This paper presents a systematic model order reduction (MOR) algorithm based on state relevance applied to an islanded microgrid with electronic power generation. MOR of such islanded microgrids may not benefit, a priori, from the well-established time-scale separation usually applied to conventional power systems, and a systematic MOR is still an open issue. The proposed algorithm uses a balanced realization of the linear system, where state variables may not have physical meaning, to obtain the states' energies. It then calculates the relevance of the original system states from those energy values. The newly proposed ``state-relevance coefficient'' should help to choose which states to consider in a reduced model in each study case. Detailed nonlinear simulation results show that the proposed algorithm is able to find the relevant states to include in the reduced model systematically, even in operation points near the stability limit, where ad-hoc MOR techniques are likely to fail. The performance of the algorithm is illustrated in a system with grid-forming converters in various scenarios but can be easily applied to other systems.

This paper presents a systematic model order reduction (MOR) algorithm based on state relevance applied to an islanded microgrid with electronic power generation. MOR of such islanded microgrids may not benefit, a priori, from the well-established time-scale separation usually applied to conventional power systems, and a systematic MOR is still an open issue. The proposed algorithm uses a balanced realization of the linear system, where state variables may not have physical meaning, to obtain the states' energies. It then calculates the relevance of the original system states from those energy values. The newly proposed ``state-relevance coefficient'' should help to choose which states to consider in a reduced model in each study case. Detailed nonlinear simulation results show that the proposed algorithm is able to find the relevant states to include in the reduced model systematically, even in operation points near the stability limit, where ad-hoc MOR techniques are likely to fail. The performance of the algorithm is illustrated in a system with grid-forming converters in various scenarios but can be easily applied to other systems.