Reduced-Order Modeling of Droop-Controlled Inverters Using Slow Coherency and Aggregation Algorithm

Abstract

By considering time-scale separation characteristics and nonlinear node voltage equation, an improved slow coherency algorithm has been proposed, and for the first time, applied to the slow coherency model widely used for modeling three-phase droop-controlled grid-connected inverters. The method works by combining two coherency clustering results, which in turn, are computed from mode matrix, representing slow mode and damping coefficient of the slow coherency model. Subsequently, fuzzy C-means spatial clustering algorithm and aggregation method are used for identifying coherent generators and building the reduced-order model, respectively. Compared to the traditional slow coherency algorithms, the presented method can accurately identify oscillating mode between power supplies under the condition of selecting any number of areas. Its range of application can moreover be expanded to cover distributed networks with multiple droop-controlled inverters. These expectations, and accuracy and effectiveness of the proposed method have eventually been verified by simulations performed with both detailed and reduced-order models under different faults.