Nuclear power device coordination control method and system based on multi-agent reinforcement learning

Abstract

The invention discloses a multi-agent reinforcement learning-based nuclear power plant coordination control method and system, and the scheme can comprise the steps: carrying out the analysis of a key equipment operation mechanism and boundary conditions, combining actual design parameters and a PI controller, and carrying out the optimization of the operation mechanism and boundary conditions, and carrying out the optimization of the operation mechanism and boundary conditions. And a system lumped parameter model easy for rapid simulation of a computer is obtained. On the basis, the mutual influence between different local control systems and the coupling relation between input and output parameters are comprehensively considered through multi-agent reinforcement learning, and overall coordination optimization of the system in the working condition conversion process of the nuclear power device is achieved according to a specific reward and punishment mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of nuclear power plants, and in particular to a coordinated control method and system for a nuclear power plant using multi-agent reinforcement learning. Background Art

[0002] With the development of modern control theory and artificial intelligence technology, the research results of advanced control algorithms in the field of nuclear power plant control strategy optimization are constantly updated, and the digital transformation of nuclear power plant control systems is in urgent need. Combining artificial intelligence methods such as deep learning and reinforcement learning with traditional control logic is one of the key ways to achieve system transformation.

[0003] A nuclear power plant is a complex system with multiple inputs and multiple outputs, nonlinearity, and time-varying. During operation, the average coolant temperature, reactor power, turbine power, and other key parameters need to be adjusted in real time according to changes in load demand. The PI controller has a simple structure, is easy to implement, has certain robustness and reliability, and has been widely used in the current nuclear power plant control system. However, the inherent limitations of the conventional PI controller mean that the dynamic quality of a complex system such as a nuclear power plant under variable operating conditions still has room for improvement. At the same time, the coupling characteristics between the local control loops make the transient response of the system poor, especially under large-scale variable operating conditions. Some controlled parameters are prone to large overshoot or long adjustment time, which is not conducive to the normal operation of the equipment and system, and in serious cases, it will also pose a challenge to the safety of the nuclear power plant. Summary of the invention

[0004] The present invention provides a nuclear power plant coordinated control method and system based on multi-agent reinforcement learning, which is used to overcome at least one technical problem existing in the prior art.

[0005] According to a first aspect of an embodiment of the present invention, a method for coordinated control of a nuclear power plant using multi-agent reinforcement learning is provided, comprising:

[0006] S1: Select several key equipment in the nuclear power plant to be analyzed; for any one of the several key equipment, establish a lumped parameter model of the any one of the key equipment based on the flow and heat transfer characteristics of the internal working fluid, the geometric structure and the operation design parameters of the any one of the key equipment and the law of conservation of mass, energy and momentum, and then obtain the lumped parameter model of each of the several key equipment;

[0007] S2: Based on the lumped parameter model of each key device obtained in step S1, the controlled parameter setting value and the PI controller are set, each local control system is constructed, and a systematic integration is performed according to the corresponding relationship between the input and output parameters of different local control systems in the local control systems to obtain the system full model of the nuclear power plant to be analyzed;

[0008] S3: Based on the full model of the nuclear power plant system constructed in step S2, an agent capable of adjusting controller parameters in real time is assigned to each of the local control systems, and at the same time, the control logic is hierarchical according to the learning mechanism of multiple agents, and is divided into a central control layer, a coordination control layer and a local control layer, so as to form a hierarchical and distributed coordination control architecture, wherein the central control layer and the coordination control layer are respectively composed of multiple Critic networks and Actor networks, and both perform strategy guidance and strategy implementation respectively by receiving data collected by the local control layer, and the local control layer is composed of the local control systems of the nuclear power plant, and performs system-level coordination optimization by receiving real-time PI controller parameters transmitted by the coordination control layer;

[0009] S4: Select the appropriate multi-agent reinforcement learning algorithm and the corresponding training execution framework, and set the action space, state space, reward function, network structure and hyperparameters in the multi-agent;

[0010] S5: Based on the full model environment of the nuclear power plant system built in step S2 and the initialized multi-agent obtained in step S4, the multi-agent is trained under different working conditions according to the hierarchical distributed coordinated control architecture formed in step S3, and the multi-agent is saved after the cumulative reward curve converges;

[0011] S6: Integrate the trained multi-agent into the whole model of the nuclear power plant system to obtain a hierarchical distributed coordinated control simulation model of the nuclear power plant;

[0012] S7: Based on the nuclear power plant hierarchical distributed coordinated control simulation model obtained in step S6, the model operation conditions are set according to the simulation requirements, the multi-agent is loaded, and the model is run;

[0013] S8: Observe the coordinated control effect through the time domain variation curves of the controlled parameters output by the oscilloscope in the model and the controller parameters in each local control system.

[0014] Preferably, it is characterized in that the several key equipment include a reactor core, a pressurizer, a steam generator, a steam turbine, a condenser and a deaerator in the nuclear power plant to be analyzed.

[0015] Preferably, it is characterized in that the training process in step S5 specifically includes:

[0016] S5.1: Randomly initialize the online network parameters of all agents

[0017] S5.2: Initialize the target network parameters of all agents by copying

[0018] S5.3: Initialize the experience replay pool

[0019] S5.4: Initialize exploration noise ε and obtain initial global state

[0020] S5.5: Agents based on Perform actions and get corresponding rewards and the state at the next moment

[0021]

[0022] S5.6: Storing Tuples Experience Replay Pool

[0023] S5.7: From the Experience Replay Pool Randomly extract N tuple samples N×(s j ,a j ,r j ,s j+1 );

[0024] S5.8: Calculate the target Q value

[0025] S5.9: Calculate and minimize the loss function of the Critic online network

[0026] And update network parameters;

[0027] S5.10: Calculate Actor Online Network Policy Gradients

[0028] And update network parameters;

[0029] S5.11: Update target network parameters using soft update strategy

[0030]

[0031] According to a second aspect of an embodiment of the present invention, a nuclear power plant coordinated control system based on multi-agent reinforcement learning is provided, comprising:

[0032] The key parameter setting value and system initial state module is used to set the key parameter setting value and system initial state of each local control system in the "stack-follow-machine" mode to meet the simulation of various working conditions of the system;

[0033] The control system module is used to simulate the required operating conditions. It is the main part of the system and covers all local control systems.

[0034] The coordination control module is used to establish the connection between each local control system. Through multi-agent training, the real-time parameters of the controller are not only dependent on the current local system, but also affected by the remaining local systems, so as to obtain the coordinated control effect during the working condition conversion process;

[0035] The key operating parameter display module is used to display the key operating parameters in the system to facilitate the observation of coordinated control effects.

[0036] One embodiment of the present specification can achieve at least the following beneficial effects: the coordinated control method and system of a nuclear power plant based on multi-agent reinforcement learning provided by the present invention, in steps S1 and S2, the simulation modeling of the real operating system is realized by the lumped parameter method, so as to solve the problem of lack of verification platform for advanced control strategies; in step S2, systematic integration is carried out according to the correspondence between the input and output parameters of different local control systems, and the obtained full model of the nuclear power plant system not only ensures a certain simulation speed and accuracy, but also facilitates the direct embedding of relevant algorithm modules.

[0037] In step 3, a coordinated control method based on multi-agent reinforcement learning for adaptive adjustment of controller parameters is proposed. At the same time, the traditional control logic is improved and layered. Through the transmission and processing of data information between the central control layer, the coordinated control layer and the local control layer, the information barrier problem existing in the system working condition conversion process is solved. The formed hierarchical distributed coordinated control architecture also provides a basis for the training of multi-agents.

[0038] The present invention comprehensively considers the mutual influence between different local control systems and the coupling relationship between input and output parameters through multi-agent reinforcement learning, and realizes the global coordinated optimization of the system during the operating condition conversion of the nuclear power plant according to a specific reward and punishment mechanism, which is of great significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 A schematic flow chart of a coordinated control method for a nuclear power plant based on multi-agent reinforcement learning provided by the present invention;

[0041] Figure 2 A schematic diagram of the implementation process of the nuclear power plant coordinated control solution provided by an embodiment of the present invention;

[0042] Figure 3 A schematic diagram of the basic structure of an intelligent agent provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of one or more embodiments of this specification clearer, the technical solutions of one or more embodiments of this specification will be clearly and completely described below in combination with the specific embodiments of this specification and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of one or more embodiments of this specification.

[0044] It should be understood that although the terms first, second, third, etc. may be used in this application document to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other.

[0045] In the embodiments of this specification, a coordinated control method of a nuclear power plant based on multi-agent reinforcement learning is provided. By analyzing the operation mechanism and boundary conditions of key equipment, combined with actual design parameters and PI controllers, a system lumped parameter model that is easy for computer rapid simulation is obtained. On this basis, the mutual influence between different local control systems and the coupling relationship between input and output parameters are comprehensively considered through multi-agent reinforcement learning, and the overall coordinated optimization of the system during the operation state conversion of the nuclear power plant is realized according to a specific reward and punishment mechanism.

[0046] In order to solve the defects in the prior art, this solution provides the following embodiments:

[0047] The present invention first provides a nuclear power plant coordinated control method based on multi-agent reinforcement learning, which may include the following contents:

[0048] Step S1: Select several key equipment in the nuclear power plant to be analyzed; for any one of the several key equipment, establish a lumped parameter model of the any one of the key equipment based on the flow and heat transfer characteristics of the internal working fluid, the geometric structure and the operation design parameters of the any one of the key equipment and the law of conservation of mass, energy and momentum, and then obtain the lumped parameter model of each of the several key equipment.

[0049] Step S2: Based on the lumped parameter model of each key equipment obtained in step S1, the controlled parameter setting value and the PI controller are set, each local control system is built, and a systematic integration is performed according to the corresponding relationship between the input and output parameters of different local control systems in the local control systems to obtain the full system model of the nuclear power plant to be analyzed.

[0050] In step S2, the specific implementation can be as follows: First, the lumped parameter models established for the key equipment of the nuclear power plant in step S1 are used. These models are constructed by applying the laws of conservation of mass, energy and momentum based on the internal working fluid flow and heat transfer characteristics, geometric structure and operation design parameters of the equipment. Then, the set values ​​of the controlled parameters and the proportional integral controller (PI controller) are configured for these lumped parameter models, in order to achieve local control of each key equipment. Next, these local control systems are integrated according to the relationship between their input and output parameters. This is done to form a full system model that can reflect the operating characteristics of the entire nuclear power plant. This process ensures that the model can maintain a certain simulation speed and ensure simulation accuracy, and it is also convenient for the access and testing of subsequent control algorithms. Through the above steps, the construction of the full system model of the nuclear power plant is completed, providing a basic platform for subsequent multi-agent coordinated control.

[0051] Step S3: Based on the full model of the nuclear power plant system constructed in step S2, an agent capable of adjusting controller parameters in real time is assigned to each of the local control systems, and at the same time, the control logic is hierarchical according to the learning mechanism of multiple agents, and is divided into a central control layer, a coordination control layer and a local control layer, so as to form a hierarchical and distributed coordination control architecture, wherein the central control layer and the coordination control layer are respectively composed of a plurality of Critic networks and Actor networks, and both perform strategy guidance and strategy implementation respectively by receiving data collected by the local control layer, and the local control layer is composed of the local control systems of the nuclear power plant, and performs system-level coordination optimization by receiving real-time PI controller parameters transmitted by the coordination control layer;

[0052] Step S4: Select a suitable multi-agent reinforcement learning algorithm and the corresponding training execution framework, and set the action space, state space, reward function, network structure and hyperparameters in the multi-agent.

[0053] In this step, the specific implementation may include the following key operations:

[0054] Select a multi-agent reinforcement learning algorithm and execution framework: First, you need to determine a multi-agent reinforcement learning algorithm suitable for the coordinated control scenario of nuclear power plants, such as Multi-Agent Deep Q-Network (MADQN), Multi-Agent Deep Deterministic Policy Gradient (MADDPG), etc., as well as the corresponding execution framework, such as TensorFlow or PyTorch.

[0055] Define the action space, state space and reward function of the multi-agent. Specifically, the action space is to define the range of actions that each agent can take, which is usually determined by their role in the control hierarchy and the controller parameters that can be adjusted. State space: clarify the environmental state that the agent needs to perceive, including but not limited to the key parameters of each local control system, the overall operating state of the system, etc. Reward function: design a reward mechanism to motivate the agent to achieve the global optimization goal. The reward should reflect the system stability, efficiency and safety during the operating condition conversion process, and punish overshoot, long adjustment time, etc. Network structure: design a specific neural network architecture for the Critic network (evaluating the value of the agent's behavior) and the Actor network (determining what action to take), including the number of layers, the number of nodes per layer, etc. Hyperparameter adjustment: including learning rate, discount factor, exploration decay rate, etc. These parameters need to be carefully adjusted to balance learning efficiency and final performance. Through the careful implementation of these steps, it is ensured that the multi-agent system can effectively learn and work together in the simulation environment of the nuclear power plant, and achieve global coordinated optimization during the operating condition conversion process.

[0056] Step S5: Based on the full model environment of the nuclear power plant system constructed in step S2 and the initialized multi-agent obtained in step S4, the multi-agent is trained under different working conditions according to the hierarchical distributed coordinated control architecture formed in step S3, and the multi-agent is saved after the cumulative reward curve converges.

[0057] Step S6: Integrate the trained multi-agent into the whole model of the nuclear power plant system to obtain a hierarchical distributed coordinated control simulation model of the nuclear power plant.

[0058] In this step, the trained and optimized multi-agent can be integrated into the previously established full model of the nuclear power plant system. This integration process enables the realization of a hierarchical distributed coordinated control architecture, which includes agents in the central control layer, the coordinated control layer, and the local control layer. In this way, a nuclear power plant simulation model capable of hierarchical distributed coordinated control can be obtained. Next, the operating conditions of the model are configured according to different simulation requirements, and the trained multi-agent is loaded into the model, and then the model is run to evaluate and optimize the coordinated control effect of the nuclear power plant.

[0059] Step S7: Based on the nuclear power plant hierarchical distributed coordinated control simulation model obtained in step S6, the model operating conditions are set according to the simulation requirements, and the multi-agent is loaded to run the model.

[0060] In this step, the operating conditions of the hierarchical distributed coordinated control simulation model of the nuclear power plant can be set according to the simulation requirements to ensure that the model can simulate the expected operating scenarios. Then, the trained and optimized multi-agent is loaded into the simulation model. After the model is running, each agent works together according to the functional role of its level (central control layer, coordinated control layer or local control layer). The Critic network of the central control layer and the Actor network of the coordinated control layer perform strategy planning and guidance based on the data collected from the local control layer, while the local control layer adjusts its control behavior based on the received real-time PI controller parameters to achieve system-level coordination and optimization of the entire nuclear power plant in different operating conditions. Finally, the time domain change curves of the controlled parameters and the controller parameters in the local control system are observed through the built-in oscilloscope tool of the model to evaluate and verify the effect of coordinated control.

[0061] Step S8: Observe the coordinated control effect through the time domain variation curves of the controlled parameters output by the oscilloscope in the model and the controller parameters in each local control system.

[0062] In this step, the oscilloscope tool in the hierarchical distributed coordinated control simulation model of nuclear power plants is used to observe and analyze the time domain change curves of the controlled parameters and the controller parameters in each local control system. The change curves of these parameters reflect the effect of coordinated control, that is, whether the system has achieved optimal adjustment globally during the conversion of different working conditions. The observation results are helpful in evaluating the effectiveness and stability of the proposed control method, and can also be used for further algorithm tuning and verification.

[0063] The present invention provides a coordinated control method for a nuclear power plant based on multi-agent reinforcement learning. In step S1 and step S2, the simulation modeling of the real operating system is realized by the lumped parameter method, so as to solve the problem of lack of verification platform for advanced control strategies. In step S3, a coordinated control method for adaptively adjusting controller parameters based on multi-agent reinforcement learning is proposed. At the same time, the traditional control logic is improved and layered. Through the transmission and processing of data information between the central control layer, the coordination control layer and the local control layer, the information barrier problem existing in the system operating state conversion process is solved. The formed hierarchical distributed coordinated control architecture also provides a basis for the training of multi-agents. It is of great significance to comprehensively consider the mutual influence between different local control systems and the coupling relationship between input and output parameters through multi-agent reinforcement learning, and realize the coordinated optimization of the overall system in the process of nuclear power plant operating state conversion according to a specific reward and punishment mechanism. In addition, since there are many devices in the nuclear power plant and the operating mechanism and structure are very complex, it is difficult to directly model the whole device, which is not conducive to the analysis of its internal mechanism characteristics. Therefore, the present invention adopts the modular modeling idea to model the relatively independent equipment and control systems in the system in modules, and then systematically integrates them according to the corresponding relationship between the input and output parameters of different local control systems. The obtained full model of the nuclear power plant system not only ensures a certain simulation speed and accuracy, but also facilitates the direct embedding of related algorithm modules, and is suitable for the subsequent coordination control method verification of the present invention.

[0064] based on Figure 1 The method, the examples of this specification also provide some specific implementation plans of the method, which are described below.

[0065] In an optional embodiment technical solution, the several key equipment include a reactor core, a pressurizer, a steam generator, a steam turbine, a condenser and a deaerator in the nuclear power plant to be analyzed.

[0066] In an optional embodiment, if Figure 2 As shown, the training process in step 5 may include:

[0067] S5.1: Randomly initialize the online network parameters of all agents

[0068] S5.2: Initialize the target network parameters of all agents by copying

[0069] S5.3: Initialize the experience replay pool

[0070] S5.4: Initialize exploration noise ε and obtain initial global state

[0071] S5.5: Agents based on Perform actions and get corresponding rewards and the state at the next moment

[0072]

[0073] S5.6: Storing Tuples Experience Replay Pool

[0074] S5.7: From the Experience Replay Pool Randomly extract N tuple samples N×(s j ,a j ,r j ,s j+1 );

[0075] S5.8: Calculate the target Q value

[0076] S5.9: Calculate and minimize the loss function of the Critic online network

[0077] And update network parameters;

[0078] S5.10: Calculate Actor Online Network Policy Gradients

[0079] And update network parameters;

[0080] S5.11: Update target network parameters using soft update strategy

[0081]

[0082] In this embodiment, by training different working conditions through a multi-agent reinforcement learning algorithm, each agent can learn how to achieve global coordinated optimization during the working condition conversion process in the simulated nuclear power plant system full model environment. This process not only takes into account the mutual influence between local control systems, but also handles the coupling relationship between input and output parameters, thereby improving the dynamic performance of the entire nuclear power plant control system. After the training is completed, when the cumulative reward curve converges, it means that the multi-agent has found a more optimized control strategy, which is then saved and applied to actual simulation or control to achieve a smoother and more efficient working condition conversion control.

[0083] The technical solution of the present invention also provides a nuclear power plant coordination control system based on multi-agent reinforcement learning, which may include:

[0084] The key parameter setting value and system initial state module is used to set the key parameter setting value and system initial state of each local control system in the "stack-follow-machine" mode to meet the simulation of various working conditions of the system;

[0085] The control system module is used to simulate the required operating conditions. It is the main part of the system and covers all local control systems.

[0086] The coordination control module is used to establish the connection between each local control system. Through multi-agent training, the real-time parameters of the controller are not only dependent on the current local system, but also affected by the remaining local systems, so as to obtain the coordinated control effect during the working condition conversion process;

[0087] The key operating parameter display module is used to display the key operating parameters in the system to facilitate the observation of coordinated control effects.

[0088] Those skilled in the art can understand that the accompanying drawings are only schematic diagrams of an embodiment, and the modules or processes in the accompanying drawings are not necessarily required to implement the present invention.

[0089] Those skilled in the art can understand that the modules in the device in the embodiment can be distributed in the device in the embodiment according to the description of the embodiment, or can be changed accordingly and located in one or more devices different from the embodiment. The modules in the above embodiment can be combined into one module, or can be further divided into multiple sub-modules.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A coordinated control method for a nuclear power plant based on multi-agent reinforcement learning, characterized in that: The method comprises: S1: Select several key equipment in the nuclear power plant to be analyzed; for any one of the several key equipment, establish a lumped parameter model of the any one of the key equipment based on the flow and heat transfer characteristics of the internal working fluid, the geometric structure and the operation design parameters of the any one of the key equipment and the law of conservation of mass, energy and momentum, and then obtain the lumped parameter model of each of the several key equipment; S2: Based on the lumped parameter model of each key device obtained in step S1, the controlled parameter setting value and the PI controller are set, each local control system is constructed, and a systematic integration is performed according to the corresponding relationship between the input and output parameters of different local control systems in the local control systems to obtain the system full model of the nuclear power plant to be analyzed; S3: Based on the full model of the nuclear power plant system constructed in step S2, an agent capable of adjusting controller parameters in real time is assigned to each of the local control systems, and at the same time, the control logic is hierarchical according to the learning mechanism of multiple agents, and is divided into a central control layer, a coordination control layer and a local control layer, so as to form a hierarchical and distributed coordination control architecture, wherein the central control layer and the coordination control layer are respectively composed of multiple Critic networks and Actor networks, and both perform strategy guidance and strategy implementation respectively by receiving data collected by the local control layer, and the local control layer is composed of the local control systems of the nuclear power plant, and performs system-level coordination optimization by receiving real-time PI controller parameters transmitted by the coordination control layer; S4: Select the appropriate multi-agent reinforcement learning algorithm and the corresponding training execution framework, and set the action space, state space, reward function, network structure and hyperparameters in the multi-agent; S5: Based on the full model environment of the nuclear power plant system built in step S2 and the initialized multi-agent obtained in step S4, the multi-agent is trained under different working conditions according to the hierarchical distributed coordinated control architecture formed in step S3, and the multi-agent is saved after the cumulative reward curve converges; S6: Integrate the trained multi-agent into the whole model of the nuclear power plant system to obtain a hierarchical distributed coordinated control simulation model of the nuclear power plant; S7: Based on the nuclear power plant hierarchical distributed coordinated control simulation model obtained in step S6, the model operation conditions are set according to the simulation requirements, the multi-agent is loaded, and the model is run; S8: Observe the coordinated control effect through the time domain variation curves of the controlled parameters output by the oscilloscope in the model and the controller parameters in each local control system.

2. The coordinated control method of a nuclear power plant based on multi-agent reinforcement learning according to claim 1 is characterized in that: The several key equipment include a reactor core, a pressurizer, a steam generator, a steam turbine, a condenser and a deaerator in the nuclear power plant to be analyzed.

3. The coordinated control method of a nuclear power plant based on multi-agent reinforcement learning according to claim 1 is characterized in that: The training process in step S5 specifically includes: S5.1: Randomly initialize the online network parameters of all agents S5.2: Initialize the target network parameters of all agents by copying S5.3: Initialize the experience replay pool S5.4: Initialize exploration noise ε and obtain initial global state S5.5: Agents based on Perform actions and get corresponding rewards and the state at the next moment S5.6: Storing Tuples Experience Replay Pool S5.7: From the Experience Replay Pool Randomly extract N tuple samples N×(s j ,a j ,r j ,s j+1 ); S5.8: Calculate the target Q value S5.9: Calculate and minimize the loss function of the Critic online network And update network parameters; S5.10: Calculate Actor Online Network Policy Gradients And update network parameters; S5.11: Update target network parameters using soft update strategy 4. A coordinated control system for a nuclear power plant based on multi-agent reinforcement learning, characterized in that: include: The key parameter setting value and system initial state module is used to set the key parameter setting value and system initial state of each local control system in the "stack-follow-machine" mode to meet the simulation of various working conditions of the system; The control system module is used to simulate the required operating conditions. It is the main part of the system and covers all local control systems. The coordination control module is used to establish the connection between each local control system. Through multi-agent training, the real-time parameters of the controller are not only dependent on the current local system, but also affected by the remaining local systems, so as to obtain the coordinated control effect during the working condition conversion process; The key operating parameter display module is used to display the key operating parameters in the system, making it easier to observe the coordinated control effects.

Images