Coordination Control for Distributed Energy Based on Multi-Agent System
ZHENG Yi,SHI YiXiang,LUO Yu,GAN ZhongXue
DOI: https://doi.org/10.1360/n972017-00659
2017-01-01
Chinese Science Bulletin
Abstract:In the future, energy utilization structure will be centralized plant combining with decentralized and distributed energy resources. With the development of clean and renewable energy, much more difficult control problem occurs simultaneously. Centralized control is suitable for existing system while not sufficient to future. We propose a multi- agent system to control distributed energy resources, which is a typical idea that information can exert influence on physical system. The most important design of the control system is communicating protocol between two energy points. Traditionally in multi-agent system, the communicating protocol is used to get to emergence of consensus, which means all the state of agent become the same. However, in energy system, what we considered is not the consensus but optimal allocation. We rewrite the multi-agent protocol to make it suitable for our demand: to get optimum not consensus and we make it under some specific assumption. In this paper firstly, we demonstrate the convergence and optimum of the algorithm whereby detailed mathematic calculation. Secondly, an ideal 8 distributed energy systems model is used to study the process generated by the protocol. Besides, different topology of distributed energy points is considered carefully and the detailed reason why the efficiency can be improved is discussed. Finally, we apply this method to a realistic system to make further study. Both the mathematic demonstration and simulation results show that the distributed algorithm can be successfully used to improve efficiency of the system. In mathematic demonstration, we use induction to prove that final result must be better than the origin only by every two points′ communication. Without centralized control, the system will evolve into a better allocation. Then in simulation, we study the process detailly. A typical overload working condition is studied. There are 8 points in the system and each point has its own load and load curve. When we set a point′s load increasing by 50%, we get the synergy of the whole system. Traditionally, centralized control will dispatch the load to other points to make sure that the overload one is not so bad. Here, without any external command, the system regulates itself automatically. The overload is shared by other points around the broken one and how much other points share depends on its ability. Furthermore, we study the influence of the topology of system. The conclusion is that with more complete information topology, which means there are connection between every two points as much as possible, the convergence speed will be faster but the final convergence point is the same. However, if the topology is unconnected, the final result is totally different with connected one. Finally, the algorithm is applied to a distributed energy system with generation, energy storage and users. The result shows even the load curve of each point is much more difficult than ideal one, the algorithm still works and the average efficiency is improved. In this paper, the most creative idea is switching multi-agent consensus algorithm to distributed control method. Although some characteristics are neglected, the result can assure the essential validity of the algorithm.