Dynamic modeling of active distribution networks using cluster analysis of field measurement data

Active distribution network (ADN) models generally rely on artificial dynamic responses generated either by simulation software or by small laboratory-scale microgrids and not on real field d ata. Therefore, they are unable to take into account changes in dynamic behavior over time in a realistic fashion. To this end, this paper presents a new approach to build aggregated ADN models based on real measurements. The method leverages the k-means++ algorithm in order to cluster the various dynamic responses. Then, a new formulation of the exponential recovery model (ERM), which deploys linear functions, is introduced for each cluster. To assess the validity of the proposed model, its output is compared with the real measurements acquired within a year in a real substation in Southern Germany with significant distributed generation (DG). The results indicate the capability of this approach to capture a wide range of different grid dynamics.