Clustering distribution feeders in the Arizona Public Service territory

This paper describes a methodology and approach used to perform clustering on distribution feeders within the service territory of Arizona Public Service (APS). This clustering process was performed in order for APS to characterize the types of feeders in their service territory and to provide a method of feeder classification to inform PV interconnection requirements in the future. The paper begins with a description of the history of clustering techniques for classifying distribution feeders. The method chosen for clustering APS feeders incorporates best practices for the clustering algorithm, stopping criteria, and variable selection. All routines were written in open-source scripting languages. Specific steps in the clustering process are described and applied to the feeders in the APS service territory showing results for each step. The paper concludes with a summary of the work.