An intelligent electric ball valve control system based on artificial intelligence

Abstract

The application relates to the technical field of gate valve control, and provides an intelligent electric ball valve control system based on artificial intelligence; the system comprises a sensor monitoring module, a data preprocessing module, a ball valve control module, a strategy optimization module, a valve execution module and an electric ball valve; the system is structurally improved on the actor network of a DDPG algorithm in the ball valve control module, a three-dimensional instruction-DDPG algorithm is constructed, and thus the accurate adjustment of valve control is realized; in the strategy optimization module, a skill neuron protection mechanism and an unbalanced experience playback mechanism are adopted, the adaptability and robustness of the control system in a complex and dynamic environment are improved; through the innovative mechanisms, the precision, stability and system response speed of electric ball valve control are remarkably improved, the shortcomings of a traditional control system under multidimensional interference are solved, and the application prospect is wide.

Description

Technical Field

[0001] This invention relates to the field of valve control technology, and in particular to an intelligent electric ball valve control system based on artificial intelligence. Background Technology

[0002] With the development of modern agriculture, agricultural irrigation methods are gradually shifting from traditional manual control to intelligent and automated processes. Electric ball valves, as a type of automated valve driven by a motor, have been widely used in agricultural irrigation systems. However, existing electric ball valve control systems still have many shortcomings in terms of intelligence: First, existing electric ball valve control systems typically employ algorithms based on classical control theory (such as PID control). Although these algorithms perform well in some stable systems, they are poorly adaptable to complex and nonlinear conditions. When facing dynamic environmental changes, these algorithms struggle to achieve precise real-time adjustment, especially when the system faces the interaction of multiple input variables, making the control effect susceptible to interference and leading to a decrease in control accuracy. Firstly, traditional systems often make decisions based on fixed sensor data and rely on simple data filtering and anomaly detection mechanisms. This data processing approach fails to fully explore the potential patterns in sensor data and lacks the fusion and analysis of multi-source and multi-dimensional data, resulting in the system's inability to efficiently cope with complex changes in operating conditions. For example, when faced with sudden pressure fluctuations, flow fluctuations, or drastic temperature changes, the system cannot make rapid adjustments, which may lead to control lag or misoperation, affecting the valve's execution accuracy. Finally, most existing systems use rules or fixed models based on prior settings for control, lacking adaptive learning capabilities. The system struggles to automatically learn and adjust its control strategy, making it impossible to guarantee optimal performance under different operating conditions. Summary of the Invention

[0003] This invention aims to overcome the problems of slow response, poor strategy adaptability, and insufficient control precision in existing electric ball valve control systems under complex environments, and provides an intelligent electric ball valve control system based on artificial intelligence. This system, by introducing a structured reinforcement learning algorithm and a strategy optimization mechanism, achieves intelligent perception, adaptive optimization, and high-precision execution of the valve control process, significantly improving the system's robustness and control efficiency. It is particularly suitable for applications such as agricultural irrigation in variable environments. Specifically, this invention includes: First, based on the traditional Deep Deterministic Policy Gradient (DDPG) algorithm, this invention introduces three types of operators—frequency, amplitude, and direction—to structurally improve the original DDPG actor network, constructing a three-dimensional instruction-DDPG algorithm. This algorithm decomposes the valve control action into frequency, amplitude, and direction... This invention achieves structured modeling and fine-grained control of the action space across three physical dimensions, overcoming the limitations of traditional DDPG in learning continuous action spaces and effectively improving generalization ability and control stability in complex multivariable systems. Secondly, to address training instability caused by policy drift during policy gradient updates, this invention further proposes a policy stability enhancement mechanism. This mechanism integrates a skill neuron protection mechanism and an imbalanced experience replay mechanism, ensuring policy stability and convergence of the FAD algorithm during long-term training. In summary, this invention, by constructing a three-dimensional structured DDPG control framework and introducing a multi-level policy stability optimization mechanism, achieves significant improvements in valve action control accuracy, system response speed, and policy adaptability. It is suitable for high-requirement, high-dynamic intelligent fluid control systems and has broad application prospects and promotional value.

[0004] This invention provides an intelligent electric ball valve control system based on artificial intelligence. The system includes a sensor monitoring module, a data preprocessing module, a ball valve control module, a strategy optimization module, a valve execution module, and an electric ball valve.

[0005] The sensor monitoring module obtains raw sensor data by collecting real-time parameter information on flow rate, pressure, temperature, and liquid level position from pressure sensors, flow sensors, temperature sensors, and liquid level sensors.

[0006] The data preprocessing module filters, normalizes, and detects anomalies in the raw sensor data to generate cleaned sensor data.

[0007] The ball valve control module constructs the DDPG algorithm. By introducing three types of operators—frequency, amplitude, and direction—the actor network of the DDPG algorithm is structurally improved to construct a three-dimensional instruction-DDPG algorithm. Based on cleaning sensor data, the three-dimensional instruction-DDPG algorithm is used to generate precise valve control commands.

[0008] The strategy optimization module employs a combination of a skill neuron protection mechanism and an imbalanced experience replay mechanism to construct a strategy stability enhancement mechanism; through this mechanism, the strategy gradient update in the 3D instruction-DDPG algorithm is optimized.

[0009] The valve actuator module controls the electric ball valve according to precise valve control commands.

[0010] Furthermore, the ball valve control module generates precise valve control commands, specifically including the following steps:

[0011] Step S1: Based on the cleaning sensor data, set the initial environmental state;

[0012] Step S2: Initialize the actor network and commentator network of the 3D instruction-DDPG algorithm; introduce frequency operator, amplitude operator and direction operator to optimize the structure of the actor network and construct the FAD actor network; generate valve control commands through the FAD actor network according to the initial state of the environment; the FAD actor network decomposes the output of the actor network into a structured form to control the frequency, amplitude and direction of the valve control commands respectively;

[0013] Step S3: Evaluate the effectiveness of valve control commands through a network of commentators;

[0014] Step S4: Transmit the valve control command to the valve execution module, execute the valve control command, and generate the updated environment state;

[0015] Step S5: Define the target state and reward function, and calculate the immediate reward based on the target state and the updated environment state using the reward function;

[0016] Step S6: Based on immediate rewards, update the commentator network according to the Bellman equation to guide the update strategy of the FAD actor network; based on the updated commentator network, update the FAD actor network through policy gradient to improve the valve control command.

[0017] Step S7: Iterate through steps S3-S6 to train the FAD actor network and generate precise valve control commands.

[0018] Furthermore, the policy optimization module optimizes the policy gradient in the 3D instruction-DDPG algorithm, specifically including the following steps:

[0019] Step B1: Record neuron activity during the training of the FAD actor network, calculate the gradient contribution of each neuron, and obtain the skill neuron score; construct the old experience pool and the new experience pool;

[0020] Step B2: Based on the skill neuron score, a gradient masking factor is introduced. During the policy gradient backpropagation process, the gradient masking factor is applied to suppress the update of the skill neuron and stabilize the propagation path of the policy gradient.

[0021] Step B3: Construct the angle-experience pool loss function. The angle-experience pool loss function dynamically selects data from the old or new experience pools to participate in training, thereby mitigating catastrophic forgetting and stabilizing the direction of the policy gradient.

[0022] Furthermore, step B3 specifically includes: introducing an imbalanced gating replay mechanism and gradient direction consistency; constructing an angle-experience pool loss function; dynamically selecting data from the old or new experience pools for training through the angle-experience pool loss function to mitigate catastrophic forgetting and stabilize the direction of the policy gradient; the imbalanced gating replay mechanism dynamically adjusts the proportion of old and new experiences in the angle-experience pool loss function by comparing the direction relationship between the current policy gradient and the old task policy gradient; specifically, in each training iteration, the policy gradients obtained based on the new and old experience pools are first calculated separately, and the angle between the two is used to determine whether the current update direction is consistent with the past.

[0023] If the angle is small (the directions are similar), it means that the current strategy update will not damage the performance of the old task. In this case, reduce the participation of old experience in order to adapt to the new task more quickly.

[0024] If the angle is large (directional conflict), the participation weight of old experience is increased to maintain the continuity of the policy gradient direction and alleviate catastrophic forgetting.

[0025] By adopting the above solution, the beneficial effects achieved by the present invention are as follows:

[0026] This invention constructs a three-dimensional instruction-DDPG algorithm based on three types of operators: frequency, amplitude, and direction. This algorithm enables the structured generation and high-precision adjustment of control commands for electric ball valves. Compared to traditional algorithms that only model the control quantity in a single dimension, the FAD structure proposed in this invention explicitly decomposes the action output space into three types of adjustable parameters. This makes the control commands more physically interpretable and more flexible in response, significantly improving the dynamic response performance and control stability of the valve under different operating conditions. This structure is particularly suitable for complex adjustment needs caused by multi-dimensional disturbances such as flow rate, pressure, and liquid level in agricultural irrigation scenarios, ensuring the precise adaptation of crops to water supply and improving overall irrigation efficiency.

[0027] The policy optimization module introduced in this invention achieves dynamic control and protection of the policy update path of the actor network by constructing a policy stability enhancement mechanism. On the one hand, the skill neuron protection mechanism can identify neurons that play a key role in the control policy and limit their gradient fluctuations, effectively avoiding policy degradation caused by critical path perturbations during training. On the other hand, through the joint design of the imbalanced experience replay mechanism and the angle-experience pool loss function, the policy gradient direction between old and new experiences is balanced, alleviating the catastrophic forgetting problem in reinforcement learning training. The design of this optimization module closely serves the electric ball valve control objective of this invention, enhancing the robustness and continuous optimization capability of the system during multi-task and long-term operation.

[0028] Through the above innovative design, this invention achieves precise control of electric ball valves based on artificial intelligence, improves the system's adaptability in dynamic, multi-source information input environments, solves the problems of slow response and error accumulation in traditional systems under multi-dimensional coupled interference, and enhances the intelligent level of water resource allocation in agricultural irrigation. The overall system has stronger generalization ability, stability and control precision, and can operate reliably for a long time in complex actual irrigation scenarios, providing strong support for efficient irrigation and water conservation in modern agriculture. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of a module of an intelligent electric ball valve control system based on artificial intelligence proposed in this invention;

[0030] Figure 2 Example 2 shows a comparison of the training effects of the three proposed FAD actor networks and actor networks. Detailed Implementation

[0031] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0032] Example 1, according to Figure 1 The present invention provides an intelligent electric ball valve control system based on artificial intelligence. The system includes a sensor monitoring module, a data preprocessing module, a ball valve control module, a strategy optimization module, a valve execution module, and an electric ball valve.

[0033] The sensor monitoring module obtains raw sensor data by collecting real-time parameter information on flow rate, pressure, temperature, and liquid level position from pressure sensors, flow sensors, temperature sensors, and liquid level sensors; the sensor monitoring module then transmits the raw sensor data to the data preprocessing module.

[0034] The data preprocessing module filters, normalizes, and detects anomalies in the raw sensor data to generate cleaning sensor data; the data preprocessing module then transmits the cleaning sensor data to the ball valve control module.

[0035] The ball valve control module constructs the DDPG algorithm. By introducing three types of operators—frequency, amplitude, and direction—the actor network of the DDPG algorithm is structurally improved to construct a three-dimensional instruction-DDPG algorithm. Based on cleaning sensor data, the three-dimensional instruction-DDPG algorithm is used to generate precise valve control commands. The ball valve control module connects the strategy optimization module and the valve execution module.

[0036] The strategy optimization module employs a combination of a skill neuron protection mechanism and an imbalanced experience replay mechanism to construct a strategy stability enhancement mechanism; through this mechanism, the strategy gradient update in the 3D instruction-DDPG algorithm is optimized.

[0037] The valve actuator module controls the opening and closing status, opening angle, opening and closing speed, opening and closing direction, multi-level adjustment position, flexible opening and closing process, and execution timing of the electric ball valve according to precise valve control commands.

[0038] Example 2, according to Figure 2 This embodiment is based on Embodiment 1. In this embodiment, the process of the ball valve control module generating precise valve control commands specifically includes the following steps:

[0039] Step S1: Based on the cleaning sensor data, set the initial environmental state;

[0040] Step S2: Initialize the actor network and commentator network of the 3D instruction-DDPG algorithm; introduce frequency, amplitude, and direction operators to optimize the structure of the actor network and construct the FAD actor network; generate valve control commands through the FAD actor network based on the initial environmental state; the FAD actor network structurally decomposes the output of the actor network to control the frequency, amplitude, and direction of the valve control commands respectively, using the following formulas:

[0041] ;

[0042] in, Indicates a time step. Indicates the initial state of the environment. Indicates the disturbance term. Indicates in The environmental state at any given time; Indicates in Valve control commands at specific times; Indicates the magnitude operator, Indicates the feedforward term. Indicates the direction operator, Indicates the frequency operator;

[0043] Step S3: Evaluate the effectiveness of valve control commands through a network of commentators;

[0044] Step S4: Transmit the valve control command to the valve execution module, execute the valve control command, and generate the updated environment state;

[0045] Step S5: Define the target state and reward function, and calculate the immediate reward based on the target state and the updated environment state using the reward function;

[0046] Step S6: Based on immediate rewards, update the commentator network according to the Bellman equation to guide the update strategy of the FAD actor network; based on the updated commentator network, update the FAD actor network through policy gradient to improve the valve control command.

[0047] Step S7: Iterate through steps S3-S6 to train the FAD actor network and generate precise valve control commands.

[0048] Example 3, according to Figure 2 This embodiment is based on Embodiment 1. In this embodiment, the process of the ball valve control module generating precise valve control commands specifically includes the following steps:

[0049] Step R1: Based on the cleaning sensor data, set the initial environmental state;

[0050] Step R2: Initialize the actor network and commentator network of the DDPG algorithm; generate valve control commands through the actor network based on the initial state of the environment;

[0051] Step R3: Evaluate the effectiveness of valve control commands through a network of commentators;

[0052] Step R4: Transmit the valve control command to the valve execution module, execute the valve control command, and generate the updated environment state;

[0053] Step R5: Define the target state and reward function, and calculate the immediate reward based on the target state and the updated environment state using the reward function;

[0054] Step R6: Based on immediate rewards, update the commentator network according to the Bellman equation to guide the update strategy of the actor network; based on the updated commentator network, update the actor network through policy gradient to improve the valve control command;

[0055] Step R7: Iterate through steps R3-R6 to train the actor network and generate precise valve control commands.

[0056] Example 4, based on Example 2, describes the process of optimizing the policy gradient in the 3D Instruction-DDPG algorithm using the policy optimization module. Specifically, this includes the following steps:

[0057] Step B1: Record neuron activity during the training of the FAD actor network, calculate the gradient contribution of each neuron, and obtain the skill neuron score; construct the old experience pool and the new experience pool;

[0058] Step B2: Based on the skill neuron scores, a gradient masking factor is introduced. During the policy gradient backpropagation process, the gradient masking factor is applied to suppress the updating of skill neurons and stabilize the propagation path of the policy gradient. The formula used is as follows:

[0059] Gradient masking factor definition:

[0060] ;

[0061] in, Represents the neuron index. Indicates the first The gradient masking factor applied by each neuron during backpropagation; This indicates the first in the FAD actor network One neuron; Indicates the first Stability score of each neuron, Indicates the gradient suppression coefficient;

[0062] Step B3: Construct the angle-experience pool loss function. The angle-experience pool loss function dynamically selects data from the old or new experience pools to participate in training, thereby mitigating catastrophic forgetting and stabilizing the direction of the policy gradient.

[0063] Example 5, based on Example 2, describes the process of optimizing the policy gradient in the 3D Instruction-DDPG algorithm using the policy optimization module. Specifically, this includes the following steps:

[0064] Step E1: Record neuron activity during the training of the FAD actor network, calculate the gradient contribution of each neuron, and obtain the skill neuron score; construct the old experience pool and the new experience pool;

[0065] Step E2: Based on the skill neuron score, a gradient masking factor is introduced. During the policy gradient backpropagation process, the gradient masking factor is applied to suppress large updates of skill neurons and stabilize the propagation path of the policy gradient.

[0066] Step E3: Construct the experience pool loss function. The experience pool loss function dynamically selects data from the old or new experience pools to participate in training, thereby mitigating catastrophic forgetting and stabilizing the direction of the policy gradient.

[0067] Example 6, based on Example 4, specifically includes step B3: introducing an imbalanced gating replay mechanism and gradient direction consistency, constructing an angle-experience pool loss function, and dynamically selecting data from the old or new experience pools to participate in training, mitigating catastrophic forgetting and stabilizing the direction of the policy gradient; the imbalanced gating replay mechanism dynamically adjusts the proportion of old and new experiences in the angle-experience pool loss function by comparing the direction relationship between the current policy gradient and the old task policy gradient; specifically, in each training iteration, the policy gradients obtained based on the new and old experience pools are first calculated separately, and the angle between the two is used to determine whether the current update direction is consistent with the past.

[0068] If the angle is small (the directions are similar), it means that the current strategy update will not damage the performance of the old task. In this case, reduce the participation of old experience in order to adapt to the new task more quickly.

[0069] If the angle is large (directional conflict), the weight of old experience is increased to maintain the continuity of the policy gradient direction and mitigate catastrophic forgetting. The formula used is as follows:

[0070] The formula for the angle-experience pool loss function is as follows:

[0071] ;

[0072] in, This represents the angle between policy gradients. This represents the policy gradient of the old experience pool. This represents the policy gradient of the new experience pool. express and dot product, express and Multiply the modulus lengths, Represents the inverse cosine function;

[0073] ;

[0074] in, This represents the angle-experience pool loss function. Represents the angle-gated function. Indicates the current environmental state. Indicates the state The following actions were taken. Indicates an immediate reward. Indicates the next state. This refers to the old experience pool. This indicates a new experience pool. This represents the loss item from the old experience pool. This represents the loss item in the new experience pool.

[0075] The present invention and its embodiments have been described above. This description is not restrictive. The accompanying drawings are only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present invention, such design should fall within the protection scope of the present invention.

Claims

1. An artificial intelligence based intelligent motor operated ball valve control system, the system comprising: The data preprocessing module generates cleaning sensor data; the system further comprises a ball valve control module, a strategy optimization module and a valve execution module; The ball valve control module constructs a three-dimensional instruction-DDPG algorithm, generates precise valve control instructions using the three-dimensional instruction-DDPG algorithm based on the cleaning sensor data; The strategy optimization module constructs a strategy stability enhancement mechanism; the strategy stability enhancement mechanism optimizes the strategy gradient in the three-dimensional instruction-DDPG algorithm; The valve execution module drives and controls the electric ball valve according to the precise valve control instructions; The ball valve control module constructs the three-dimensional instruction-DDPG algorithm, specifically including: constructing a DDPG algorithm; structurally improving the DDPG algorithm by introducing frequency, amplitude and direction operators, thereby constructing a three-dimensional instruction-DDPG algorithm; The ball valve control module generates precise valve control instructions, specifically including the following steps: Step S1: based on the cleaning sensor data, set the initial state of the environment; Step S2: initialize the actor network and critic network of the three-dimensional instruction-DDPG algorithm; introduce the frequency operator, amplitude operator and direction operator to optimize the structural composition of the actor network, construct the F-A-D actor network; generate valve control instructions through the F-A-D actor network according to the initial state of the environment; Step S3: evaluate the valve control instructions through the critic network; Step S4: pass the valve control instructions to the valve execution module, execute the valve control instructions, and generate an updated environment state; Step S5: define the target state and the reward function, calculate the immediate reward through the reward function according to the target state and the updated environment state; Step S6: based on the immediate reward, update the critic network according to the Bellman equation, and then update the F-A-D actor network through the strategy gradient to improve the valve control instructions; Step S7: iterate steps S3-S6 to train the F-A-D actor network and generate precise valve control instructions.

2. The intelligent electric globe valve control system based on artificial intelligence according to claim 1, characterized in that: The strategy stability enhancement mechanism is jointly constructed by adopting the skill neuron protection mechanism and the unbalanced experience replay mechanism.

3. The intelligent electric globe valve control system based on artificial intelligence according to claim 1, characterized in that: The F-A-D actor network structurally splits the output of the actor network to control the frequency, amplitude and direction of the valve control instructions.

4. The intelligent electric globe valve control system based on artificial intelligence according to claim 1, characterized in that: The strategy optimization module optimizes the strategy gradient in the three-dimensional instruction-DDPG algorithm, specifically including the following steps: Step B1: record the neuron activity during the training process of the F-A-D actor network, calculate the gradient contribution of each neuron to obtain the skill neuron score; construct an old experience pool and a new experience pool; Step B2: according to the skill neuron score, introduce a gradient masking factor to inhibit the update of the skill neurons during the backpropagation of the strategy gradient, and stabilize the propagation path of the strategy gradient; Step B3: construct an angle-experience pool loss function, dynamically select data in the old experience pool or the new experience pool for training through the angle-experience pool loss function, alleviate catastrophic forgetting, and stabilize the direction of the strategy gradient.

Images