AN INTELLIGENT CONTROL HYDROPOWER SYSTEM BASED ON MULTI-AGENT THEORY

In order to achieve the potential of existing power stations, increase the efficiency of waterpower, and change the running of power stations from an individual plant basis to centralization, many people have researched the automation of power stations over the years with excellent results. However, in reality there are no power stations that have been fully automated. This paper proposes a control system based on Multi-Agent Theory, including a management agent, communication agent, control agent, echo agent, and so on. These agents are able to harmonize automation and intelligence to provide effective control of the running of a hydropower station. The end goal is a real control system that can operate independently of human intervention.


INTRODUCTION
Power systems are the most important and most extensive of the modern social pillar industries.Because of the ease and diversity of electrical energy production, transmission, distribution, and use, electrical energy is increasingly playing an important role in all aspects of social life and has become a major energy source in the development of modern society.However, because today's modern industrial and social development requires large amounts of electricity, there are serious waste problems in the production of electrical energy because of the absence of good control systems.In order to solve this problem of wasted resources, many researchers have proposed that hydropower stations should put economic operations in practice and have researched this academic and technical problem in-depth.Many algorithms for the optimal distribution load of power stations and automatic control methods of operation have been produced, but they do not combine the positive results of two aspects: intervention and correction for imperfect communication are still needed for coordination of the automation process.
Along with the development of science and technology, man's understanding of his own behavior and the social colony is constantly deepened.Scholars in the field of distributed artificial intelligence have abstracted divisions of the collaborative nature of social behavior into Agent and Multi-Agent Theories.Generally speaking, an agent is a kind of independently-calculated entity or procedure that can perceive the environment under a specific social environment and can flexibly and independently operate to achieve a series of design objectives.
A Multi-Agent System (MAS) (Xiao & Cai, 2006;Zhou & Gao, 2004;Zhou & Sun, 2003;Jennings et al., 1998;Data Science Journal, Volume 6, Supplement, 30 July 2007Shen, 1998) refers to a computer system that uses multiple agents to complete certain tasks or achieve certain goals by cooperation.These agents cooperate to solve problems above their individual capacities; they are autonomous and distributed operations.There are cooperation and the sharing of services between every two agents.The objectives and conduct often have contradictions and conflicts.Through competition or consultation to solve these problems, the agents complete a task.A multi-agent system requires that the exchanges among agents in the system have the capability of intelligence or auto-organic activities, such as reasoning, planning, learning, etc.
Once the Multi-Agent Theory was put forward, it was quickly and widely used in various fields of study and achieved fruitful results.However, in the field of electrical systems closely linked with automation and control, the combination and application of multi-agent and automatic control theory are very few, particularly for hydropower station operations.Thus, creating a hydropower station operating control system based on the multi-agent systems theory, to control the operations, will provide an important basis for achieving intelligent control.

INTELLIGENT CONTROL SYSTEM OF CONTROL SYSTEM OF HYDROPOWER BASED ON MULTI-AGENT THEORY
In an intelligent control system for hydropower operation based on Multi-Agent Theory, in accordance with the requirements of control, the system is divided into a number of agent subsystems and agent modules based on functionality.These subsystems and modules are scattered geographically and are relatively independent of the function of each module, but each module's sole function is to serve the entire system.Intelligent modules in the system share system resources and communication, as well as coordinate with each other to control the whole system.Furthermore, an intelligent control system of hydropower stations operating on a multi-agent theory is different from the traditional system of modular and simple distributed control.The traditional modular system only requires each system module to have the ability to do conventional calculations and control rather than to be autonomous and intelligent.Simple distributed control only emphasizes that functions will be scattered, but each part after scattering does not have mutual communication and coordination functions.Only the control module with independence, intelligence, and communication coordination is, in the real sense, an intelligent control agent.

Model of an intelligent control system of hydropower based on Multi-Agent Theory
A control system uses different control types.According to the operational characteristics of hydropower stations, the whole system uses hierarchical control types, with each subsystem under the central control.
According to the requirements of hydropower operational control, the station operational control system based on Multi-Agent Theory can be divided into three types: a communication module to establish contact with the network and receive and send information from and to the network; the agent federation module composed of several sub-modules (agents) to achieve economic operation of the station, by determining the optimal load distribution among the units through coordination and communication; the federal agent module, which controls each unit and ultimately puts the optimized parameters into practice for each individual unit.In each unit, the federal agent composed of several functional sub-agents to control the operation of the units, simultaneously working with the upper level agent.The specific model is shown in Figure 1.

The categories and basis functions of agents in the system
In the model, the entire system can be divided into three categories: communication agent, management agent, and function agent.The communication agent is primarily used to coordinate communications among agents.
In a multi-agent system, every agent communicates with the others, and if we do not categorize like agents together to simplify communications among agents, the entire system would be a very complex communications network.The use of the communication agent makes the entire system of communications very simple, rapid, and efficient.The communication agent receives information, classifies accepted information, and sends the information to agents with different functionalities.
The management agent is the central control unit of the entire system.The entire system has three levels of management agents.The top management agent is used to coordinate and manage the optimal load distribution system and the optimal start-and-stop-sequence-confirm system.It is very easy to locate in the level chart of system.There are two levels of management agents in an optimal load distribution system: the optimal start-and-stop-sequence-confirm system internal management agent and the unit control system management agent.The main function of these two levels is to coordinate the functions among multi agents in each subsystem, distribute tasks to each function agent, and receive the results of their completed tasks.Then they analyze the results and ultimately determine whether the subsystem task has been completed, simultaneously transferring the messages that need to be sent to the other subsystems by the communication agent.
The final category is a functional agent.In addition to a management agent, which plays a control role in each subsystem's internal workings, there are a number of functional agent modules.Each functional agent has its own task in accordance with the different task name corresponding to it.Every functional agent receives its task from the management agent, accomplishes the task, and then returns the results to the management agent.

The composition of the agents
Each agent is composed of a generic agent kernel and many functional modules.The agent kernel consists of an internal database, communications coordinator, blackboards, and an executive.Among these, the internal database includes the information about itself (agent-self), the target muster, the world model, and so on.The agent communication processor provides the mechanism (language and codes) with which the agent communicates with the world and other agents.The blackboard supports the agent's internal communications Data Science Journal, Volume 6, Supplement, 30 July 2007 among the various functional modules.The executive completes the assigned tasks and acts as the functional module's implementation control.One agent can have many functional modules.These modules are pre-compiled and executable files, are relatively independent entities, and express some functional properties and functions also exhibited by the agent, which can have fully parallel implementation after being launched by the executive with work coordinated by the blackboard.The next subsection looks at the core structure of the agent.

Internal database of the agent
The internal database of an agent includes a description of the agent-self, a description of the state of the world, and a description of the state of the other agents.

Model of the agent world
Agents are located in a group called the agent world.Every agent needs to have a description of the agent world in order to communicate and collaborate with other agents.The description is not a complete description and only describes the sensory information of the agent.This information is located in the agent's internal database.
It is not pre-defined by system development personnel, but is built during the agent's active use.

Executor of the agent
The function of this part is similar to the process management in an operating system, but its management is done with functional modules.Each agent executes the same control loop.

Joints linking the agent with functional modules
Communication between the agent kernel and the function modules is completed through the blackboard, which provides a set of standard programming joints for communication between the kernel and the functional module and allows for convenient communication between the agent and the modules.

The communication and coordination between agents in the system
In the whole system, the functions of the communications agent are the simplest.Those functions, which link communication between the upper and lower federal agents, receive and classify sent information, so this agent's communication with the other agents is simple and not elaborate.According to the level structure chart, (Figure 1) the main communication and coordination tasks focus on load optimal distribution.as discussed below.

2.4.1
Communications and coordination between the load optimal distribution system and the unit agent federation.
The main function of the load optimal distribution system is, after receiving the load task from the management agent, to determine the optimal load distribution and then send the results to each unit's agent federation.After the agent federation of each unit has received the information, the status of every unit is measured to determine whether it can perform the task assigned by the load distribution system.If it cannot, then this information is returned to the load distribution system.Because every hydropower station consists of at least two units, Data Science Journal, Volume 6, Supplement, 30 July 2007

Figure 1 .
Figure 1.System level structure chart