Efficient automatic feed line feed flow control method

Abstract

The invention provides an efficient automatic feed line feed flow control method, and belongs to the technical field of automatic animal husbandry. With an advanced hierarchical control strategy, a global controller is designed to accurately determine feed flow in an automated stockline to achieve optimal feed distribution. By accurately controlling the feed flow, the feed saturation in each discharge box is minimized on the premise of ensuring sufficient livestock feed, so that feed waste is avoided. And the local controller is used as a supplement of the global controller, and adjusts the feed flow of each discharge box in real time according to a reference value provided by the global controller. Compared with a global controller, the local controller is more focused on local details and can deal with internal changes and dynamic characteristics of the system more flexibly. Although the local controller may not comprehensively understand all dynamic characteristics of the system, the complexity and nonlinear characteristics of the system can be better adapted through real-time monitoring and adjustment, so that the stability and efficiency of the system are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of automated animal husbandry, and in particular relates to a highly efficient automated feed line feed flow control method. Background Art

[0002] In animal husbandry production, automated feed line systems play a vital role. They effectively manage and control the delivery of feed to ensure that animals get the right amount of feed, thereby improving production efficiency and ensuring animal health. However, traditional automated feed line systems usually use material flow control methods based on fixed rules or experience, which have certain limitations: 1. Control methods based on fixed rules or experience guidance lack intelligence and adaptability. This makes it difficult to adapt to different production environments and changes in demand. Because these methods cannot make flexible adjustments based on real-time changing conditions, the system cannot cope with changing demands and environmental factors.

[0003] 2. Traditional methods require a deep understanding of the inherent dynamics of the system, which increases the complexity of system design and maintenance. Since fixed rules often fail to accurately capture the dynamic characteristics of the system, more manual intervention and adjustments are required. Moreover, in a dynamic environment, this method is more difficult to achieve accurate feed delivery, which may lead to instability and waste of delivery.

[0004] 3. Traditional methods cannot flexibly adjust feed delivery strategies to adapt to different production needs and the actual needs of animals. This will result in the inability to maximize production efficiency and may affect the health of animals because it is impossible to accurately supply feed according to the actual needs of animals. Summary of the invention

[0005] In view of the above problems, the present invention provides an efficient automatic feed line feed flow control method, comprising the following process: Determine a hierarchical control strategy, and construct a global controller model and a local controller model; the global controller model simulates the minimized feed saturation in the discharge box, that is, the optimal storage capacity of the discharge box; the local controller model controls the valve according to the result of minimizing the feed saturation to realize the automatic operation of the feed line; The hierarchical control strategy is as follows: the expected minimum saturation of each discharge box is used as the reference target of the local controller. The local controller needs to take the real-time status of each discharge box as feedback, including feed volume, internal flow and feed saturation, and obtain the actual opening of the valve. The control effect is judged through deep reinforcement learning and corresponding adjustments are made. The global controller model is established based on the linear state equation of the discharge box. Based on the global controller model, the feed flow rate in the automated feed line is determined to achieve optimal feed distribution, that is, the feed saturation in each discharge box is minimized by controlling the feed flow rate to provide reference to the local controller. Among them, the local controller model is trained based on deep reinforcement learning; based on the trained local controller model, the gate is operated to control the flow of feed in the automated feed line to complete the automated feed line feeding.

[0006] Preferably, the global controller model is specifically: Measuring feed volume , from The discharge box to the Flow rate of each discharge box is considered as internal flow, and external inflow is represented by , while the external outflow is represented by Indicates the volume of feed stored in the discharge box and internal flow The relationship between is linear, so the box system consists of the triple It indicates that, To limit the conditions: is the number of discharge boxes, Represents internal flow The borders of Indicates the weight of each part; To include time disturbances or external inflows; is the weight of external outflow, so that ; The global controller model consists of the linear state equations of each discharge box in the system, as shown in equation (1): (1) in, and They are The collection of upstream and downstream discharge boxes of a discharge box.

[0007] Preferably, the local controller model is specifically: Get the minimum saturation of feed in the entire automated feed line system , a local feed flow controller is established based on deep reinforcement learning. In deep reinforcement learning, in order to obtain the optimal result, a state space and action space are established, and the state transition process of the environment is determined. Represents the time step The state of Represents the time step The action of time, Indicates in status Execute an action The immediate reward obtained after that, the state space The volume of feed in the discharge box 、Internal flow of the discharge box , feed saturation Composition, state space It is expressed as a vector, as shown in formula (4): (4) The action space represents the set of actions that the controller can take. It is expressed as the opening of the valve, expressed as a value between 0 and 1, where 0 represents fully closed and 1 represents fully open; According to the set state space and action space, the local feed flow controller environment modeling is shown in formula (5): (5) in, is the state transition function of the environment, is the random disturbance in the environment.

[0008] Preferably, the real-time feed volume of each discharge box is obtained by installing a frequency modulated continuous wave radar sensor; the internal flow between each discharge box, that is, the actual internal flow, is obtained by an electromagnetic flowmeter; the feed saturation of each discharge box is calculated based on the feed volume, discharge box capacity and internal flow.

[0009] Preferably, the method of achieving the minimized feed saturation in each discharge box is specifically as follows: Based on the global controller model, for each time step , determine the The discharge box to the Optimal internal flow of the discharge box , thereby minimizing the feed saturation in each feed box, and constructing a method for quantifying the feed saturation at time The feed saturation of the entire automated feed line system , as shown in formula (2): (2) in, It's time Time Volume of feed in a discharge box, It's time Time The capacity of a discharge box; by solving formula (3): (3) Get the minimum saturation of feed in the entire automated feed line system , and then get the optimal internal flow .

[0010] Preferably, according to the local feed flow controller environment model, a reward function is designed to guide the behavior of the local feed flow controller through reinforcement learning, and the reward function Should be consistent with the optimization goal, which is to maintain the optimal internal flow , the reward function is shown in formula (6): (6) in, To balance the internal flow of the discharge box and the optimal internal flow Parameters.

[0011] Preferably, a Q-value function is constructed to learn the action-value function to guide the selection of actions; the Q-value function represents the state Next action The expected cumulative reward, that is, the Q value function is based on the cumulative reward function And update; based on the Q value function of the local controller model For training, the loss function is defined by each experience tuple The expected value of the sample is calculated as is the next state; the loss function is shown in formula (7): (7) in, Indicates the expected operator, is the discount factor, is the parameter of the target network, which is used to calculate the target Q value, that is, the expected cumulative reward in the future. are the parameters of the main network, used to evaluate the Q-values ​​of taking different actions in a given state.

[0012] Preferably, the local feed flow controller model is trained based on the loss function, and the specific steps are: S1, initialize the experience tuple , parameters of the main network and target network parameters And the state space ; S2, select actions based on the current Q network ; S3, perform actions in the environment , observe the next state , get instant rewards ; S4, Store the experience tuple; S5, sampling batch data from the experience tuple and calculating the loss of the batch data according to formula (7); S6, use stochastic gradient descent to update the Q value network parameters and perform state Updates; S7, repeat steps S2-S6. When the total reward of the model does not increase significantly, it proves that the model has converged and the training ends. After training, the local feed flow controller model is obtained , as shown in formula (8): (8) in, The current status, including feed volume in the discharge box, internal flow, and feed saturation; For the current state The action selected below is the valve opening; are the trained local controller model parameters, that is, the trained deep Q network parameters.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The present invention can flexibly adjust the amount of feed delivery according to the real-time changing environment and demand conditions by introducing intelligent control strategies, thereby achieving improved production efficiency and quality. Through the hierarchical control strategy, the system can be adjusted according to the control requirements of different levels to achieve precise control of feed delivery, meet the needs of animals to the greatest extent, and improve the adaptability and stability of the system. This will bring more efficient and intelligent feed line management and feed delivery solutions to animal husbandry production.

[0014] In the hierarchical control strategy, the global controller is designed to accurately determine the feed flow in the automated feed line to achieve optimal feed distribution. By accurately controlling the feed flow, the feed saturation in each discharge box is minimized while ensuring sufficient livestock feed, thereby avoiding feed waste and improving breeding efficiency. As a supplement to the global controller, the local controller is responsible for adjusting the feed flow of each discharge box in real time according to the reference value provided by the global controller to ensure the stable operation of the system. Compared with the global controller, the local controller focuses more on local details and can respond more flexibly to changes and dynamic characteristics within the system. Although the local controller may not have a comprehensive understanding of all the dynamic characteristics of the system, it can better adapt to the complexity and nonlinear characteristics of the system through real-time monitoring and adjustment, thereby improving the stability and efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a block diagram of the hierarchical control strategy of the present invention.

[0016] Figure 2 It is the overall logic block diagram of the automated feed line feed flow control method of the present invention. DETAILED DESCRIPTION

[0017] The present invention is further described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] In modern animal husbandry, it is crucial to ensure that farmed animals receive an appropriate supply of feed. To achieve this goal, the present invention adopts an efficient automated feed line system. The system adopts an advanced hierarchical control strategy to ensure the effective distribution and utilization of feed.

[0019] In the hierarchical control strategy, the global controller is designed to accurately determine the feed flow in the automated feed line to achieve optimal feed distribution. By accurately controlling the feed flow, the feed saturation in each discharge box is minimized while ensuring sufficient livestock feed, thereby avoiding feed waste and improving breeding efficiency. As a supplement to the global controller, the local controller is responsible for adjusting the feed flow of each discharge box in real time according to the reference value provided by the global controller to ensure the stable operation of the system. Compared with the global controller, the local controller focuses more on local details and can respond more flexibly to changes and dynamic characteristics within the system. Although the local controller may not have a comprehensive understanding of all the dynamic characteristics of the system, it can better adapt to the complexity and nonlinear characteristics of the system through real-time monitoring and adjustment, thereby improving the stability and efficiency of the system.

[0020] This hierarchical control strategy not only makes full use of the advantages of the global controller and the local controller, but also provides a flexible and intelligent management method to meet the needs of modern animal husbandry for efficient and sustainable development. Through continuous optimization and improvement, this automated feed line system will become an important tool for improving animal husbandry production efficiency and economic benefits, and contribute more possibilities and opportunities to the development of the breeding industry.

[0021] The block diagram of the hierarchical control strategy is as follows: Figure 1 As shown in the figure, the expected minimum saturation of each discharge box is used as the reference target of the local controller. The local controller needs to take the real-time status of each discharge box as feedback. The real-time status includes feed volume, internal flow and feed saturation, and obtain the actual opening of the valve. Through deep reinforcement learning, the control effect is judged and corresponding adjustments are made.

[0022] Based on the above hierarchical control strategy, the present invention provides an automated feed line feed flow control method based on a hierarchical control strategy, comprising the following process: Determine the hierarchical control strategy and construct the global controller model and the local controller model; the global controller model simulates the minimum feed saturation in the discharge box, that is, the optimal storage volume of the discharge box; the local controller model controls the valve according to the result of minimizing the feed saturation to realize the automatic operation of the feed line; The global controller model is established based on the linear state equation of the discharge box. Based on the global controller model, the feed flow rate in the automated feed line is determined to achieve optimal feed distribution, that is, the feed saturation in each discharge box is minimized by controlling the feed flow rate to provide reference to the local controller. Among them, the local controller model is trained based on deep reinforcement learning; based on the trained local controller model, the gate is operated to control the flow of feed in the automated feed line to complete the automated feed line feeding.

[0023] Regarding the establishment of the global controller model: frequency modulated continuous wave (FMCW) radar sensors are installed in each individual discharge box to accurately measure the feed volume The process by which feeds are transported within a system is called a flow. Thus, a system may contain external or internal flows. The discharge box to the Flow rate of each discharge box is considered an internal flow. External inflows are , while the external outflow is represented by Taking into account the volume of feed stored in the discharge box and internal flow The relationship between is considered linear, so the bin system consists of the triple It indicates that, To limit the conditions: is the number of discharge boxes, Represents internal flow The borders of Indicates the weight of each part; To include time disturbances or external inflows; is the weight of external outflow, so that ; The global controller model consists of the linear state equations of each discharge box in the system, as shown in equation (1): (1) in, and They are A collection of upstream and downstream discharge boxes of a discharge box; Achieving minimum feed saturation in each discharge box, the process includes: Based on the global controller model, for each time step , determine the The discharge box to the Optimal internal flow of the discharge box , thereby minimizing the feed saturation in each feed box, and constructing a method for quantifying the feed saturation at time The feed saturation of the entire automated feed line system , as shown in formula (2): (2) in, It's time Time Volume of feed in a discharge box, It's time Time The capacity of a discharge box; by solving formula (3): (3) Get the minimum saturation of feed in the entire automated feed line system , and then get the optimal internal flow .

[0024] Get the minimum saturation of feed in the entire automated feed line system , a local feed flow controller is established based on deep reinforcement learning. In deep reinforcement learning, in order to obtain the optimal result, a state space and action space are established, and the state transition process of the environment is determined. Represents the time step The state of Represents the time step The action of time, Indicates in status Execute an action The immediate reward obtained after that, the state space The volume of feed in the discharge box 、Internal flow of the discharge box , feed saturation Composition, state space It is expressed as a vector, as shown in formula (4): (4) The action space represents the set of actions that the controller can take, usually expressed as a continuous space or a discrete space; in this invention, the action It can be expressed as the opening of the valve, usually expressed as a value between 0 and 1, where 0 means fully closed and 1 means fully open.

[0025] According to the set state space and action space, the local feed flow controller environment modeling is shown in formula (5): (5) in, is the state transition function of the environment, is the random disturbance in the environment.

[0026] According to the local feed flow controller environment model, a reward function is designed to guide the behavior of the reinforcement learning local feed flow controller. The reward function It should be consistent with the optimization goal. In the present invention, the optimization goal is to maintain the optimal internal flow , therefore, the reward function is as shown in formula (6): (6) in, To balance the internal flow of the discharge box and the optimal internal flow Parameters.

[0027] Construct a Q-value function to learn the action-value function to guide the selection of actions. The Q-value function represents the state Next action The expected cumulative reward, that is, the Q value function is based on the cumulative reward function Based on the Q value function, the local controller model is updated. For training, the loss function is defined by each experience tuple The expected value of the sample is calculated as is the next state; the loss function is shown in formula (7): (7) in, Indicates the expected operator, is the discount factor, is the parameter of the target network, which is used to calculate the target Q value, that is, the expected cumulative reward in the future. are the parameters of the main network, used to evaluate the Q-values ​​of taking different actions in a given state.

[0028] Based on the loss function, the local feed flow controller model is trained. The specific steps are: 1) Initialize the experience tuple , parameters of the main network and target network parameters And the state space ; 2) Select actions based on the current Q network ; 3) Perform actions in the environment , observe the next state , get instant rewards ; 4) Store the experience tuple; 5) Sample batch data from the experience tuple and calculate the loss of the batch data according to formula (7); 6) Use stochastic gradient descent to update the Q value network parameters and perform state Updates; 7) Repeat steps 2)-6). When the total reward of the model does not increase significantly, it proves that the model has converged and the training ends; After training, the local feed flow controller model is obtained , as shown in formula (8): (8) in, The current status, including feed volume in the discharge box, internal flow, and feed saturation; For the current state The action selected below is the valve opening; are the trained local controller model parameters, that is, the trained deep Q network parameters.

[0029] Among them, the real-time feed volume of each discharge box is obtained by installing a frequency modulated continuous wave radar sensor (FMCW); the internal flow between each discharge box, that is, the actual internal flow, is obtained by an electromagnetic flowmeter; the feed saturation of each discharge box is calculated based on the feed volume, discharge box capacity and internal flow.

[0030] The system operation process based on this method: Controller deployment: Deploy trained local feed flow controllers to control units in automated feed line systems.

[0031] State acquisition: During the control process, the controller needs to obtain the current state information of the system through sensors, including the volume of feed in the feed box, the internal flow of feed flow, and feed saturation.

[0032] Action selection: Based on the current state information, the global controller obtains the desired minimized feed saturation in the feed box. Based on this saturation, the local controller uses the trained model to select the optimal action.

[0033] Valve control: The action value output by the controller is used to operate the opening of the valve to adjust the feed flow.

[0034] Feedback and Adjustment: During the control process, the system collects real-time feedback information, such as feed volume, internal flow of feed flow, feed saturation, etc. This feedback information can be used to evaluate the performance of the controller and make adjustments as needed.

[0035] Continuous operation: The local controller model will continuously select actions based on the system status information and control the gate to realize the feeding operation of the automated feed line. In this way, the system can maintain a stable feed flow, minimize feed saturation, and complete the automated feeding process in continuous adjustment.

[0036] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0037] Although the above describes the specific implementation methods of the present invention, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. An efficient automatic feed line feed flow control method, characterized in that: The process includes: Determine a hierarchical control strategy, and construct a global controller model and a local controller model; the global controller model simulates the minimized feed saturation in the discharge box, that is, the optimal storage capacity of the discharge box; the local controller model controls the valve according to the result of minimizing the feed saturation to realize the automatic operation of the feed line; The hierarchical control strategy is as follows: the expected minimum saturation of each discharge box is used as the reference target of the local controller. The local controller needs to take the real-time status of each discharge box as feedback, including feed volume, internal flow and feed saturation, and obtain the actual opening of the valve. The control effect is judged through deep reinforcement learning and corresponding adjustments are made. The global controller model is established based on the linear state equation of the discharge box. Based on the global controller model, the feed flow rate in the automated feed line is determined to achieve optimal feed distribution, that is, the feed saturation in each discharge box is minimized by controlling the feed flow rate to provide reference to the local controller. Among them, the local controller model is trained based on deep reinforcement learning; based on the trained local controller model, the gate is operated to control the flow of feed in the automated feed line to complete the automated feed line feeding.

2. The efficient automatic feed line feed flow control method according to claim 1, characterized in that: The global controller model is specifically: Measuring feed volume , from The discharge box to the Flow rate of each discharge box is considered as internal flow, and external inflow is represented by , while the external outflow is represented by Indicates the volume of feed stored in the discharge box and internal flow The relationship between is linear, so the box system consists of the triple It indicates that, To limit the conditions: is the number of discharge boxes, Represents internal flow The borders of Indicates the weight of each part; To include time disturbances or external inflows; is the weight of external outflow, so that ; The global controller model consists of the linear state equations of each discharge box in the system, as shown in equation (1): （1） in, and They are The collection of upstream and downstream discharge boxes of a discharge box.

3. The efficient automatic feed line feed flow control method according to claim 1, characterized in that: The local controller model is specifically: Get the minimum saturation of feed in the entire automated feed line system , a local feed flow controller is established based on deep reinforcement learning. In deep reinforcement learning, in order to obtain the optimal result, a state space and action space are established, and the state transition process of the environment is determined. Represents the time step The state of Represents the time step The action of time, Indicates in status Execute an action The immediate reward obtained after that, the state space The volume of feed in the discharge box 、Internal flow of the discharge box , feed saturation Composition, state space It is expressed as a vector, as shown in formula (4): （4） The action space represents the set of actions that the controller can take. It is expressed as the opening of the valve, expressed as a value between 0 and 1, where 0 represents fully closed and 1 represents fully open; According to the set state space and action space, the local feed flow controller environment modeling is shown in formula (5): （5） in, is the state transition function of the environment, is the random disturbance in the environment.

4. The efficient automatic feed line feed flow control method according to claim 1, characterized in that: The real-time feed volume of each discharge box is obtained by installing a frequency modulated continuous wave radar sensor; the internal flow between each discharge box, that is, the actual internal flow, is obtained by an electromagnetic flowmeter; the feed saturation of each discharge box is calculated based on the feed volume, discharge box capacity and internal flow.

5. A highly efficient automated feed line feed flow control method as claimed in claim 2, characterized in that: The method of achieving the minimized feed saturation in each discharge box is specifically as follows: Based on the global controller model, for each time step , determine the The discharge box to the Optimal internal flow of the discharge box , thereby minimizing the feed saturation in each feed box, and constructing a method for quantifying the feed saturation at time The feed saturation of the entire automated feed line system , as shown in formula (2): （2） in, It's time Time Volume of feed in a discharge box, It's time Time The capacity of the discharge box; By solving formula (3): （3） Get the minimum saturation of feed in the entire automated feed line system , and then get the optimal internal flow .

6. The efficient automatic feed line feed flow control method according to claim 3, characterized in that: According to the local feed flow controller environment model, a reward function is designed to guide the behavior of the reinforcement learning local feed flow controller. The reward function Should be consistent with the optimization goal, which is to maintain the optimal internal flow , the reward function is shown in formula (6): （6） in, To balance the internal flow of the discharge box and the optimal internal flow Parameters.

7. The efficient automatic feed line feed flow control method according to claim 1, characterized in that: Construct a Q-value function to learn the action-value function to guide the selection of actions; the Q-value function represents the state Next action The expected cumulative reward, that is, the Q value function is based on the cumulative reward function And update; based on the Q value function of the local controller model For training, the loss function is defined by each experience tuple The expected value of the sample is calculated as is the next state; the loss function is shown in formula (7): （7） in, Indicates the expected operator, is the discount factor, is the parameter of the target network, which is used to calculate the target Q value, that is, the expected cumulative reward in the future. are the parameters of the main network, used to evaluate the Q-values ​​of taking different actions in a given state.

8. The efficient automatic feed line feed flow control method according to claim 1, characterized in that: The local feed flow controller model is trained based on the loss function, and the specific steps are as follows: S1, initialize the experience tuple , parameters of the main network and target network parameters And the state space ; S2, select actions based on the current Q network ; S3, perform actions in the environment , observe the next state , get instant rewards ; S4, Store the experience tuple; S5, sampling batch data from the experience tuple and calculating the loss of the batch data according to formula (7); S6, use stochastic gradient descent to update the Q value network parameters and perform state Updates; S7, repeat steps S2-S6. When the total reward of the model does not increase significantly, it proves that the model has converged and the training ends. After training, the local feed flow controller model is obtained , as shown in formula (8): （8） in, The current status, including feed volume in the discharge box, internal flow, and feed saturation; For the current state The action selected below is the valve opening; are the trained local controller model parameters, that is, the trained deep Q network parameters.

Images