Low Voltage Network Clustering for High Renewable Penetration Studies–An Isolated Network Case Study

Photovoltaic (PV) hosting capacity determination for power systems often entails power network simulation work that is onerous due to the typically large number of medium voltage (MV)/low voltage (LV) feeders, even for small isolated diesel networks. Dividing the power system into several parts has proven itself useful as a means of simulating the system as a whole. The availability of consistent data containing important attributes (e.g., network configuration, load and customer data) for each feeder and identifying the relevant attributes is crucial in clustering. These attributes form the input, on which the clustering is performed. This paper presents the observation of clustering a unique remote network in Western Australia in which PV penetration has been significantly increased over the last decade. K-means clustering methodology is utilised in this work to assign the feeders into different groups. Thirteen different attributes for each feeder were available as a basis for the clustering. This paper also identified the suitable representative feeders for each cluster, which can be used for system-wide simulation work.