Coordination in Microgrid Networks Using Multi-agent Systems/Modeling and Coordination of interconnected microgrids using distributed artificial intelligence approaches

As renewable sources penetrate the current electrical system to relief global warming and energy shortage, microgrid (MG) emerges to reduce the impact of intermittent generation on the utility grid. Additionally, it improves the automation and intelligence of the power grid with plug-and-play characteristics. Inserting more MGs into a distribution network promotes the development of the smart grid. Thus MG networks existing in the power system are in prospect. Coordinating them could gain a system with high reliability, low cost, and strong resistance to electrical faults. Achieving these profits relies on developed technologies of communication, control strategy, and corresponding algorithms. Dispatching power in distributed MGs while coordinating elements within the individual MG demands a decentralized control system, in which the multi-agent system possesses advantages. It is applied to the MG network for establishing a physically and electrically distributed system. Based on the multi-agent system, this thesis mainly studies the coordination control in the MG network and its modeling. It aims at promoting control performance in terms of efficiency, reliability, economic benefit and scalability. According to the holonic structure, two methods are considered to enable the system scalability, including the coordination with neighboring MGs and within the extensive coordinating area. Then a simulation platform is established to validate the proposed approaches with Python language. The control strategies for coordination between MGs and their neighbors are proposed to maintain the complete load supply and global security operation while minimizing the generation cost. Centralized control in the coordination group is applied for economic energy management. It uses a Newton-Raphson method to dispatch power among neighboring MGs by simplifying the relationship between MG generation cost and its output power into a quadratic function. An average consensus theorem is adopted to calculate the caused network power flow, and the results are compared with the maximal capacity on the line to keep safe operation. To improve the economic benefits, the approximated relationship between MG output power and the caused generation cost is improved in another strategy. It builds a market for neighboring power trade. Here, the bid is derived from the average generation cost on maximal output. This method maintains the operation privacy of individual MG. Power flow calculation is simplified to be proportional to the angle difference between the two terminates of the connecting line. Both strategies are tested on a 13-node, and 34-node network. Their performance shows that both approaches possess scalability and could economically compensate for the lack of load supply in faulted MG. For the control strategy with higher reliability and profit, the coordination strategy within a selected extensive area of MGs is proposed. Expanding the coordination area based on neighboring MGs provides more energy sources to the demanded MG. It ensures enough power to compensate imbalance and offers more choices for power dispatching. The selection of the coordination area adopts the distributed evolutionary algorithm programming to the multi-agent system to accelerate the calculation speed. Quadratic programming in Gurobi is used to solve the power dispatching problem. Another genetic algorithm is also adopted to solve the problem of optimal power dispatching with a quadratic generation cost for microturbine. The performance of this strategy is tested, and the results show that it has comprehensive advantages on reliability, scalability, and profit compared with centralized and market methods. (author)