Intelligent efficient vertical axis wind turbine with variable flaps and propeller pitches

Abstract

An intelligent efficient vertical axis wind turbine with variable flaps and propeller pitches comprises blades, blade sliding rings, a main body sliding ring, a motor, a connecting rod, a connecting disc, an intermediate shaft, a motor connecting shaft and a rolling bearing. The blade comprises a flap, a flap connecting rod, a connecting rod, a rudder arm, a steering engine, a blade shaft and a blade front part, wherein the flap, the flap connecting rod, the connecting rod, the rudder arm and the rudder mechanism form a parallelogram-shaped connecting rod mechanism; the steering engine is a driving rocker, rotates by a certain angle under the driving of electric power, and can drive a driven rocker composed of a flap and a flap connecting rod to rotate by a certain angle. The invention further provides a method for controlling the variable flaps of the vertical axis wind turbine through the deep reinforcement learning algorithm. According to the invention, the pitching moment generated after the flap is driven by the steering engine to deflect is indirectly regulated and controlled. Under driving of methods such as deep reinforcement learning, the flap angle can be intelligently adjusted in a closed-loop mode along with the rotor phase and the incoming flow direction, and real-time optimization of the lift resistance and the direction of the blade is achieved.

Description

Technical Field

[0001] The invention relates to the technical field of wind power, and in particular to a high-efficiency vertical axis wind turbine with intelligent variable flaps and blade pitches. Background Art

[0002] At present, wind turbines are mainly divided into vertical axis wind turbines and horizontal axis wind turbines. For small urban environments and areas with changeable wind directions, vertical axis wind turbines have certain advantages over horizontal axis wind turbines because they can automatically adapt to changes in wind direction without adjusting direction. However, due to the constant change in the blade angle of attack, the blades of vertical axis wind turbines face serious dynamic stall problems, causing an increase in unstable loads and a serious decrease in power generation efficiency. In addition, compared with horizontal axis wind turbines, the self-starting performance of vertical axis wind turbines is relatively poor, and the motor needs to be driven to a certain speed before it can work stably. In response to these two prominent problems, researchers proposed to suppress dynamic stall by adjusting the blade pitch angle, thereby improving the power generation efficiency of the wind turbine. In current experimental research, the pitch angle is mainly adjusted in an open-loop manner, and the control instructions have poor adaptability to the rotor phase and inflow conditions, making it difficult to maintain a high power generation efficiency in complex meteorological conditions and wind farm resources (Zhang Lixun, Liang Yingbin, Jiao Qifei, Zhang Song, Zou Yupeng. A vertical axis wind turbine pitch control mechanism, application number: CN111997835A, 2020). The method of real-time control of the pitch angle is currently mainly combined with numerical simulation, setting the model in advance, and the numerical simulation calculation parameter settings are mostly based on the working conditions under a certain Reynolds number, and the universality is poor (Zhang Qiang, Xu Tao, Weng Yujie, Yue Minnan, Miao Weipao, Liu Qingsong, Li Chun, Jia Wenzhe. Vertical axis wind turbine pitch control method based on tip speed ratio and phase angle changes, publication number: CN118273878A, 2024). In addition, the pitch angle is mostly adjusted by installing a motor in the middle to achieve the deflection of the blades, which requires higher control power and longer starting and braking times (Reference [1] Han Z, Chen H, Chen Y, et al. Aerodynamic performance optimization of vertical axis wind turbine with straight blades based on synergic control of pitch and flap [J]. Sustainable Energy Technologies and Assessments, 2023 (Jun.): 57.). Summary of the invention

[0003] In view of the problems existing in the prior art, in order to maximize the net power generation of the wind turbine, the present invention proposes a high-efficiency vertical axis wind turbine with intelligent variable flaps and pitch, the device comprises a blade 1, a blade slip ring 2, a main body slip ring 3, a motor 4, a connecting rod 5, a connecting plate 6, an intermediate shaft 7, a motor connecting shaft 8, and a rolling bearing 9; wherein

[0004] The intermediate shaft 7 is a vertical solid cylindrical long rod, which is connected to the rotating shaft of the motor 4 below it through a set screw, and the intermediate shaft 7 can rotate with the rotating shaft of the motor 4;

[0005] The connecting rod 5 is a solid long rod placed horizontally;

[0006] The connecting disc 6 is a flat cylinder with a small cylindrical boss on the center. The connecting disc 6 and the center of the boss are provided with threaded holes that penetrate from top to bottom. The intermediate shaft 7 passes through the holes from bottom to top. The connecting disc 6 and the intermediate shaft 7 are fixed by a fastening device and rotate synchronously with the intermediate shaft 7. The upper surface of the connecting disc 6 is provided with fixing devices arranged radially at intervals of 120° at the central angle, which are used to fix the connecting disc 6 and the three connecting rods 5. The connecting disc 6 drives the blades 1 connected to the ends of the connecting rods 5 to rotate synchronously. The intermediate shaft 7 rotates synchronously with the motor shaft, and the connecting disc 6 rotates synchronously with the intermediate shaft 7.

[0007] Two upper and lower connecting plates 6 are arranged above and below the intermediate shaft, and are used to fix the inner ends of three connecting rods 5 arranged at intervals of 120 degrees on their upper surfaces; the outer ends of the three connecting rods 5 correspond to each other in pairs, and the upper and lower connecting rods 5 are used to fix one blade 1; six identical connecting rods 5 are arranged horizontally, and a total of three blades 1 are fixed; when the connecting plate 6 rotates with the motor shaft, the connecting rod 5 rotates with the connecting plate 6;

[0008] The blade 1 comprises three blades 1 placed vertically and evenly spaced apart, and the blade 1 is sandwiched between two upper and lower connecting rods 5;

[0009] The blade slip ring 2 is located below the blade 1 and connected to the connecting rod 5. The outer rotor part of the blade slip ring 2 is fixed to the blade 1 so that the two can rotate synchronously. The outer rotor part of the blade slip ring 2 is connected to the connecting rod 5 through the rolling bearing 9 and can rotate freely on the connecting rod 5. One end of the wire of the stator part inside the blade slip ring 2 is connected to the stator part, and after passing through the stator part, it continues to extend along the connecting rod 5 to connect with the wire of the rotor part of the main slip ring 3.

[0010] The main body slip ring 3 is located near the lower end of the intermediate shaft 7, above the motor 4 and below the lower connection plate 6; but it maintains a certain distance from the bottom end of the intermediate shaft 7, and the distance is used to arrange the wires at the lower end of the slip ring; the rotor part of the main body slip ring 3 is fixed on the intermediate shaft 7 and rotates with the intermediate shaft 7. The wires at the rotor are fixed on the intermediate shaft 7 and connected to the wires of the blade slip ring 2; the wires at the stator are led out along the upper surface of the motor 4 and connected to the external power supply;

[0011] The motor connecting shaft 8 is used to connect the intermediate shaft 7 and the main shaft of the motor 4. The intermediate shaft 7 and the motor connecting shaft 8, and the main shaft of the motor 4 and the motor connecting shaft 8 are connected by fastening devices respectively so that they can rotate together;

[0012] The motor 4 is provided with a Hall sensor and a driving circuit board.

[0013] In one embodiment of the present invention, a plurality of threaded holes are radially arranged at intervals of 120° at the central angle on the upper surface of the connecting disk 6 for connecting the connecting disk 6 and the connecting rod 5 via bolts; holes are also punched at corresponding positions of the connecting rod 5 to allow the bolts to pass through.

[0014] In another embodiment of the present invention, the cross section of the connecting rod 5 is elliptical, the ratio of the major axis to the minor axis is 3-8, and the length of the major axis ranges from 10-30 mm; the spanwise positions where the connecting rod 5 is connected to the blade 1 are at both ends of the blade 1.

[0015] In a specific embodiment of the present invention, the major axis length of the connecting rod 5 is 20 mm, and the material is aluminum alloy; the chord-wise position where the connecting rod 5 is connected to the blade 1 is 25% of the chord length.

[0016] In another embodiment of the present invention, the blade 1 adopts a NACA four-digit airfoil, with a typical chord length ranging from 0.1 to 10 m and an aspect ratio ranging from 2 to 5; and a movable flap is installed on the blade 1.

[0017] In another specific embodiment of the present invention, the blade 1 has an aspect ratio of 3 and is made of high-performance plastic.

[0018] In another embodiment of the present invention, one end of the wire of the stator part inside the blade slip ring 2 is connected to the stator part, and after passing through the stator part, it is tied with a rolled belt along the connecting rod 5 and continues to extend to connect with the wire of the rotor part of the main slip ring 3.

[0019] Provided is a blade, which is suitable for the above-mentioned high-efficiency vertical axis wind turbine with intelligent variable flaps and pitch, comprising a flap 101, a flap connecting rod 102, a connecting rod 103, a rudder arm 104, a steering gear 105, a blade shaft 106, and a blade front part 107; wherein the flap 101, the flap connecting rod 102, the connecting rod 103, the rudder arm 104 and the steering gear 105 constitute a parallelogram connecting rod mechanism; the steering gear 105 is an active rocker, which rotates a certain angle under electric drive, and can drive the slave rocker composed of the flap 101 and the flap connecting rod 102 to rotate a certain angle;

[0020] The flap 101 is hingedly connected to the front portion 107 of the blade by a flexible material;

[0021] The flap connecting rod 102 is in a T-shape as a whole, and is fixed to the flap 101 through two threaded holes on the "one" part of the T-shape, forming a follower rocker in the connecting rod mechanism; and the other end is hingedly connected to the connecting rod 103 through a through hole;

[0022] The connecting rod 103 is an oblong shape, and has a hole on the left and right, respectively connected to the rudder arm 104 and the flap connecting rod 102 by bolts;

[0023] The rudder arm 104 is in the shape of an oblong, one end of which is connected to the connecting rod 103 by a bolt, and the other end of which is connected to the steering gear shaft of the steering gear 105;

[0024] The servo 105 is fixed to the front middle position of the blade front part 107 by bolt connection; the power line and signal line of the servo 105 are connected to the wires of the stator part of the blade slip ring 2;

[0025] The rotation of the steering gear shaft drives the rudder arm 104 and the flap connecting rod 102 to move, thereby causing the flap 101 to swing;

[0026] The blade shaft 106 is a round rod slightly longer than the extension, which passes through the blade 1 in the extension direction and extends from the upper and lower ends of the blade; the blade shaft 106 is connected to the upper and lower connecting rods 5 through bearings, and the blade shaft 106 can rotate freely; the blade shaft 106 is fixedly connected to the blade front part 107, and the two rotate synchronously; the lower end of the blade shaft 106 is a cylinder with a large diameter, and the blade 1 is axially fixed through the fixed connection between the two; a circular hole is opened in the center of the bottom end of the cylinder, and the blade slip ring 2 can be inserted. The blade shaft 106 is fixedly connected to the blade slip ring 2 by a set screw, so that the blade slip ring 2 can rotate synchronously with the blade shaft 105; a rolling bearing is installed at each end of the blade shaft 106, so that the blade shaft 106 can drive the blade 1 to rotate freely.

[0027] In another specific embodiment of the present invention, the flap 101 of the blade has a length of 0.25 chord lengths, is located at 0.75 chord lengths, and is made of high-strength resin; the flexible material at the connecting hinge is polyimide; the material of the flap connecting rod 102, the connecting rod 103 and the rudder arm 104 is a high-strength alloy; the chord-wise position of the blade shaft 106 is at 0.25 chord lengths from the leading edge.

[0028] A method for controlling the variable flaps of a vertical axis wind turbine using a deep reinforcement learning algorithm is also provided, as follows:

[0029] The speed and rotor phase information of the current vertical axis wind turbine are obtained through a laser velocimeter or encoder, and the rotor phase of the wind turbine is input into the intelligent agent as the state s; the instantaneous torque of the wind turbine is measured by a torque sensor, and the instantaneous energy consumed by the servo is calculated by the current and voltage of the servo. The net power of the wind turbine can be obtained by subtracting the instantaneous energy consumed by the servo from the wind turbine power generation; the net power of the wind turbine is input into the intelligent agent from the environment, and is used as a reward in deep reinforcement learning to optimize the control law, that is, through feedback on the reward, the intelligent agent gradually adjusts the strategy to increase the frequency of behaviors that receive high rewards, while reducing the occurrence of low-reward behaviors;

[0030] The agent has two modules: the actor and the critic. The actor module makes the corresponding action a according to the current state s, while the critic module calculates the value of the action a taken under the current state s. In the Markov decision process, the agent starts from state s and performs action a according to strategy π. The reward G t, thereby obtaining the action-state value function; in the actor module, there are two neural network models, the policy network and the target policy network; and in the critic module, there are two neural network models, the deep Q network and the target Q network. These four neural network models are built through neural networks; the state s and reward r are input from the environment to the policy network of the actor module, and the updated policy network and the target policy network are optimized by using the gradient ascent method to calculate the loss function; the soft update method is used to update the parameters in the policy network to the target policy network, that is, the updated parameters of the policy network and the original parameters of the target policy network are mixed in a certain proportion to form a new average value as the new parameter; at the same time, the state s and Action a is input into the critic module, and the value of taking action a in state s is calculated through the deep Q network, and the action corresponding to the expectation of making the reward value larger is selected; in order to update the deep Q network during the training process, the target Q network is used to calculate the loss function using the gradient descent method to minimize the loss function, so that the deep Q network approaches the target Q network; the intelligent agent uses the control law after initialization or the control law obtained by training to output the servo PWM signal in real time according to the different rotor phases of the wind turbine to control the rotation angle of the servo, where the rotor phase is the state s in the DDPG algorithm. After optimization by the DDPG algorithm, the best PWM signal is output at different rotor phases to obtain the best control law;

[0031] At the same time, the current state s, the action a taken for state s, and the reward r obtained are shuffled and stored in the experience replay unit as experience samples, and experience samples are randomly extracted from the experience replay unit and input into the intelligent agent for training; according to the expectation of the reward obtained by taking action a under different states s, the action that maximizes the total reward value is finally obtained, that is, the optimal control law, and it is output to the servo to continuously adjust the servo rotation angle so as to control the angle change of the flap in real time.

[0032] In the present invention, the adjustment of the blade pitch angle is not directly controlled by the motor, but indirectly regulated by the pitch torque generated by the deflection of the flap driven by the servo. Driven by methods such as deep reinforcement learning, the flap angle can be adjusted in a closed loop intelligently with the rotor phase and the incoming flow direction to achieve real-time optimization of the size and direction of the blade lift and drag. Compared with traditional methods, the control power is lower and the control strategy is more adaptable to changes in complex incoming flow conditions, so it can have higher power generation efficiency and self-starting performance.

[0033] The advantages of the present invention are as follows:

[0034] (1) High efficiency: The variable flap and pitch control laws of traditional vertical axis wind turbines are relatively simple, and it is difficult to achieve real-time control with the rotor phase. The present invention uses an intelligent algorithm to optimize the lift and drag of the airfoil and the direction of aerodynamic force in real time under different wind turbine rotor phases and incoming flow conditions, improve the self-starting performance of the vertical axis wind turbine, improve the power generation efficiency of the vertical axis wind turbine, and have good environmental adaptability.

[0035] (2) Strong self-starting and braking capabilities: Traditional variable pitch control methods consume a lot of energy and require a relatively long starting and braking time. The present invention uses a variable flap method to control the change of the pitch angle, which has a small load and low energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The overall structure of the vertical axis wind turbine is shown;

[0037] Figure 2 The specific structure diagram of the blade is shown;

[0038] Figure 3 The specific method of intelligent flap control is shown. DETAILED DESCRIPTION

[0039] The present invention will be described in detail below with reference to the accompanying drawings.

[0040] Figure 1 The overall structure of the vertical axis wind turbine of the present invention is shown. The device mainly comprises blades 1, blade slip rings 2, main body slip rings 3, motors 4, connecting rods 5, connecting plates 6, intermediate shafts 7, motor connecting shafts 8, and rolling bearings 9.

[0041] The intermediate shaft 7 is a vertical solid cylindrical long rod, which is connected to the rotating shaft of the motor 4 below it through a set screw. The intermediate shaft 7 can rotate along with the rotating shaft of the motor 4.

[0042] The connecting rod 5 is a solid long rod placed horizontally.

[0043] The connecting plate 6 is a flat cylinder with a small cylindrical boss on the center. The connecting plate 6 and the center of the boss are provided with threaded holes that penetrate from top to bottom. The intermediate shaft 7 passes through the holes from bottom to top. The connecting plate 6 and the intermediate shaft 7 can be fixed with set screws and rotate synchronously with the intermediate shaft 7. Three groups of threaded holes are arranged radially on the upper surface of the connecting plate 6 at a center angle of 120°, which are used to connect the connecting plate 6 and the connecting rod 5 by bolts (therefore, holes are also punched at the corresponding positions of the connecting rod 5 to allow the bolts to pass through). The connecting plate 6 drives the blade 1 connected to the end of the connecting rod 5 to rotate synchronously. The intermediate shaft 7 can rotate synchronously with the motor shaft, and the connecting plate 6 can rotate synchronously with the intermediate shaft 7. The intermediate shaft 7 can be made of high-strength metal material (preferably stainless steel).

[0044] Two upper and lower connecting plates 6 are arranged above and below the intermediate shaft, which are used to fix the inner ends of three connecting rods 5 arranged at intervals of 120 degrees on their upper surfaces, and the two are fixed by bolts, for example. The outer ends of the three connecting rods 5 correspond to each other in pairs, and the upper and lower connecting rods 5 are used to fix one blade 1. Six completely identical connecting rods 5 are arranged horizontally, and a total of three blades 1 are fixed. The spanwise position of the connection between the connecting rod 5 and the blade 1 is preferably at both ends of the blade 1, and the chordwise position is preferably at 25% of the chord length. When the connecting plate 6 rotates with the motor shaft, the connecting rod 5 can rotate with the connecting plate 6. The cross-section of the connecting rod 5 is designed to be elliptical, with a ratio of the major axis to the minor axis of 3-8, and the typical range of the major axis length is 10-30mm (preferably 20mm). The typical material of the connecting rod 5 is aluminum alloy.

[0045] The blade 1 adopts a NACA four-digit airfoil, and the typical chord length ranges from 0.1 to 10 m. The larger the chord length, the greater the airflow captured and the greater the power generation obtained. The aspect ratio ranges from 2 to 5, preferably 3. The typical material of the blade is high-performance plastic. The blade 1 is equipped with a movable flap, and its specific structure is described below. In a specific embodiment of the present invention, three blades 1 are used that are placed vertically and evenly spaced. If necessary, the number of connecting rods 5 can also be adjusted to place multiple blades 1. The blade 1 is clamped between the upper and lower connecting rods 5.

[0046] The blade slip ring 2 is located below the blade 1 and connected to the connecting rod 5. It is mainly used to prevent the servo wire from being entangled during the free rotation of the blade 1. The outer rotor part of the blade slip ring 2 can be fixed to the blade 1 by means of set screws or gluing, so that the two can rotate synchronously; at the same time, the outer rotor part of the blade slip ring 2 is connected to the connecting rod 5 through a rolling bearing 9 and can rotate freely on the connecting rod 5. One end of the wire of the stator part inside the blade slip ring 2 is connected to the stator part, and after passing through the stator part, it is tied along the connecting rod 5 with a rolling belt, and continues to extend to connect with the wire of the rotor part of the main slip ring 3. The wire of the stator part is tied by the rolling belt and rotates synchronously with the intermediate shaft 7, so it can be ensured that the wire of the stator part will not be entangled on the intermediate shaft 7.

[0047] The main body slip ring 3 is located near the lower end of the intermediate shaft 7, above the motor 4 and below the lower connection plate 6. However, it maintains a certain distance from the bottom end of the intermediate shaft 7, and this distance is used to arrange the wires at the lower end of the slip ring. The main body slip ring 3 is used to prevent the wires from getting entangled during the rotation of the wind turbine. The rotor part of the main body slip ring 3 is fixed to the intermediate shaft 7 by set screws and rotates with the intermediate shaft 7. The wires at the rotor can be fixed to the intermediate shaft 7 by rolling strips or the like, and connected to the wires of the blade slip ring 2; and the wires at the stator can be led out along the upper surface of the motor 4 and connected to an external power supply.

[0048] The motor connecting shaft 8 is used to connect the intermediate shaft 7 and the main shaft of the motor 4. The intermediate shaft 7 and the motor connecting shaft 8, and the main shaft of the motor 4 and the motor connecting shaft 8 are connected by set screws respectively so that they can rotate together. The typical material of the motor connecting shaft 8 is a high-strength alloy.

[0049] The motor 4 is provided with a Hall sensor and a driving circuit board, which can be used to measure the rotation speed and power supply current of the vertical axis wind turbine respectively. The torque can be obtained by calculating the product of the power supply current and the motor torque coefficient, and then the torque and the rotation speed are multiplied to obtain the power, thereby realizing the measurement of the average power output of the wind turbine. The instantaneous power of the wind turbine can be measured by a torque sensor.

[0050] The blades 1, blade slip rings 2, main body slip rings 3, motors 4, connecting rods 5, connecting plates 6, intermediate shafts 7, motor connecting shafts 8, and rolling bearings 9 are all components well known to those skilled in the art.

[0051] Figure 2 (a) The upper figure shows a bottom view of the blade 1, and the lower figure shows a top view of the blade 1. The blade 1 mainly includes a flap 101, a flap connecting rod 102, a connecting rod 103, a rudder arm 104, a steering gear 105, a blade shaft 106, and a blade front 107. The flap 101, the flap connecting rod 102, the connecting rod 103, the rudder arm 104 and the steering gear 105 form a parallelogram connecting rod mechanism. The steering gear 105 is the active rocker. When it rotates a certain angle under electric drive, it can drive the slave rocker composed of the flap 101 and the flap connecting rod 102 to rotate a certain angle.

[0052] The typical length of flap 101 is 0.25 chord length, the typical position is at 0.75 chord length, and the typical material is high-strength resin. In order to allow flap 101 to swing around the hinge at 0.75 chord length, flap 101 is hinged to blade front 107 through flexible material. Typical flexible materials for connecting hinges are polyimide and the like. Flap connecting rod 102 is T-shaped as a whole, and is fixed to flap 101 through two threaded holes on the "one" part of the T-shaped, forming a driven rocker in the linkage mechanism. The other end is hinged to connecting rod 103 through a through hole. Connecting rod 103 is oblong, and has a hole on the left and right, respectively, which is connected to rudder arm 104 and flap connecting rod 102 by bolts. The shape of the rudder arm 104 is also approximately oblong (the rudder arm 104 is shown in the figure as thick at one end and thin at the other end, but it is not a necessary condition, as long as the rudder arm 104 can complete its function), one end is connected to the connecting rod 103 by bolts, and the other end is connected to the steering gear shaft of the steering gear 105. The typical materials of the flap connecting rod 102, the connecting rod 103 and the rudder arm 104 are high-strength alloys, preferably stainless steel. The steering gear 105 is fixed to the front position of the blade front part 107 by bolts, and is responsible for controlling the rotation angle of the flap. The power line and signal line of the steering gear 105 are all connected to the electric wires of the stator part of the blade slip ring 2. The rotation of the steering gear shaft drives the rudder arm 104 and the flap connecting rod 102 to move, so that the flap 101 can swing around the 0.75 chord length, generating lift changes, so that the deflection angle of the blade 1 relative to the wind turbine changes.

[0053] like Figure 2 As shown in (b), the blade shaft 106 is a round rod slightly longer than the span, which passes through the blade 1 in the span direction and extends from the upper and lower ends of the blade. The blade shaft 106 is connected to the upper and lower connecting rods 5 by bearings, and the blade shaft 106 can rotate freely. The typical position of the blade shaft 106 in the chord direction is 0.25 chord length away from the leading edge. The blade shaft 106 is fixedly connected to the front part 107 of the blade, and the two can rotate synchronously. The lower end of the blade shaft 106 is a cylinder with a larger diameter, and the blade 1 can be axially fixed by the fixed connection between the two. A circular hole is opened in the center of the bottom end of the cylinder, and the blade slip ring 2 can be inserted. The blade shaft 106 is fixedly connected to the blade slip ring 2 by a set screw, so that the blade slip ring 2 can rotate synchronously with the blade shaft 105. A rolling bearing is installed at each end of the blade shaft 106, so that the blade shaft 106 can drive the blade 1 to rotate freely.

[0054] Under different rotor phases and incoming flow conditions, the swing angle of the flaps can be controlled by controlling the rotation angle of the servo 105, thereby controlling the pitch angle of the wind turbine, optimizing the lift and drag of the airfoil and the direction of aerodynamic force in real time, so that the blades can reach the optimal angle of attack, improve the self-starting performance of the vertical axis wind turbine, and improve the power generation efficiency of the vertical axis wind turbine.

[0055] Figure 3The specific method of using deep reinforcement learning algorithm to control the variable flap of vertical axis wind turbine is shown. In order to solve the continuous control problem, the deep deterministic policy gradient algorithm is adopted. Among them, the control law is expressed as a pulse width modulation (PWM) control signal output to the servo under a certain rotor phase.

[0056] In the deep reinforcement learning algorithm, an environment needs to be set up, in which various required state quantities are included, such as wind turbine rotor phase, wind turbine net power, etc. The speed and rotor phase information of the current vertical axis wind turbine are obtained by a laser velocimeter or encoder, and the wind turbine rotor phase is input into the intelligent agent as the state s (Mnih V, Kavukcuoglu K, Silver D, et al. Playing Atari with Deep Reinforcement Learning [J]. Computer Science, 2013. DOI: 10.48550 / arXiv.1312.5602.). The instantaneous torque of the wind turbine is measured by a torque sensor, and the instantaneous energy consumed by the steering gear is calculated by the current and voltage of the steering gear (this calculation method is well known to those skilled in the art). The net power of the wind turbine can be obtained by subtracting the instantaneous energy consumed by the steering gear from the wind turbine power generation. The net power of the wind turbine is input into the agent from the environment and used as a reward to optimize the control law in deep reinforcement learning. That is, through feedback on the reward, the agent can gradually adjust its strategy to increase the frequency of behaviors that receive high rewards while reducing the occurrence of low-reward behaviors.

[0057] The agent has two modules: the actor and the critic. The actor module makes the corresponding action a according to the current state s, for example, outputs the PWM signal of the servo according to the phase of the wind turbine rotor, while the critic module calculates the value of the action a taken under the current state s, and calculates the reward G obtained by the agent starting from state s and executing action a according to strategy π in the Markov decision process. t, thereby obtaining the action-state value function. Among them, in the actor module, there are two neural network models, the policy network and the target policy network; and in the critic module, there are two neural network models, the deep Q network and the target Q network. These four neural network models are built by neural networks, for example, they can be built using a fully connected neural network structure. Input the state s and reward r from the environment to the policy network of the actor module, and use the gradient ascent method (which is well known to those skilled in the art) to calculate the loss function for the updated policy network and the target policy network for optimization. The parameters in the policy network are updated to the target policy network using the soft update method, that is, the updated parameters of the policy network and the original parameters of the target policy network are mixed into a new average value as the new parameter in a certain ratio (for example, 0.005:0.095). At the same time, the state s and action a are input into the critic module, and the value obtained by taking action a in the case of state s is calculated by the deep Q network (which is well known to those skilled in the art), and the action corresponding to the expectation of making the reward value larger is selected. To update the deep Q network during the training process, as shown below, the target Q network can be used to calculate the loss function using a gradient descent method (which is well known to those skilled in the art) to minimize the loss function, thereby making the deep Q network approach the target Q network. The agent uses the control law after initialization or the control law obtained by training to output the servo PWM signal in real time according to the different rotor phases of the wind turbine to control the rotation angle of the servo, where the rotor phase is the state s in the DDPG algorithm. After optimization by the DDPG algorithm, the best PWM signal is output at different rotor phases to obtain the best control law.

[0058] At the same time, the current state s, the action a taken for state s, and the reward r are shuffled and stored in the experience replay unit as experience samples, and experience samples are randomly extracted from the experience replay unit and input into the intelligent agent for training. The purpose of shuffling the order of experience samples and random sampling is to eliminate the association between experience samples in order to solve the problem that the neural network in the intelligent agent is difficult to successfully fit. According to the expectation of the reward obtained by taking action a under different states s, the action that can maximize the total reward value, that is, the optimal control law, is finally obtained and output to the servo to continuously adjust the rotation angle of the servo so as to control the angle change of the flap in real time.

[0059] The specific steps of the algorithm are as follows:

[0060] Step 1: Initialize the system parameters, including state s, action a, reward r, control law and other parameters.

[0061] Step 2: Interact with the environment, obtain the wind turbine rotor phase state s and wind turbine net power reward r from the environment, package the state, action and reward recorded by the agent and the environment in each state into experience samples, and record them in the experience replay unit in a random order. If the experience replay unit is full, randomly extract experience samples to update the network parameters in the agent.

[0062] Step 3: Randomly extract experience samples from the experience replay unit and input them into the agent, where the critic module calculates the value of taking action a in state s, which is the reward G obtained by the agent in the Markov decision process when it starts from state s and performs action a according to strategy π. t The expectation uses the target Q network to calculate the loss function using the gradient descent method (which is well known to those skilled in the art) to minimize the loss function and make the deep Q network approach the target Q network. This is the method for updating the critic network. The actor module uses the gradient ascent method (which is well known to those skilled in the art) to update the policy network and select the action corresponding to the expectation that makes the reward value larger, that is, output the PWM signal to the servo.

[0063] Step 4: Enter a new state and repeat steps 2 and 3. If the net power change of the wind turbine converges or the deep reinforcement learning reaches the set number of steps, the learning ends and the obtained PWM signal action a is the optimal control law.

[0064] To sum up, the core of the present invention is to utilize the angle change control of the flaps to realize the pitch angle control of the vertical axis wind turbine, and adopt an intelligent algorithm to autonomously optimize the flap angle changes under different rotor phases, so as to realize real-time optimization of the lift and drag of the airfoil and the aerodynamic direction, so as to achieve the purpose of improving the self-starting performance of the vertical axis wind turbine and improving the power generation efficiency of the vertical axis wind turbine.

Claims

1. A high-efficiency vertical axis wind turbine with intelligent variable flaps and pitch, characterized in that: The device comprises a blade (1), a blade slip ring (2), a main body slip ring (3), a motor (4), a connecting rod (5), a connecting plate (6), an intermediate shaft (7), a motor connecting shaft (8), and a rolling bearing (9); wherein The intermediate shaft (7) is a vertical solid cylindrical long rod, which is connected to the rotating shaft of the motor (4) below it through a set screw, and the intermediate shaft (7) can rotate along with the rotating shaft of the motor (4); The connecting rod (5) is a solid long rod placed horizontally; The connecting disk (6) is a flat cylinder with a small cylindrical boss at the center. The connecting disk (6) and the center of the boss are provided with a threaded hole that passes through from top to bottom. The intermediate shaft (7) passes through the threaded hole from bottom to top. The connecting disk (6) and the intermediate shaft (7) are fixed by a fastening device and rotate synchronously with the intermediate shaft (7). The upper surface of the connecting disk (6) is provided with fixing devices arranged radially at a center angle of 120°, which are used to fix the connecting disk (6) and the three connecting rods (5). The connecting disk (6) drives the blades (1) connected to the ends of the connecting rods (5) to rotate synchronously. The intermediate shaft (7) rotates synchronously with the motor shaft, and the connecting disk (6) rotates synchronously with the intermediate shaft (7). Two upper and lower connecting plates (6) are arranged above and below the intermediate shaft, and are used to fix the inner ends of three connecting rods (5) arranged at intervals of 120 degrees on their upper surfaces; the outer ends of the three connecting rods (5) correspond to each other in pairs, and the upper and lower connecting rods (5) are used to fix one blade (1); six identical connecting rods (5) are arranged horizontally, and a total of three blades (1) are fixed; when the connecting plate (6) rotates with the motor shaft, the connecting rods (5) rotate with the connecting plate (6); The blade (1) comprises three blades (1) arranged vertically and evenly spaced apart, and the blade (1) is sandwiched between two upper and lower connecting rods (5); The blade slip ring (2) is located below the blade (1) and connected to the connecting rod (5). The outer rotor part of the blade slip ring (2) is fixed to the blade (1) so that the two can rotate synchronously. The outer rotor part of the blade slip ring (2) is connected to the connecting rod (5) through a rolling bearing (9) and can rotate freely on the connecting rod (5). One end of the wire of the stator part inside the blade slip ring (2) is connected to the stator part, passes through the stator part, and then continues to extend along the connecting rod (5) to connect with the wire of the rotor part of the main body slip ring (3). The main body slip ring (3) is located near the lower end of the intermediate shaft (7), above the motor (4) and below the lower connection plate (6); but it maintains a certain distance from the bottom end of the intermediate shaft (7), and the distance is used to arrange the wires at the lower end of the slip ring; the rotor part of the main body slip ring (3) is fixed on the intermediate shaft (7) and rotates with the intermediate shaft (7), the wires at the rotor are fixed on the intermediate shaft (7) and connected to the wires of the blade slip ring (2); the wires at the stator are led out along the upper surface of the motor (4) and connected to an external power supply; The motor connecting shaft (8) is used to connect the intermediate shaft (7) and the main shaft of the motor (4). The intermediate shaft (7) and the motor connecting shaft (8), and the main shaft of the motor (4) and the motor connecting shaft (8) are connected respectively by fastening devices so that they can rotate together. The motor (4) is provided with a Hall sensor and a driving circuit board.

2. The high-efficiency vertical axis wind turbine with intelligent variable flaps and pitch according to claim 1, characterized in that: The upper surface of the connecting plate (6) is provided with a plurality of groups of threaded holes arranged radially at a central angle of 120°, which are used to connect the connecting plate (6) and the connecting rod (5) by bolts; holes are also punched at corresponding positions of the connecting rod (5) to allow the bolts to pass through.

3. The high-efficiency vertical axis wind turbine with intelligent variable flaps and pitch according to claim 1, characterized in that: The cross section of the connecting rod (5) is elliptical, the ratio of the major axis to the minor axis is 3-8, and the length of the major axis ranges from 10 to 30 mm; the spanwise positions where the connecting rod (5) is connected to the blade (1) are at the two ends of the blade (1).

4. The high-efficiency vertical axis wind turbine with intelligent variable flaps and pitch as claimed in claim 3, characterized in that: The long axis length of the connecting rod (5) is 20 mm, and the material is aluminum alloy; the chord-wise position where the connecting rod (5) is connected to the blade (1) is at 25% of the chord length.

5. The high-efficiency vertical axis wind turbine with intelligent variable flaps and pitch according to claim 1, characterized in that: The blade (1) adopts a NACA four-digit airfoil, with a typical chord length range of 0.1-10m and an aspect ratio range of 2-5; and a movable flap is installed on the blade (1).

6. The high-efficiency vertical axis wind turbine with intelligent variable flaps and pitch according to claim 5, characterized in that: The blade (1) has an aspect ratio of 3 and is made of high-performance plastic.

7. The high-efficiency vertical axis wind turbine with intelligent variable flaps and pitch according to claim 1, characterized in that: One end of the electric wire of the stator part inside the blade slip ring (2) is connected to the stator part, and after passing through the stator part, it is tied with a rolling belt along the connecting rod (5) and continues to extend to connect with the electric wire of the rotor part of the main body slip ring (3).

8. A blade, which is suitable for a high-efficiency vertical axis wind turbine with intelligent variable flaps and pitch as claimed in any one of claims 1 to 8, characterized in that: The blade (1) comprises a flap (101), a flap connecting rod (102), a connecting rod (103), a rudder arm (104), a steering gear (105), a blade shaft (106), and a blade front portion (107); wherein the flap (101), the flap connecting rod (102), the connecting rod (103), the rudder arm (104), and the steering gear (105) form a parallelogram connecting rod mechanism; the steering gear (105) is an active rocker, which rotates a certain angle under electric drive, and can drive a slave rocker composed of the flap (101) and the flap connecting rod (102) to rotate a certain angle; The flap (101) is hingedly connected to the front part of the blade (107) through a flexible material; The flap connecting rod (102) is in a T-shape as a whole, and is fixed to the flap (101) through two threaded holes on the "one" part of the T-shape to form a follower rocker in the connecting rod mechanism; and the other end is hingedly connected to the connecting rod (103) through a through hole; The connecting rod (103) is in an oblong shape, and has a hole on the left and right sides respectively connected to the rudder arm (104) and the flap connecting rod (102) by bolts; The rudder arm (104) is in the shape of an oblong, one end of which is connected to the connecting rod (103) by a bolt, and the other end of which is connected to the steering gear shaft of the steering gear (105); The steering gear (105) is fixed to the front middle position of the blade front part (107) by bolt connection; the power line and signal line of the steering gear (105) are connected to the electric wires of the stator part of the blade slip ring (2); The steering gear shaft rotates to drive the rudder arm (104) and the flap connecting rod (102) to move, thereby causing the flap (101) to swing; The blade shaft (106) is a round rod slightly longer than the extension, which penetrates the blade (1) in the extension direction and extends from the upper and lower ends of the blade; the blade shaft (106) is connected to the upper and lower connecting rods (5) through bearings, and the blade shaft (106) can rotate freely; the blade shaft (106) is fixedly connected to the front part (107) of the blade, and the two rotate synchronously; the lower end of the blade shaft (106) is a cylinder with a large diameter, and the blade (1) is axially fixed by the fixed connection between the two; a circular hole is opened at the center of the bottom end of the cylinder, and the blade slip ring (2) can be inserted. The blade shaft (106) is fixedly connected to the blade slip ring (2) by a set screw, so that the blade slip ring (2) can rotate synchronously with the blade shaft 105; a rolling bearing is installed at each end of the blade shaft (106), so that the blade shaft (106) can drive the blade (1) to rotate freely.

9. A blade as described in claim 8, wherein the flap (101) is 0.25 chord length long, located at 0.75 chord length, and is made of high-strength resin; the flexible material at the connecting hinge is polyimide; the flap connecting rod (102), connecting rod (103) and rudder arm (104) are made of high-strength alloy; the chord-wise position of the blade shaft (106) is 0.25 chord length away from the leading edge.

10. A method for controlling the variable flaps of a vertical axis wind turbine using a deep reinforcement learning algorithm, characterized in that: The details are as follows: The speed and rotor phase information of the current vertical axis wind turbine are obtained through a laser velocimeter or encoder, and the rotor phase of the wind turbine is input into the intelligent agent as the state s; the instantaneous torque of the wind turbine is measured by a torque sensor, and the instantaneous energy consumed by the servo is calculated by the current and voltage of the servo. The net power of the wind turbine can be obtained by subtracting the instantaneous energy consumed by the servo from the wind turbine power generation; the net power of the wind turbine is input into the intelligent agent from the environment, and is used as a reward in deep reinforcement learning to optimize the control law, that is, through feedback on the reward, the intelligent agent gradually adjusts the strategy to increase the frequency of behaviors that receive high rewards, while reducing the occurrence of low-reward behaviors; The agent has two modules: the actor and the critic. The actor module makes the corresponding action a according to the current state s, while the critic module calculates the value of the action a taken under the current state s. In the Markov decision process, the agent starts from state s and performs action a according to strategy π. The reward G t , thereby obtaining the action-state value function; in the actor module, there are two neural network models, the policy network and the target policy network; and in the critic module, there are two neural network models, the deep Q network and the target Q network. These four neural network models are built through neural networks; the state s and reward r are input from the environment to the policy network of the actor module, and the updated policy network and the target policy network are optimized by using the gradient ascent method to calculate the loss function; the soft update method is used to update the parameters in the policy network to the target policy network, that is, the updated parameters of the policy network and the original parameters of the target policy network are mixed in a certain proportion to form a new average value as the new parameter; at the same time, the state s and Action a is input into the critic module, and the value of taking action a in state s is calculated through the deep Q network, and the action corresponding to the expectation of making the reward value larger is selected; in order to update the deep Q network during the training process, the target Q network is used to calculate the loss function using the gradient descent method to minimize the loss function, so that the deep Q network approaches the target Q network; the intelligent agent uses the control law after initialization or the control law obtained by training to output the servo PWM signal in real time according to the different rotor phases of the wind turbine to control the rotation angle of the servo, where the rotor phase is the state s in the DDPG algorithm. After optimization by the DDPG algorithm, the best PWM signal is output at different rotor phases to obtain the best control law; At the same time, the current state s, the action a taken for state s, and the reward r obtained are shuffled and stored in the experience replay unit as experience samples, and experience samples are randomly extracted from the experience replay unit and input into the intelligent agent for training; according to the expectation of the reward obtained by taking action a under different states s, the action that maximizes the total reward value is finally obtained, that is, the optimal control law, and it is output to the servo to continuously adjust the servo rotation angle so as to control the angle change of the flap in real time.

Images