Automatic control system for condensed water regulating valve of nuclear power plant

Abstract

The invention relates to the technical field of nuclear power plant automatic control and fluid regulation, in particular to a nuclear power plant condensation water regulating valve automatic control system which comprises a sensor data acquisition and synchronization module which acquires and synchronizes parameters in a nuclear power plant condensation water system; the reinforcement learning control module reads parameters of the sensor data acquisition and synchronization module, and generates a regulating valve control strategy by using a reinforcement learning algorithm; the real-time prediction and control decision module controls and predicts the opening degree of the condensed water regulating valve according to the regulating strategy of the reinforcement learning control module; and the anomaly detection and safety protection module reads the parameters acquired by the data acquisition and synchronization module, monitors the parameters and gives an alarm according to an anomaly monitoring result. According to the system, valve oscillation and unreasonable adjustment in a traditional method can be avoided while a valve adjustment strategy is learned and optimized in real time, and the precision and stability of condensed water flow adjustment are improved.

Description

Technical Field

[0001] This invention relates to the field of nuclear power plant automation control and fluid regulation technology, and in particular to an automatic control system for condensate regulating valves in nuclear power plants. Background Technology

[0002] As energy facilities with high safety and reliability requirements, nuclear power plants require precise control of multiple critical process loops during operation, among which the regulation of the deaerator tank water level in the condensate system is particularly important. Traditionally, nuclear power plants use empirical formulas or PID control algorithms to adjust the opening of condensate regulating valves to ensure stable deaerator water levels. However, these traditional methods often cannot effectively handle nonlinear characteristics under complex operating conditions, such as load fluctuations during unit peak shaving or valve oscillations within specific flow ranges. Especially when condensate flow is low or fluctuates significantly, precise valve adjustment becomes crucial for achieving stable deaerator water level control.

[0003] In existing technologies, valve regulation mainly relies on manual adjustment and control strategies based on static empirical formulas, typically assuming a linear or simple piecewise linear relationship between valve flow rate and opening degree. However, this assumption ignores the complexity of the valve's dynamic response in actual operation. For example, under conditions of high temperature, high pressure, variable load, and equipment wear, a significant nonlinear relationship may exist between flow rate and opening degree, leading to a substantial reduction in the control accuracy of traditional methods. Especially under sudden operating conditions, this can cause valve control instability and affect system safety. Summary of the Invention

[0004] This invention provides an automatic control system for condensate regulating valves in nuclear power plants, which addresses the problem that existing valve control systems neglect the complexity of the dynamic response of valves during actual operation.

[0005] The technical solution of the present invention is as follows:

[0006] This invention proposes an automatic control system for condensate regulating valves in nuclear power plants. The system includes a sensor data acquisition and synchronization module, a reinforcement learning control module, a real-time prediction and control decision module, and an anomaly detection and safety protection module. The sensor data acquisition and synchronization module acquires and synchronizes parameters in the nuclear power plant's condensate system. The reinforcement learning control module reads the parameters from the sensor data acquisition and synchronization module and generates a regulating valve control strategy using a reinforcement learning algorithm. The real-time prediction and control decision module controls and predicts the opening degree of the condensate regulating valve based on the regulation strategy from the reinforcement learning control module. The anomaly detection and safety protection module reads the parameters acquired by the data acquisition and synchronization module, monitors the parameters, and issues an alarm for anomaly detection results.

[0007] In some embodiments, sensors are installed at pipeline nodes and regulating valve locations in the condensate system of a nuclear power plant, and redundant sensors are configured. The sensor data acquisition and synchronization module acquires the sensor parameters and transmits the data via industrial Ethernet or fieldbus protocols.

[0008] In some embodiments, the sensor data acquisition and synchronization module acquires condensate flow rate, deaerator water level, regulating valve opening, regulating valve pressure, and regulating valve temperature. The sensor data acquisition and synchronization module acquires condensate flow rate and regulating valve opening at a frequency of 1 Hz or higher.

[0009] In some embodiments, the reinforcement learning control module generates a control strategy for the regulating valve by learning formula Q. The construction of formula Q includes defining a state space, defining an action space, designing a reward function, and updating the learning formula. The defined state space includes a state space s. t This includes the deaerator water level, condensate flow rate, valve opening of the three regulating valves, temperature, and pressure at time t; the defined action space includes action space a. t This indicates the adjustment of the opening of three regulating valves for time t, with each action space a. t This represents the increment of the opening of the three valves; the reward function is designed as shown in formula (1):

[0010]

[0011] In the formula:

[0012] L t L represents the deaerator water level at time t. ref The target water level for the deaerator. Let i be the ideal valve opening degree of the regulating valve at time t; ||(u i (t)∈Unfavorable interval) is an indicator function for whether the valve opening at time t is in the unfavorable interval; a, b, g are weighting coefficients; the learning formula and the updating formula are as shown in formula (2):

[0013] Q(s t ,a t )←Q(s t ,a t )+h[t t +g maxQ(s t+1 ,a')-Q(s t ,a t (2)

[0014] In the formula:

[0015] Q is the learning formula, t t For immediate reward, that is, r at time t calculated by the reward function in formula (1). tg is the discount factor, h is the learning rate, and s is the learning rate. t Let s be the state space at time t. t+1 Let a be the state space at time t next time step. t Let be the action space at time t, and let a′ represent the optimal action at the next time step t+1.

[0016] In some embodiments, the system further includes a water level dynamic modeling module, which simulates the water level change of the deaerator water tank and compares and corrects it with the water level data of the deaerator water tank collected by the sensor data acquisition and synchronization module. The water level dynamic equation of the deaerator water tank simulated by the water level dynamic modeling module is as shown in formula (3):

[0017]

[0018] In the formula, L(t) is the deaerator water level, and F in (t) represents the flow rate of condensate entering the deaerator, F out (t) represents the flow rate of condensate flowing out of the deaerator, and A represents the cross-sectional area of ​​the deaerator's water tank.

[0019] In some embodiments, the real-time prediction and control decision module uses a PID control algorithm to control and predict the valve opening. The opening control output of the PID control algorithm is shown in formula (4):

[0020]

[0021] In the formula:

[0022] e(t) is the difference between the target water level and the actual water level at time t, K p ,K i ,K d Here are the proportional, integral, and derivative gain coefficients, ucontrol(t) is the control output at time t, and e(t′) is the error signal at time t′.

[0023] In some embodiments, the real-time prediction and control decision module uses a model predictive control method to control and predict valve opening. The model predictive control method establishes a model of the condensate regulating valve of the nuclear power plant, and predicts the performance index of the system response for a future period through the model in each control cycle. The model predictive control method uses the minimum error between the predicted system response performance index and the target performance index as the objective function to solve the valve opening control method; the specific objective function is as shown in formula (5):

[0024]

[0025] In the formula, L(t+k) is the deaerator water level at the future time t+k, and L refLet t be the target water level, Δu(t+k) be the change in valve opening at time t+k in the future, l1 and l2 be weighting coefficients used to balance the accuracy of water level control and the stability of valve adjustment, and N be the predicted time length.

[0026] In some embodiments, the model predictive control method establishes a condensate regulating valve model for a nuclear power plant based on the water level change equation and the flow characteristics of the regulating valve, using condensate flow rate, regulating valve opening, regulating valve pressure, and regulating valve temperature as influencing factors; the model predictive control method predicts the future system response for a period of 10 to 20 seconds; the performance indicators of the system response predicted by the model predictive control method are the fluctuations in deaerator water level and condensate flow rate.

[0027] In some embodiments, the anomaly detection and safety protection module sets a safety threshold for a parameter and compares it with the parameter acquired by the data acquisition and synchronization module, and implements an alarm strategy based on the comparison result. The alarm strategy includes issuing a warning to the operator when the parameter approaches the boundary of the safety threshold; issuing a warning to the operator when the parameter continuously deviates from the normal range and exceeds the predetermined tolerance; adjusting the control strategy of the regulating valve and activating redundant equipment or switching to a backup channel in the nuclear power plant to ensure the operation of the nuclear power plant's condensate system; and entering an emergency protection phase when the parameter shows excessive valve opening, excessive flow, or abnormal water level, switching the system to a safe mode and adopting safety strategies such as starting a backup valve or bypass pump, conservative regulating valve adjustment strategy, and forced valve closure or shutdown. The conservative regulating valve adjustment strategy specifically involves reducing the amplitude of valve opening changes and reducing the frequency of valve adjustment.

[0028] In some embodiments, the system also includes a visualization and alarm management module, which provides a real-time monitoring interface for the system. The visualization and alarm management module displays the parameters acquired by the sensor data acquisition and synchronization module and the predicted parameters of the real-time prediction and control decision module through line graphs, bar charts and dashboards. The visualization and alarm management module also displays the alarm strategies and alarm status of the anomaly detection and safety protection module.

[0029] The implementation of this invention has the following beneficial effects:

[0030] This invention aims to provide an automatic control method and system for condensate regulating valves in nuclear power plants based on reinforcement learning algorithms. This system, by combining reinforcement learning with a dynamic water level control model, solves the problems of unstable valve opening regulation, nonlinear flow changes, and response lag during peak shaving processes in traditional regulation methods under complex operating conditions. The system can learn and optimize valve regulation strategies in real time while avoiding valve oscillations and unreasonable adjustments in traditional methods, thus improving the accuracy and stability of condensate flow regulation. This method is applicable to online regulation and automatic control in nuclear power plants, effectively improving system safety and economy, reducing manual intervention, and ensuring stable control of deaerator water level under various operating conditions. Attached Figure Description

[0031] Figure 1 This is a flowchart of an automatic control system for a condensate regulating valve in a nuclear power plant, as proposed in an embodiment of the present invention. Detailed Implementation

[0032] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] The condensate system of a nuclear power plant is primarily responsible for collecting the water condensed from the exhaust steam of the turbines. This condensate, after purification and treatment, is then transported to the deaerator water tank. The deaerator water tank is a crucial container in the nuclear power plant's thermal system, requiring a large amount of water to maintain normal operation. Condensate, as the primary water source for the deaerator water tank, ensures that the tank has sufficient water for deaeration operations.

[0034] like Figure 1 As shown, this invention proposes an automatic control system for condensate regulating valves in nuclear power plants. The system includes a sensor data acquisition and synchronization module, a reinforcement learning control module, a water level dynamic modeling module, a real-time prediction and control decision module, an anomaly detection and safety protection module, and a visualization and alarm management module.

[0035] The sensor data acquisition and synchronization module serves as the system's data entry point. This module acquires and synchronizes various parameters in the nuclear power plant's condensate system in real time, including condensate flow rate, deaerator water level, regulating valve opening, pressure, and temperature. Sensors (such as flow meters, pressure sensors, and temperature sensors) are installed at key pipeline nodes and regulating valve locations. The sensor data acquisition and synchronization module performs high-frequency acquisition of these critical parameters to ensure real-time performance and data integrity. For data transmission, the module uses industrial Ethernet or fieldbus protocols to ensure stability and data confidentiality in a high-security environment.

[0036] To improve data acquisition accuracy, the sensor data acquisition and synchronization module selects to collect key signals such as condensate flow rate and valve opening at high frequencies (e.g., above 1Hz), and can dynamically adjust the sampling frequency according to different operating conditions. When the system detects abnormal operating conditions (e.g., valve vibration or abnormal pressure), it automatically increases the sampling frequency to more accurately capture early signs of faults. In addition, the system is equipped with redundant sensors for critical equipment to ensure automatic switching to backup sensors in the event of sensor failure or malfunction, and the fault information is recorded in the log.

[0037] The reinforcement learning control module is the core computational component of the system, responsible for optimizing the valve's regulation strategy based on real-time data input. This module uses reinforcement learning algorithms (such as Deep Q-Network, DQN) to train the system's decision-making process, selecting the optimal valve opening control strategy. The reinforcement learning control module learns the strategy given by formula Q to control the regulating valve. The specific construction of formula Q is as follows:

[0038] 1. Definition of state space

[0039] Assuming there are three control valves in the system (LCA20AA201, LCA20AA202, LCA20AA203), then the state space s t This includes the deaerator water level L(t), condensate flow rate, the valve openings of the three regulating valves u(t) = [u1(t), u2(t), u3(t)], and other key parameters of the three regulating valves (such as pressure, temperature, etc.). The definition of the state space can be extended according to actual needs, where t is time.

[0040] 2. Definition of action space

[0041] Action space a t This indicates the adjustment of the opening degree of each control valve. The opening degrees of the three control valves are u1(t), u2(t), and u3(t), and each action a... tThis represents the increment or absolute value of the valve opening (which can be a discrete step adjustment or a continuous adjustment), where t is time.

[0042] 3. Design of the reward function

[0043] reward function r t The reward function is used to measure the system's performance and aims to optimize the regulation strategy by maximizing the reward. Based on water level stability, valve opening smoothness, and avoidance of nonlinear regions, the reward function is designed as follows (1):

[0044]

[0045] In the formula:

[0046] L t L represents the deaerator water level at time t. ref The target water level; The ideal valve opening at time t; ||(u i (t)∈Unfavorable interval) is an indicator function for whether the valve opening at time t is in the unfavorable interval (e.g., the valve opening is in the unfavorable interval if it is between 43% and 47%); a, b, and g are weighting coefficients, which are obtained through experimental optimization.

[0047] 4. Q. Learn and update formulas

[0048] The Q-learning algorithm is used for optimization, with the goal of continuously updating the state-action value function Q(s). t ,a t The optimal strategy is found by using Q-learning. The update formula for Q-learning is as follows: Formula (2):

[0049] Q(s t ,a t )←Q(s t ,a t )+h[t t +g maxQ(s t+1 ,a')-Q(s t ,a t (2)

[0050] In the formula:

[0051] Q is the learning formula, t t For immediate reward, that is, r at time t calculated by the reward function in formula (1). t g is the discount factor, h is the learning rate, and s is the learning rate. t Let s be the state space at time t. t+1 Let a be the state space at time t next time step. t Let be the action space at time t, and let a′ represent the optimal action at the next time step t+1.

[0052] 5. Strategy Output

[0053] After the reinforcement learning model is trained, the control strategy of the regulating valve, which is continuously updated according to the learning formula Q, is used to adjust the opening of the regulating valve in real time to achieve precise control of the water level and stable operation of the valve.

[0054] The water level dynamic modeling module is responsible for simulating the water level changes in the deaerator tank using a mathematical model and comparing and correcting it with actual operating data. This module uses the water level dynamic equation for calculation as shown in formula (3):

[0055]

[0056] In the formula, L(t) is the deaerator water level, and F in (t) represents the flow rate of condensate entering the deaerator, F out (t) represents the condensate flow rate out of the deaerator, and A is the cross-sectional area of ​​the deaerator's water tank. This equation describes changes in water level, taking into account the effects of condensate flow rate and outflow rate, and monitors water level changes in real time.

[0057] The real-time prediction and control decision module performs real-time prediction and control based on the regulation strategy generated by the reinforcement learning control module and the water level dynamic model.

[0058] This module adjusts and predicts the valve opening using a PID control algorithm or model predictive control (MPC). It can smoothly adjust the flow rate during valve regulation and implement safety redundancy measures to control the valve in case of anomalies. Specifically, the module uses a PID control algorithm to control the valve opening, as follows: the PID control algorithm is a proportional-integral-derivative control algorithm, and the control output of the PID control algorithm is determined by the following formula (4):

[0059]

[0060] In the formula:

[0061] e(t) represents the water level error at time t (i.e., the difference between the target water level and the actual water level); K p ,K i ,K d The proportional, integral, and derivative gain coefficients are determined through system debugging. ucontrol(t) is the control output at the current time t, i.e., the result of the PID controller output (used to adjust the valve opening), and e(t′) is the error signal at time t′.

[0062] This module employs Model Predictive Control (MPC) for valve opening control. Specifically, MPC predicts valve actions over a future period by solving an optimization problem, ensuring optimal system performance in the future, and is particularly suitable for rapidly changing operating conditions. MPC calculates the optimal valve opening sequence using an optimization problem solver (such as Quadratically Constrained Programming QCP). The basic framework of MPC can be summarized in the following steps:

[0063] System modeling: Modeling the future behavior of the system through its dynamic equations. In the condensate control system of a nuclear power plant, the dynamic model is usually established based on the water level change equation and the flow characteristics of the control valve, taking into account the changes in condensate flow rate, valve opening, and other influencing factors (such as pressure, temperature, etc.).

[0064] Predicting future behavior: In each control cycle, MPC uses a system model to predict the performance metrics of the system response over a future period (called the prediction time domain, typically 10–20 seconds). This process considers the current state of the system and possible changes in the control input to predict the future trend of the system's response.

[0065] Optimization problem solving: MPC uses an optimization algorithm to minimize the performance index error between the system response and the target in the future prediction time domain as the objective function. The performance index error can be the deaerator water level error, condensate flow fluctuation, etc., to solve a set of control decisions (i.e., valve opening control methods). The optimization algorithm for the optimization problem usually includes constraints, such as the physical opening limit of the valve, the flow rate change limit, and the water level safety range limit. The objective function of MPC can be expressed by the following formula (5):

[0066]

[0067] In the formula: L(t+k) is the deaerator water level at the future time t+k, L ref Let t be the target water level, Δu(t+k) be the change in valve opening at time t+k in the future, l1 and l2 be weighting coefficients used to balance the accuracy of water level control and the stability of valve adjustment, and N be the length of the prediction time domain.

[0068] The anomaly detection and safety protection module is a crucial component of the automatic control system for condensate regulating valves in nuclear power plants. It is responsible for real-time monitoring of various critical parameters during system operation and timely detection of potential anomalies. This module reads parameters collected by the sensor data acquisition and synchronization module, ensuring the system maintains stability and safety under complex and variable operating conditions by monitoring multiple variables such as water level, flow rate, valve opening, pressure, and temperature. Its primary purpose is to respond rapidly upon detecting anomalies, preventing unsafe or unstable situations, avoiding potential equipment damage or system failure, and maximizing the protection of the safe and stable operation of the nuclear power plant.

[0069] The core of anomaly detection is the real-time analysis and comparison of various system parameters. When the system enters normal operating condition, the changes between various parameters follow certain patterns and expected ranges. Based on this, the module compares real-time data with historical data by setting safety thresholds for parameters such as valve opening, water level, and flow rate to assess whether the current operating conditions are consistent with normal operation. For example, drastic fluctuations in water level, excessive adjustment of valve opening, or rapid changes in flow rate may all be abnormal manifestations, and the system will analyze these changes according to set rules.

[0070] When the system detects that the valve control deviation exceeds the preset range or a sudden change occurs, the anomaly detection and safety protection module will immediately activate the alarm mechanism and take corresponding safety protection measures. Alarms are divided into multiple levels, with the specific level depending on the severity of the anomaly and its impact on system operational safety.

[0071] (1) Early warning stage

[0072] During system operation, if certain parameters (such as water level, flow rate, valve opening, etc.) approach anomaly thresholds (i.e., reach the boundary of safety thresholds), the anomaly detection and safety protection module will issue a warning to the operator. Anomalies at the warning level are usually characterized by significant parameter changes, but not yet reaching a level that endangers system safety. In this case, the system will issue an alarm to the operator through the user interface or dashboard, indicating that further monitoring or preventative measures may be necessary. At this point, the operator can manually adjust valve openings, monitor water levels, or check equipment operating status to prevent the problem from escalating further.

[0073] (2) Warning stage

[0074] If parameters continuously deviate from the normal range and exceed the predetermined tolerance, the anomaly detection and safety protection module will issue a warning to the operator. This tolerance can be a quantitative range, set as a combination of two dimensions: deviation from the normal range (±5%) and duration (e.g., exceeding 10 seconds). Triggering either of these dimensions is considered exceeding the predetermined tolerance, which is used to prevent overreaction to occasional minor disturbances. In this situation, in addition to triggering an alarm to notify the operator, the system will automatically perform emergency procedures, such as slowing down the rate of valve opening changes to prevent further instability caused by frequent valve operations. Simultaneously, the system will activate redundant equipment or switch to a backup channel to ensure the operation of the nuclear power plant's condensate system. During this phase, the automatic control system will adjust the valve control algorithm according to the set safety strategy, gradually reducing control deviations and minimizing risks to system safety.

[0075] (3) Emergency Protection Phase

[0076] Upon detecting extremely serious anomalies, such as excessive valve opening, excessive flow, or abnormal water level, the anomaly detection and safety protection module will enter the emergency protection phase. At this time, the module will immediately trigger the advanced protection mechanism, automatically switching to safety mode. Safety mode refers to a controlled and restricted operating state that the system automatically enters when it detects extremely serious anomalies to ensure the safety of nuclear power plant equipment and personnel. In this mode, the system prioritizes safety, restricts unnecessary operations, and takes conservative or shutdown measures to prevent the accident from escalating or equipment damage. The following protective measures will be implemented:

[0077] Activate backup valves or bypass system: The system will immediately start backup valves or bypass pumps to share the load of the upstream valves, reduce the working pressure of abnormal valves, and thus avoid equipment failure or damage.

[0078] Adjusting the control strategy: In an emergency, the system may switch to a more conservative control strategy, such as reducing the range of valve opening changes and reducing the frequency of valve adjustments, to avoid over-adjustment or oscillation in the system.

[0079] Forced valve closure or shutdown: If an anomaly exceeds the safety tolerance range and cannot be adjusted to a safe level, the system will forcibly close the valve or shut down the equipment to protect the safety of the equipment and personnel. This process is usually accompanied by safety warnings and emergency notifications, and operators can take further action based on the system prompts.

[0080] The visualization and alarm management module is the front-end display of the system, providing a real-time monitoring interface to help operators understand the current system status and predicted results. Through the graphical interface, operators can monitor key parameters such as valve opening, flow rate, and water level in real time, and receive timely alarm notifications when the system detects anomalies.

[0081] The system displays key parameters such as valve opening, actual flow rate, predicted flow rate, and pressure difference in real time through line charts, bar charts, and dashboards. For sudden changes in flow rate or water level exceeding limits, the charts will highlight these changes with color changes and warning signs to help operators respond promptly.

[0082] The system is designed with a multi-level alarm mechanism. Based on the magnitude and duration of the error, alarms are divided into early warning, warning, and emergency levels, and corresponding emergency measures are automatically triggered to ensure that the system can take timely action in abnormal situations to prevent accidents. When the system's anomaly detection and safety protection module triggers an alarm, the visualization and alarm management module will display the alarm status and level.

[0083] Through the coordinated operation of the above modules, this system can accurately adjust valve opening and predict and respond to flow changes in real time in the complex and highly safety-required environment of nuclear power plants, significantly improving the safety and stability of the overall system.

[0084] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. An automatic control system for condensate regulating valves in a nuclear power plant, characterized in that, The system includes a sensor data acquisition and synchronization module, a reinforcement learning control module, a real-time prediction and control decision module, and an anomaly detection and safety protection module. The sensor data acquisition and synchronization module acquires and synchronizes parameters in the condensate system of the nuclear power plant. The reinforcement learning control module reads the parameters from the sensor data acquisition and synchronization module and uses a reinforcement learning algorithm to generate a control strategy for the regulating valve. The real-time prediction and control decision module controls and predicts the opening degree of the condensate regulating valve according to the regulation strategy of the reinforcement learning control module. The anomaly detection and safety protection module reads the parameters acquired by the data acquisition and synchronization module, monitors the parameters, and issues an alarm for anomaly monitoring results.

2. The automatic control system for a condensate regulating valve in a nuclear power plant according to claim 1, characterized in that, The system installs sensors at pipeline nodes and regulating valve locations in the nuclear power plant condensate system, and redundant sensors are configured. The sensor data acquisition and synchronization module collects the parameters of the sensors, and transmits the data via industrial Ethernet or fieldbus protocols.

3. The automatic control system for a condensate regulating valve in a nuclear power plant according to claim 2, characterized in that, The sensor data acquisition and synchronization module collects condensate flow rate, deaerator water level, regulating valve opening, regulating valve pressure, and regulating valve temperature. The sensor data acquisition and synchronization module collects condensate flow rate and regulating valve opening at a frequency of 1 Hz or higher.

4. The automatic control system for a condensate regulating valve in a nuclear power plant according to claim 3, characterized in that, The reinforcement learning control module generates a control strategy for the regulating valve by learning formula Q. The construction of the learning formula Q includes defining a state space, defining an action space, designing a reward function, and updating the learning formula. The defined state space includes a state space s. t This includes the deaerator water level, condensate flow rate, valve opening of the three regulating valves, temperature, and pressure at time t; the defined action space includes action space a. t This indicates the adjustment of the opening of three regulating valves for time t, with each action space a. t This represents the increment of the opening of the three valves; the reward function is designed as shown in formula (1): In the formula: L t L represents the deaerator water level at time t. ref The target water level for the deaerator. Let i be the ideal valve opening degree of the regulating valve at time t; ||(u i (t)∈Unfavorable interval) is an indicator function indicating whether the valve opening at time t is in the unfavorable interval; a, b, g are weighting coefficients; Learn the formula and update the formula as shown in formula (2): Q(s t ,a t )←Q(s t ,a t )+h[t t +gmaxQ(s t+1 ,a')-Q(s t ,a t )] (2) In the formula: Q is the learning formula, t t For immediate reward, that is, r at time t calculated by the reward function in formula (1). t g is the discount factor, h is the learning rate, and s is the learning rate. t Let s be the state space at time t. t+1 Let a be the state space at time t next time step. t Let be the action space at time t, and let a′ represent the optimal action at the next time step t+1.

5. The automatic control system for a condensate regulating valve in a nuclear power plant according to claim 1, characterized in that, The system also includes a water level dynamic modeling module, which simulates the water level changes in the deaerator water tank and compares and corrects them with the water level data of the deaerator water tank collected by the sensor data acquisition and synchronization module. The water level dynamic equation of the deaerator water tank simulated by the water level dynamic modeling module is as shown in formula (3): In the formula, L(t) is the deaerator water level, and F in (t) represents the flow rate of condensate entering the deaerator, F out (t) represents the flow rate of condensate flowing out of the deaerator, and A represents the cross-sectional area of ​​the deaerator's water tank.

6. The automatic control system for a condensate regulating valve in a nuclear power plant according to claim 4, characterized in that, The real-time prediction and control decision module uses a PID control algorithm to control and predict the valve opening. The opening control output of the PID control algorithm is shown in formula (4): In the formula: e(t) is the difference between the target water level and the actual water level at time t, K p ,K i ,K d Here are the proportional, integral, and derivative gain coefficients, ucontrol(t) is the control output at time t, and e(t′) is the error signal at time t′.

7. The automatic control system for a condensate regulating valve in a nuclear power plant according to claim 4, characterized in that, The real-time prediction and control decision module uses a model predictive control method to control and predict valve opening. This method establishes a model of the condensate regulating valve in a nuclear power plant. In each control cycle, the model predicts the system response performance indicators for a future period. The objective function of this method is to minimize the error between the predicted system response performance indicators and the target performance indicators, thus solving for the valve opening control method. The specific objective function is shown in formula (5): In the formula, L(t+k) is the deaerator water level at the future time t+k, and L ref Let t be the target water level, Δu(t+k) be the change in valve opening at time t+k in the future, l1 and l2 be weighting coefficients used to balance the accuracy of water level control and the stability of valve adjustment, and N be the predicted time length.

8. An automatic control system for a condensate regulating valve in a nuclear power plant according to claim 7, characterized in that, The model predictive control method is based on the water level change equation and the flow characteristics of the regulating valve. It establishes a condensate regulating valve model for a nuclear power plant using condensate flow rate, regulating valve opening, regulating valve pressure, and regulating valve temperature as influencing factors. The model predictive control method predicts the future system response for a period of 10 to 20 seconds. The performance indicators of the system response predicted by the model predictive control method are the fluctuations in deaerator water level and condensate flow rate.

9. An automatic control system for a condensate regulating valve in a nuclear power plant according to claim 1, characterized in that, The anomaly detection and safety protection module sets a safety threshold for the parameter and compares it with the parameter collected by the data acquisition and synchronization module. Based on the comparison result, an alarm strategy is implemented. The alarm strategy includes issuing a warning to the operator when the parameter approaches the boundary of the safety threshold; issuing a warning to the operator when the parameter continuously deviates from the normal range and exceeds the predetermined tolerance; adjusting the control strategy of the regulating valve and activating redundant equipment or switching to a backup channel in the nuclear power plant to ensure the operation of the nuclear power plant's condensate system; and entering an emergency protection phase when the parameter indicates excessive valve opening, excessive flow, or abnormal water level. The system switches to a safe mode and adopts safety strategies such as starting a backup valve or bypass pump, conservative regulating valve adjustment strategy, and forced valve closure or shutdown. The conservative regulating valve adjustment strategy specifically involves reducing the amplitude of valve opening changes and reducing the frequency of valve adjustments.

10. An automatic control system for a condensate regulating valve in a nuclear power plant according to claim 9, characterized in that, The system also includes a visualization and alarm management module, which provides a real-time monitoring interface for the system. The visualization and alarm management module displays the parameters acquired by the sensor data acquisition and synchronization module and the predicted parameters of the real-time prediction and control decision module through line charts, bar charts, and dashboards. The visualization and alarm management module also displays the alarm strategies and alarm status of the anomaly detection and safety protection module.

Images