A water flow energy collection device and method

Abstract

The present application provides a water flow energy collection device and method. The device provided by the present application includes a sensing component, a motion component, and a control component; the motion component is in the water flow energy collection area, and the sensing component is arranged around the motion component according to a preset position; the control component includes a pre-trained control model; a jet injection device penetrates the motion device; the motion device is a hollow cylindrical motion tube; the cylindrical motion tube is connected to the guide rail device through a connector and a connecting rod; the jet injection device includes a water pump, a water pipe, an internal water flow channel, and a plurality of jet ports; the angles between the directions of the jet ports and the water flow direction are 90°, 180°, and 270° respectively; the internal water flow channel connects each jet port; the water pump is connected to the internal water flow channel through the water pipe; the control model controls the mass flow rate of the water ejected at the jet port. The device and method provided by the present application improve the collection efficiency of water flow energy.

Description

Technical Field

[0001] This application relates to the technical field of water flow energy collection, and particularly relates to a water flow energy collection device and method. Background Art

[0002] Since the 20th century, the demand for electricity by humans has almost doubled every decade. Currently, the main forms of power generation are hydroelectric power generation, thermal power generation, and nuclear power generation. The status of the three main forms varies due to the different energy compositions of each country, but thermal power generation is the main form in the world, accounting for more than 60%-70% of the total power generation. With the continuous consumption of fossil energy and the problem of waste disposal in nuclear power generation, hydroelectric power generation has become increasingly important.

[0003] Currently, the method of hydroelectric power generation is to build different types of hydropower stations, and the water energy utilized is mainly the potential energy stored in the water body, converting the potential energy of water into electrical energy. A water flow energy collection device is a device widely used in hydroelectric power generation. The water flow energy collection device collects the energy brought by the impact of water flow on the moving device. The water flow causes the Karman vortex street effect on the moving device, making the moving device perform vertical movement on the same horizontal plane perpendicular to the water flow. The transmission mechanism transmits this vertical movement to the generator to achieve power generation.

[0004] Since the moving device in the existing water flow energy collection device only moves with the vortex-induced resonance effect brought by the water flow and lacks active control of the water flow smoothness, the energy collection efficiency is relatively low. Summary of the Invention

[0005] This application provides a water flow energy collection device and method, which can be used to solve the technical problem of low water flow energy collection efficiency in the existing technology.

[0006] In a first aspect, an embodiment of this application provides a water flow energy collection device, and the device includes:

[0007] A sensing component, a moving component, and a control component;

[0008] The moving component is in the water flow energy collection area, and the sensing component is arranged around the moving component according to a preset position; the control component is electrically connected to the moving component and the sensing component;

[0009] The control component includes a pre-trained control model;

[0010] The moving component includes a jet injection device, a moving device, a guide rail device, a transmission device, a power generation device, and a support device;

[0011] The jet injection device penetrates through the motion device; the motion device is vertically arranged in the water flow energy collection area, in the oncoming flow direction of the water flow;

[0012] The motion device is a hollow cylindrical motion tube;

[0013] The cylindrical motion tube is connected to the guide rail device through a connector and a connecting rod; the connector includes an upper connector and a lower connector;

[0014] The guide rail device is installed at the upper and lower ends of the motion device; the transmission device is installed on the guide rail device, one end of the transmission device is connected to the power generation device, and the other end of the transmission device is connected to the motion device;

[0015] The motion device and the power generation device are arranged at the support device;

[0016] The jet injection device includes a water pump, a water pipe, an internal water flow channel, and a plurality of jet ports; the angles between the directions of the jet ports and the water flow direction are 90°, 180°, and 270° respectively, and the jet ports are all arranged at the edge of the motion device; the internal water flow channel connects each jet port and penetrates through the motion device; the water pump is connected to the internal water flow channel through the water pipe;

[0017] The control model controls the mass flow rate of the water flow ejected from the jet port.

[0018] Combined with the first aspect, in an implementable manner of the first aspect, the sensing component includes a plurality of sensors. Among them, the sensors in the first part are evenly arranged around the cylindrical motion tube at intervals of every 10° of radian and form two sensor rings; the sensors in the second part are densely arranged in the direction perpendicular to the oncoming flow of the cylindrical motion tube; the sensors in the third part are arranged in a square area far from the oncoming flow inlet.

[0019] Combined with the first aspect, in an implementable manner of the first aspect, the guide rail device includes a guide rail and a spring; one end of the spring is connected to the motion device, and the other end is connected to the connecting rod; a chute is provided in the middle of the guide rail; the motion device is restricted to move in the chute.

[0020] Combined with the first aspect, in an implementable manner of the first aspect, the transmission device includes a gear and a rack; the rack is connected to the lower connector through a thin connecting rod; the rack meshes with the gear.

[0021] Combined with the first aspect, in an implementable manner of the first aspect, the support device includes a guide rail base, a generator base, and a water pump base.

[0022] Second aspect, an embodiment of the present application provides a method for collecting water flow energy. The method is applied to the water flow energy collection device described above. The device includes a sensing component, a motion component, and a control component; the motion component includes a jet injection device, a motion device, a guide rail device, a transmission device, a power generation device, and a support device; the method includes:

[0023] Place the motion component in the water flow energy collection area in a predetermined direction;

[0024] The sensing component collects the water flow velocity at each sensor and sends the water flow velocity to the control component;

[0025] The control component inputs the water flow velocity into a pre-trained control model to obtain the target mass flow rate and issues a control instruction; the control instruction is used to instruct the jet injection device to eject the target mass flow rate.

[0026] In combination with the second aspect, in a feasible implementation manner of the second aspect, the control model is a neural network model, and the parameters of the neural network model are optimized through deep reinforcement learning.

[0027] In combination with the second aspect, in a feasible implementation manner of the second aspect, the control model includes 4 layers;

[0028] Among them, the first layer is the input layer, and the input layer corresponds to the water flow velocities of 151 sensors;

[0029] The second layer is the first hidden layer, and the number of nodes is 512;

[0030] The third layer is the second hidden layer, and the number of nodes is 512;

[0031] The fourth layer is the output layer, and the number of nodes is 3, which respectively correspond to the target mass flow rates of the water flows at three jet ports on the same cross-section.

[0032] In combination with the second aspect, in a feasible implementation manner of the second aspect, during the training process of the control model, the learning rate is set to 0.001, and the maximum number of iterations is 2000.

[0033] In combination with the second aspect, in a feasible implementation manner of the second aspect, the deep reinforcement learning adopts a local policy optimization algorithm. The jet ports eject 80 times in each iteration. The algorithm optimizer adopts an adam optimizer, and the parameter frequency in the neural network model is set to be updated once every 20 times; the reward function of the reinforcement learning is:

[0034] r = |x|

[0035] |·| refers to the absolute value, and x is the displacement value of the cylindrical motion tube

[0036] The cost function for updating the neural network by reinforcement learning is as follows:

[0037]

[0038] θ represents the parameters of the actual control policy, and θ′ represents the parameters of the control policy for training;

[0039] E (st，at) is the expected reward value of the entire control process;

[0040] P θ (a t |s t ) refers to the probability that the neural network model executes action a t under the state s t ;

[0041] P θ′ (a t |s t ) refers to the probability that the neural network model executes action a t under the state s t ;

[0042] A θ′ is a parameter for evaluating the pros and cons of the θ′ control policy.

[0043] The device provided by the embodiment of the present application is provided with three groups of jet ports in three directions of the moving device, as well as corresponding internal water flow pipes and a Laval-shaped jet port. The internal water flow pipes provide the jet source for the jet ports, and the water flow ejected from the jet ports will change the force of the external water flow acting on the moving device, accelerating the movement of the moving device, thereby improving the energy collection rate of the entire energy collection device. The method provided by the embodiment of the present application makes minor modifications to the device and uses the neural network model to control the mass flow rate of the water flow at the jet ports, so as to make the energy provided to the generator more significant. Description of the Drawings

[0044] Figure 1 is a schematic structural diagram of a water flow energy collection device provided by the embodiment of the present application;

[0045] Figure 2 is a front view structural diagram of a moving component provided by the embodiment of the present application;

[0046] Figure 3 is a rear view structural diagram of a moving component provided by the embodiment of the present application;

[0047] Figure 4 is a sectional view taken along the A-A direction of a jet injection device provided by the embodiment of the present application;

[0048] Figure 5 A cross-sectional view of a jet injection device B-B provided by an embodiment of the present application;

[0049] Figure 6 A sectional view taken along line A-A of a connector provided by an embodiment of the present application;

[0050] Figure 7 A sectional view taken along line B-B of a connector provided by an embodiment of the present application;

[0051] Figure 8 A schematic flow chart of a method for collecting water flow energy provided by an embodiment of the present application;

[0052] Figure 9 A schematic diagram of a neural network structure provided by an embodiment of the present application;

[0053] Figure 10 A schematic diagram showing the degree of convergence of the learning process of a neural network provided by an embodiment of the present application. Detailed implementation manners

[0054] To make the objectives, technical solutions, and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the accompanying drawings.

[0055] First, the following will be combined with Figure 1 to introduce the possible system architectures applicable to the embodiments of the present application.

[0056] Please refer to Figure 1 , which exemplarily shows a schematic structural diagram of a water flow energy collection device provided by an embodiment of the present application. As Figure 1 shown, the device provided by the embodiment of the present application includes the following parts: a sensing component, a motion component, and a control component. Figure 1 In, the solid ring is the motion component in the embodiment of the present application. In the present application, the control component is a computer capable of performing control functions. The control component can be arranged near the motion component or can be remotely controlled. Figure 1 All the points in are the sensing components.

[0057] In the embodiment of the present application, the motion component is in the water flow energy collection area, and the sensing components are arranged around the motion component according to preset positions. The control component is electrically connected to the motion component and the sensing components.

[0058] Among them, the control component includes a pre-trained control model.

[0059] As Figure 2 shown, it is a front view structural diagram of a motion device provided by an embodiment of the present application.

[0060] AsFigure 3 As shown, it is a rear view structural schematic diagram of a motion device provided by an embodiment of the present application.

[0061] Next, in combination with Figure 2 and Figure 3 the motion device provided by the embodiment of the present application will be elaborated.

[0062] In the embodiment of the present application, the motion component includes a jet injection device, a motion device, a guide rail device, a transmission device, a power generation device, and a support device;

[0063] The jet injection device penetrates the motion device. The motion device is vertically arranged in the water flow energy collection area, in the oncoming flow direction of the water flow.

[0064] The motion device is a hollow cylindrical motion tube 3. As Figure 3 shown, it is an internal sectional view of a motion device provided by an embodiment of the present application.

[0065] The cylindrical motion tube 3 is connected to the guide rail device through a connector 6 and a connecting rod 7.

[0066] The connector 6 includes an upper connector and a lower connector. As Figure 6 shown, it is an A-A sectional view of a connector provided by an embodiment of the present application. As Figure 7 shown, it is a B-B sectional view of a connector provided by an embodiment of the present application.

[0067] The guide rail device is installed at the upper and lower ends of the motion device. The transmission device is installed on the guide rail device. One end of the transmission device is connected to the power generation device, and the other end of the transmission device is connected to the motion device.

[0068] The motion device and the power generation device are arranged at the support device.

[0069] As Figure 4 shown, it is an A-A sectional view of a jet injection device provided by an embodiment of the present application. Figure 5 It is a B-B sectional view of a jet injection device provided by an embodiment of the present application.

[0070] The jet injection device includes a water pump 1, a water pipe 2, an internal water flow channel 4, and a plurality of jet ports 5. The angles between the directions of the jet ports 5 and the water flow direction are 90°, 180°, and 270° respectively. The jet ports 5 are all arranged at the edge of the motion device. The internal water flow channel 4 connects each jet port 5 and penetrates the motion device. The water pump 1 is connected to the internal water flow channel 4 through the water pipe 2. The water pump 1 provides a certain mass flow rate of water, which is transmitted through the water pipe 2 to the internal water flow channel 4 and ejected through the acceleration motion of the jet ports 5.

[0071] In a feasible implementation of the embodiments of the present application, at each of the angles of 90°, 180°, and 270° with respect to the direction of water flow, four jet ports 5 are provided, and the shape of each jet port 5 is a Laval shape.

[0072] It should be noted that in the prior art, the moving device is a hollow or solid cylinder, and no internal water flow channel 4 and multiple jet ports 5 are provided inside the moving device. That is, the moving device in the prior art can only move passively with the water flow, so the collected energy is limited.

[0073] In the embodiments of the present application, the guide rail device includes a guide rail 13 and a spring 8. One end of the spring 8 is connected to the moving device, and the other end is connected to a connecting rod 12. A chute is provided in the middle of the guide rail 13. The moving device is restricted to move in the chute.

[0074] The transmission device includes a gear 11 and a rack 10. The rack 10 is connected to the lower end connector through a thin connecting rod 9. The rack 10 meshes with the gear 11. The gear 11 drives the generator 19 to generate electricity through the transmission of a gear shaft 18 and a key 17.

[0075] The support device includes a guide rail base 14, a generator base 15, and a water pump base 16. Among them, the guide rail base 14 supports the cylindrical moving pipe 3, the generator base 15 supports the generator 19, and the water pump base 16 supports the water pump 1.

[0076] The generator device includes a generator 19, a gear shaft 18, and a key 17. The gear shaft 18 transmits the rotational motion of the gear 11 to the generator 19 through the key 17 for power generation.

[0077] In the embodiments of the present application, both the spring 8 and the engagement of the gear 11 and the rack 10 are coupled to the moving device through a connector 6. The back-and-forth movement of the moving device will continuously compress and stretch the spring to ensure the back-and-forth movement of the moving device, thereby achieving the purpose of cyclic energy collection. At the same time, the back-and-forth movement of the moving device will also drive the connecting rod 6 and the rack 10 to move, and the movement is transmitted to the gear through the engagement of the rack 10 and the gear 11. A slider is installed at one end of the said connector to ensure the movement of the connector on the guide rail. The said guide rail device includes two guide rail rods with slider grooves in the middle, which are used to connect the upper and lower connecting rods 12 on the one hand, and provide space for the movement of the connector 12 and the rack 10 on the other hand.

[0078] In the implementation of the present application, the sensing component includes multiple sensors. Among them, the sensors in the first part are evenly arranged around the cylindrical moving pipe 3 at intervals of every 10° of radian, and two sensor rings are formed. The sensors in the second part are densely arranged in the direction perpendicular to the oncoming flow of the cylindrical moving pipe 3. The sensors in the third part are arranged in a square area far from the oncoming flow inlet.

[0079] The working process of the device provided by the embodiment of the present application is as follows: The sensing component collects the water flow velocity at a preset position of the moving component and transmits the water flow velocity to the control component. The control component uses the control model to control the target mass flow rate ejected from the jet orifice according to the water flow velocity.

[0080] Water flows from one side to impact the moving device. Due to the vortex-induced resonance effect, the moving device will move back and forth in the direction parallel to the central axis of the spring. And the jet ejection device will eject a certain mass flow rate of water through the water pump 1, water pipe 2, internal water flow channel 4 and jet orifice 5. The spring 8 is coupled with the moving device through the connector 6 to ensure the stability of the back-and-forth movement of the moving device. The back-and-forth movement of the moving device and the connector 6 is transmitted to the gear 11 meshing with the rack 10 through the rack 10. The rotational movement of the gear 11 is transmitted through the gear shaft 18 and the key 17, and the movement is transmitted to the generator 19 to generate electricity.

[0081] The control model controls the mass flow rate corresponding to the water flow ejected at the jet orifice 5.

[0082] In the embodiment of the present application, three groups of a total of 12 jet orifices 5 are provided in three directions where the included angle between the moving device and the water flow direction is 90°, 180°, and 270° respectively. The internal water flow channel 4 provides the jet source for the jet orifice 5. The water flow ejected from the jet orifice 5 changes the force exerted by the external water flow on the moving device, accelerates the movement of the moving device, and the moving device provides more energy to the generator, thereby improving the energy collection rate of the entire energy collection device.

[0083] The device provided by the embodiment of the present application is provided with three groups of jet orifices in three directions of the moving device, as well as corresponding internal water pipes and a Laval-shaped jet orifice. The internal water pipes provide the jet source for the jet orifices. The water flow ejected from the jet orifices will change the force exerted by the external water flow on the moving device, accelerate the movement of the moving device, thereby improving the energy collection rate of the entire energy collection device. The method provided by the embodiment of the present application makes minor modifications to the device and uses a neural network model to control the water mass flow rate of the jet orifices, so that the energy provided to the generator is more significant.

[0084] The embodiment of the present application also provides a method for collecting water flow energy. The method for collecting water flow energy provided by the embodiment of the present application needs to use the water flow energy collection device provided by the foregoing embodiment. For the specific device structure, refer to the foregoing description and will not be elaborated here one by one.

[0085] As Figure 8 shown is a schematic flow chart of a method for collecting water flow energy provided by the embodiment of the present application. The method provided by the embodiment of the present application includes the following steps:

[0086] Step S801: Place the motion component in the water flow energy collection area in a predetermined direction.

[0087] In the embodiments of the present application, the motion component is placed perpendicular to the oncoming flow direction of the water flow.

[0088] Step S802: The sensing component collects the water flow velocities at each sensor and sends the water flow velocities to the control component.

[0089] The method provided in the embodiments of the present application is to make the water flow generate a vortex effect with the water ejected from the jet orifice, accelerate the movement of the motion component, and thus provide more energy for the power generation device. The water ejected from the jet orifice is related to the water flow velocity, so it is necessary to collect the original water flow velocity.

[0090] Specifically, the positions where the sensor components are arranged have been described in detail in the device provided in the embodiments of the present application, and will not be elaborated here.

[0091] Step S803: The control component inputs the water flow velocity into a pre-trained control model, obtains the target mass flow rate, and issues a control instruction.

[0092] The control instruction is used to instruct the jet injection device to eject the target mass flow rate.

[0093] The control model provided in the embodiments of the present application is a neural network model. The parameters of the neural network model are optimized through deep reinforcement learning. The model construction of the neural network and reinforcement learning is developed based on the deep learning development framework (tensorforce) in artificial intelligence, and can be realized according to the settings of the structural parameters of the neural network and the algorithm of reinforcement learning set as follows.

[0094] As Figure 9 shown, it is a schematic diagram of the structure of reinforcement learning and neural network provided in the embodiments of the present application.

[0095] Specifically, the control model includes 4 layers.

[0096] Among them, the first layer is the input layer, and the input layer corresponds to the water flow velocities of 151 sensors.

[0097] The second layer is the first hidden layer, and the number of nodes is 512.

[0098] The third layer is the second hidden layer, and the number of nodes is 512.

[0099] The fourth layer is the output layer, and the number of nodes is 3, which respectively correspond to the target mass flow rates of the water flows of the three jet orifices on the same cross-section.

[0100] During the training process of the control model, the learning rate is set to 0.001, and the maximum number of loops is 2000.

[0101] In the embodiments of the present application, deep reinforcement learning adopts a local policy optimization algorithm. In each cycle, the jet orifice sprays 80 times. The algorithm optimizer adopts the adam optimizer, and the parameter frequency in the neural network model is set to be updated once every 20 times. The reward function of the reinforcement learning is as follows:

[0102] r = |x|

[0103] |·| represents the absolute value, and x is the displacement value of the cylindrical motion tube

[0104] The cost function for the reinforcement learning to update the neural network is as follows:

[0105]

[0106] θ represents the parameter of the actual control policy, which represents the weight parameter of each node of the optimal neural network model in this instance. θ′ is the parameter of the control policy for training, which refers to the weight parameter of each node of the neural network in the process of controlling the jet in this instance.

[0107] E (st，at) is the expected reward value of the entire control process.

[0108] P θ (a t |s t ) refers to the probability that the neural network model executes the action a t under the control policy θ in the state s t , that is, the probability distribution of the mass flow rate of the next jet made by the neural network in the determined state of the cylinder after the jet is ejected.

[0109] P θ′ (a t |s t ) refers to the probability that the neural network model executes the action a t under the control policy θ′ in the state s t .

[0110] A θ′ is a parameter for evaluating the pros and cons of the control policy θ′.

[0111] To further illustrate the method provided by the present application, the present application will be further described below in conjunction with an embodiment.

[0112] As Figure 10 shown, it is a schematic diagram of the convergence degree of the learning process of a neural network provided by the embodiments of the present application. As Figure 10The single reward value shown represents the quality of each process of updating the neural network during each traversal, and the average value emphasizes the overall change process of the reward value. As the average reward value tends to be stable horizontally, it means that the update of the neural network by reinforcement learning tends to be stable, which can be regarded as the neural network having achieved the optimal control effect.

[0113] After the jet stability control by deep reinforcement learning, the rewards of reinforcement learning tend to converge and stabilize, indicating that the control of the jet on the cylindrical motion tube tends to be stable. After the stability control, the size of the water flow energy collection is calculated. In the embodiments of the present application, Newton's laws of motion and the kinetic energy equation are used to calculate the output and input sizes of the energy of the cylindrical motion tube during the motion process. The input energy includes the kinetic energy of the water jet ejected from the jet orifice, and the output energy is the energy collected by the motion of the cylindrical motion tube being transmitted to the generator through the transmission mechanism. The specific method is as follows:

[0114]

[0115] In the formula, W jet[i] The input energy of the i-th jet, Q i [t] The mass flow rate of the i-th jet, ρ jet The density of the jet, l The area of the jet orifice, dt is the time, that is, 0.0005 s.

[0116] In the embodiments of the present application, the energy of the external incoming flow is used as the reference energy to reference the sizes of the input and output energies.

[0117] Specifically, In, W ref Is the reference energy, D is the diameter of the cylindrical motion tube, ρ is the density of the incoming flow, U is the velocity of the incoming flow, and dy is the calculus in the y direction of the cylinder. There are six indicators for evaluating the energy collection situation. Specifically, the jet energy E j , The energy W c Collected under jet control, the energy W b Collected without jet control, the energy difference between controlled and uncontrolled W i =W c -W b -E j , The energy W i / E j Guided by the jet and the ratio of the energy difference to the energy collected without jet control W i / W b . Among the above six evaluation indicators, E j , W c , W b And W i =W c -W b -E jis the true value representing the ratio of the magnitude of the energy during the energy harvesting process compared to the oncoming flow energy, and W i / E j and W i / W b respectively represent the improvement in the energy harvesting rate by jet control. As can be seen from Table 1, the cylindrical moving tube gains energy 26.82 times that of the jet input energy by increasing the jet, and the additional gained energy is 3.57 times higher than that of the cylindrical moving tube without jet control.

[0118] Table 1: Evaluation Table of Energy Harvesting

[0119] <![CDATA[Jet energy, E j > 0.0098 <![CDATA[Energy collected under jet control, W c > 0.3462 <![CDATA[Energy collected without jet control, W b > 0.07350 <![CDATA[Difference between controlled and uncontrolled energy, W i = W c - W b - E j > 0.2629 <![CDATA[Jet-guided energy, W i / E j > 26.82 <![CDATA[Ratio of energy difference to collected energy without jet control, W i / W b > 3.57

[0120] The method provided in the embodiments of the present application is provided with three groups of jet ports, corresponding internal water flow pipes, and a Laval-shaped jet port in three directions of the moving device. The internal water flow pipes provide the jet source for the jet ports, and the water jets ejected from the jet ports will change the force exerted by the external water flow on the moving device, accelerating the movement of the moving device, thereby improving the energy harvesting rate of the entire energy harvesting device. The method provided in the embodiments of the present application makes minor modifications to the equipment and uses a neural network model to control the mass flow rate of the water flow at the jet ports, so as to make the energy provided to the generator more significant.

[0121] Those skilled in the art can clearly understand that the technologies in the embodiments of the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solutions in the embodiments of the present application, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present application.

[0122] For the same and similar parts among the various embodiments in this specification, reference can be made to each other. In particular, for the embodiments of the service construction device and the service loading device, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the descriptions in the method embodiments.

[0123] The above-described embodiments of the present application do not constitute a limitation to the protection scope of the present application.

Claims

1. A water flow energy collection device, characterized in that, The device includes: a sensing component, a motion component, and a control component; the motion component is in a water flow energy collection area, and the sensing component is arranged around the motion component according to a preset position; the control component is electrically connected to the motion component and the sensing component; the control component includes a pre-trained control model; the motion component includes a jet injection device, a motion device, a guide rail device, a transmission device, a power generation device, and a support device; the jet injection device penetrates the motion device; the motion device is vertically arranged in the oncoming flow direction of the water flow in the water flow energy collection area; the motion device is a hollow cylindrical motion tube (3); the cylindrical motion tube (3) is connected to the guide rail device through a connector (6) and a connecting rod (7); the connector (6) includes an upper connector and a lower connector; the guide rail device is installed at the upper and lower ends of the motion device; the transmission device is installed on the guide rail device, one end of the transmission device is connected to the power generation device, and the other end of the transmission device is connected to the motion device; the motion device and the power generation device are arranged at the support device; the jet injection device includes a water pump (1), a water pipe (2), an internal water flow channel (4), and a plurality of jet orifices (5); the angles between the directions of the jet orifices (5) and the water flow direction are 90°, 180°, and 270° respectively, and the jet orifices (5) are all arranged at the edge of the motion device; the internal water flow channel (4) connects each jet orifice (5) and penetrates the motion device; the water pump (1) is connected to the internal water flow channel (4) through the water pipe (2); the control model controls the mass flow rate of the water flow ejected at the jet orifice (5).

2. The device according to claim 1, wherein, the sensing component includes a plurality of sensors. Among them, the sensors in the first part are evenly arranged around the cylindrical motion tube (3) at intervals of every 10° radian and form two sensor rings; the sensors in the second part are densely arranged in the direction perpendicular to the oncoming flow of the cylindrical motion tube (3); the sensors in the third part are arranged in a square area far from the oncoming flow inlet.

3. The device according to claim 2, wherein the guide rail device includes a guide rail (13) and a spring (8), one end of the spring (8) is connected to the motion device, and the other end is connected to a connecting rod (12); a chute is arranged in the middle of the guide rail (13); the motion device is restricted to move in the chute.

4. The device according to claim 3, characterized in that, the transmission device includes a gear (11) and a rack (10); the rack (10) is connected to the lower connector through a thin connecting rod (9); the rack (10) meshes with the gear (11).

5. The device according to claim 4, characterized in that, the support device includes a guide rail base (14), a generator base (15), and a water pump base (16).

6. A method for collecting water flow energy, characterized in that, The method is applied to the water flow energy collection device according to any one of claims 1 to 5. The device includes a sensing component, a motion component, and a control component; the motion component includes a jet injection device, a motion device, a guide rail device, a transmission device, a power generation device, and a support device; the method includes: Place the motion component in the water flow energy collection area in a predetermined direction; The sensing component collects the water flow velocities at each sensor and sends the water flow velocities to the control component; The control component inputs the water flow velocities into a pre-trained control model to obtain a target mass flow rate and issues a control instruction; the control instruction is used to instruct the jet injection device to inject the target mass flow rate.

7. The method according to claim 6, characterized in that, The control model is a neural network model, and the parameters of the neural network model are optimized by deep reinforcement learning.

8. The method according to claim 7, wherein The control model includes 4 layers; Among them, the first layer is the input layer, and the input layer corresponds to the water flow velocities of 151 sensors; The second layer is the first hidden layer, and the number of nodes is 512; The third layer is the second hidden layer, and the number of nodes is 512; The fourth layer is the output layer, and the number of nodes is 3, which respectively correspond to the target mass flow rates of the water flows at three jet ports on the same cross-section.

9. The method according to claim 7, characterized in that, During the training process of the control model, the learning rate is set to 0.001, and the maximum number of iterations is 2000.

10. The method according to claim 7, characterized in that, The deep reinforcement learning adopts a local policy optimization algorithm. The jet ports are injected 80 times in each iteration. The algorithm optimizer adopts an adam optimizer, and the parameter frequency in the neural network model is set to be updated once every 20 times; The reward function of the reinforcement learning is: r＝|x| |·| represents the absolute value, and x is the displacement value of the cylindrical motion tube The cost function for the reinforcement learning to update the neural network is: θ represents the parameters of the actual control strategy, and θ′ represents the parameters of the control strategy for training; E (st，at) is the expected reward value of the entire control process; P θ (a t |s t ) refers to the probability that the neural network model executes action a under the θ control strategy in state s t ; t ​ P θ′ (a t |s t ) refers to the probability that the neural network model executes action a under the θ′ control strategy in state s t ; t ​ A θ′ is a parameter for evaluating the advantages and disadvantages of the θ′ control strategy.

Images