Microgrid working conditions identification based on cluster analysis. A case study from lambda microgrid

This article presents the application of cluster analysis (CA) to data proceeding from a testbed microgrid located at Sapienza University of Rome. The microgrid consists of photovoltaic (PV), battery storage system (BESS), emergency generator set, and different types of load with a real-time energy management system based on supervisory control and data acquisition. The investigation is based on the area-related approach - the CA algorithm considers the input database consisting of data from all measurement points simultaneously. Under the investigation, different distance measures (Euclidean, Chebyshev, or Manhattan), as well as an approach to the optimal number of cluster selections. Based on the investigation, the four different clusters that represent working conditions were obtained using methods to define an optimal number of clusters. Cluster 1 represented time with high PV production; cluster 2 represented time with relatively low PV production and when BESS was charged; cluster 3 represents time with relatively high PV production and when BESS was charged; cluster 4 represents time without PV production. Additionally, after the clustering process, a deep analysis was performed in relation to the working condition of the microgrid.