Fast Stability Enhancement of Inverter-Based Microgrids using Machine Learning Techniques

With the increasing penetration of inverter-based distributed generators (DGs), the stability problem of microgrid (MG) become challenging due to the intermittent features of DGs. To solve this challenge, this paper proposes a machine learning based method to enhance the system stability in an online fashion. During the design process, we first construct an optimization model to generate the optimal control parameters (i.e. the cut-off frequencies of the low-pass filters (F
pi
, F
qi
)) required for stability improvement. Then, regarding the active and reactive power injection at bus i (P
i
^inj
, Q
i
^inj
) as input variables and F
pi
, F
qi
as output variables, we utilize neural network to learn the mapping relationship between them. In order to reduce the dependence on global information, we utilize improved Relevant Features (Relief-F) algorithm to select the important local measurements, such that the designed controller can be implemented in a local manner. Compared with the traditional offline optimization-based method, the calculation speed of proposed data-driven method is much faster and therefore can adapt to fast time-varying system conditions in MG. The effectiveness and advantage of the proposed algorithm is tested in a modified 9-bus microgrid.