MAS Based Adaptive Coordinated Protection System

The safe operation of power system is the foundation of national economic development and social stability. However, under the background of increasingly complex modern power grid, the traditional relay protection exposes more and more problems, and constantly threatens the safety of power grid. In recent years, with the development of artificial intelligence, distributed computing and modern communication technologies, it provides an opportunity to improve the performance of traditional relay protection, wherein the multi-Agent technology and adaptive technology have become the research hotspots. This paper analyzes and improves the problems existing in the traditional adaptive protection based on multi-Agent system,and builds a new kind of multi-Agent grid adaptive coordinated protection system.The specific content summarized as follows:Traditional grid adaptive coordination of protection based on multi-Agent system is simply put the multi-Agent combined with adaptive protection. Researchers focus on the adaptive protection technology, and paid little attention to the Agent’s autonomy and adaptability, which leads to a low adaptive degree of multi-Agent system. From this perspective, the paper introduces the reinforcement learning of adaptive Agent and constructs a more adaptive multi-Agent system. In addition,aiming at the problem that the distributed single Agent protection information is less,and the centralized Agent decision center is too large, this paper constructs a hierarchical structure of multi Agent system. The system structure making the decision by a wide-area protection Agent which has been centralized the wide area information. Station-area protection Agent is responsible for information coordination, and as a backup protection. Interval protection Agent is responsible for collecting information and performing the action. When the communication is failed,the interval protection Agent will use the traditional protection method to realize the fault remove.Real-time analysis of network topology is the premise and key for adaptive protection. In view of the problem that the traditional topology analysis algorithm can not effectively utilize the distributed parallel computing, a parallel topology analysis algorithm based on node merging and hierarchical method is proposed.When in the normal operation of power network, using parallel fusion nodes adjacent table, to static electric Island grid topology analysis, electrical island adjacent toform and hierarchy; When the network topology changes, according to the node where the loop chain table attribute and parallel depth-first search method to update the local network topology and the electrical Island adjacency list. Finally, the simulation is carried out on the quad core AMD computer through the MATLAB parallel tool kit, which verifies the correctness and rapidity of the method.To further improve the system adaptability, this paper combines the adaptive hierarchical structure of multi agent system and the traditional adaptive protection together. And in Netbeans integrated development environment is developed multiAgent system based on JADE, through monitoring platform and the dynamic display of the visual interface, verified the system messaging processing accuracy and the reliability of protection control.