Parallel double-agent water level control method and system for steam generator

Abstract

The application discloses a kind of steam generator parallel double intelligent agent water level control method and system, belong to nuclear power plant instrumentation and artificial intelligence technical field.The method establishes steam generator simulation model, constructs including water level PID controller and flow PI controller Parallel double-channel basic control framework;Water level intelligent agent and flow intelligent agent are constructed and are independently trained respectively, so that water level intelligent agent learns water level regression and false water level suppression strategy, so that flow intelligent agent learns water supply flow regulation and flow smooth control strategy;Two intelligent agents after training are deployed to Parallel double-channel basic control framework, respectively output water level PID parameter increment and flow PI parameter increment, update corresponding controller parameter, and the output of two channels is added to generate water supply flow regulation instruction, to control steam generator water level.The application can improve water level control response speed and stability.

Description

Technical Field

[0001] This invention belongs to the field of interdisciplinary technology of nuclear power plant instrumentation and control and artificial intelligence, specifically relating to a method and system for parallel dual-intelligent agent water level control of a steam generator. Background Technology

[0002] Steam generator water level control in nuclear power plants is one of the key systems for ensuring the safe operation of nuclear power units. Due to the severe "false water level" phenomenon (i.e., the non-minimum phase characteristic of water level rising and then falling when the load increases), large hysteresis, strong nonlinearity, and time-varying characteristics of steam generators, traditional control strategies face many challenges.

[0003] The fundamental challenge of false water level phenomena: Steam generator water level control has long faced the challenge of "false water level" phenomena, where the water level signal measured by the water level sensor deviates from the actual water volume, leading to measurement errors. This phenomenon mainly stems from fluid dynamics coupling, time delay effects, and the nonlinear characteristics of the system.

[0004] 1) Complex mechanism: When the feedwater flow rate decreases, the natural circulation driving force inside the steam generator decreases, causing the water level to drop in the descending section, resulting in the so-called "water level contraction"; while under low power or high load conditions, changes in steam flow rate will trigger the phenomenon of "water level expansion". These transient processes cause the water level signal to rise and then fall or fall and then rise in a short period of time, exhibiting the non-minimum phase characteristic of water level rising and then falling when the load increases.

[0005] 2) Time Delay and Hysteresis: The steam generator has a complex structure and exhibits significant thermal inertia and hydraulic time delay. During water level deviation adjustment, the system response speed often lags behind the actual disturbance, making it difficult for the adjustment signal to compensate for the error of the false water level in a timely manner.

[0006] 3) Nonlinearity and time-varying characteristics: The heat transfer efficiency and natural circulation resistance of the steam generator vary significantly with the power level, and the system exhibits strong nonlinearity and time-varying characteristics, which further increases the difficulty of controller design.

[0007] Currently, the control strategies mainly used in industry (such as three-impulse control) are no longer sufficient to meet the dual requirements of safety and economy for modern nuclear power units, mainly manifested in the following aspects of pain points: 1) Inherent defects of cascade control structures: Traditional three-impulse control typically employs a cascade structure with an outer loop for water level and an inner loop for feedwater flow. While cascade control can suppress disturbances to some extent, its adjustment speed is limited by the bandwidth of the inner loop (flow control). When the outer loop (e.g., power load) experiences significant disturbances or false water levels, the outer loop response is limited, resulting in system response lag and limited adjustment accuracy. Furthermore, the cascade structure leads to complex parameter tuning, severe coupling between the inner and outer loops, and difficulty in balancing speed and stability.

[0008] 2) Limitations of Single-Agent Control: With the introduction of artificial intelligence technology, although controllers based on deep reinforcement learning (DRL) possess self-learning capabilities, fundamental conflicts still exist in water level and flow control tasks. A single agent struggles to simultaneously ensure the long-term macroscopic stability of water levels and the rapid microscopic tracking of water flow. Attempting to learn all features with a single network often results in poor policy generalization ability, making it difficult to achieve optimal performance under complex operating conditions (such as sudden load changes or pump station failures).

[0009] 3) Multi-agent collaboration challenge: Schemes that introduce multi-agents (such as MADDPG, parallel DRL) for control usually rely on complex explicit communication mechanisms or shared reward functions for joint training. However, they are difficult to converge in industrial simulation environments such as Simulink, have extremely high computational costs, and are difficult to implement.

[0010] In summary, existing technologies urgently need to overcome the following bottlenecks: 1) Decoupling control structure: How to break the strong coupling between water level and flow control, and achieve physical or logical decoupling between the two, thereby reducing system complexity.

[0011] 2) Independent training and cooperative execution: How to achieve independent training of multiple agents (reducing convergence difficulty and computational cost) and avoid conflicts through effective cooperative mechanisms (such as constraint-based allocation and priority scheduling) to achieve the safety of parallel control.

[0012] 3) Engineering verification: How to achieve rapid and stable verification in industrial standard simulation environments such as Simulink, and provide clear safety boundaries and fault tolerance mechanisms to meet the engineering implementation requirements of nuclear power plants. Summary of the Invention

[0013] The technical problem to be solved by this invention is to address the shortcomings of the prior art by providing a parallel dual-agent water level control method and system for steam generators. Under a parallel PID control architecture, two independently trained Deep Deterministic Policy Gradient (DDPG) agents are used to adaptively tune parameters for the macroscopic trend of water level and the microscopic dynamics of flow, respectively. The final control command is output by summing the results. This invention addresses the technical problems in existing nuclear power plant steam generator water level control, such as obvious false water level phenomena, large system lag and strong nonlinear time-varying behavior, limitations of traditional cascade control due to inner loop bandwidth, difficulty for a single agent to simultaneously achieve macroscopic water level stability and microscopic flow stability, and difficulties in convergence and complex engineering deployment of multi-agent joint training. The invention allows the water level agent and the flow agent to be trained completely independently in the Simulink environment without interfering with each other, while complementing each other in overall water level regulation and local flow suppression. This effectively overcomes the false water level phenomenon and improves the robustness and response speed of the system over a wide power range.

[0014] The present invention adopts the following technical solution: A method for controlling the water level of a steam generator in parallel with two intelligent agents includes the following steps: S1. Establish a steam generator simulation model, and construct a parallel dual-channel basic control framework based on the steam generator simulation model. The parallel dual-channel basic control framework includes a water level control channel and a feedwater flow control channel. The water level control channel includes a water level PID controller, and the feedwater flow control channel includes a flow PI controller. S2. Construct a parallel dual-agent architecture, which includes a water level agent and a flow rate agent. The water level agent is used to tune the control parameters of the water level PID controller, and the flow rate agent is used to tune the control parameters of the flow rate PI controller. S3. The water level agent is independently trained in a simulation environment, and a reward function is set for the water level regression speed and false water level suppression, so that the water level agent learns the water level control strategy. S4. The flow agent is independently trained in a simulation environment, and a reward function for the stability of water supply flow is set so that the flow agent learns the flow control strategy. S5. Deploy the trained water level agent and flow rate agent into the parallel dual-channel basic control framework. The water level agent outputs the increment of the water level PID parameter, and the flow rate agent outputs the increment of the flow rate PI parameter. S6. Update the control parameters of the water level PID controller incrementally according to the water level PID parameter, and update the control parameters of the flow rate PI controller incrementally according to the flow rate PI parameter. Sum the output of the water level control channel and the output of the water supply flow control channel to generate a water supply flow regulation command, and control the water level of the steam generator according to the water supply flow regulation command.

[0015] Preferably, in step S1, the steam generator simulation model is a linear variable parameter steam generator simulation model; the linear variable parameter steam generator simulation model is established in the following way: A basic model of a steam generator at different power points was established using the Irving transfer function model set. The basic models of the steam generator at different power points were then weighted and fused using the triangular membership function to form a steam generator simulation model covering the power range of 5% to 100%.

[0016] Preferably, in step S1, the water level PID controller calculates the output of the water level control channel based on the water level deviation, the flow rate PI controller calculates the output of the water supply flow control channel based on the steam-water compatibility deviation, and the initial parameters of the water level PID controller and the flow rate PI controller are obtained using the Ziegler-Nichols tuning method.

[0017] Preferably, the water level agent is a DDPG agent, and the state space of the water level agent includes water level error, water level change rate, normalized power, water supply flow rate and water supply flow rate change rate, and the action space is the water level PID parameter increment, which includes the proportional parameter increment, integral parameter increment and derivative parameter increment.

[0018] Preferably, the flow agent is a DDPG agent, and the state space of the flow agent includes the steam-water adaptation error, the rate of change of water flow, the normalized power, the water level and the rate of change of water level, and the action space is the flow PI parameter increment, which includes the proportional parameter increment and the integral parameter increment.

[0019] Preferably, in step S3, independently training the water level agent includes: A simplified water level control model is constructed in a simulation environment. The parameters of the flow PI controller are fixed, and it is assumed that the water flow control channel can track the set value, so that the water level agent learns water level regression and false water level suppression strategies.

[0020] Preferably, in step S4, independently training the traffic agent includes: In the simulation environment, the water level setpoint is fixed, the parameters of the water level PID controller are kept constant, and a load step change is injected into the steam generator simulation model, so that the flow agent learns the nonlinear characteristics of water flow regulation and the flow stabilization control strategy.

[0021] Preferably, after steps S3 and S4, a training evaluation step is also included: If the water level agent or flow rate agent reaches the convergence threshold within the preset maximum number of iterations, and the performance index variance is less than the variance threshold in N consecutive verification rounds, then the training is considered successful and the network parameters of the corresponding agent are saved; if the preset maximum number of iterations is reached but the agent still fails to converge, or the performance fluctuation exceeds the preset range, then the training is considered to have failed and the retraining mechanism is triggered.

[0022] Preferably, the retraining mechanism includes: The Actor-Critic neural network weights of the agent that was determined to have failed training are restored to their initial state. The exploration noise parameters of the DDPG algorithm are reset, historical data of the simulation environment is cleared and perturbation samples are re-injected, and then independent training is re-executed for the agent that was determined to have failed training.

[0023] Secondly, embodiments of the present invention provide a parallel dual-agent water level control system for a steam generator, used to execute the aforementioned parallel dual-agent water level control method for a steam generator, including: The model building module is used to build the simulation model of the steam generator; The basic control module is used to construct the parallel dual-channel basic control framework based on the steam generator simulation model. The parallel dual-channel basic control framework includes a water level control channel and a feedwater flow control channel. A dual-agent training module includes a water level agent training unit and a flow rate agent training unit. The water level agent training unit is used to train the water level agent independently, and the flow rate agent training unit is used to train the flow rate agent independently. The evaluation and retraining module is used to determine whether the water level agent and the flow rate agent meet the convergence conditions, and to trigger the retraining mechanism when the convergence conditions are not met. The collaborative control module is used to deploy the trained water level agent and flow rate agent into the parallel dual-channel basic control framework. It updates the corresponding control parameters according to the increment of the water level PID parameter and the increment of the flow rate PI parameter, and sums the output of the water level control channel and the output of the water supply flow control channel to generate a water supply flow regulation command.

[0024] Thirdly, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for parallel dual-agent water level control of a steam generator.

[0025] Fourthly, embodiments of the present invention provide a computer-readable storage medium including a computer program, which, when executed by a processor, implements the steps of the above-described method for parallel dual-agent water level control of a steam generator.

[0026] Fifthly, a chip includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the above-described parallel dual-agent water level control method for a steam generator.

[0027] In a sixth aspect, embodiments of the present invention provide an electronic device including a computer program, wherein when the computer program is executed by the electronic device, it implements the steps of the above-described method for parallel dual-smart agent water level control of a steam generator.

[0028] Compared with the prior art, the present invention has at least the following beneficial effects: A parallel dual-agent water level control method for a steam generator is proposed. This method involves establishing a steam generator simulation model, constructing a parallel dual-channel basic control framework, training a water level agent and a flow rate agent separately, and then fusing their outputs to form a feedwater flow rate regulation command, thus creating a complete parallel dual-agent water level control process. The method decouples the macroscopic water level regulation task from the microscopic feedwater flow rate regulation task in steam generator water level control. The water level agent focuses on handling false water levels, large lag, and water level regression problems, while the flow rate agent focuses on handling feedwater flow rate fluctuations and high-frequency disturbances. Deploying the two agents in parallel avoids the problem of the outer loop being limited by the inner loop bandwidth in traditional cascade control, and also avoids policy conflicts caused by a single agent simultaneously learning both water level and flow rate objectives. Finally, the control command is formed by summing the outputs, enabling the two channels to coordinate at the execution end, which is beneficial for improving the response speed, stability, and robustness of water level control.

[0029] Furthermore, it accurately characterizes the nonlinear and time-varying characteristics of the steam generator over a wide power operating range. The heat transfer efficiency, natural circulation resistance, spurious water level characteristics, and dynamic response speed of the steam generator differ under low, medium, and full power conditions, making it difficult for a single fixed-parameter model to cover all operating conditions. By weighted fusion of Irving models at different power points, the corresponding model parameters can be obtained in real time based on the current power, making the simulation environment closer to the real-world object. This provides a training foundation with wide power and variable operating conditions for subsequent agent training, which is beneficial for improving the generalization ability and engineering applicability of the agent strategy.

[0030] Furthermore, PID / PI controllers are mature, interpretable, and easily verifiable control structures in nuclear power plant control systems. Usable initial parameters can be quickly obtained through Ziegler-Nichols tuning. The intelligent agent does not directly replace the traditional controller; instead, it outputs incremental values ​​based on the basic parameters for adaptive tuning. This retains the stability fallback capability of the traditional controller while utilizing reinforcement learning to improve dynamic performance under complex operating conditions. This helps reduce the safety risks associated with direct intelligent control takeover and aligns with the conservative and verifiable engineering requirements of nuclear power plant control systems.

[0031] Furthermore, this enables the water level agent to establish control strategies based on the macroscopic dynamics of the water level. Water level error and rate of change reflect the current degree of deviation and its trend; normalized power reflects the dynamic differences of the system under different power levels; and feedwater flow rate and its rate of change reflect important driving factors of water level changes. DDPG is suitable for handling continuous action spaces and can directly output the increments of PID proportional, integral, and derivative parameters, achieving continuous adaptive tuning. By focusing the water level agent on adjusting the water level PID parameters, its ability to learn from spurious water levels, hysteresis responses, and water level regression rates is enhanced, avoiding the reduction in strategy effectiveness caused by it simultaneously undertaking the task of local flow stability control.

[0032] Furthermore, the flow agent is specifically designed to learn the local regulation patterns of feedwater flow. The steam-water matching error reflects the degree of matching between feedwater and steam flow, the feedwater flow rate change reflects flow fluctuations and valve action trends, normalized power adapts to different power conditions, and water level and its rate of change provide a macroscopic state reference. The continuous output of the DDPG is suitable for incremental PI parameter tuning, enabling adaptive adjustment of flow channel parameters according to operating conditions. By limiting the flow agent to handling flow stability and suppressing high-frequency disturbances, the burden on the water level agent is reduced, and it acts as a stabilizer in parallel control, mitigating the risk of drastic fluctuations in feedwater regulation commands.

[0033] Furthermore, the water level control task is decoupled from the local dynamics of flow, reducing training complexity. False water levels in steam generators manifest as short-term discrepancies between the water level measurement signal and the actual water volume. If trained in a fully coupled system, the agent is easily affected by flow fluctuations, controller coupling, and power changes simultaneously, leading to unclear learning objectives. By simplifying the water level control model and assuming ideal flow channel tracking, the water level agent can focus on learning the mapping relationship between steam flow changes and water level response, improving its ability to correctly adjust the direction and speed of water level return during false water level periods.

[0034] Furthermore, the flow control task is separated from the macroscopic water level regulation, allowing the flow agent to focus on local rapid dynamics. The feedwater flow channel needs to respond quickly to load changes while avoiding excessive valve movement, high-frequency flow oscillations, or overshoot. Joint training with a water level agent would increase convergence difficulty due to reward conflicts and state coupling. By fixing the water level setpoint and water level PID parameters, the flow agent can learn the nonlinear flow regulation law under load steps in a relatively independent training environment, improving the smoothness and disturbance rejection capability of the feedwater flow output, and providing stable flow channel support for final parallel coordinated control.

[0035] Furthermore, the success of training is determined by the maximum number of iterations, the convergence threshold, and the variance of the performance metric over N consecutive validation rounds. A retraining mechanism is triggered when these conditions are not met, improving the reliability and engineering usability of the agent's training results. Reinforcement learning training is inherently stochastic; good performance in a single round does not guarantee policy stability. This is especially crucial in safety-critical applications like nuclear power plant steam generators, where policy fluctuations under continuous validation conditions must be carefully monitored. By simultaneously setting the convergence threshold and performance metric variance, policies that achieve local gains but lack stability are filtered out, preventing the deployment of networks with large fluctuations or insufficient convergence in the control system. Providing clear admission criteria for agent parameter storage and system integration enhances the repeatability and safety boundaries of the control policy.

[0036] Furthermore, the Actor-Critic neural network weights of the failed agent are restored to their initial state, the DDPG exploration noise parameters are reset, historical data from the simulation environment is cleared, and perturbation samples are re-injected. Then, independent training is re-executed. This addresses issues that may arise during reinforcement learning training, such as policies getting stuck in local optima, insufficient exploration, or biased training data. DDPG employs an Actor-Critic structure; if the early exploration direction is unfavorable, the network may converge to a poorly performing policy. Historical data from the simulation environment may also cause training to be continuously affected by abnormal trajectories. By restoring the weights, resetting the exploration noise, and re-injecting perturbation samples, the policy search space can be reopened, allowing the agent to relearn better control laws. This enhances the fault tolerance and automation of the training process, reduces the cost of repeated manual parameter tuning, and improves the efficiency of engineering implementation.

[0037] It is understood that the beneficial effects of the second to sixth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0038] In summary, this invention achieves parallel decoupling, independent training, and output summation control of the water level agent and the flow rate agent, taking into account water level regression, spurious water level suppression, and feedwater flow stability, reducing the complexity of joint training, overcoming the bandwidth limitation of cascade control, and improving the response speed, stability, and engineering feasibility of steam generator water level control.

[0039] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the process of the present invention; Figure 2 A schematic diagram of the construction of a steam generator model weighted by triangular membership functions; Figure 3 This is an overall architecture diagram of a water level control system based on parallel dual-agent DDPG collaborative tuning provided in an embodiment of the present invention; Figure 4 A flowchart for training two agents independently; Figure 5 A schematic diagram of a computer device provided in an embodiment of the present invention; Figure 6 This is a block diagram of a chip provided according to an embodiment of the present invention.

[0041] Among them, 60. Computer equipment; 61. Processor; 62. Memory; 63. Computer program; 600. Electronic device; 610. Processing unit; 620. Storage unit; 6201. Random access memory unit; 6202. Cache memory unit; 6203. Read-only memory unit; 6204. Program / utility; 6205. Program module; 630. Bus; 640. Display unit; 650. Input / output interface; 660. Network adapter; 700. External device. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0043] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0044] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0045] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this invention generally indicates that the preceding and following objects have an "or" relationship.

[0046] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0047] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0048] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0049] This invention provides a parallel dual-agent water level control method for steam generators. The actual steam generator is abstracted into a model, and a parallel PID control architecture is constructed in a Simulink simulation environment. Two independent Deep Deterministic Policy Gradient (DDPG) agents are used to adaptively tune parameters for the macroscopic trend of water level and the microscopic dynamics of flow rate, respectively. The water level agent focuses on learning the water level regression rate and suppressing spurious water levels through a simplified model, while the flow rate agent learns flow stability and noise suppression under a fixed water level setpoint. The two agents output the final control command through a signal fusion mechanism, achieving decoupling of physical features and implicit collaboration. This invention supports completely independent training of the agents, avoiding the complexity of multi-agent joint training, and significantly improves the response speed, stability, and robustness of the water level control system. It is suitable for precise water level control of steam generators under complex operating conditions in nuclear power plants.

[0050] Please see Figure 1 The present invention discloses a method for controlling the water level of a steam generator in parallel with two intelligent agents, comprising the following steps: S1. Based on the specific object, abstract and construct a simulation model of the steam generator, and build a parallel dual-channel basic control framework to form a basic control model. S101, Model Building Please see Figure 2 This study selects a commonly used nuclear power-grade steam generator as the research object and establishes a basic mechanistic model of the steam generator using the Irving transfer function model. Triangular membership functions are designed to perform weighted interpolation on the Irving model at different power points (5%-100% rated power), constructing a linear variable-parameter (LPV) steam generator simulation model covering the full power operating range to simulate the nonlinear and time-varying characteristics of the actual object. Specifically, triangular membership functions are designed using the transfer function parameters of the Irving model at different power points (5%, 15%, 30%, 50%, 100%). a i ( P For the current power P Perform fuzzy weighting to generate real-time model parameters:

[0051] in, G i ( s ) is the first i The Irving transfer function model at a single power point accurately reflects the dynamic characteristics across the entire power range, particularly the "false water level" phenomenon.

[0052] S102, Basic Framework Setup Please see Figure 3 ,exist Figure 2 Based on the simulation model, a parallel dual-channel control architecture is constructed: the water level channel includes a water level PID controller, used to calculate the basic water level control quantity based on the water level deviation; the flow rate channel includes a flow rate PI controller, used to calculate the basic flow rate control quantity based on the flow rate deviation. The initial parameters of the two controllers at different power points are tuned using the Ziegler-Nichols (ZN) method. K p1 base , K i1 base , K p2 base , K i2 base , K d2 base This forms a parallel control framework with basic stability; the output signals of the two channels are weighted and summed to jointly drive the water supply regulating valve.

[0053] The following are the designs of some specific devices within this framework: Water level PID controller: Input is water level deviation e h = h set -h Output u L base .

[0054] Flow PI controller: Input is soda / water compatibility deviation e q = q set -q Output u F base .

[0055] Tuning at each power point using the ZN method K p1 base , K i1 base , K p2 base , K i2 base ， K d2 base As a basic parameter.

[0056] Initial value of the final water supply flow rate adjustment command U base = u L base +u F base .

[0057] S2. Design and train the traffic intelligence agent. S201. Three key elements for designing Agent-F intelligent agents in reinforcement learning.

[0058] State space: including steam-water adaptation error, water flow rate change rate, normalized power, water level and water level change rate; Action space: Increment Δ of the output flow rate PI parameter K F =[Δ K p1 Δ K i1 ]; Reward function: It primarily rewards the stability of the water flow and penalizes high-frequency fluctuations and large movements in the water flow.

[0059] State:

[0060] Action: The range is limited to ±50% of the basic parameters.

[0061] Reward: ; in, R 1F It is a piecewise reward function based on the soft drink compatibility deviation. λ 1 represents a positive reward. λ 2 represents a small positive reward. λ 3 represents a negative punishment; R 2F It is an extremely large penalty term that exceeds the safety boundary, and is a fixed constant; R 3F It is the output fluctuation penalty term, which mainly determines whether the output fluctuates by the number of zeros of the output derivative.

[0062]

[0063] S202. Train the designed agent accordingly.

[0064] Based on the constructed wide-power-range steam generator simulation model, a simplified flow training sub-environment was built in the Simulink environment: a fixed water level setpoint was used, and continuous step changes in load were artificially injected; the parameters of the water level PID in the water level channel were kept constant at the initial values ​​of the model in step S1 to simulate slow regulation in a macroscopic background; Agent-F was started for DDPG training, which focused on learning the nonlinear characteristics of feedwater flow regulation and the suppression strategy for high-frequency disturbances; training continued until the flow fluctuation variance and valve action frequency met the preset indicators, and Agent-F was saved.

[0065] Training Agent-F: A flow local loop is built in Simulink. With a fixed water level setpoint, continuous step changes in load are artificially injected, decreasing from 100% FP to 75% FP, 50% FP, 40% FP, and 30% FP every 300 seconds. Agent-F is trained until it can effectively suppress flow fluctuations and smoothly regulate the feedwater flow. The training time is approximately 200 episodes.

[0066] S3. Design and train the water level agent. S301. Three key elements for designing Agent-L intelligent agents in reinforcement learning.

[0067] State space: includes water level error, water level change rate, normalized power, water supply flow rate and its change rate; Action space: Increment Δ of the output water level PID parameter K L =[Δ K p2 Δ K i2 Δ K d2 ]; Reward function: Focus on rewarding the speed at which the water level returns to the set value, and severely punish reverse operations during the "false water level" period.

[0068] State:

[0069] Action: The range is limited to ±60% of the basic parameters.

[0070] Reward: ,in, R 1F It is a piecewise reward function based on the water level adaptation deviation. λ 1 represents a positive reward. λ 2 represents a small positive reward. λ 3 represents a negative punishment; R 2LIt is an excessive penalty term for improper adjustment when the liquid level exceeds the safety boundary or is a false liquid level; it is a fixed constant. R 3L It is the output fluctuation penalty term, which mainly determines whether the output fluctuates by the number of zeros of the output derivative.

[0071]

[0072] S302. Train the designed agent accordingly.

[0073] Based on the constructed wide-power-range steam generator simulation model, a simplified water level training sub-environment was built in the Simulink environment: the parameters of the fixed-flow PI controller were the initial values ​​in the model of step S1, assuming that the flow loop could ideally track; Agent-L was started for DDPG training, so that it could focus on learning the mapping relationship between steam flow mutation and water level response, and master the macroscopic control strategy to overcome "false water level"; training continued until the water level overshoot and adjustment time met the preset indicators, and Agent-L was saved.

[0074] Training Agent-L: A simplified loop is built in Simulink, locking the flow rate PI parameters of the flow path and retaining only the water level dynamics. Continuous step changes in load are applied, decreasing sequentially from 100% FP to 75% FP, 50% FP, 40% FP, and 30% FP every 300 seconds. Agent-L is trained until it can stably and rapidly regulate the water level and effectively suppress false water level readings. The training time is approximately 200 episodes.

[0075] The two training processes are completely independent and require no interaction.

[0076] S4, System Integration The trained Agent-L and Agent-F are deployed into a complete parallel control system; the system status is collected in real time and input into the two agents respectively; Agent-F outputs ΔK. F Update the flow PI parameter and calculate the flow channel control quantity. u F Agent-L outputs ΔK L Update the PID parameters for water level and calculate the control quantity for the water level channel. u L Summation mechanism adopted U = u L + u F The final water supply flow rate regulation command is synthesized to drive the regulation of the water supply flow rate, thereby achieving adaptive and precise control of the water level.

[0077] A 25% power step reduction condition was set as the test condition. Simulation results show that when the water level shows a false rise, Agent-L immediately outputs a large opening command to overcome the false water level; Agent-F simultaneously suppresses the resulting flow overshoot. Compared with traditional cascaded PID, the maximum water level deviation is reduced by 3.4%, the settling time is shortened by 35.8%, and there is no high-frequency oscillation.

[0078] In another embodiment of the present invention, a parallel dual-agent water level control system for a steam generator is provided. This system can be used to implement the above-mentioned parallel dual-agent water level control method for a steam generator. Specifically, the parallel dual-agent water level control system for a steam generator includes a model building module, a basic control module, a dual-agent training module, an evaluation and retraining module, and a collaborative control module.

[0079] The model building module is used to establish the simulation model of the steam generator. The basic control module is used to construct the parallel dual-channel basic control framework based on the steam generator simulation model. The parallel dual-channel basic control framework includes a water level control channel and a feedwater flow control channel. A dual-agent training module includes a water level agent training unit and a flow rate agent training unit. The water level agent training unit is used to train the water level agent independently, and the flow rate agent training unit is used to train the flow rate agent independently. The evaluation and retraining module is used to determine whether the water level agent and the flow rate agent meet the convergence conditions, and to trigger the retraining mechanism when the convergence conditions are not met. The collaborative control module is used to deploy the trained water level agent and flow rate agent into the parallel dual-channel basic control framework. It updates the corresponding control parameters according to the increment of the water level PID parameter and the increment of the flow rate PI parameter, and sums the output of the water level control channel and the output of the water supply flow control channel to generate a water supply flow regulation command.

[0080] This invention provides a terminal device comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or corresponding function. The processor described in this embodiment can be used in the operation of a parallel dual-agent water level control method for a steam generator, including: A steam generator simulation model was established, and a parallel dual-channel basic control framework was constructed based on the model. This framework includes a water level control channel and a feedwater flow control channel. The water level control channel includes a water level PID controller, and the feedwater flow control channel includes a flow rate PI controller. A parallel dual-agent architecture was constructed, comprising a water level agent and a flow rate agent. The water level agent is used to tune the control parameters of the water level PID controller, and the flow rate agent is used to tune the control parameters of the flow rate PI controller. The water level agent was independently trained in a simulation environment, and reward functions were set for water level regression velocity and spurious water level suppression, enabling the agent to learn the water level control strategy. The flow control agent is independently trained in a simulation environment, and a reward function for the stability of the feedwater flow is set to enable the flow control agent to learn the flow control strategy. The trained water level agent and flow control agent are deployed in the parallel dual-channel basic control framework. The water level agent outputs the increment of the water level PID parameter, and the flow control agent outputs the increment of the flow rate PI parameter. The control parameters of the water level PID controller are updated according to the increment of the water level PID parameter, and the control parameters of the flow rate PI controller are updated according to the increment of the flow rate PI parameter. The outputs of the water level control channel and the feedwater flow control channel are summed to generate a feedwater flow regulation command, and the steam generator water level is controlled according to the feedwater flow regulation command.

[0081] Please see Figure 5 The terminal device is a computer device. In this embodiment, the computer device 60 includes a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When executed by the processor 61, the computer program 63 implements the parallel dual-agent water level control method for the steam generator in this embodiment. To avoid repetition, these details are not elaborated here. Alternatively, when executed by the processor 61, the computer program 63 implements the functions of each model / unit in the parallel dual-agent water level control system for the steam generator in this embodiment. To avoid repetition, these details are not elaborated here.

[0082] Computer device 60 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. Computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art will understand that... Figure 5 This is merely an example of computer device 60 and does not constitute a limitation on computer device 60. It may include more or fewer components than shown, or combine certain components, or different components. For example, computer device may also include input / output devices, network access devices, buses, etc.

[0083] The processor 61 may be a Central Processing Unit (CPU), or other general-purpose processors, graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0084] The memory 62 can be an internal storage unit of the computer device 60, such as a hard disk or memory of the computer device 60. The memory 62 can also be an external storage device of the computer device 60, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the computer device 60.

[0085] Furthermore, the memory 62 may include both internal storage units of the computer device 60 and external storage devices. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 can also be used to temporarily store data that has been output or will be output.

[0086] Please see Figure 6 The terminal device is an electronic device 600, which is manifested in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including storage unit 620 and processing unit 610), a display unit 640, etc.

[0087] The storage unit stores program code, which can be executed by the processing unit 610 to perform the steps described in the method section of this specification according to various exemplary embodiments of the present invention. For example, the processing unit 610 can perform actions such as... Figure 1 The steps are shown in the figure.

[0088] Storage unit 620 may include readable media in the form of volatile storage units, such as random access memory (RAM) 6201 and / or cache memory 6202, and may further include read-only memory (ROM) 6203.

[0089] Storage unit 620 may also include a program / utility 6204 having a set (at least one) program module 6205, such program module 6205 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0090] Bus 630 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the multiple bus structures.

[0091] Electronic device 600 can also communicate with one or more external devices 700 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 600, and / or with any device that enables electronic device 600 to communicate with one or more other computing devices (e.g., router, modem). This communication can be performed via input / output interface 650. Furthermore, electronic device 600 can also communicate with one or more networks (e.g., local area network, wide area network, and / or public network, such as the Internet) via network adapter 660. Network adapter 660 can communicate with other modules of electronic device 600 via bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0092] This invention also provides a storage medium, specifically a computer-readable storage medium, which is a memory device in a terminal device for storing programs and data. It is understood that the computer-readable storage medium here can include both built-in storage media in the terminal device and extended storage media supported by the terminal device; it can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor, which can be one or more computer programs (including program code). More specific examples of the computer-readable storage medium include: an electrical connection with one or more wires, a portable disk, a hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical fiber, portable compact disk read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0093] Computer-readable storage media also include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable storage medium can also be any readable medium other than a readable storage medium that can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium can be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, radio frequency, etc., or any suitable combination thereof.

[0094] Program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0095] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the parallel dual-agent water level control method for steam generators in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor to perform the following steps: A steam generator simulation model was established, and a parallel dual-channel basic control framework was constructed based on the model. This framework includes a water level control channel and a feedwater flow control channel. The water level control channel includes a water level PID controller, and the feedwater flow control channel includes a flow rate PI controller. A parallel dual-agent architecture was constructed, comprising a water level agent and a flow rate agent. The water level agent is used to tune the control parameters of the water level PID controller, and the flow rate agent is used to tune the control parameters of the flow rate PI controller. The water level agent was independently trained in a simulation environment, and reward functions were set for water level regression velocity and spurious water level suppression, enabling the agent to learn the water level control strategy. The flow control agent is independently trained in a simulation environment, and a reward function for the stability of the feedwater flow is set to enable the flow control agent to learn the flow control strategy. The trained water level agent and flow control agent are deployed in the parallel dual-channel basic control framework. The water level agent outputs the increment of the water level PID parameter, and the flow control agent outputs the increment of the flow rate PI parameter. The control parameters of the water level PID controller are updated according to the increment of the water level PID parameter, and the control parameters of the flow rate PI controller are updated according to the increment of the flow rate PI parameter. The outputs of the water level control channel and the feedwater flow control channel are summed to generate a feedwater flow regulation command, and the steam generator water level is controlled according to the feedwater flow regulation command.

[0096] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0097] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0098] To further verify the control effect of this invention over a wide power operating range, linear power increase / decrease tests were conducted in a steam generator simulation model. The test conditions included a power reduction process linearly decreasing from 100% FP to 30% FP, and a power increase process linearly increasing from 30% FP to 100% FP, used to simulate the dynamic water level control process of a nuclear power unit during power regulation over a wide power range. Under these conditions, the steam generator model parameters change with the power level, and the water level response, feedwater flow response, and spurious water level influence all change accordingly. Therefore, this can be used to verify the adaptability of the control method over a wide power range.

[0099] Simulation results show that during the linear reduction of power from 100% FP to 30% FP, the present invention, under the coordinated regulation of the water level agent and the flow rate agent, achieves a 17.8% reduction in maximum water level deviation and a 17.9% reduction in settling time compared to traditional cascade PID control, without high-frequency oscillations. These results demonstrate that during power reduction, the water level agent can adjust the water level PID parameters based on water level errors and trends, while the flow rate agent can adjust the flow rate PI parameters based on changes in feedwater flow. This allows the water level control channel and the feedwater flow control channel to participate in regulation together in a parallel structure, thereby improving the water level tracking effect during power reduction.

[0100] During the linear increase from 30%FP to 100%FP, compared with traditional cascade PID control, this invention reduces the maximum water level deviation by 3.5%, shortens the settling time by 29.1%, and eliminates high-frequency oscillations. This result demonstrates that during the power increase process, as the power level continuously changes, the water level agent and flow rate agent can still respectively tune the parameters of the macroscopic water level trend and the dynamic water flow rate, enabling the system to maintain a relatively fast water level recovery speed and a relatively stable dynamic response.

[0101] The above-mentioned linear power increase / decrease tests further demonstrate that the present invention is not only applicable to a single power point or a single step disturbance condition, but also exhibits good control performance during the tested power reduction processes from 100%FP to 30%FP and power increase processes from 30%FP to 100%FP. Compared with traditional cascade PID control, the present invention avoids water level regulation being completely limited by the inner loop response of flow by connecting the water level control channel and the feedwater flow control channel in parallel; through independent training of the water level agent and the flow agent, the two agents respectively form control strategies for water level regression, spurious water level suppression, and feedwater flow stability; and by summing the outputs of the two channels to generate the final feedwater flow regulation command, the overall water level regulation and the local flow regulation form a synergy in the complete system.

[0102] Therefore, under the above simulation conditions, the present invention can reduce the maximum water level deviation and shorten the adjustment time while avoiding high-frequency oscillations, demonstrating its good dynamic response capability and operational stability under wide power linear rise and fall conditions. This result further verifies the effectiveness of the parallel dual-agent architecture compared to traditional cascade PID control in steam generator water level control.

[0103] In summary, the parallel dual-smart agent water level control method and system for a steam generator of the present invention has the following significant advantages: The dual advantages complement each other, overcoming bandwidth limitations: the water level agent excels at handling large time lags and feedforwards (false water levels), while the flow rate agent excels at handling high-frequency disturbances and nonlinearities in flow. The parallel structure allows the advantages of both to be directly superimposed. The water level agent can directly issue commands based on the steam flow feedforward without waiting for the inner loop response, overcoming the limitation of cascade control by the inner loop bandwidth and significantly improving the dynamic response speed.

[0104] Training is extremely simple and highly practical for engineering applications: It completely solves the problems of difficult convergence and complex debugging in multi-agent joint training. Two agents can be developed in parallel and debugged independently without sharing rewards or communication mechanisms, which greatly reduces the difficulty of algorithm implementation and computing power requirements, and is easy to deploy in industrial simulation platforms (such as Simulink).

[0105] Enhanced resistance to false water levels: The water level agent can specifically target the characteristics of "false water levels" through reinforcement learning, outputting aggressive yet accurate macroscopic control quantities; while the flow agent, acting as a "stabilizer," automatically suppresses the resulting drastic flow fluctuations. In a parallel structure, the two achieve implicit synergy through the natural separation of physical time scales, realizing a perfect combination of "offense" and "defense."

[0106] High fault tolerance and safety: If one agent fails, another agent can still provide basic control, and the system will not completely lose control. Meanwhile, the deviation tuning method based on PID parameters retains the fallback function of the basic PID controller, aligning with the defense-in-depth safety concept of nuclear power plants.

[0107] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0108] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0109] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0110] In the embodiments provided by this invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0111] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0112] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0113] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random-access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0114] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0115] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0116] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0117] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for parallel dual-agent water level control in a steam generator, characterized in that, Includes the following steps: S1. Establish a steam generator simulation model, and construct a parallel dual-channel basic control framework based on the steam generator simulation model. The parallel dual-channel basic control framework includes a water level control channel and a feedwater flow control channel. The water level control channel includes a water level PID controller, and the feedwater flow control channel includes a flow PI controller. S2. Construct a parallel dual-agent architecture, which includes a water level agent and a flow rate agent. The water level agent is used to tune the control parameters of the water level PID controller, and the flow rate agent is used to tune the control parameters of the flow rate PI controller. S3. The water level agent is independently trained in a simulation environment, and a reward function is set for the water level regression speed and false water level suppression, so that the water level agent learns the water level control strategy. S4. The flow agent is independently trained in a simulation environment, and a reward function for the stability of water supply flow is set so that the flow agent learns the flow control strategy. S5. Deploy the trained water level agent and flow rate agent into the parallel dual-channel basic control framework. The water level agent outputs the increment of the water level PID parameter, and the flow rate agent outputs the increment of the flow rate PI parameter. S6. Update the control parameters of the water level PID controller incrementally according to the water level PID parameter, and update the control parameters of the flow rate PI controller incrementally according to the flow rate PI parameter. Sum the output of the water level control channel and the output of the water supply flow control channel to generate a water supply flow regulation command, and control the water level of the steam generator according to the water supply flow regulation command.

2. The method for parallel dual-agent water level control of a steam generator according to claim 1, characterized in that, In step S1, the steam generator simulation model is a linear variable parameter steam generator simulation model; the linear variable parameter steam generator simulation model is established in the following way: A basic model of a steam generator at different power points was established using the Irving transfer function model set. The basic models of the steam generator at different power points were then weighted and fused using the triangular membership function to form a steam generator simulation model covering the power range of 5% to 100%.

3. The method for parallel dual-agent water level control of a steam generator according to claim 1, characterized in that, In step S1, the water level PID controller calculates the output of the water level control channel based on the water level deviation, and the flow rate PI controller calculates the output of the feedwater flow control channel based on the steam-water compatibility deviation. The initial parameters of the water level PID controller and the flow rate PI controller are obtained using the Ziegler-Nichols tuning method.

4. The method for parallel dual-agent water level control of a steam generator according to claim 1, characterized in that, The water level agent is a DDPG agent. The state space of the water level agent includes water level error, water level change rate, normalized power, water supply flow rate and water supply flow rate change rate. The action space is the water level PID parameter increment, which includes proportional parameter increment, integral parameter increment and derivative parameter increment.

5. The method for parallel dual-agent water level control of a steam generator according to claim 1, characterized in that, The flow intelligence agent is a DDPG agent. The state space of the flow intelligence agent includes the steam-water adaptation error, the rate of change of water flow, the normalized power, the water level and the rate of change of water level. The action space is the flow PI parameter increment, which includes the proportional parameter increment and the integral parameter increment.

6. The method for parallel dual-agent water level control of a steam generator according to claim 1, characterized in that, Step S3, independently training the water level agent includes: A simplified water level control model is constructed in a simulation environment. The parameters of the flow PI controller are fixed, and it is assumed that the water flow control channel can track the set value, so that the water level agent learns water level regression and false water level suppression strategies.

7. The method for parallel dual-agent water level control of a steam generator according to claim 1, characterized in that, Step S4, independently training the traffic agent includes: In the simulation environment, the water level setpoint is fixed, the parameters of the water level PID controller are kept constant, and a load step change is injected into the steam generator simulation model, so that the flow agent learns the nonlinear characteristics of water flow regulation and the flow stabilization control strategy.

8. The method for parallel dual-agent water level control of a steam generator according to claim 1, characterized in that, Following steps S3 and S4, a training evaluation step is also included: If the water level agent or flow rate agent reaches the convergence threshold within the preset maximum number of iterations, and the performance index variance is less than the variance threshold in N consecutive verification rounds, then the training is considered successful and the network parameters of the corresponding agent are saved; if the preset maximum number of iterations is reached but the agent still fails to converge, or the performance fluctuation exceeds the preset range, then the training is considered to have failed and the retraining mechanism is triggered.

9. The method for parallel dual-agent water level control of a steam generator according to claim 8, characterized in that, The retraining mechanism includes: The Actor-Critic neural network weights of the agent that was determined to have failed training are restored to their initial state. The exploration noise parameters of the DDPG algorithm are reset, historical data of the simulation environment is cleared and perturbation samples are re-injected, and then independent training is re-executed for the agent that was determined to have failed training.

10. A parallel dual-intelligent agent water level control system for a steam generator, characterized in that, The method for implementing the parallel dual-agent water level control method for a steam generator according to any one of claims 1 to 9 includes: The model building module is used to build the simulation model of the steam generator; The basic control module is used to construct the parallel dual-channel basic control framework based on the steam generator simulation model. The parallel dual-channel basic control framework includes a water level control channel and a feedwater flow control channel. A dual-agent training module includes a water level agent training unit and a flow rate agent training unit. The water level agent training unit is used to train the water level agent independently, and the flow rate agent training unit is used to train the flow rate agent independently. The evaluation and retraining module is used to determine whether the water level agent and the flow rate agent meet the convergence conditions, and to trigger the retraining mechanism when the convergence conditions are not met. The collaborative control module is used to deploy the trained water level agent and flow rate agent into the parallel dual-channel basic control framework. It updates the corresponding control parameters according to the increment of the water level PID parameter and the increment of the flow rate PI parameter, and sums the output of the water level control channel and the output of the water supply flow control channel to generate a water supply flow regulation command.

Images