Small-signal stability enhancement of islanded microgrids via domain-enriched optimization

A domain-enriched optimization algorithm is developed to enhance the dynamic resilience of islanded microgrids, with a specific emphasis on improving small-signal stability. A novel eigenvalue-oriented objective function and accompanying constraints are designed to optimize the controller parameters for the power-electronic interfaces of distributed energy resources (DERs), which are critically important to the system's dynamic resilience. To solve the resulting non-smooth and non-convex optimization problem, we introduce an auxiliary loss term based on microgrid domain knowledge, which takes the form of a multivariate polynomial in the optimization variables. We refer to our proposed method as Domain-Enriched Navigation (DEN), which combines the gradient of this domain-enriched loss term as a navigation strategy with an evolutionary algorithm. By incorporating the navigated update, we successfully address initialization sensitivity and slow convergence commonly observed in ordinary evolutionary algorithms. Numerical tests on typical islanded microgrids validate the effectiveness of our approach in enhancing the small-signal stability. Our results demonstrate a notable improvement of over 1 dB in the objective function, along with much faster computation than state-of-the-art alternatives.