Dynamic networking of islanded regional multi-microgrid networks based on graph theory and multi-objective evolutionary optimization

With the rapid development of microgrids, dynamic regional multi-microgrid networks are emerging as an efficient and flexible solution in smart distribution networks. To optimize the dynamic networking of multi-microgrid for better economics and reliability, this article proposes an evolutionary optimization method based on graph theory. This article first reviews the need for dynamic networking in multi-microgrids. Then a dynamic networking model is proposed based on graph theory and solved by the nondominated sorting genetic algorithm II, which can provide a set of Pareto-optimal solutions efficiently. The proposed method is tested by a regional multi-microgrid network in an industrial park in Northern China. The results showed that the daily operating costs of the regional multi-microgrid can be reduced by 17.4% with the proposed dynamic networking method. When faults are considered, the daily operation costs can be reduced by 11.0%, and the system average interruption frequency index value is reduced from 0.27 to 0.19. The results also demonstrated the efficiency of the proposed algorithm over other evolutionary computing methods in terms of both runtime and convergence with promising applications in real world scenarios. Through simulations in the IEEE 33-node system, different operation programs are provided according to the preferences of operators under normal conditions. The operating cost and active power loss are reduced under failure conditions, which shows the effectiveness of the proposed method. The work in this article is expected to provide some reference for the application of regional multi-microgrid.

Automated summary

This article introduces an evolutionary optimization method based on graph theory for optimizing dynamic regional multi-microgrid networks. Test results in a industrial park in Northern China demonstrate that the proposed method can significantly reduce daily operating costs and improve system reliability. The algorithm shows high efficiency in terms of runtime and convergence, with promising applications in real world scenarios.