The Simulation Of Multi-agent Systems In The Water Pipe Network Flow Control Applications

In this paper, a Multi-Agent System for Flow-Controlling of Tap-Water Pipeline Network (GWLKS) is constructed based on its physically distributed property in cities and towns, together with the Multi-Agent System Theory in Distributed Artificial Intelligence and Water Engineering Theory, trying to solve the complicate problem of its control. The System's architecture integrates hierarchy structure and coalition structure with network structure according to the structure property of the Pipeline Network. In the system, the water-source nodes the conjunction nodes and the end nodes of a certain Pipeline Network are designed to SourceAgent TubenetAgent UserAgent respectively. TmanagerAgent, designed as the manager, is selected among the Agents in the same circuit. The virtue node of a multi-source Pipeline Network is designed to VirtueAgent according to the concept of "Virtue Circuit". And the relative knowledge on Water Engineering Theory is embedded in these Agents in the system, so it can do the flow-control for the Pipeline Network with the Multi-Agent System. Five typical examples are given and the results indicate that: The system can decompose a complicate problem to different Agents and search the resolution cooperatively in order to reducing the complexity. It can distribute intelligently and control automatically the tube-flow for different types of Pipeline Network quickly and accurately. It can estimate the total supply more accurately by getting users' needs at different moment autonomously, then can control the supply of each water-source. It can meet users' needs with different grade as well as possible at the prerequisite of guaranteeing the tubes' constraint, such as the max-flow, etc.. It's running results can also help us design Pipeline Network and do with troubles in the network. These tests indicate that the system is achieved to the predicated purposes basically.