An irrigation amount compensation method, and a uav irrigation operation method and system

Abstract

The application discloses an irrigation amount compensation method and an unmanned aerial vehicle irrigation operation method and system, and relates to the technical field of agricultural irrigation. The method first identifies a target irrigation unit through position coordinates and calculates required nozzle flow, and then obtains target atomization parameters and droplet distribution data containing initial state variables of droplets in combination with an atomization power model. After adjusting the nozzle according to the target atomization parameters, real-time parameters such as environmental wind speed and irrigation height are collected and input into the drift power model to predict the drift vector of the droplet cloud, and the target spray angle against drift is calculated to adjust the nozzle. During the spraying process, the change vectors of the flight speed, the irrigation height and the environmental wind speed are continuously monitored, if any change vector exceeds the disturbance threshold, the drift prediction is re-performed, the dynamic closed-loop optimization is realized, otherwise, the spray angle is fine-tuned according to the change vector, and the accuracy of droplet deposition is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of agricultural irrigation technology, in particular to an irrigation amount compensation method and an unmanned aerial vehicle irrigation operation method and system. BACKGROUND

[0002] The traditional agricultural unmanned aerial vehicle irrigation system mainly relies on preset routes and fixed flow for operation, which has low intelligence and cannot respond to the spatial heterogeneity of crop water demand and the dynamic changes of environmental factors during operation, often leading to uneven irrigation, water resource waste or insufficient irrigation, etc. The patent application with publication number CN107336830A discloses an intelligent unmanned aerial vehicle for agricultural production with precise irrigation. The unmanned aerial vehicle compensates for the wind direction through the end effector (nozzle). However, the compensation mechanism of this scheme mainly relies on the direct feedback of instantaneous wind resistance, and the control model is relatively simple. Essentially, it is a kind of macroscopic directional correction of "outgoing water flow". This method fails to establish and utilize the whole-process dynamics model of the droplets from separation to deposition, and thus cannot predict and compensate for the complex motion trajectory of droplets of different particle sizes in the air, resulting in that the control of drift cannot reach the level of precision and refinement.

[0003] The patent application with publication number CN112906477A discloses an irrigation prescription map inversion method based on unmanned aerial vehicle spectral data. The method inverses the irrigation prescription map by combining different irrigation technologies with historical meteorological irrigation data, real-time thermal infrared, multi-spectral high-resolution remote sensing image data, and soil moisture information data from ground sensors. However, this method focuses on decision planning before irrigation, and its technical scope stops at the generation of the prescription map, without covering the regulation of key dynamic factors (such as wind-induced drift) that affect the effect during operation execution. Therefore, under the action of wind, the sprayed water droplets may also fail to accurately deposit in the preset target area, and ultimately the irrigation uniformity cannot be guaranteed. Therefore, an integrated solution that can integrate real-time water demand sensing, dynamic drift prediction and compensation, and self-adaptive closed-loop calibration is needed to truly realize the whole-process precise irrigation from sensing to execution. SUMMARY

[0004] In order to solve the defects existing in the prior art, the present application proposes an irrigation amount compensation method and an unmanned aerial vehicle irrigation operation method and system. The method forms a dynamic closed loop by integrating real-time drift prediction and path compensation, crop water demand sensing, variable irrigation decision-making and operation effect evaluation, effectively improving the irrigation uniformity and water resource utilization rate.

[0005] The technical scheme of the present application is implemented as follows:

[0006] An irrigation amount compensation method, comprising the following steps:

[0007] Step 1: calibrate the target irrigation unit according to the position coordinates, and extract the target irrigation amount of the target irrigation unit;

[0008] Step 2: calculate the nozzle flow rate according to the flight speed, spraying width and target irrigation amount, and obtain the target atomization parameters and droplet distribution data according to the nozzle flow rate and atomization power model;

[0009] Step 3: adjust the nozzle according to the target atomization parameters, collect the irrigation height and environmental wind speed, input the irrigation height, environmental wind speed, flight speed and droplet distribution data into the drift power model to generate the drift vector of the droplet cloud;

[0010] Step 4: calculate the target spray angle compensated for drift based on the drift vector and irrigation height, and adjust the spray angle of the nozzle to the target spray angle;

[0011] Step 5: collect the change vector of the flight speed, irrigation height or environmental wind speed in real time, if any change vector is less than the corresponding disturbance threshold, go to step 6, otherwise return to step 3;

[0012] Step 6: update the adjustment vector of the spray angle according to the change vector, adjust the spray angle of the nozzle by using the adjustment vector, and return to step 5.

[0013] In the present application, in step 2, the unit irrigation total amount D is calculated according to the area S of the target irrigation unit and the target irrigation amount D , and the nozzle flow rate Q is , wherein d1 is the spraying width of the unmanned aerial vehicle, and v1 is the flight speed.

[0014] In the present application, in step 2, the target atomization parameters are output by querying the nozzle characteristic database with the nozzle flow rate and the median value of the target droplet particle size as inputs, and the corresponding droplet distribution data is obtained based on the target atomization parameters, wherein the nozzle characteristic data includes a mapping table between the atomization parameters, nozzle flow rate and median value of droplet particle size of each nozzle.

[0015] In the present application, in step 3, the drift power model includes a state variable function and a drift vector function, the state variable x(t f ) at the terminal time t f when the droplet settles to the ground or the crop canopy is estimated by the state variable function, and then the drift vector y of the droplet is estimated in combination with the reference terminal state variable x ref of the droplet under the ideal condition of no wind field, and the average of the drift vectors of multiple droplets is the drift vector y' of the droplet cloud.

[0016] In the present application, in step 4, the drift vector y' is decomposed into a drift component y1=(y x , y y), the horizontal azimuth angle θ and the pitch angle α of the target spray angle are calculated, and the calculation formula is: θ = arctan2 (y x ,y y ) + π, α = arctan (h / ||y1||), wherein y x is the component of the offset vector y1 in the X-axis direction, y y is the component of the offset vector y1 in the Y-axis direction, arctan2 is a four-quadrant arctangent function, π is a reverse compensation, ||y1|| is the length of the offset component y1 in the horizontal plane, and h is the irrigation height.

[0017] In the present application, in step 6, the variation vector includes a variation vector of flight speed, a variation vector of environmental wind speed, and a variation vector of irrigation height, the adjustment vector includes a horizontal azimuth angle adjustment amount and a pitch angle adjustment amount, the variation vector is converted to the nozzle coordinate system, and the horizontal azimuth angle adjustment amount and the pitch angle adjustment amount of the nozzle are generated again.

[0018] An unmanned aerial vehicle irrigation operation method, comprising the following steps:

[0019] Step 100: obtaining vegetation index data of a target area and generating a first water demand distribution map, the first water demand distribution map comprising a plurality of irrigation units and theoretical irrigation amounts of the irrigation units;

[0020] Step 200: determining an operation boundary of each irrigation unit and performing path planning on the operation boundary to generate a global operation path of the unmanned aerial vehicle;

[0021] Step 300: calculating an evapotranspiration loss rate in an irrigation period through real-time meteorological data and an evapotranspiration model, and determining a target irrigation amount of each irrigation unit in combination with the evapotranspiration loss rate and the theoretical irrigation amount;

[0022] Step 400: obtaining an irrigation instruction based on the global operation path and the target irrigation amounts of the irrigation units, and the unmanned aerial vehicle performing the irrigation instruction according to the irrigation amount compensation method;

[0023] Step 500: obtaining vegetation index data of the target area again to obtain a second water demand distribution map, obtaining an irrigation uniformity parameter according to the first water demand distribution map and the second water demand distribution map, and updating the drift force model.

[0024] In the present application, in step 500, the irrigation uniformity parameter is calculated through the difference between pixel values in the first water demand distribution map and the second water demand distribution map.

[0025] In the present application, in step 500, the covariance matrix of process noise in the drift power model is adjusted according to the irrigation uniformity parameter, wherein when the irrigation uniformity parameter is lower than a preset threshold, the diagonal element value in the covariance matrix is increased.

[0026] An unmanned aerial vehicle irrigation operation system according to the unmanned aerial vehicle irrigation operation method, comprising:

[0027] A data acquisition unit configured to acquire vegetation index data;

[0028] An irrigation setting unit configured to generate a theoretical irrigation amount and a target irrigation amount;

[0029] A path setting unit configured to generate a global operation path of the unmanned aerial vehicle;

[0030] A device control unit configured to generate an irrigation instruction;

[0031] An unmanned aerial vehicle configured to execute the irrigation instruction, wherein

[0032] The data acquisition unit generates a first water requirement distribution map and a second water requirement distribution map,

[0033] The device control unit calibrates the drift power model based on the first water requirement distribution map and the second water requirement distribution map,

[0034] The device control unit inserts the updated drift power model into the irrigation instruction.

[0035] The irrigation amount compensation method, the unmanned aerial vehicle irrigation operation method and the system according to the present application have the following beneficial effects: the present application first obtains the initial motion characteristics of the fog droplet cloud through the atomization power model, combines the flight speed, the environmental wind speed, the irrigation height and the drift power model to predict the offset vector of the fog droplet cloud, calculates the target spray angle according to the offset vector to perform feedforward compensation. When the irrigation height, the flight speed or the environmental wind speed is greatly changed, the system immediately re-executes the drift calculation process to update the spray angle to match the current working condition. For small continuous changes, the adjustment vector is calculated through the change vector calculation, and the spray angle is fine-tuned and corrected through the adjustment vector. This hierarchical compensation mechanism not only ensures the rapid response capability when the working condition changes suddenly, but also realizes accurate control in the steady state working condition. Further, the process noise in the drift power model is updated according to the water requirement distribution map before and after irrigation, so as to realize online self-correction of the model parameters, and continuously improve the prediction accuracy and adaptability of the drift power model under different field environments and weather conditions. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 The flowchart of the irrigation amount compensation method of the present application;

[0037] Figure 2 Hydraulic control diagram for the nozzle of the present application;

[0038] Figure 3 Schematic diagram of the flow field of the unmanned aerial vehicle rotor of the present application;

[0039] Figure 4 Unmanned aerial vehicle spraying diagram parallel to the moving direction of the present application;

[0040] Figure 5 Unmanned aerial vehicle spraying diagram perpendicular to the moving direction of the present application;

[0041] Figure 6 Deposition density distribution curve diagram of different environmental wind speeds at the irrigation height of the present application;

[0042] Figure 7 Deposition density distribution curve diagram of different environmental wind speeds at the irrigation height of the present application;

[0043] Figure 8 Simulation diagram of the trajectory of the movement of the droplets in the preferred state of the present application;

[0044] Figure 9 Flowchart of the unmanned aerial vehicle irrigation operation method for implementing the irrigation amount compensation method of the present application;

[0045] Figure 10 Block diagram of the unmanned aerial vehicle irrigation operation system of the present application.

[0046] Reference numerals in the drawings: water tank 100, gate valve 200, electric pump 300, throttle valve 400, nozzle 500, flow collector 601, pressure collector 602, overflow valve 603, controller 700. DETAILED DESCRIPTION

[0047] In order to more clearly understand the purpose, technical scheme and advantages of the present application, the present application is described and explained below in combination with the drawings and examples.

[0048] Variable irrigation is one of the core techniques of modern precision agriculture, which improves resource utilization, promotes crop growth and reduces environmental impact by supplying water or liquid chemicals to different areas of farmland as needed. Existing UAV variable irrigation technology still has significant limitations. During the operation, the deposition of liquid chemicals (or water) is affected by the comprehensive influence of environmental wind speed, flight speed, irrigation height, and liquid drop characteristics, etc., and will drift, resulting in spatial deviation between the liquid drop landing point and the preset irrigation unit, i.e. the deposition amount of liquid drops at different positions is different. This deviation makes the actual irrigation area not match the target water demand area, seriously affecting the precision and uniformity of irrigation. Secondly, the traditional irrigation decision-making model is usually based on static theoretical calculation, and does not fully consider the dynamic changes of crop evapotranspiration caused by changes in meteorological conditions (such as temperature, humidity, wind speed, and sunshine) during the irrigation period, resulting in a deviation between the theoretical irrigation amount and the actual water demand, affecting the irrigation effect. In addition, the existing technology lacks an effective closed-loop feedback and calibration mechanism. After a job is completed, the system cannot evaluate and optimize its core model through the job effect, resulting in insufficient adaptability and robustness of the system in changing environments, making it difficult to achieve long-term, stable precision irrigation.

[0049] In order to overcome the defects of the above prior art, the present application introduces a drift force model, compensates the spray angle by calculating the offset vector, and makes a slight adjustment to the spray angle in combination with real-time flight parameters. After the operation, the system compares the water demand distribution map before and after irrigation to evaluate the uniformity, and if it does not meet the standard, it automatically calibrates the drift force model and starts a self-optimization cycle. The present application effectively solves the problem of inaccurate irrigation caused by liquid drop drift, dynamically compensates for crop evapotranspiration loss, and thus significantly improves the overall precision, uniformity and environmental adaptability of variable irrigation. Embodiment one

[0050] Reference Figures 1 to 5 The irrigation amount compensation method of the present application detailed in this embodiment includes the following steps:

[0051] Step 1: Determine the target irrigation unit according to the position coordinates, and extract the target irrigation amount of the target irrigation unit. When the UAV flies to the target area, the positioning system (such as GNSS) of the UAV returns the position coordinates in real time, which are matched with the digital map to determine the target irrigation unit (such as the B5 area) being flown over. Then, the target irrigation amount preset for the target irrigation unit is extracted from the digital map, which is calculated in advance based on real-time weather data and evapotranspiration model.

[0052] Step 2: Calculate the nozzle flow rate according to the flight speed, spray width and target irrigation amount, and obtain the target atomization parameters and droplet distribution data according to the nozzle flow rate and atomization power model. Calculate the unit irrigation total amount Q = K * (P * A)1 / 2 wherein d1 is the swath of the UAV, and v1 is the flight speed. Figure 2 The hydraulic control diagram of the nozzle of the present application, the UAV is equipped with a water tank 100, which stores liquid medium such as water or pesticide. The water tank 100 is connected to an electric pump 300 through a gate valve 200, and the electric pump 300 provides high-pressure liquid medium to the nozzle 500 through a throttle valve 400. A flow collector 601 and a pressure collector 602 measure the state of the liquid medium. An overflow valve 603 is provided on the side of the electric pump 300 to limit the system pressure. The controller 700 adjusts the nozzle flow and atomization parameters through the throttle valve 400 and the electric pump 300, and controls the spray angle of the nozzle 500.

[0053] The operation process of the atomization power model includes two stages: the first stage takes the nozzle flow and the target droplet size median as input, and outputs the target atomization parameters by querying the nozzle characteristic database, and the second stage obtains the corresponding droplet distribution data based on the target atomization parameters. The droplet distribution data includes the initial state variables of the droplet cloud and the droplets in the droplet cloud. The nozzle characteristic data stores a mapping table between the atomization parameters, the nozzle flow and the droplet size median of each nozzle. The nozzle can be a hydraulic nozzle, a centrifugal nozzle or an air flow nozzle. The atomization parameters of the hydraulic nozzle are the spray pressure and the nozzle model, the atomization parameters of the centrifugal nozzle are the rotating speed of the rotating disc and the inlet pressure of the nozzle, and the atomization parameters of the air flow nozzle are the liquid supply pressure, the atomization air pressure and the atomization air flow. The preferred method for obtaining droplet distribution data according to the atomization power model is specifically referred to in Example 2.

[0054] The target droplet size median is determined according to the operation type, and the operation type at least includes: water application, weed control, sterilization, insect control and foliar fertilizer. For example, medium-large droplets (Dv0.5: 250-350 microns) are used for weed control, which can reduce the drift damage to adjacent crops; medium-fine droplets (Dv0.5: 100-200 microns) are used for sterilization and insect control, which can improve the penetration and uniformity of droplets on the back of the leaf surface and the canopy; medium droplets (Dv0.5: 200-300 microns) are used for foliar fertilizer, which can ensure the deposition amount and absorption efficiency; and large droplets (Dv0.5: 300-400 microns) are selected for water application, which can ensure that the water can quickly settle in the target irrigation unit and reduce the evaporation of water.

[0055] Step 3: Adjust the nozzle according to the target atomization parameters, collect the irrigation height and the environmental wind speed, and input the irrigation height, the environmental wind speed, the flight speed and the droplet distribution data into the drift power model to generate the drift vector of the droplet cloud. For example, Figures 3 to 5As shown, the droplet distribution data is affected by moving direction and environmental wind speed, etc., and needs to be predicted according to a drift force model. The drift force model is constructed based on computational fluid dynamics and a discrete phase model, and a three-dimensional complex wind field under the superposition of unmanned aerial vehicle rotor and natural environmental wind is numerically simulated by solving Navier-Stokes equations. On this basis, the motion trajectory of droplets of different particle sizes in the three-dimensional complex wind field is solved by using the discrete phase model and the Lagrangian particle tracking method, so as to obtain high-precision droplet deposition distribution benchmark data. Finally, taking the droplet deposition distribution benchmark data as the fitting target, the key unknown parameters in the drift force model are solved by system identification method (such as nonlinear least squares method), so as to ensure that the prediction result of the drift force model is consistent with the high-fidelity simulation result.

[0056] For small particle size droplets, the drift force model is , the drift force model includes a state variable function and an offset vector function. Wherein, the state variable x is the three-dimensional space position and three-dimensional velocity vector of the droplet, A is a state matrix containing air resistance and droplet properties, u1 is the superposition flow field of unmanned aerial vehicle rotor flow field and natural air flow, u1 changes with flight speed v1, environmental wind speed v2 and irrigation height h, u1=F(h, v1, v2), u2 is the gravity field, b1 and b2 are input vectors of u1 and u2 respectively, b2 is constant, n is process noise, which is used to represent unmodeled dynamics and random disturbance, C is the output matrix, t f is the terminal time when the droplet settles on the ground or the crop canopy, x ref is the reference terminal state variable of the droplet under ideal conditions without wind field, y is the offset vector of the droplet, and the average of the offset vectors of multiple droplets is the offset vector y' of the droplet cloud, which is described in detail in Example Three. In the drift force model, the initial state variable x0 of the state variable x is provided by the droplet distribution data output by the atomization dynamic model. For large particle size droplets, the drift force model needs to consider the aerodynamic resistance in the non-Stokes region, which is in the form of a nonlinear state space, and the state variable function is .

[0057] Step 4: Based on the offset vector and the irrigation height, the target spray angle of the compensation drift is calculated, and the spray angle of the nozzle is adjusted to the target spray angle. The offset vector y' is decomposed into the offset component y1=(y x ,y y ) in the horizontal plane, the horizontal azimuth angle θ and the pitch angle α of the target spray angle are calculated, and the calculation formula is: θ=arctan2(y x ,y y )+π, α=arctan(h / ||y1||), wherein y x is the component of the offset vector y1 in the X-axis direction.y is the component of the offset vector y1 in the Y-axis direction, arctan2 is the four quadrant arctangent function, π is the inverse compensation, ||y1|| is the length of the offset component y1 in the horizontal plane, and h is the irrigation height.

[0058] Step 5: Real-time acquisition of the variation vectors of the flight speed, the irrigation height, or the environmental wind speed. If any variation vector is less than the corresponding disturbance threshold, proceed to Step 6, otherwise return to Step 3. The disturbance threshold of the flight speed is usually ±(0.3-0.7) m / s. A variation less than 0.3 m / s has little effect on the trajectory of the droplets, while a variation exceeding 0.7 m / s indicates that the flight state has changed significantly. The disturbance threshold of the irrigation height is usually ±(0.3-0.5) m, and the disturbance threshold of the environmental wind speed is usually ±(0.5-1) m / s. The disturbance thresholds can be adjusted within a reasonable range according to the specific model and operation requirements.

[0059] Step 6: Update the adjustment vector of the spray angle according to the variation vector, adjust the spray angle of the nozzle using the adjustment vector, and return to Step 5. First, add the variation vectors of the flight speed and the environmental wind speed to obtain a comprehensive wind speed variation vector in the geographic coordinate system, then convert the comprehensive wind speed variation vector to the body coordinate system of the UAV and extract its lateral component. Based on the size and direction of the lateral component, generate a horizontal azimuth adjustment amount (e.g., not more than ±5°) through proportional mapping; at the same time, generate a pitch angle adjustment amount (e.g., not more than ±3°) based on the variation vector of the irrigation height. The horizontal azimuth adjustment amount and the pitch angle adjustment amount are the adjustment vector. Embodiment Two

[0060] Reference Figure 2 In this embodiment, the nozzle 500 is, for example, a centrifugal nozzle, and the atomization parameters include the rotating disc speed R and the nozzle inlet pressure P1. The liquid flows out of the water tank 100, passes through the gate valve 200, the electric pump 300, and the throttle valve 400 in sequence, and finally reaches the nozzle 500. The controller 700, as the core control unit, actively controls the start-stop of the electric pump 300, the opening degree of the throttle valve 400, and the on-off of the gate valve 200, thereby comprehensively regulating the nozzle flow and the nozzle inlet pressure. The pressure relief valve 603 constitutes a pressure safety and stability circuit. The flow collector 601 and the pressure collector 602 feed the real-time monitored nozzle flow Q and nozzle inlet pressure P1 back to the controller 700, and finally the liquid enters the nozzle with a rotating disc speed of R under the working conditions set by the controller 700, breaks and atomizes under the action of strong centrifugal force, and forms droplets that detach from the nozzle 500.

[0061] The particle size distribution of the mist droplets is obtained. The atomization dynamics model first needs to determine the particle size distribution S1(d2) of the mist droplets. In this embodiment, the atomization dynamics model is represented by, for example, a Rosin-Rammier distribution model where γ is a characteristic particle size, k1 is a distribution width parameter, the larger k1 is, the more concentrated the particle size distribution is, and S1(d2) represents the probability density of the mist droplets with a particle size of d2. The distribution width parameter is directly determined by the pre-stored mapping relationship between the nozzle flow rate Q and the rotating disc speed R collected by the controller 700 in real time. The mapping relationship can be established through a pre-calibration experiment and stored in the controller 700.

[0062] The velocity distribution of the mist droplets is obtained. After obtaining the particle size distribution, the atomization dynamics model assigns a corresponding reference velocity v3 to each particle size d2 based on the physical law that the average velocity of small particle size mist droplets is higher than that of large particle size mist droplets. The reference velocity is determined by a basic velocity term and a fluctuation velocity. The basic velocity term is determined by the rotating disc speed R, and the fluctuation velocity is determined by the particle size d2. The reference velocity is the initial value of the mist droplet velocity.

[0063] The direction distribution of the mist droplets is obtained. The direction of the mist droplets when they leave the centrifugal nozzle is determined by the structural characteristics of the nozzle. In this embodiment, the rotating disc has a smooth edge, and the mist droplets mainly fly along the tangential direction of the rotating disc. The jet direction θ1 of the mist droplets is modeled as a normal distribution centered on the theoretical tangential direction (for example, 90 degrees), and the direction distribution is where the standard deviation σ of the jet direction angle is a fixed parameter related to the design of the nozzle and can be determined by a small number of experiments, and θ0 is the theoretical tangential direction.

[0064] The mist droplet distribution data is obtained. The second stage of the mist droplet dynamics model first randomly samples a large number of mist droplets with different particle sizes according to the particle size distribution S1(d2), then calculates the corresponding reference velocity v3 according to the particle size d2 obtained by each sampling and the current rotating disc speed R, and finally independently samples the jet direction θ1 from the direction distribution S2(θ1). The three-tuple (d2, v3, θ1) constitutes a sample point, and a large number of sample points can be randomly extracted to form a mist droplet cloud that matches the current spray flow rate and atomization parameters. The mist droplet cloud is the mist droplet distribution data. Embodiment Three

[0065] Reference Figures 6 to 8 This embodiment further discloses a preferred method of constructing a drift dynamics model and calculating a drift vector based on the drift dynamics model. The flight wind speed, the environmental wind speed, and the irrigation height all significantly affect the deposition distribution of the mist droplets, such as Figure 6 and Figure 7As shown, faster ambient wind speed and higher irrigation height can exacerbate the drift effect of fog droplets, changing the center position of fog droplet cloud deposition. The drift dynamic model used in this embodiment can effectively capture the influence of flight wind speed, ambient wind speed and irrigation height on fog droplet drift.

[0066] Parameterization of the drift dynamics model. Differentiate the state variable x and expand according to Newton's second law to obtain the state derivative. , where, [p x ,p y ,p z ] T For the three-dimensional spatial position of the droplets, [a x ,a y ,a z ] T Let a be the acceleration of the droplet, [v] x ,v y ,v z ] T Let v be the velocity of the droplets. The motion of the droplets in the air is mainly caused by gravity F. g and air resistance F d The effect, the resulting acceleration a=F g / m1+F d / m1, where m1 is the mass of the droplet. F g / m1 is defined as the input vector u2, and the acceleration due to air resistance follows Stokes' law: F d / m1=(w1-v) / τ, where w1 is the local airflow velocity. Defining w1 as the input vector u1, the acceleration is expressed as a=(u1-v) / τ+u2. Substituting the acceleration a into the state derivative and rearranging it into state-space form yields the standard state equation. Where u2 = [0, 0, -g] T , , , , τ is the relaxation time constant of the droplet, g is the gravitational acceleration, and I is the velocity of the droplet. 3×3 It is an identity matrix.

[0067] Determining the parameters of the drift dynamics model. In the standard equation of state, the relaxation time constant τ of the state matrix A and the input vector u1=F(h,v1,v2) are unknown parameters, requiring the construction of a high-fidelity CFD-DPM model for trajectory simulation. For example... Figure 8As shown, a geometric model incorporating the UAV rotor is established. By solving the Navier-Stokes equations, the CFD flow field of the UAV under different combinations of flight speed v1, ambient wind speed v2, and irrigation height h is simulated. Droplets of different particle sizes d2 are injected into this CFD flow field. The Lagrange particle tracking method (i.e., discrete phase model DPM) is used to solve the trajectory of the droplets, obtaining the droplet motion at discrete time steps t1, t2, ..., t K The sequence of state variables For a fixed set of data (h, v1, v2, d2), a nonlinear least squares method is established to optimize the loss function. By minimizing L1(τ), the optimal relaxation time constant for the current particle size is obtained. This is then used to iterate through all possible particle sizes, establishing a mapping relationship τ=f(d2) between the relaxation time constant τ and the particle size d2. Here, K is the total number of time steps sampled during the entire motion process from droplet release to sedimentation in the trajectory simulation, and x... mod1 (t k ,τ) are the state variables of the standard state equation at the k-th discrete time step. The initial relaxation time constant is τ0 = ρd2. 2 / 18μ, where ρ is the droplet density and μ is the aerodynamic viscosity. Simultaneously, the average wind speed vector at the initial droplet position is extracted from the CFD flow field as the true value of w1, thus obtaining u1. Through multivariate regression analysis, the functional relationship u1=F(h,v1,v2) is established.

[0068] Integration of drift dynamic models. Once the model parameters are determined, the standard equation of state is fully defined. To characterize the dynamic effects not fully described by the standard equation of state (such as droplet evaporation and collision) and random disturbances, process noise *n* is introduced into the standard equation of state to form state variable functions. Numerical integration is performed using the standard state equations after determining the model parameters to obtain the sequence of state variables. Furthermore, the residual sequence ε(t) is obtained. k )=x CFD (t k )-x mod2 (t k The covariance matrix of the process noise n is obtained by statistically analyzing the sample covariance of the residual sequence ε. The output matrix C of the drift kinetic model is defined as [I 3×3 ,0 3×3 ], thus obtaining the drift force model .

[0069] Calculation of the offset vector. Output of the drift force model: y = C(x(t) f )-x ref ) represents the offset vector of a single droplet, where t fThe output matrix C is used to extract the final three-dimensional spatial position of the droplets at the moment of their deposition onto the ground or crop canopy, referencing the terminal state variable x. ref The offset vector y' of the fog cloud is obtained by integrating the state variable function under the windless condition (u1=0). The offset vector y' of the fog cloud is obtained by taking the average of the offset vectors y of all fog droplets in the fog cloud. Example 4

[0070] Reference Figure 9 A drone irrigation operation method for implementing the irrigation volume compensation method includes the following steps:

[0071] Step 100: Acquire vegetation index data for the target area and generate a first water requirement distribution map. The first water requirement distribution map includes multiple irrigation units and the theoretical irrigation amount for each irrigation unit. Vegetation index data for the target area is collected using a drone equipped with a multispectral sensor. The vegetation index data includes the Normalized Difference Vegetation Index (NDVI) or the Chlorophyll Absorption Ratio (CRI). In this embodiment, the multispectral sensor collects reflectance data in at least the red and near-infrared bands. Based on the reflectance data, the NDVI is calculated as NDVI = (NIR - Red) / (NIR + Red), where NIR is the near-infrared reflectance and Red is the red reflectance.

[0072] By spatially interpolating the calculated normalized differential vegetation index and combining it with crop coefficients and soil moisture characteristics, a mapping model between the vegetation index and crop water requirement was established. Based on this mapping model, the entire target area was divided into several grids, and the theoretical irrigation amount q1=(1-V / V) was calculated for each grid. max Ne, where V is the average normalized differential vegetation index value of this grid, V max The maximum value of the normalized differential vegetation index is defined as N, where N is an empirical coefficient and e is the reference crop evapotranspiration. Irrigation thresholds are dynamically set based on crop type, growth stage, and water supply conditions. Grids with theoretical irrigation amounts greater than the irrigation thresholds are retained as irrigation units, ultimately generating a first water demand distribution map containing the spatial distribution of all irrigation units and their corresponding theoretical irrigation amounts.

[0073] Step 200: Determine the operational boundary of each irrigation unit and perform path planning on the operational boundary to generate the global operational path of the UAV. Taking the geographic center of any irrigation unit as the center and combining the coordinate position of the irrigation unit, a rectangular or convex polygon region is generated as the operational boundary of the irrigation unit through a spatial analysis algorithm. The global operational path is solved using a traversal algorithm or a genetic algorithm, with the optimization objective of minimizing the total operational path length or the total operational time.

[0074] Step 300: Calculate the evapotranspiration loss rate in the irrigation period through real-time meteorological data and an evapotranspiration model, and determine the target irrigation amount of each irrigation unit by combining the evapotranspiration loss rate and the theoretical irrigation amount. The real-time meteorological data includes air temperature, relative humidity, solar radiation, and environmental wind speed. The evapotranspiration model, for example, uses the Penman-Monteith formula recommended by FAO to calculate the evapotranspiration loss rate through the real-time meteorological data, and then combines the theoretical irrigation amount to obtain the target irrigation amount of each irrigation unit.

[0075] Step 400: Obtain irrigation instructions based on the global operation path and the target irrigation amount of each irrigation unit, and the UAV executes the irrigation instructions according to the irrigation amount compensation method. The operation boundary of the global operation path is fused with the target irrigation amount data of the corresponding irrigation unit, and a set of instruction sets are compiled and generated. The instruction sets include spatial path control signals and flow control signals, and the flow control signals are strictly spatio-temporally synchronized with the spatial path. When the UAV determines that it has entered the operation boundary of an irrigation unit through the real-time positioning system, a flow control signal matching the target irrigation amount of the irrigation unit is immediately sent out.

[0076] Step 500: Obtain vegetation index data of the target area again to obtain a second water demand distribution map, obtain an irrigation uniformity parameter according to the first water demand distribution map and the second water demand distribution map, and update the drift motion model. The irrigation uniformity parameter is calculated by the difference between the pixel values in the first water demand distribution map and the second water demand distribution map. Specifically, the irrigation uniformity parameter is wherein, o j is the pixel difference value of the jth pixel between the first water demand distribution map and the second water demand distribution map, is the average value of all pixel difference values, and J is the total number of pixel difference values. The covariance matrix of the process noise in the drift motion model is adjusted according to the irrigation uniformity parameter. When the irrigation uniformity parameter is lower than a preset threshold, the diagonal element value in the covariance matrix is increased. Embodiment Five

[0077] With reference to Figure 10The unmanned aerial vehicle irrigation operation system according to the unmanned aerial vehicle irrigation operation method of the application comprises a data acquisition unit, an irrigation setting unit, a path setting unit, a device control unit, an unmanned aerial vehicle and an external database. The data acquisition unit is configured to acquire vegetation index data, and the data acquisition unit comprises, for example, a camera module, a remote sensing detection module, a spectrum detection module and the like. The irrigation setting unit is configured to generate a theoretical irrigation amount and a target irrigation amount. The path setting unit is configured to generate a global operation path of the unmanned aerial vehicle, and the path setting unit adopts, for example, a shortest path algorithm to perform path planning. The device control unit is configured to generate an irrigation instruction. The unmanned aerial vehicle is configured to execute the irrigation instruction and feed back a state to the device control unit. The external database is configured to acquire real-time weather data. The data acquisition unit generates a first water requirement distribution map and a second water requirement distribution map. The device control unit calibrates the drift force model based on the first water requirement distribution map and the second water requirement distribution map. The device control unit inserts the updated drift force model into the irrigation instruction. In a specific embodiment, when the variation vector is less than a corresponding disturbance threshold, the device control unit inserts an updated spray angle in a dormant section of the irrigation instruction.

[0078] The above merely describes preferred embodiments of the application and is not intended to limit the application. Any modification, equivalent replacement and improvement within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A method for compensating irrigation volume, characterized in that, Includes the following steps: Step 1: Locate the target irrigation unit based on its location coordinates and extract the target irrigation amount for the target irrigation unit; Step 2: Calculate the nozzle flow rate based on the flight speed, spray width, and target irrigation volume; obtain the target atomization parameters and droplet distribution data based on the nozzle flow rate and atomization dynamics model. Step 3: Adjust the nozzle according to the target atomization parameters, collect irrigation height and ambient wind speed, and input the irrigation height, ambient wind speed, flight speed and droplet distribution data into the drift force model to generate the offset vector of the droplet cloud; Step 4: Calculate the target spray angle to compensate for drift based on the offset vector and irrigation height, and adjust the nozzle spray angle to the target spray angle; Step 5: Collect the change vectors of flight speed, irrigation altitude or ambient wind speed in real time. If any change vector is less than the corresponding disturbance threshold, proceed to step 6; otherwise, return to step 3. Step 6: Update the spray angle adjustment vector based on the change vector, adjust the nozzle spray angle using the adjustment vector, and return to Step 5. In step 3, the drift dynamic model includes a state variable function and an offset vector function. First, the state variable function is used to estimate the terminal time t when the droplets settle to the ground or crop canopy. f The state variable x(t) f Furthermore, considering the ideal condition of no wind field, the reference terminal state variable x of the fog droplets... ref Estimate the offset vector y of the fog droplets. The average of the offset vectors of multiple fog droplets is the offset vector y' of the fog cloud. For small-diameter fog droplets, the drift kinetic model is: In the drift dynamics model, the state variable x represents the three-dimensional spatial position and three-dimensional velocity vector of the droplet; A is the state matrix containing air resistance and droplet properties; u1 is the superimposed flow field of the UAV rotor and natural airflow, which changes with flight speed v1, ambient wind speed v2, and irrigation height h, u1 = F(h, v1, v2); u2 is the gravity field; b1 and b2 are the input vectors of u1 and u2, respectively, with b2 remaining constant; n is the process noise; and C is the output matrix. In this drift dynamics model, the initial state variable x0 of the state variable x is provided by the droplet distribution data output from the atomization dynamics model. For large-diameter droplets, the drift dynamics model needs to consider aerodynamic drag outside the Stokes region, exhibiting a nonlinear state-space form. The state variable function is... .

2. The irrigation compensation method according to claim 1, characterized in that, In step 2, the total irrigation amount per unit is calculated based on the area S of the target irrigation unit and the target irrigation amount D. Then the nozzle flow rate Where d1 is the drone's spray width and v1 is its flight speed.

3. The irrigation compensation method according to claim 1, characterized in that, In step 2, the nozzle flow rate and the median target droplet size are taken as inputs. The target atomization parameters are output by querying the nozzle characteristic database. Then, the corresponding droplet distribution data is obtained based on the target atomization parameters. The nozzle characteristic data includes a mapping table between the atomization parameters of each nozzle, the nozzle flow rate and the median droplet size.

4. The irrigation compensation method according to claim 1, characterized in that, In step 4, the offset vector y' is decomposed into offset components y1=(y on the horizontal plane) x ,y y ), calculate the horizontal azimuth angle θ and the pitch angle α of the target spray angle, using the formula: θ = arctan2(y x ,y y )+π,α=arctan(h / ||y1||), where y x Let y1 be the component of the offset vector y1 along the X-axis. y Let y1 be the component of the offset vector y1 in the Y-axis direction, arctan2 be the arctangent function in the four quadrants, π be the reverse compensation, ||y1|| be the magnitude of the offset component y1 on the horizontal plane, and h be the irrigation height.

5. The irrigation volume compensation method according to claim 1, characterized in that, In step 6, the variation vectors include the variation vectors of flight speed, ambient wind speed, and irrigation altitude. First, the variation vectors of flight speed and ambient wind speed are added in the geographic coordinate system to obtain the comprehensive wind speed variation vector. Then, the comprehensive wind speed variation vector is transformed to the UAV body coordinate system and its lateral component is extracted. Based on the magnitude and direction of the lateral component, a horizontal azimuth adjustment is generated through proportional mapping. At the same time, a pitch adjustment is generated based on the variation vector of irrigation altitude. The horizontal azimuth adjustment and pitch adjustment are adjustment vectors.

6. A method for unmanned aerial vehicle (UAV) irrigation operations, characterized in that, Includes the following steps: Step 100: Obtain vegetation index data for the target area and generate a first water demand distribution map, which includes multiple irrigation units and the theoretical irrigation volume of each irrigation unit; Step 200: Determine the operational boundary of each irrigation unit, and perform path planning on the operational boundary to generate the global operational path of the UAV; Step 300: Calculate the evapotranspiration loss rate during the irrigation cycle using real-time meteorological data and evapotranspiration models, and determine the target irrigation amount for each irrigation unit by combining the evapotranspiration loss rate and theoretical irrigation amount; Step 400: Based on the global operation path and the target irrigation amount of each irrigation unit, an irrigation instruction is obtained, and the UAV executes the irrigation instruction according to the irrigation amount compensation method described in claim 1; Step 500: Obtain vegetation index data for the target area again to obtain a second water demand distribution map. Based on the first and second water demand distribution maps, obtain irrigation uniformity parameters and update the drift dynamic model. Irrigation uniformity parameters are calculated based on the differences in pixel values ​​between the first and second water demand distribution maps. o j Let j be the pixel difference between the first and second water demand distribution maps. J is the average of all cell differences, and J is the total number of cell differences. The covariance matrix of process noise in the drift force model is adjusted according to the irrigation uniformity parameter. When the irrigation uniformity parameter is lower than a preset threshold, the diagonal element values ​​in the covariance matrix are increased.

7. A drone irrigation operation system for implementing the drone irrigation operation method of claim 6, characterized in that, include: The data acquisition unit is configured to collect vegetation index data; The irrigation setting unit is configured to generate theoretical irrigation volume and target irrigation volume; The path setting unit is configured to generate a global operational path for the drone; The equipment control unit is configured to generate irrigation commands; The drone is configured to execute the irrigation instructions, wherein, The data acquisition unit generates a first water demand distribution map and a second water demand distribution map. The equipment control unit calibrates the drifting force model based on the first water demand distribution map and the second water demand distribution map. The device control unit inserts the updated drift force model into the irrigation command.

Images