Efficient oil-electric hybrid plant protection unmanned aerial vehicle system

Abstract

The invention discloses an efficient oil-electric hybrid plant protection unmanned aerial vehicle system, and relates to the technical field of plant protection unmanned aerial vehicles, and the system comprises a sensor module, a control center module, a power switching module, a fuel engine and a motor. A fuel-electric power mixing proportion instruction is accurately generated, a power switching module flexibly adjusts the power mixing proportion of a fuel engine and an output shaft of a motor according to the instruction, and in a high-torque demand scene of upwind climbing, the fuel engine outputs high power, the strong power demand is met, and the motor is efficiently driven in a low-power-consumption steady-state flight scene. Meanwhile, the power of the electric motor is reasonably allocated, it is ensured that the unmanned aerial vehicle stably operates under the complex working condition, and energy consumption and operation cost are greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of plant protection unmanned aerial vehicles, and particularly to an efficient oil-electric hybrid plant protection unmanned aerial vehicle system. Background Art

[0002] With the acceleration of the agricultural modernization process, as an efficient and convenient agricultural plant protection tool, plant protection unmanned aerial vehicles have been widely used in the prevention and control of large-area crop diseases and insect pests, spraying of pesticides and fertilizers, etc. Compared with the traditional manual plant protection method, plant protection unmanned aerial vehicles can greatly improve the operation efficiency, reduce the labor cost, and at the same time reduce the direct contact between operators and pesticides, protecting the health of personnel. However, with the continuous expansion of agricultural production scale and the increasing requirements for precision agriculture, the existing plant protection unmanned aerial vehicle technology has gradually revealed some limitations.

[0003] At present, most traditional plant protection unmanned aerial vehicles are relatively backward in energy utilization and power distribution mechanisms. On the one hand, unmanned aerial vehicles with a single energy supply are limited by the energy reserve, and the endurance is seriously insufficient, making it difficult to complete continuous operations in large areas of farmland. Frequent charging or refueling operations greatly affect the operation efficiency. On the other hand, even for unmanned aerial vehicles using an oil-electric hybrid mode, only a fixed proportion of power output scheme is used. During the actual flight process, the unmanned aerial vehicle faces various and complex flight postures such as climbing against the wind, turning, hovering, etc. At the same time, the continuous spraying of pesticides or fertilizers causes the load to change continuously. The fixed proportion of power output can neither flexibly adapt to the power requirements under different working conditions nor lead to serious energy waste, and it is impossible to achieve an optimal balance between energy utilization efficiency and operation efficiency. In addition, the existing technology lacks an effective mechanism for predicting the change of rotor aerodynamic load and quickly adjusting the power in the face of natural wind interference in complex farmland environments, especially gust impacts, resulting in poor flight stability of the unmanned aerial vehicle and making it difficult to ensure the accurate and uniform spraying of pesticides or fertilizers.

[0004] In summary, the existing plant protection unmanned aerial vehicle technology has obvious deficiencies in terms of the flexibility of energy distribution, the adaptability to complex flight working conditions and load changes, and the ability to cope with natural wind interference. These problems restrict the application effect and development potential of plant protection unmanned aerial vehicles in modern agricultural production. It is urgent to develop a new type of plant protection unmanned aerial vehicle system with the ability of dynamic energy distribution and load change prediction. Summary of the Invention

[0005] The objective of the present invention is to make up for the deficiencies of the prior art and provide an efficient fuel-electric hybrid plant protection UAV system, which can intelligently switch between fuel and electric power through dynamic energy distribution. In high-torque scenarios where strong thrust is required for climbing against the wind, the fuel engine outputs full power to ensure power; while in low-power steady-state flights such as horizontal uniform flight, the electric motor operates efficiently to reduce fuel consumption. Precise power distribution and load prediction ensure that the UAV can operate stably under various flight postures and load conditions.

[0006] To solve the above technical problems, the present invention provides the following technical solution: An efficient fuel-electric hybrid plant protection UAV system, which consists of: a sensor module, a control center module, a power switching module, a fuel engine, and an electric motor;

[0007] The sensor module includes a torque sensor, a flight attitude sensor, and a spraying load sensor, which are used to collect the torque value T of the UAV's power output in real time r , the attitude data A, and the liquid medicine injection reaction force data L;

[0008] The control center module is composed of a feedforward prediction sub-module, a feedback correction sub-module, and a dynamic distribution sub-module. The feedforward prediction sub-module processes sensor data and is used to predict the torque demand value T d , the feedback correction sub-module calibrates the prediction error in real time, and the dynamic distribution sub-module generates a fuel-electric power mixing ratio instruction by fusing the real-time flight attitude data and the prediction error through a bionic neural network;

[0009] The power switching module is connected to the output shafts of the fuel engine and the electric motor, and dynamically adjusts the fuel-electric power mixing ratio R according to the instructions of the control center module;

[0010] The fuel engine and the electric motor are used as power output devices, and provide corresponding power according to different working conditions under the control of the control center module.

[0011] Furthermore, the torque sensor in the sensor module is embedded in the root substrate of the rotor, and the micro-strain on the rotor surface is measured in real time through the change of the electron tunneling current to predict the torque fluctuation ΔT;

[0012] The flight attitude sensor is used to monitor the attitude data A during the flight of the UAV in real time;

[0013] The spraying load sensor is used to obtain the remaining amount of pesticides in real time, so as to reflect the liquid medicine injection reaction force data L generated by the change of the spraying load.

[0014] Even further, the torque fluctuation ΔT of the torque sensor is: dy, μ is the air dynamic viscosity, u is the flow velocity on the rotor surface, y is the radial coordinate of the rotor, used to determine different position points on the rotor, t0 is the starting time.

[0015] Furthermore, the feed-forward prediction sub-module processes the data collected by the sensor module, generates a torque demand value using the fused data, and predicts the torque demand change trend of the UAV. The torque demand value is calculated through where T d is the predicted torque demand value, that is, the torque magnitude required by the UAV during operation, a 0 is the constant term, a 1 , a 2 , a 3 , a i are the weight coefficients corresponding to the data respectively, reflecting the influence degree of different data on the torque demand. ΔT is the torque fluctuation value, A is the flight attitude data vector, L is the liquid medicine injection reaction force data, X i is the environmental data, and □ is the error term, used to represent the deviation between the predicted value and the actual value.

[0016] Furthermore, the feedback correction sub-module compares the actually measured torque value T r with the torque demand value T d predicted by the feed-forward prediction sub-module, calculates the error E e between them, and adjusts the weight coefficients a e in the feed-forward prediction model according to the calculated error E 0 , a 1 , a 2 , a 3 , a i . Among them, the error E e = T r - T d . The adjustment of the weight coefficients adjusts the weight coefficients in the feed-forward prediction through gradient descent, and its update process is: where is the updated weight coefficient value, is the weight coefficient value before update, α is the learning rate, controlling the step size of each weight update, is the partial derivative of the square of the error with respect to the weight coefficient a j .

[0017] Furthermore, the in the feedback correction sub-module Taking the partial derivative with respect to it gives:

[0018] When j = 0,

[0019] When j = 1,

[0020] When j = 2,

[0021] When j = 3,

[0022] When j = i, i = 4, …, n,

[0023] By continuously iterating and updating the weight coefficients, the UAV system can operate stably.

[0024] Furthermore, the dynamic allocation sub-module uses a bionic neural network to fuse real-time flight attitude, torque prediction, and liquid medicine injection reaction force data, and outputs the oil-electric power distribution ratio. The bionic neural network inputs the integrated data into the bionic neural network. During the training process, based on historical data, the network learns the relationship between the input data and the optimal oil-electric power mixing ratio under different working conditions. For the currently input data set, the neural network conducts information transmission and calculation through the weight connections between neurons in each layer. After weighted summation and activation processing, the input data is analyzed and feature extracted, so as to evaluate the oil-electric power mixing ratio under the current working condition. This bionic neural network includes an input layer, several hidden layers, and an output layer.

[0025] Furthermore, the working principle of the bionic neural network in the dynamic allocation sub-module is as follows:

[0026] Calculation from the input layer to the hidden layer: The input layer receives the integrated data, that is, the torque demand prediction value T d , the error E e and the flight attitude data A, and represents it as the input vector X = [T d , E e , A]. The neurons in the input layer and the neurons in the hidden layer are connected through the weight matrix W ih . i represents the index of the neurons in the input layer, h represents the index of the neurons in the hidden layer. The input z h of the h-th neuron in the hidden layer = ∑ i W ih X i + b h , where X i is an element in the input vector X, b h is the bias of the h-th neuron in the hidden layer. After being processed by the activation function f(z) = max(0, z), the output y h of the h-th neuron in the hidden layer = f(z h );

[0027] Hidden layer to output layer calculation: The hidden layer and the output layer are connected by the weight matrix W ho Connecting, h represents the index of the hidden layer neuron, o represents the index of the output layer neuron, and the input z of the output layer neuron o is: z o = ∑ h W ho y h + b o , where b o is the bias of the output layer neuron. After passing through the activation function, the output layer directly outputs the hybrid ratio command R of the electric and fuel power. The activation function takes z o as the input of the σ function, that is where R represents the proportion of the motor power, 1 - R is the proportion of the fuel engine power, and the output value range is between 0 and 1, representing the proportion of the motor power, 1 - R is the proportion of the fuel engine power. During the training process, by adjusting the weight matrix W ih and W ho as well as the bias b h 、b o , the bionic neural network can output the hybrid ratio command of the electric and fuel power that conforms to the actual working conditions.

[0028] Compared with the prior art, the efficient electric - fuel hybrid plant protection UAV system has the following beneficial effects:

[0029] First, the present invention fuses real - time flight attitude and predicted torque demand change data through a bionic neural network to accurately generate the hybrid ratio command of the electric and fuel power. According to this command, the power switching module flexibly adjusts the power hybrid ratio of the output shafts of the fuel engine and the motor. In the high - torque demand scenario of climbing against the wind, the fuel engine outputs high power to meet the strong power demand. In the low - power - consumption steady - flight scenario, the motor drives efficiently, reducing fuel consumption. At the same time, the power of the electric motor is reasonably allocated to ensure the stable operation of the UAV under complex working conditions, greatly reducing energy consumption and operation costs.

[0030] Second, the control center module of the present invention realizes the precise regulation of the power system based on the flight attitude and load data collected by the sensors, ensuring that the UAV can fly smoothly under various flight attitudes and load conditions. This makes the spraying process of pesticides or fertilizers more stable and uniform, effectively improving the quality of plant protection operations and ensuring the stable operation of the UAV under complex meteorological conditions.

[0031] Other advantages, objectives, and features of the present invention will be described to some extent in the subsequent description, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. Brief Description of the Drawings

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0033] Figure 1 It is an operation flow chart of an efficient oil-electric hybrid plant protection UAV system;

[0034] Figure 2 It is a working principle diagram of the control center module. Detailed Embodiments

[0035] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following will, in conjunction with the drawings and preferred embodiments, detail the specific embodiments, structures, features and their effects of the present invention as follows.

[0036] Embodiment 1

[0037] This embodiment elaborates in detail the specific operation process of an efficient oil-electric hybrid plant protection UAV system. The system collects data in real time through the sensor module, the control center module processes the data using specific algorithms and generates instructions, and the power switching module adjusts the oil-electric power mixing ratio according to the instructions to achieve the efficient operation of the UAV under different working conditions.

[0038] The sensor module includes a torque sensor, a flight attitude sensor, and a spraying load sensor, which are used to collect the torque value T of the UAV power output in real time r , attitude data A, and liquid medicine injection reaction force data L. Among them, the torque sensor uses the change of electron tunneling current to measure the micro-strain on the surface of the rotor in real time. When the surface of the rotor is subjected to force and generates micro-strain, it will cause the electron tunneling current to change. By monitoring and analyzing the current change, the micro-strain situation on the surface of the rotor can be obtained in real time, and predict the torque fluctuation ΔT, where μ is the air dynamic viscosity, which reflects the degree of air resistance to the movement of the rotor, u is the flow velocity on the surface of the rotor, and y is the radial coordinate of the rotor, which is used to determine different position points on the rotor. Because the air dynamic forces received by different positions of the rotor are different, by determining different position points, the torque fluctuation can be calculated more accurately, t 0is the starting time, and here we take the current moment. The integration time is 0.05 seconds, that is, to predict the torque fluctuation in the next 50 ms. The torque sensor transmits the torque fluctuation data measured and calculated in real time to the feedforward prediction sub-module of the control center module; during the flight of the UAV, the flight attitude sensor continuously and real-time monitors the attitude data A of the UAV, including attitude information such as the angular velocity, pitch angle, roll angle, and yaw angle of the UAV. When the UAV performs a turning operation, the attitude sensor will quickly capture the changes in angular velocity and acceleration and convert these changes into attitude data A. These attitude data will be transmitted to the control center module in real time to provide important flight state information for the feedforward prediction sub-module and the dynamic allocation sub-module; the spraying load sensor obtains the remaining amount of pesticides in real time to reflect the data L of the liquid medicine injection reaction force generated by the change in spraying load. The remaining amount of pesticides is calculated by measuring the pressure change or weight change in the pesticide storage container. When the pesticide is sprayed out of the nozzle of the UAV, the pressure or weight in the storage container will decrease accordingly. The sensor converts this change into an electrical signal to obtain the data L of the liquid medicine injection reaction force. When the spraying speed of the pesticide increases, the pressure in the storage container drops faster, the pressure change detected by the sensor is greater, and the value of L will also change accordingly. These data are also transmitted to the feedforward prediction sub-module of the control center module in real time.

[0039] As Figure 2 shown, the control center module consists of a feedforward prediction sub-module, a feedback correction sub-module, and a dynamic allocation sub-module. The feedforward prediction sub-module receives the torque fluctuation data ΔT, flight attitude data A, and liquid medicine injection reaction force data L from the sensor module, as well as environmental data X i (such as environmental wind speed and wind direction), fuses and processes these data, and uses to generate a torque demand value T d , where a 0 is a constant term, and a 1 , a 2 , a 3 , a i are the weight coefficients corresponding to the data respectively, reflecting the influence degree of different data on the torque demand. For example, in a headwind environment, the influence of torque fluctuation ΔT on the torque demand is relatively large, and the value of a 1 will be relatively large accordingly. A is the flight attitude data vector, which contains multiple attitude information of the pitch angle, roll angle, and yaw angle of the UAV; L is the liquid medicine injection reaction force data, reflecting the influence of spraying load on the torque demand, and X i is the environmental data, and □ is the error term, used to represent the deviation between the predicted value and the actual value. The feedforward prediction sub-module calculates the torque demand value T according to d, predict the changing trend of the torque demand of the drone and transmit this value to the feedback correction sub-module and the dynamic allocation sub-module. The feedback correction sub-module measures the actual torque value T r and the torque demand value T d predicted by the feedforward prediction sub-module, calculate the error E e between them, where E r = T d - T e . This error reflects the accuracy of the feedforward prediction. The larger the error, the greater the deviation between the predicted value and the actual value. According to the calculated error E 0 , the feedback correction sub-module uses gradient descent to adjust the weight coefficients of the feedforward prediction. The weight coefficients a 1 , a 2 , a 3 , a i are updated as follows where is the updated weight coefficient value, is the weight coefficient value before update, α is the learning rate, which controls the step size of each weight update, is the partial derivative of the squared error with respect to the weight coefficient a j . Taking the partial derivative with respect to it according to the of the feedforward prediction sub-module, we get: when j = 0, when j = 1, when j = 2, when j = 3, when j = i (i = 4,..., n), . The feedback correction sub-module continuously iterates to update the weight coefficients, making the predicted value closer to the actual value, thereby improving the accuracy of the feedforward prediction model and ensuring the stable operation of the drone system. Each updated weight coefficient is fed back to the feedforward prediction sub-module for the next torque demand prediction calculation; the dynamic allocation sub-module uses a bionic neural network to fuse the real-time flight attitude data A, the torque prediction value T d and the liquid medicine injection reaction force data L (indirectly included through the error E e ), and outputs the oil-electric power distribution ratio R. The bionic neural network includes an input layer, a hidden layer, and an output layer. Among them, the input layer receives the integrated data, that is, the torque demand prediction value T d , the error E e and the flight attitude data A, and represents them as an input vector X = [T d , E e , A]. The neurons in the input layer are connected to the neurons in the hidden layer through the weight matrix W ih . i represents the index of the neurons in the input layer, h represents the index of the neurons in the hidden layer, and the input z of the h-th neuron in the hidden layerh = ∑ i W ih X i + b h , where X i is an element in the input vector X, and b h is the bias of the h-th neuron in the hidden layer. The role of the bias b h is to provide an additional learnable parameter for the neuron, increasing the fitting ability. After being processed by the activation function f(z) = max(0, z), the output y of the h-th neuron in the hidden layer is obtained h = f(z h ). The role of the activation function f(z) is to perform a non-linear transformation on the input of the neuron, enabling the neural network to learn more complex functional relationships. The hidden layer and the output layer are connected by the weight matrix W ho . Here, h represents the index of the hidden layer neuron, o represents the index of the output layer neuron, and the input z of the output layer neuron o = ∑ h W ho y h + b o , where b o is the bias of the output layer neuron. After being processed by the activation function , the oil-electric power hybrid ratio command R is directly output. Taking z o as the input of the σ function, that is . Here, R represents the proportion of the electric motor power, and 1 - R is the proportion of the fuel engine power, and the output value range is between 0 and 1. During the training process, by adjusting the weight matrices W ih and W ho as well as the biases b h and b o , the bionic neural network can output the oil-electric power hybrid ratio command that conforms to the actual working conditions. The training process uses historical data to adjust these parameters by minimizing the error between the actual oil-electric power hybrid ratio and the oil-electric power hybrid ratio predicted by the neural network, enabling the neural network to gradually learn the relationship between the input data and the optimal oil-electric power hybrid ratio under different working conditions. The dynamic allocation sub-module transmits the generated oil-electric power hybrid ratio command R to the power switching module.

[0040] The power switching module is connected to the output shafts of the fuel engine and the electric motor. According to the fuel-electric power mixing ratio instruction R transmitted by the dynamic distribution sub-module of the control center module, it dynamically adjusts the fuel-electric power mixing ratio. When the value of R is relatively large, it means that the electric motor needs to provide more power. The power switching module will adjust the working states of the fuel engine and the electric motor, so that the electric motor outputs a larger power, and at the same time correspondingly reduces the power output of the fuel engine; on the contrary, when the value of R is relatively small, the fuel engine will output more power, and the power output of the electric motor will be correspondingly reduced. For example, in the high-torque demand scenario of the unmanned aerial vehicle (UAV) climbing against the wind, the R value output by the dynamic distribution sub-module is relatively small. After receiving the instruction, the power switching module controls the fuel engine to output high power to meet the strong power demand; while in the low-power steady-state flight scenarios such as horizontal uniform flight, the R value is relatively large. The power switching module controls the electric motor to drive efficiently, reduces the fuel consumption of the fuel engine, and at the same time reasonably allocates the power of the electric motor to ensure the stable operation of the UAV under complex working conditions. In this way, the power switching module realizes the intelligent switching of fuel-electric power of the UAV under different working conditions, improves the energy utilization efficiency, and reduces the operation cost.

[0041] As power output devices, the fuel engine and the electric motor provide corresponding power according to different working conditions under the control of the control center module. When the power switching module adjusts the fuel-electric power mixing ratio, the fuel engine and the electric motor work together according to the instruction. In the high-torque demand scenario, such as climbing against the wind, the fuel engine, under the control of the power switching module, generates high-temperature and high-pressure gas by burning fuel, pushes the piston to move, and then drives the crankshaft to rotate, outputting high power. In this process, the fuel injection system of the fuel engine will increase the fuel injection volume according to the instruction of the power switching module to improve the output power of the engine. The electric motor plays a major role in the low-power steady-state flight scenario, such as horizontal uniform flight. Under the control of the power switching module, the electric motor converts electrical energy into mechanical energy to drive the rotor of the UAV to rotate. The speed and torque of the electric motor can be adjusted by controlling the input current and voltage. The power switching module accurately controls the input current and voltage of the electric motor according to the instruction of the dynamic distribution sub-module to make it output appropriate power to ensure the stable flight of the UAV. At the same time, in the transitional working condition, the fuel engine and the electric motor cooperate with each other to jointly provide power to ensure the smooth operation of the UAV.

[0042] In summary, in this embodiment, each module of the high-efficiency hybrid fuel-electric plant protection UAV system collaborates closely. The sensor module collects data in real time and accurately, providing a reliable information source for the control center module. Through the coordinated work of the feedforward prediction sub-module, feedback correction sub-module, and dynamic allocation sub-module in the control center module, and by using bionic neural network technology, accurate prediction of torque demand and intelligent generation of the hybrid fuel-electric power ratio are achieved. The power switching module accurately adjusts the power output of the fuel engine and the electric motor according to the instructions of the control center module, enabling the UAV to operate efficiently under different working conditions. The fuel engine and the electric motor provide power stably according to the control of the power switching module, ensuring the flight and operation of the UAV. Through the coordinated operation of these modules, the UAV system realizes efficient utilization of energy, reduces operation costs, improves the quality of plant protection operations, and ensures the stable operation of the UAV under complex weather conditions.

[0043] Embodiment 2

[0044] As Figure 1 shown, this embodiment provides the specific steps for a high-efficiency hybrid fuel-electric plant protection UAV system to perform a pesticide spraying task:

[0045] (1) Real-time data collection

[0046] During the flight operation of the UAV, the torque sensor continuously monitors the micro-strain situation at the root of the rotor and converts it into torque fluctuation data;

[0047] The flight attitude sensor continuously collects attitude information such as the pitch angle, roll angle, and yaw angle of the UAV, as well as the corresponding angular velocity data, comprehensively reflecting the changes in the flight attitude of the UAV;

[0048] The spraying load sensor measures the remaining amount of pesticides in real time to obtain the data of the reaction force of the liquid medicine injection generated by the change in the spraying load;

[0049] (2) Predict torque demand

[0050] The feedforward prediction sub-module of the control center module collects the torque fluctuation data, flight attitude data, and reaction force data of the liquid medicine injection from the sensor module;

[0051] The feedforward prediction sub-module predicts the torque demand value required for the next operation of the UAV and simultaneously predicts the change trend of the torque demand;

[0052] (3) Error calculation and model correction

[0053] The feedback correction sub-module obtains the actually measured torque value, compares it with the torque demand value predicted by the feedforward prediction sub-module, and calculates the error between the two;

[0054] According to the calculated error, the feedback correction sub-module adjusts the parameters in the feed-forward prediction model according to a specific algorithm to improve the prediction accuracy and make the predicted value closer to the actual value;

[0055] (4) Generate power distribution instructions

[0056] The dynamic distribution sub-module receives the predicted torque demand value from the feed-forward prediction sub-module, the error calculated by the feedback correction sub-module, and the real-time flight attitude data collected by the flight attitude sensor;

[0057] The dynamic distribution sub-module uses a bionic neural network to fuse and analyze these data, evaluate the optimal fuel-electric power hybrid ratio under the current working conditions, and generate the corresponding fuel-electric power hybrid ratio instruction;

[0058] (5) Power output regulation

[0059] The power switching module receives the fuel-electric power hybrid ratio instruction generated by the dynamic distribution sub-module;

[0060] According to the instruction, the power switching module adjusts the power hybrid ratio of the output shafts of the fuel engine and the motor, so that the fuel engine and the motor output power according to different working conditions to meet the flight and operation requirements of the UAV;

[0061] (6) Continuous cyclic optimization

[0062] During the entire operation process of the UAV, the steps of data collection, prediction, correction, instruction generation, and power regulation are continuously cycled;

[0063] With the changes in the operation environment and the UAV state, each module continuously adjusts its work, continuously optimizes the fuel-electric power distribution and power output, and ensures the efficient and stable operation of the UAV.

[0064] The above is only a preferred embodiment of the present invention, and it is not a limitation to the present invention in any form. Although the present invention has been disclosed above with the preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to the equivalent embodiments with equivalent changes within the scope of the technical solution of the present invention. However, as long as it does not depart from the content of the technical solution of the present invention, any brief modifications, equivalent changes and modifications made to the above embodiments according to the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A highly efficient hybrid plant protection drone system, characterized in that: The system consists of: sensor module, control center module, power switching module, fuel engine, and electric motor; The sensor module includes a torque sensor, a flight attitude sensor, and a spray load sensor, which are used to collect the torque value T of the UAV power output in real time. r , posture data A and liquid injection reaction force data L; The control center module is composed of a feedforward prediction submodule, a feedback correction submodule and a dynamic allocation submodule. The feedforward prediction submodule processes sensor data to predict the torque demand value T d , the feedback correction submodule calibrates the prediction error in real time, and the dynamic allocation submodule fuses the real-time flight attitude data and the prediction error through a bionic neural network to generate an oil-electric power hybrid ratio instruction; The power switching module connects the fuel engine and the motor output shaft, and dynamically adjusts the oil-electric power mixing ratio R according to the command of the control center module; The fuel engine and the electric motor serve as power output devices, and provide corresponding power according to different working conditions under the control of the control central module.

2. The high-efficiency hybrid plant protection UAV system according to claim 1, characterized in that: The torque sensor in the sensor module is embedded in the rotor root substrate, and measures the micro-strain on the rotor surface in real time through the change of electron tunneling current to predict the torque fluctuation ΔT; The flight attitude sensor is used to monitor the attitude data A of the drone in real time during flight; The spray load sensor is used to obtain the remaining amount of pesticide in real time, so as to reflect the liquid spray reaction force data L generated by the spray load change.

3. The high-efficiency hybrid plant protection UAV system according to claim 2, characterized in that: The torque fluctuation ΔT of the torque sensor is: μ is the aerodynamic viscosity, u is the flow velocity on the rotor surface, y is the rotor radial coordinate, which is used to determine different positions on the rotor, and t0 is the starting time.

4. The high-efficiency hybrid plant protection UAV system according to claim 1, characterized in that: The feedforward prediction submodule fuses the data collected by the sensor module, generates a torque demand value using the fused data, and predicts the torque demand change trend of the UAV. The torque demand value is calculated by Calculate, where T d is the predicted torque demand value, that is, the torque required by the drone during operation, a0 is a constant term, a1, a2, a3, a i are the weight coefficients of the corresponding data, reflecting the influence of different data on the torque demand, ΔT is the torque fluctuation value, A is the flight attitude data vector, L is the liquid injection reaction force data, X i is the environmental data, and □ is the error term, which is used to represent the deviation between the predicted value and the actual value.

5. The high-efficiency hybrid plant protection UAV system according to claim 1, characterized in that: The feedback correction submodule converts the actual measured torque value T r The torque demand value T predicted by the feedforward prediction submodule d Compare and calculate the error E between the two e , according to the calculated error E e , the weight coefficients a0,a1,a2,a3,a in the feedforward prediction i Adjustment is performed, wherein the error E e =T r -T d , the weight coefficient is adjusted by gradient descent to adjust the weight coefficient in the feedforward prediction model, and the updating process is: in, is the updated weight coefficient value, is the weight coefficient value before updating, α is the learning rate, which controls the step size of each weight update. is the square of the error with respect to the weight coefficient a j The partial derivative of .

6. The high-efficiency hybrid plant protection UAV system according to claim 5, characterized in that: The feedback correction submodule According to the feedforward prediction submodule Taking partial derivatives we get: When j = 0, When j = 1, When j = 2, When j = 3, When j=i, i=4,…,n, By continuously iterating and updating the weight coefficients, the UAV system can operate stably.

7. The high-efficiency hybrid plant protection UAV system according to claim 1, characterized in that: The dynamic allocation submodule adopts a bionic neural network to fuse real-time flight attitude, torque prediction and liquid injection reaction force data, and outputs the oil-electric power allocation ratio. The bionic neural network inputs the integrated data into the bionic neural network. During the training process, the network learns the relationship between input data and the optimal oil-electric power mixing ratio under different working conditions based on historical data. For the current input data set, the neural network transmits and calculates information through the weight connection between each layer of neurons. After weighted summation and activation processing, the input data is analyzed and features are extracted, thereby evaluating the oil-electric power mixing ratio under the current working condition. The bionic neural network includes an input layer, several hidden layers and an output layer.

8. The high-efficiency hybrid plant protection UAV system according to claim 7, characterized in that: The working principle of the bionic neural network in the dynamic allocation submodule is: Input layer to hidden layer calculation: The input layer receives the integrated data, that is, the torque demand prediction value T d , Error E e and flight attitude data A, which is represented as input vector X = [T d ,E e ,A], the input layer neurons and the hidden layer neurons are connected by the weight matrix W ih connect, i represents the input layer neuron index, h represents the hidden layer neuron index, and the input z of the hth neuron in the hidden layer h =∑ i W ih X i +b h , where X i is the element in the input vector X, b h is the bias of the hth neuron in the hidden layer, and is processed by the activation function f(z)=max(0,z) to obtain the output y of the hth neuron in the hidden layer h =f(z h ); Calculation from hidden layer to output layer: The hidden layer and the output layer are connected by the weight matrix W ho connection, h represents the hidden layer neuron index, o represents the output layer neuron index, and the input z of the output layer neuron o is: z o =∑ h W ho y h +b o , where b o is the bias of the output layer neuron. After being processed by the activation function, the output layer directly outputs the oil-electric power mixing ratio command R. The activation function z o As the input of the σ function, Among them, R represents the proportion of electric motor power, and 1-R represents the proportion of fuel engine power. The output value range is between 0 and 1, which represents the proportion of electric motor power, and 1-R represents the proportion of fuel engine power. During the training process, by adjusting the weight matrix W ih and W ho and bias b h 、b o , so that the bionic neural network can output the oil-electric power mixing ratio command that meets the actual working conditions.

Images