term anomalies and long-term trends. They are prone to produce misadjusted signals or system oscillations under interference conditions, affecting the stability of the control system and the physiological safety of crops.

[0005] Furthermore, although some current systems support remote monitoring or regular data uploads, they fail to establish a unified feedback window and model backtracking path, resulting in the inability of perception data to feed back into the strategy model and insufficient system self-optimization capabilities.

[0006] In summary, the existing water and fertilizer control systems still have significant shortcomings in the real-time performance, adjustment accuracy, anti-disturbance ability and adaptive optimization ability of the control model. It is urgent to build a data feedback closed-loop control module that integrates multi-source information, has a memory feedback mechanism and dynamic strategy self-update capability, so as to solve the key problems in precision agriculture such as response lag between root zone status and irrigation strategy, unstable control, and uneven resource allocation. Summary of the Invention

[0007] The purpose of the present invention is to provide a crop root system intelligent irrigation system and method, which solves the problems of the existing water and fertilizer control system such as lack of real-time perception of root zone status, regulation lag, and lack of adaptive feedback capability.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: A crop root system intelligent irrigation system, comprising: The data acquisition module is used to collect soil environmental parameters and root spatial distribution information through a multimodal soil sensor array and underground computed tomography deployed at different soil depths. Based on the collected information, a perception matrix and a root density weight matrix are generated, and the two matrices are multiplied element by element to generate a root zone state vector. a root zone state modeling module, connected to the data acquisition module, configured to receive the root zone state vector and crop growth stage data, perform training and prediction using a federated learning modeling approach, and output a future root zone state vector; an irrigation strategy decision module, connected to the root zone state modeling module, for taking the future state vector as input, constructing a Markov decision process model, calculating the optimal irrigation strategy and outputting an irrigation control decision; an irrigation path decision module, connected to the irrigation strategy decision module, for calculating the optimal irrigation path and outputting irrigation path control instructions based on a preset irrigation network topology and path weight matrix in combination with the irrigation control decision; an execution module, connected to the irrigation path decision module, generating an irrigation scheduling plan based on the system energy storage status data and the drought degree of each plot, the crop growth stage, and the energy feedback potential, and controlling the microfluidic device to execute the irrigation task of the target area according to the irrigation scheduling plan; A data feedback closed-loop module is connected to the execution module and is used to collect water and fertilizer absorption feedback information of the root zone in real time, update the perception matrix and the root zone state vector, and return the updated root zone state vector to the root zone state modeling module to build a closed-loop control system of perception, modeling, decision-making, execution and feedback.

[0009] Preferably, in the data acquisition module, the multimodal soil sensor array includes a water potential sensor, a conductivity sensor, a pH sensor, a temperature sensor and an ion selective electrode sensor. The sensor array is arranged at multiple depth levels in the vertical direction to collect the water potential, conductivity, pH value, temperature and nitrogen, phosphorus and potassium ion concentrations of each soil layer respectively. The collected data are classified by layer to form a data group, and finally the data group is converted into a perception matrix for characterizing the soil environmental status of the root zone.

[0010] Preferably, the data acquisition module further includes an underground computed tomography device and an image recognition processing unit, wherein the image recognition processing unit is used to segment the scanned image into root areas, calculate the root volume density value of each depth layer based on the segmentation result, summarize the root density data group by level, and finally convert the root density data group into a root density weight matrix; The root density weight matrix and the perception matrix are element-by-element multiplied by the Hadamard product to obtain the root zone state vector. The calculation formula of the element-by-element multiplication is: R ij =F ij w i ; Among them, F ij is the jth environmental parameter of the i-th layer in the perception matrix; w i is the root density weight of the layer; The root zone state vector is used to reflect the resource response capabilities of different soil layers and root system structures.

[0011] Preferably, the root zone state modeling module includes an edge computing node and a federated modeling server. The edge computing node is used to locally receive the root zone state vector and crop growth stage data and perform preliminary model training. The crop growth stage data includes sowing time, growth stage identifier, current plant height, biomass estimation value, and daily accumulated temperature data. The root zone state modeling module uses a federated learning modeling method for training and prediction. The step of outputting the future root zone state vector includes: The edge node performs preliminary training of the state prediction model based on local historical data and periodically sends model gradient parameters to the federated modeling server; The federated modeling server aggregates model parameters of multiple edge nodes using a weighted average algorithm and updates the global model; The global model synchronously transmits parameters to the edge nodes after each round of update. Finally, each edge node jointly models the input root zone state vector and crop growth stage data based on the updated model, and finally outputs the future state vector of the root zone; The root zone future state vector is used to describe the predicted state of the soil environment at a future moment.

[0012] Preferably, in the irrigation strategy decision module, the step of calculating the optimal irrigation strategy and outputting the irrigation control decision includes: The irrigation strategy decision module constructs a Markov decision process model based on the future state vector of the root zone. The Markov decision process model is expressed as: M=(S,A,P,R,γ); Where S is the state space, representing the root zone environment state at different time points; A is the action space, representing the combination of various irrigation amounts and irrigation time intervals; P is the transition probability; R is the reward function; and γ is the discount factor, which is used to measure the influence of future rewards on the current decision. In the Markov decision process model, strategy selection is performed through a reward function, and the functional form of the reward function is: R(s,a)=-(λ1·C w(a)+λ2·C e (a)+λ3·D(s,s * )); Among them, C w (a) is the water cost of the irrigation operation; C e (a) is the energy consumption cost; D(s,s * ) is the state s and the target state s * Deviation; λ1, λ2 and λ3 are weighting coefficients; In order to achieve the long-term optimal irrigation control goal, the irrigation strategy decision module calculates the value function of each state according to the Bellman expectation equation: the calculation formula is: Among them, P(s ′ |s,a) means taking action a in state s and transitioning to state s ′ The probability of s; V(s) is the value function of state s, which represents the maximum expected cumulative reward that can be obtained from state s under the optimal strategy; V(s ′ ) is the subsequent state s ′ The value function represents the value of the state s ′ The maximum expected cumulative reward that can be obtained from the departure; Based on the above value function, the value iteration method is used to solve the optimal policy function, which is: π * (s)=argmax a∈A Q(s,a); Among them, π * (s) is the optimal policy function, which selects the optimal action under state s; Q(s,a) is the action-value function, which represents the expected cumulative reward obtained after taking action a under state s; argmax is the action that maximizes the expression in the brackets; after solving, the future state vector of the root zone is mapped to the current state, and the optimal policy function selects the irrigation action and outputs the corresponding irrigation control decision.

[0013] Preferably, the irrigation path decision module constructs a path weight matrix based on the irrigation network topology, and uses the Dijkstra algorithm to calculate the irrigation path that satisfies the shortest energy consumption or minimum delay starting from the current irrigation source node, and generates a path control instruction; The irrigation network topology is composed of a node set and a connection relationship set. The elements in the node set represent water source nodes, branch nodes, and terminal water outlet nodes in the irrigation pipe network. The elements in the connection relationship set represent that there are pipe connections between nodes. The elements in the path weight matrix represent the path energy consumption and time delay between nodes, and the path energy consumption and time delay are calculated based on the parameters of the actual irrigation pipe diameter, length, head loss and solenoid valve response time; The generated control instructions include the opening and closing sequence of the nodes involved in the path and their control elements; The generated control instructions are called by the irrigation control execution unit to complete the path guidance.

[0014] Preferably, the execution module includes a battery status acquisition unit and a priority calculation unit. The priority calculation unit generates an irrigation scheduling plan based on the drought degree of each plot, the crop growth stage, and the energy feedback potential. The specific steps include: obtaining the root zone status parameters corresponding to each irrigation unit, including the current water potential, water deficit, evapotranspiration prediction value, and crop growth stage parameters; A multi-factor irrigation priority scoring function is constructed based on the obtained parameters. The function form is: P i =α1·F d (i)+α2·F g (i)+α3·F e (i); Among them, P i is the priority score of the i-th irrigation unit; F d (i) is the water deficit index function, which represents the deviation between the current soil moisture state and the set lower limit; F g (i) is the crop growth stage function, reflecting the sensitivity weight of the key growth period; F e (i) is the evapotranspiration prediction function, which estimates the water consumption rate per unit time based on future meteorological conditions; α1, α2, and α3 are weight coefficients, which are set according to the target control strategy; Sort by priority score of each unit to obtain a scheduling order list of irrigation units; Taking into account the water source capacity, power load, and path accessibility of the irrigation system, a specific irrigation scheduling plan is generated based on the scheduling sequence list. The plan includes the start time, expected irrigation duration, and corresponding control path for each irrigation unit.

[0015] Preferably, the execution module further includes an adjustable proportional valve assembly and a micro pump array for accurately controlling the ratio of water to fertilizer according to an irrigation scheduling plan.

[0016] Preferably, the data feedback closed-loop module includes a root zone sensor feedback channel and a data fusion unit, which is used to collect the root zone absorption response signal after irrigation and return the fused data to the root zone state modeling module to complete the data closed-loop update.

[0017] A method for intelligent irrigation of crop root systems, comprising the following steps: S1, obtain multi-layer soil environmental parameters of the crop root zone through the data acquisition module and generate a future state prediction vector; S2. The irrigation strategy decision module flattens the state vector, constructs a Markov decision process model, calculates the state value function based on the Bellman expectation equation, further derives the optimal strategy function, and outputs the optimal irrigation action; S3. The irrigation path decision module constructs a path weight matrix based on the current irrigation network topology, uses the Dijkstra algorithm to calculate the shortest energy consumption or minimum delay path from the irrigation source node to the target node, and generates path control instructions; S4. The execution module constructs an irrigation priority scoring function based on the real-time root zone status and crop water requirements. It calculates and ranks the priorities of each irrigation unit based on factors such as water deficit level, growth stage, and evapotranspiration prediction, and generates an irrigation scheduling plan based on resource constraints. S5. Input the optimal irrigation action, path control instruction, and irrigation scheduling plan as an irrigation path control instruction, and drive the irrigation actuator to sequentially complete the target irrigation operation; S6. Update the root zone status in real time after irrigation is executed, and feed the monitoring data back to the system for subsequent strategy optimization and model update.

[0018] In summary, the present invention includes at least one of the following beneficial technical effects: 1. This invention achieves precise generation of control signals by introducing a target-measured deviation-driven model and combining it with a dual-factor regulation function of moisture and conductivity. This enables the microfluidic execution module to adjust the irrigation flow and fertilization ratio in real time according to the actual needs of the root zone, thereby ensuring that the rhizosphere environment is always close to the optimal physiological range.

[0019] 2. The present invention enhances the robustness of the control signal to abnormal data and environmental disturbances by setting a deviation integral memory model and a threshold suppression mechanism. Before the control strategy is executed, the system automatically evaluates the cumulative error intensity and screens whether to trigger the adjustment instruction based on the memory factor, thereby avoiding frequent misadjustments or system oscillations caused by short-term disturbances and improving the stability and reliability of long-term operation.

[0020] 3. By adopting standardized preprocessing and multi-source information fusion mechanisms, the present invention enables the feedback control model to have wide adaptability across regions and seasons. After uniformly normalizing and standardizing various types of sensor data, it effectively alleviates the interference of different sensor types, sampling accuracy and external environmental conditions on the control algorithm, providing a data foundation for unified modeling and generalized deployment.

[0021] 4. By constructing a feedback window and a control effect backtracking mechanism, the present invention forms a self-evolutionary closed loop of "execution-feedback-reoptimization", which has strong long-term adaptive capabilities. The control system evaluates the effectiveness of the control strategy based on the feedback error after each round of irrigation, and iteratively corrects the gain coefficient and ratio parameters, so that the water and fertilizer application strategy gradually fits the plant growth needs and soil response characteristics.

[0022] 5. Through the tightly coupled design of the execution module and the data feedback module, the present invention establishes a high-resolution closed-loop path from the root zone status to the water and fertilizer output. The system can dynamically generate regional-level control parameters based on the feedback data of each region, and coordinate with the main control system to match the scheduling rhythm, thereby achieving energy-saving operation and optimal utilization of system resources while ensuring control accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a system module diagram of the present invention; Figure 2 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0024] The following is combined with Figure 1 -Attached Figure 2 , the present invention is described in further detail.

[0025] The present invention provides a crop root system intelligent irrigation system, such as Figure 1 As shown, the crop root system intelligent irrigation system may include the following modules: The data acquisition module is used to collect soil environmental parameters and root spatial distribution information through a multimodal soil sensor array and underground computed tomography deployed at different soil depths. Based on the collected information, a perception matrix and a root density weight matrix are generated, and the two matrices are multiplied element by element to generate a root zone state vector. a root zone state modeling module, connected to the data acquisition module, configured to receive the root zone state vector and crop growth stage data, perform training and prediction using a federated learning modeling approach, and output a future root zone state vector; an irrigation strategy decision module, connected to the root zone state modeling module, for taking the future state vector as input, constructing a Markov decision process model, calculating the optimal irrigation strategy and outputting an irrigation control decision; an irrigation path decision module, connected to the irrigation strategy decision module, for calculating the optimal irrigation path and outputting irrigation path control instructions based on a preset irrigation network topology and path weight matrix in combination with the irrigation control decision; an execution module, connected to the irrigation path decision module, generating an irrigation scheduling plan based on the system energy storage status data and the drought degree of each plot, the crop growth stage, and the energy feedback potential, and controlling the microfluidic device to execute the irrigation task of the target area according to the irrigation scheduling plan; A data feedback closed-loop module is connected to the execution module and is used to collect water and fertilizer absorption feedback information of the root zone in real time, update the perception matrix and the root zone state vector, and return the updated root zone state vector to the root zone state modeling module to build a closed-loop control system of perception, modeling, decision-making, execution and feedback.

[0026] The following is a detailed description of each component in the system of the present invention.

[0027] For the data acquisition module: In this embodiment, the data acquisition module is used to perform real-time monitoring and multi-dimensional perception of the soil environmental status of the crop root zone, providing data support for subsequent modules such as root density analysis, root zone status modeling, and irrigation strategy decision-making.

[0028] In this embodiment, the data acquisition module is primarily deployed within the soil of the crop root zone. Preferably, a multi-point sensor array, either embedded or inserted, can be employed, positioned at varying depths and horizontal positions within the soil profile of the crop root zone. Sensor selection can be tailored to the soil properties to be measured and the target crop type to ensure representative and practical data.

[0029] The sensing content includes the following parameters: volumetric moisture content of the root zone soil, soil water potential, soil electrical conductivity, soil temperature, and root activity signals (such as rhizosphere electrical impedance).

[0030] The above parameters are collected in real time by the corresponding sensor modules. Preferably, volumetric moisture content and water potential parameters are obtained through the time domain reflectometry principle, soil conductivity is measured by electrodes, temperature parameters are sensed by thermistors, and rhizosphere signals can be realized based on impedance spectrum analysis.

[0031] In this embodiment, the multidimensional data collected by the data acquisition module is transmitted back to the central data node in real time in the form of a time series. Preferably, the transmission method can be wireless communication (such as LoRa, ZigBee, NB-IoT) or wired communication (RS485, CAN bus, etc.). The specific communication method is determined by the complexity of the application scenario environment and the density of the deployment to ensure the stability and real-time performance of the data transmission.

[0032] To ensure spatial continuity and dynamic response of the sensed data, this embodiment preferably arranges the sensor array at multiple depth levels along the vertical direction of the soil profile. Transversely, it can be viewed as a two-dimensional matrix distribution, forming a two-dimensional multi-point sensing network. This arrangement enables continuous monitoring of soil environmental conditions at different depths and in different areas, avoiding the information island effect caused by single-point measurement.

[0033] In terms of data processing, in this embodiment, the raw sensor data collected by the data acquisition module is first pre-processed, including: De-noising and filtering of collected values; Interpolation and completion of missing data; Outlier removal and correction.

[0034] Preferably, the filtering method may be sliding mean filtering, median filtering or wavelet denoising. For missing data, linear interpolation or spline interpolation may be used to fill in the missing data based on the data trends of adjacent time points or adjacent spatial points.

[0035] After completing the preprocessing, the data acquisition module stores the soil environmental parameters of each monitoring point in a time series format, forming the following time series matrix: Among them, M t is the multi-point multi-dimensional environmental monitoring matrix at time t; is the collected value of the jth environmental parameter at the i-th monitoring point at time t; k is the number of monitoring points; and n is the number of monitoring parameter dimensions.

[0036] The timing matrix serves as the data basis of the system of the present invention and is directly transferred to subsequent modules.

[0037] In this embodiment, the data acquisition module also analyzes and quantifies the spatial distribution characteristics of roots within the crop root zone, outputting a root density distribution function that can be used for state modeling and irrigation strategy optimization. Based on data input from the sensing system, including rhizosphere electrical signal responses and electrical impedance change signals, this module uses an algorithmic model to identify and reconstruct root distribution states.

[0038] In this embodiment, the data acquisition module collects the rhizosphere electrical impedance change information and multi-frequency electrical response signals to extract multi-channel time series features that can reflect the root activity and distribution range; For the root distribution identification method based on electrical signals, in this embodiment, an array of collection points is arranged at different depths and horizontal positions in the soil, and the electrical response spectrum is obtained through multi-frequency impedance measurement. Preferably, the root density inference model can be in the following form: where ρ(x, z) is the root density function value at coordinate (x, z) in the soil profile; ΔZ(x, z, ω) is the impedance change at that point at different frequencies ω; and f(·) is a nonlinear fitting function obtained through calibration and training, which is used to map electrical signal changes to root density.

[0039] To improve the spatial accuracy of root density estimation, this example uses a Bayesian interpolation method to continuously complete sparse data points and construct a continuous root density field. This density field is spatially represented as a two-dimensional or three-dimensional grid structure, with each cell corresponding to a root density value, enabling a detailed depiction of the spatial distribution of the root system.

[0040] Furthermore, in this embodiment, to avoid misjudgment of electrical responses due to soil heterogeneity, a soil conductivity baseline correction strategy is preferably introduced. Specifically, blank response data is collected in a root-free area as a correction baseline. The actual measured values ​​are then normalized to eliminate the influence of factors such as soil moisture, thereby improving the robustness of root density estimation.

[0041] In this embodiment, the root density distribution result is output in the form of a tensor and is uniformly stored in the following structure: R = {ρ ij ∣i=1,…,M; j=1,…,N}; Among them, ρ ij is the root density value at the i, j position in the root zone grid, M and N are the number of grid points in the horizontal and vertical sections, respectively.

[0042] The tensor structure data can be directly used as one of the input parameters of the root zone state modeling module to construct a root density weighted state prediction model.

[0043] Preferably, in this embodiment, a root density weight function w is introduced ij , in order to strengthen the focus on high root density areas during the modeling phase, the function is defined as follows: The above-mentioned weight function can be used as an attention mechanism factor in the subsequent state modeling module to guide the model to give priority to root-active areas or areas with high water use efficiency in decision-making, thereby achieving accurate matching of water and fertilizer resources and regional differentiated delivery.

[0044] Furthermore, the data acquisition module in this embodiment supports a periodic update mechanism. That is, during continuous system operation, the data acquisition module can periodically reacquire image or electrical signal data and update root distribution information using a sliding time window mechanism. This supports dynamic root zone state modeling and improves the system's adaptability to changes in crop growth stages.

[0045] For the Root Zone State Modeling module: In this embodiment, the root zone state modeling module is configured to model and predict the water state, environmental response and dynamic evolution trend of the crop root zone based on the multi-source soil environment data and root density distribution information provided by the data acquisition module, so as to provide data support and target reference for the irrigation strategy decision module.

[0046] The modeling object of the module preferably includes the water content variation trend of each monitoring position point in the crop root zone, the response relationship of root water absorption activity, the migration evolution characteristics of soil water in the spatial and temporal dimensions, and the water gradient variation law driven by the environment.

[0047] In this embodiment, the root zone state modeling module adopts a multi-physical quantity fusion modeling method to jointly model the spatial structure, temporal dynamics of the root zone environment and the root density distribution. The parameter matrix M obtained by the perception module is used as the initial input of the model. t As the initial input of the model, the parameter matrix is in the form of: The matrix records the n types of environmental parameter data collected at time t at the k spatial monitoring points, including soil water potential, volumetric water content, conductivity, temperature, etc.

[0048] At the same time, the root density tensor provided by the data acquisition module and the weight function jointly constitute the root zone structure weight factor in the model. By explicitly introducing the spatial structure information into the modeling function, the response ability of the model to the heterogeneity of root distribution can be improved.

[0049] Preferably, the core prediction framework of the root zone state modeling module is constructed based on a weighted state evolution equation, which is mathematically expressed as follows: wherein, is the volumetric water content of the grid i, j position at time t; is the water potential of the position; D(θ) is the water diffusion coefficient function related to the water content; α, β are the regulation coefficients for balancing water diffusion and root water absorption; ρ ij is the root density value, reflecting the water absorption potential of the crop at the position; f(·) is the root water absorption function, which can preferably use an empirical or fitted water potential response function, such as the Feddes type or Gardner type function.

[0050] The state evolution formula is established based on the coupling of the improved two-dimensional Richardson equation and the root water absorption theory, and has the ability to describe the migration and consumption process of water in the heterogeneous soil-root medium.

[0051] To simulate the dynamic changes of soil moisture on a time scale, this embodiment uses a discrete time stepping mechanism, combined with a finite difference or finite volume numerical solution to perform numerical calculations of state evolution, and updates the following state vector at each time step: Among them, Θ t is the set of overall state tensors of the root zone at time t, including the state variables of all spatial cells at that time; is the state variable value in the grid cell of the i-th row and the j-th column at time t, preferably the volumetric moisture content, water potential or other soil state parameters; i is the ordinal index of the spatial grid point in the horizontal direction (such as the X-axis direction), ranging from 1 to M; j is the ordinal index of the spatial grid point in the vertical direction (such as the Z-axis direction), ranging from 1 to N; M is the number of spatial divisions of the root zone model in the horizontal direction, that is, the number of horizontal grids; N is the number of spatial divisions of the root zone model in the vertical direction, that is, the number of vertical grids; t is a discrete time index, used to represent a specific time point during the model operation.

[0052] The above state vector represents the soil moisture state information at each location in the complete two-dimensional root zone grid, which serves as an important basis for the subsequent generation of irrigation criteria.

[0053] In addition, in order to improve the modeling accuracy and adapt to the characteristics of different soil types and crop varieties, this embodiment preferably introduces a parameter adaptive adjustment mechanism, that is, the diffusion coefficient function and water absorption function in the model are fitted or fine-tuned according to historical observation data, thereby enhancing the generalization and robustness of the model.

[0054] Preferably, the fitting process can be implemented based on the least squares optimization method or the Bayesian posterior estimation method. The objective function is defined as follows: in, is the measured value; is the model simulation value; is the parameter function to be estimated.

[0055] In this embodiment, the root zone state modeling module supports a periodic state update mechanism, combining continuous sensor data streams with the latest root density distribution map to achieve dynamic modeling and real-time updates of the root zone state. This mechanism is preferably based on a sliding time window algorithm. Specifically, within each period, it selects the latest n time series samples for modeling iteration, eliminating outdated data to ensure that the model output reflects the current root zone state.

[0056] For the irrigation strategy decision module: In this embodiment, the irrigation strategy decision module generates optimized irrigation plans and decisions based on the root zone moisture status information output by the root zone state modeling module, combined with real-time meteorological data and soil environmental parameters. This module intelligently adjusts irrigation volume and timing by comprehensively considering multiple dimensions of information, including soil moisture, root water absorption capacity, and crop water requirements, to achieve precise water management.

[0057] In this embodiment, the input data of the irrigation strategy decision module comes from the following parts: Root zone moisture status information: provided by the root zone state modeling module, including the moisture status of each location in the root zone, such as volumetric moisture content, soil water potential, etc. Root distribution and density information: from the data acquisition module, providing information on the distribution of crop roots at different depths and locations; meteorological data: including environmental factors such as precipitation, evaporation intensity, and temperature, updated in real time; Crop water requirement model: Generates the corresponding crop water requirement curve based on parameters such as crop variety, climate conditions, soil type, etc.

[0058] In this embodiment, the core function of the irrigation strategy decision module is to generate an irrigation plan by comprehensively analyzing various input information, establishing a multi-objective optimization model, and then generating an irrigation plan. The optimization objectives of this plan may include maximizing water resource utilization efficiency, ensuring the water required for crop growth, and avoiding excessive water waste.

[0059] To achieve the above goals, this embodiment adopts a decision framework based on constrained optimization, which is specifically formulated as the following optimization problem: Where I is the decision vector, which includes the allocation plan of irrigation amount; w ij is the weight value of the i,j grid point in the root zone, which is usually related to the root density in the area; The target moisture content for the location, usually preset based on crop water requirements, meteorological conditions, and soil type; is the moisture content of the location at the current moment (from the root zone state modeling module); I ij is the irrigation amount, which represents the amount of water replenished at the i, j grid point under the irrigation strategy; is the irrigation amount at the corresponding position at the previous moment; λ is the control coefficient, which is used to balance the difference between the irrigation amount adjustment and the target moisture content.

[0060] The goal of this optimization problem is to minimize the moisture difference at each location while taking into account the smoothness of the irrigation amount, that is, to avoid frequent large adjustments as much as possible to reduce system instability and water waste.

[0061] Based on the constraint optimization decision-making framework, the following constraints are considered in the irrigation strategy decision-making process in this embodiment: First, the moisture changes in each region should satisfy the physical law of water conservation. Specifically, the increase and decrease of soil moisture is the result of the combined effects of precipitation, evaporation, root water absorption, and irrigation, which can be converted into the mathematical formula: Among them, S ij is the water uptake rate of the roots, which is usually related to root density and water potential; Soil water diffusion coefficient, which depends on soil type and moisture content.

[0062] Second, the irrigation volume of each area must meet a certain maximum constraint, that is, the irrigation volume of each location cannot exceed the maximum irrigation volume that the system can provide to avoid excessive irrigation or waste of resources. The mathematical formula is: 0≤I ij ≤I max ; Among them, I max The maximum irrigable volume for each location.

[0063] Third, based on crop water requirements and soil properties, irrigation strategies should prioritize areas with greater water deficits, taking root density into account. This constraint can be achieved by introducing a weighting factor that can be dynamically adjusted based on root density.

[0064] After the optimization process is completed, the irrigation strategy decision module generates a set of irrigation allocation plans. The irrigation amount for each location will be used as the output of the system and handed over to the execution module for specific operations.

[0065] Furthermore, to ensure the sustainability and adaptability of irrigation strategies, this embodiment incorporates a dynamic adjustment mechanism. During system operation, irrigation strategies are iteratively updated based on real-time information such as new meteorological data, soil environmental changes, and changes in crop water requirements. By regularly re-optimizing irrigation plans, the system can adapt to the varying needs of crops during growth, ensuring a stable growing environment and the rational use of water resources.

[0066] For the irrigation path decision module: In this embodiment, the irrigation path decision module is used to formulate the optimal irrigation path allocation plan based on the generated irrigation strategy, combined with the field irrigation network structure, pipeline topology, node status and execution unit capabilities, to achieve precise delivery and dynamic control of irrigation water.

[0067] This module constructs an irrigation path network using a graph structure modeling method, abstracting the pipeline structure in the actual irrigation system into a directed graph. The mathematical representation of a directed graph is: G=(V,E); Where V={v1,v2,…,v n} is a node set in the irrigation system, including water source nodes, distribution nodes, control valve nodes and terminal execution nodes; E={e ij} is the connectivity relationship between nodes, recording the accessible path of irrigation water flow; Each edge e ij With the corresponding capacity constraint C ij and flow resistance parameter r ij , reflecting the maximum water delivery capacity of the pipeline and the energy loss per unit water volume respectively.

[0068] In this embodiment, the irrigation path decision module takes the minimum path hydraulic cost as the optimization goal and establishes a path optimization problem based on the following objective function: Among them, P is the path set from the water source node to each execution point; r ij is the edge e in the path ij Unit hydraulic resistance; q ij is the irrigation water flow along this edge in the path.

[0069] Under this optimization framework, path selection must not only consider the shortest topology, but also comprehensively consider path smoothness, water pressure stability, and terminal execution order, thereby ensuring the spatial balance of irrigation effects and the efficient use of water resources.

[0070] To achieve the above functions, this embodiment adopts a constrained multi-source shortest path scheduling algorithm and introduces dynamic flow constraints and a time window coordination mechanism. Its core constraints include: The node capacity constraint has the mathematical form: Among them, C i For node v i The maximum processing flow of a device reflects its physical or execution capacity limit.

[0071] The path continuity constraint has the mathematical form: Ensure that there is a physically reachable path from the primary water source node v0 to all selected execution nodes.

[0072] Irrigation scheduling synchronization constraints: For multiple areas that need to be irrigated within the same time window, the system needs to coordinate their execution order, preferably completing the path scheduling in different sub-periods while meeting the time limit to avoid excessive instantaneous flow load on the system. By defining the synchronization window W t ={P1,P2,...,P k}, controls the maximum number of concurrent irrigation paths in each cycle, adjusts the irrigation load, and is converted into mathematical form as: Among them, Q max Indicates the maximum carrying capacity of the system per unit time. To implement this path scheduling mechanism, this embodiment uses a weight-guided A* variant algorithm for path search. The algorithm uses a comprehensive cost function as the evaluation basis: f(n)=g(n)+h(n); Among them, g(n) is the accumulated cost (such as water consumption) from the starting node to the current node n; h(n) is the estimated remaining cost from the current node n to the target node; f(n) is the total cost function, which means that the smallest node will be searched and expanded first to improve path efficiency and global coordination.

[0073] In addition, this embodiment supports path backflow and conflict avoidance mechanisms. To prevent a pipe segment in the pipeline network from being repeatedly occupied by multiple paths or carrying excessive traffic at the same time, the system implements a "soft blocking" mechanism for conflicting paths, that is, dynamically adjusting the weights r of the conflicting edges during the path planning stage. ij , and set the priority function: Among them, δ ij =1 indicates that there is a path occupying edge e in the current cycle ij , otherwise it is 0; γ is the adjustment factor, which is used to increase the cost of the conflicting path and force the system to replan the path to avoid congestion.

[0074] To enhance the versatility and flexibility of the path decision module, this embodiment also supports an automatic topology learning mechanism. That is, after deployment, the system can construct an effective accessibility map of the actual path network through historical operation data, and correct deviations in the topology model caused by valve status, flow control, etc., so that the path decision is closer to the actual physical network at the structural model layer.

[0075] For the execution module: In this embodiment, the execution module is used to coordinate the time relationship and energy consumption distribution relationship between the energy storage device and the irrigation execution scheduling in the intelligent irrigation system, so as to achieve timing optimization and energy efficiency management of irrigation tasks.

[0076] This module is particularly suitable for distributed agricultural irrigation systems with off-grid operation capabilities or equipped with multiple energy supply forms (such as solar energy, battery packs, and water pump load balancing management).

[0077] In this embodiment, the design of the execution module is based on the following technical requirements: The dynamic states of the energy storage unit at different times, such as available power, remaining capacity, and charge and discharge efficiency, need to be considered; Need to work with the irrigation path scheduling module to achieve priority sorting based on energy status; The irrigation execution rhythm needs to be adjusted dynamically to avoid scheduling failure due to insufficient energy supply.

[0078] To achieve the above functions, the following formal model is established for the energy storage scheduling process in this embodiment: Assume that the system has multiple energy storage units S in each time step t, where each unit s K All have the following variable state parameters, and the mathematical form of the energy storage unit is: S={s1,s2,...,s K}; Among them, the variable state parameters are: Energy storage unit s k The remaining power at time t; Energy storage unit s k The maximum available power at time t; Energy storage unit s k Maximum energy storage capacity; are the discharge efficiency and charge efficiency parameters respectively; is the control variable for whether the energy storage unit participates in energy supply at the current moment,

[0079] Based on the above formal model, the energy allocation of the system in each scheduling cycle must meet the following energy balance constraints: in, is the irrigation amount of the root zone unit (i, j) at time t; ψ ij It is the energy consumption coefficient required for unit irrigation volume, which is related to the pump head, water pressure load, etc.

[0080] This energy balance constraint ensures that before each irrigation is executed, the system can evaluate whether the current energy storage level can support the irrigation task. If the energy storage is insufficient, the scheduling priority will be automatically adjusted or the irrigation task will be postponed.

[0081] Based on the above formal model, the following relationship model is established for the change of energy storage state: in, Whether the energy storage unit is in a charging state at time t; is the charging power input value, usually from the photovoltaic or mains system; Δt is the scheduling time step.

[0082] This model ensures that the change in the energy storage unit's charge is controlled by the actual discharge and charging processes, and can automatically enter the charging state when there is surplus energy in the system (such as strong photovoltaic output during the day).

[0083] In this embodiment, the execution module combines irrigation scheduling information with the energy storage system status to build a joint optimization model to determine the most reasonable irrigation task allocation and energy storage unit call strategy within a given time. The optimization problem objective function is as follows: in, For irrigation demand matching, the intention is to minimize the deviation between the root zone water status and the target value; This is an energy storage protection item, which aims to retain more power for subsequent cycles to avoid overdraft in advance; ω ij and μ are weight parameters, which can be flexibly adjusted according to crop priorities and power system availability.

[0084] In this embodiment, based on the derived irrigation task allocation and energy storage unit deployment strategy, the execution module is used to precisely control the combined water and fertilizer application in the irrigation system. By integrating a microscale flow regulation mechanism, a dynamic adaptation algorithm, and a root zone state perception model, the on-demand supply of water and fertilizer resources is dynamically matched to the rhizosphere environment, ensuring precise water and fertilizer supply and regulation targets at different crop growth stages.

[0085] In this embodiment, the execution module mainly includes: Micro-proportional fertilizer mixer (or micro-static mixer); Electronically controlled proportional valve group; flow meter and pressure sensor array; Multi-parameter sensing unit for crop root zone (monitoring moisture, temperature, and electrical conductivity); Central control algorithm unit (can be integrated into the master node or distributed nodes).

[0086] During the operation of the module, the current root zone environmental status data, including soil moisture content θ, is first collected through the perception layer. t , conductivity σ t and temperature T t .

[0087] The above data are sent to the control center via wireless communication and combined with the irrigation path scheduling results to serve as input for the generation of micro-control strategies.

[0088] In this embodiment, for different crop stages and regional soil types, the target root zone state vector is defined as Θ target =[θ target,σ target ], the system uses the target value as a reference in each control cycle to calculate the water and fertilizer combination required for the current root zone (W t ,F t ).

[0089] The calculation of water and fertilizer requirements is based on the following simultaneous model: W t =f w (θ target -θ t ,T t ,ρ soil ); F t =f f (σ target -σ t ,C f ,k leach ); Among them, f w is the moisture correction function, taking into account the soil water retention capacity and evaporation rate; f f is a function of fertilizer dissolution and migration, taking into account current fertilizer efficiency, soil type, and potential leaching losses; C f The concentration of fertilizer applied; k leach It is a regulating factor for the infiltration of fertilizers into the deep layers of the soil; ρ soil Correct parameters for soil bulk density or texture of the current plot.

[0090] In this embodiment, based on the above-mentioned simultaneous model, the microfluidic execution mechanism is implemented by the adjustable ratio mixing unit and the micro proportional valve controller. The system controls the proportion of fertilizer in the unit volume of irrigation liquid by the proportional setting parameter, as shown in the following formula: This parameter is transmitted as a control signal to the electronically controlled proportional valve group, which controls the flow ratio of clean water to fertilizer solution. Simultaneously, a micro flow meter provides real-time feedback and calibration to ensure that the actual ratio remains consistent with the set value.

[0091] During the execution process, the microfluidic device maintains a stable flow rate per unit time through a closed-loop control strategy, sets the continuous liquid supply rate within the irrigation cycle, and dynamically adjusts the fertilizer-water ratio to match the root zone absorption rate.

[0092] The micro-control process adopts the following closed-loop regulation mechanism: Among them, λ and η are control adjustment gains; Indicates the real-time liquid supply rate; The basic irrigation flow value preset for the system is used as the initial liquid supply rate.

[0093] If the deviation is greater than the threshold, the system automatically adjusts the valve opening and the fertilizer and water flow set values ​​to achieve a rapid response in the local root zone.

[0094] For the data feedback closed-loop module: In this embodiment, the data feedback closed-loop module is used to build a dynamic adjustment and control mechanism based on perception data. By continuously collecting, analyzing and feeding back crop growth environment data, closed-loop correction of subsystem operation strategies such as irrigation intensity, water-fertilizer ratio, and energy storage scheduling is achieved.

[0095] The design of this module emphasizes the real-time and closed-loop nature of system perception, judgment, and execution, and combines historical data to form long-term adaptive optimization capabilities.

[0096] During the operation cycle, the data feedback closed-loop module receives real-time information from multiple sensor nodes within the irrigation area, including: Root zone soil moisture data θ t ; Soil electrical conductivity (indicating salt concentration) σ t ; Soil or root zone temperature T t ; Physiological states of crops such as canopy leaf temperature and leaf area index; Irrigation unit valve control status and execution flow Information related to system energy consumption, such as energy storage status, execution power, and photovoltaic input.

[0097] In this embodiment, in order to improve the consistency and stability of the model response, the original data is standardized and the following transformation model is uniformly adopted: Among them, x t is the raw sensor reading (such as θ t ,σ t wait); is the historical period mean of this type of data; σ x is the historical period standard deviation.

[0098] This standardization process is used to eliminate the differences in dimensions and distributions among different data sources, and to improve the versatility and generalization performance of subsequent feedback control models.

[0099] In this embodiment, the closed-loop control core is dynamically adjusted based on the deviation between the sensed value and the target value to minimize the system execution deviation and improve the decision response accuracy.

[0100] Assume that the key state feedback of the system in each control cycle is: θ t is the real-time soil moisture; t is the soil conductivity; θ target , σ target is the desired target state of the system.

[0101] Then the control correction value Δu in each feedback cycle is t It can be given by the following error driving function: Δu t =K w ·(θtarget-θ t )+K f ·(σ tar g et -σ t ); Among them, K w is the moisture error feedback coefficient; K f is the conductivity error feedback coefficient; Δu t It is the control strategy correction amount, representing the irrigation amount, water-fertilizer ratio or energy storage scheduling strategy that needs to be adjusted.

[0102] In this embodiment, Δu t Can act on multiple control objects, for example: Microfluidic device: Adjusting the water-fertilizer ratio Execution module: Modify the energy storage discharge ratio or scheduling order; Irrigation rhythm control: dynamically delay or advance the irrigation cycle.

[0103] In order to avoid control disturbances caused by short-term sensor anomalies or environmental fluctuations, this embodiment introduces the "regulation memory mechanism" and "deviation integral accumulation" method to achieve smoothness and robustness of system response.

[0104] The deviation integral model is as follows: I t =β·I t-1 +(1-β)·[(θtarget-θ t ) 2 +κ·(σ tar g et -σ t ) 2 ]; Among them, I t is the current cumulative deviation intensity; β is the integral memory factor, which is used to control the degree of retention of historical deviations; κ is the balance weight parameter of moisture and conductivity deviations.

[0105] The system evaluates I before executing control t The amplitude, and only in I tThe actual control action is only performed when the control threshold is exceeded to ensure that the control signal has the ability to "suppress noise".

[0106] Please see the attached Figure 2 , an embodiment of the present invention provides a method for intelligent irrigation of crop root systems, comprising the following steps: S1, obtain multi-layer soil environmental parameters of the crop root zone through the root zone information perception module and generate a future state prediction vector; S2. The irrigation strategy decision module flattens the state vector, constructs a Markov decision process model, calculates the state value function based on the Bellman expectation equation, further derives the optimal strategy function, and outputs the optimal irrigation action; S3. The irrigation path decision module constructs a path weight matrix based on the current irrigation network topology, uses the Dijkstra algorithm to calculate the shortest energy consumption or minimum delay path from the irrigation source node to the target node, and generates path control instructions; S4. The irrigation scheduling module constructs an irrigation priority scoring function based on the real-time root zone status and crop water requirements. It calculates and ranks the priorities of each irrigation unit based on factors such as water deficit level, growth stage, and evapotranspiration prediction, and generates an irrigation scheduling plan based on resource constraints. S5. Inputting the optimal irrigation action, path control instructions, and irrigation scheduling plan into the irrigation control execution module to drive the irrigation actuators to sequentially complete the target irrigation operations; S6. Update the root zone status in real time after irrigation is executed, and feed the monitoring data back to the system for subsequent strategy optimization and model update.

[0107] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A crop root system intelligent irrigation system, characterized in that: include: The data acquisition module is used to collect soil environmental parameters and root spatial distribution information through a multimodal soil sensor array and underground computed tomography deployed at different soil depths. Based on the collected information, a perception matrix and a root density weight matrix are generated, and the two matrices are multiplied element by element to generate a root zone state vector. a root zone state modeling module, connected to the data acquisition module, configured to receive the root zone state vector and crop growth stage data, perform training and prediction using a federated learning modeling approach, and output a future root zone state vector; an irrigation strategy decision module, connected to the root zone state modeling module, for taking the future state vector as input, constructing a Markov decision process model, calculating the optimal irrigation strategy and outputting an irrigation control decision; an irrigation path decision module, connected to the irrigation strategy decision module, for calculating the optimal irrigation path and outputting irrigation path control instructions based on a preset irrigation network topology and path weight matrix in combination with the irrigation control decision; an execution module, connected to the irrigation path decision module, generating an irrigation scheduling plan based on the system energy storage status data and the drought degree of each plot, the crop growth stage, and the energy feedback potential, and controlling the microfluidic device to execute the irrigation task of the target area according to the irrigation scheduling plan; A data feedback closed-loop module is connected to the execution module and is used to collect water and fertilizer absorption feedback information of the root zone in real time, update the perception matrix and the root zone state vector, and return the updated root zone state vector to the root zone state modeling module to build a closed-loop control system of perception, modeling, decision-making, execution and feedback.

2. The intelligent crop root irrigation system according to claim 1, characterized in that: In the data acquisition module, the multimodal soil sensor array includes a water potential sensor, a conductivity sensor, a pH sensor, a temperature sensor, and an ion-selective electrode sensor. The sensor array is arranged vertically at multiple depth levels to collect the water potential, conductivity, pH value, temperature, and nitrogen, phosphorus, and potassium ion concentrations of each soil layer. The collected data are classified by layer to form a data group, and finally the data group is converted into a perception matrix for characterizing the soil environmental status of the root zone.

3. The crop root system intelligent irrigation system according to claim 1, characterized in that: The data acquisition module also includes an underground computer tomography device and an image recognition processing unit, wherein the image recognition processing unit is used to segment the scanned image into root areas, calculate the root volume density value of each depth layer based on the segmentation result, summarize the root density data group by level, and finally convert the root density data group into a root density weight matrix; The root density weight matrix and the perception matrix are element-by-element multiplied by the Hadamard product to obtain the root zone state vector. The calculation formula of the element-by-element multiplication is: R ij =F ij ·w i ; Among them, F ij is the jth environmental parameter of the i-th layer in the perception matrix; w i is the root density weight of the layer; The root zone state vector is used to reflect the resource response capabilities of different soil layers and root system structures.

4. The intelligent irrigation system for crop roots according to claim 1, characterized in that: The root zone state modeling module includes an edge computing node and a federated modeling server. The edge computing node is used to locally receive the root zone state vector and crop growth stage data and perform preliminary model training. The crop growth stage data includes sowing time, growth stage identifier, current plant height, biomass estimation value and daily accumulated temperature data. The root zone state modeling module uses a federated learning modeling approach for training and prediction, and the steps of outputting the future state vector of the root zone include: The edge node performs preliminary training of the state prediction model based on local historical data and periodically sends model gradient parameters to the federated modeling server; The federated modeling server aggregates model parameters of multiple edge nodes using a weighted average algorithm and updates the global model; The global model synchronously transmits parameters to the edge nodes after each round of update. Finally, each edge node jointly models the input root zone state vector and crop growth stage data based on the updated model, and finally outputs the future state vector of the root zone; The root zone future state vector is used to describe the predicted state of the soil environment at a future moment.

5. The intelligent crop root irrigation system according to claim 1, characterized in that: In the irrigation strategy decision module, the steps of calculating the optimal irrigation strategy and outputting the irrigation control decision include: The irrigation strategy decision module constructs a Markov decision process model based on the future state vector of the root zone. The Markov decision process model is expressed as: M=(S,A,P,R,γ); Where S is the state space, representing the root zone environment state at different time points; A is the action space, representing the combination of various irrigation amounts and irrigation time intervals; P is the transition probability; R is the reward function; and γ is the discount factor, which is used to measure the influence of future rewards on the current decision. In the Markov decision process model, strategy selection is performed through a reward function, and the functional form of the reward function is: R(s,a)=-(λ1·C w (a)+λ2·C e (a)+λ3·D(s,s * )); Among them, C w (a) is the water cost of the irrigation operation; C e (a) is the energy consumption cost; D(s,s * ) is the state s and the target state s * Deviation; λ1, λ2 and λ3 are weighting coefficients; In order to achieve the long-term optimal irrigation control goal, the irrigation strategy decision module calculates the value function of each state according to the Bellman expectation equation: the calculation formula is: Among them, P(s ′ |s,a) means taking action a in state s and transitioning to state s ′ The probability of s; V(s) is the value function of state s, which represents the maximum expected cumulative reward that can be obtained from state s under the optimal strategy; V(s ′ ) is the subsequent state s ′ The value function represents the value of the state s ′ The maximum expected cumulative reward that can be obtained from the departure; Based on the above value function, the value iteration method is used to solve the optimal policy function, which is: π * (s)=argmax a∈A Q(s,a); Among them, π * (s) is the optimal policy function, which selects the optimal action under state s; Q(s,a) is the action-value function, which represents the expected cumulative reward obtained after taking action a under state s; argmax is the action that maximizes the expression in the brackets; after solving, the future state vector of the root zone is mapped to the current state, and the optimal policy function selects the irrigation action and outputs the corresponding irrigation control decision.

6. The intelligent crop root irrigation system according to claim 1, characterized in that: The irrigation path decision module constructs a path weight matrix based on the irrigation network topology, and uses the Dijkstra algorithm to calculate the irrigation path that meets the shortest energy consumption or minimum delay starting from the current irrigation source node, and generates path control instructions; The irrigation network topology is composed of a node set and a connection relationship set. The elements in the node set represent water source nodes, branch nodes, and terminal water outlet nodes in the irrigation pipe network. The elements in the connection relationship set represent that there are pipe connections between nodes. The elements in the path weight matrix represent the path energy consumption and time delay between nodes, and the path energy consumption and time delay are calculated based on the parameters of the actual irrigation pipe diameter, length, head loss and solenoid valve response time; The generated control instructions include the opening and closing sequence of the nodes involved in the path and their control elements; The generated control instructions are called by the irrigation control execution unit to complete the path guidance.

7. The crop root system intelligent irrigation system according to claim 1, characterized in that: The execution module includes a battery status acquisition unit and a priority calculation unit. The priority calculation unit generates an irrigation scheduling plan based on the drought degree of each plot, the crop growth stage, and the energy feedback potential. The specific steps include: Obtain the root zone status parameters corresponding to each irrigation unit, including current water potential, water deficit, evapotranspiration prediction value and crop growth stage parameters; A multi-factor irrigation priority scoring function is constructed based on the obtained parameters. The function form is: P i =α1·F d (i)+α2·F g (i)+α3·F e (i); Among them, P i is the priority score of the i-th irrigation unit; F d (i) is the water deficit index function, which represents the deviation between the current soil moisture state and the set lower limit; F g (i) is the crop growth stage function, reflecting the sensitivity weight of the key growth period; F e (i) is the evapotranspiration prediction function, which estimates the water consumption rate per unit time based on future meteorological conditions; α1, α2, and α3 are weight coefficients, which are set according to the target control strategy; Sort by priority score of each unit to obtain a scheduling order list of irrigation units; Taking into account the water source capacity, power load, and path accessibility of the irrigation system, a specific irrigation scheduling plan is generated based on the scheduling sequence list. The plan includes the start time, expected irrigation duration, and corresponding control path for each irrigation unit.

8. The intelligent crop root irrigation system according to claim 1, characterized in that: The execution module also includes an adjustable proportional valve assembly and a micropump array for accurately controlling the ratio of water and fertilizer according to the irrigation scheduling plan.

9. The intelligent crop root irrigation system according to claim 1, characterized in that: The data feedback closed-loop module includes a root zone sensor feedback channel and a data fusion unit, which is used to collect the root zone absorption response signal after irrigation and return the fused data to the root zone state modeling module to complete the data closed-loop update.

10. A method for intelligent irrigation of crop root systems, based on the system according to any one of claims 1 to 9, characterized in that: The following steps are involved: S1, obtain multi-layer soil environmental parameters of the crop root zone through the data acquisition module and generate a future state prediction vector; S2. The irrigation strategy decision module flattens the state vector, constructs a Markov decision process model, calculates the state value function based on the Bellman expectation equation, further derives the optimal strategy function, and outputs the optimal irrigation action; S3. Through the irrigation path decision module, a path weight matrix is ​​constructed according to the current irrigation network topology. The Dijkstra algorithm is used to calculate the shortest energy consumption or minimum delay path from the irrigation source node to the target node, and a path control instruction is generated. S4. The execution module constructs an irrigation priority scoring function based on the real-time root zone status and crop water requirements. It calculates and ranks the priorities of each irrigation unit based on factors such as water deficit level, growth stage, and evapotranspiration prediction, and generates an irrigation scheduling plan based on resource constraints. S5. Input the optimal irrigation action, path control instruction, and irrigation scheduling plan as an irrigation path control instruction, and drive the irrigation actuator to sequentially complete the target irrigation operation; S6. Update the root zone status in real time after irrigation is executed, and feed the monitoring data back to the system for subsequent strategy optimization and model update.

Images