A Data-Driven Feeder Selection Method for Distribution System Planning Studies

Distribution system operators across all jurisdictions depend on simulation models to analyze multiple scenarios and derive investment strategies. Relying on century old antiquated assessments can lead to incorrect and onerous decisions, handicapping electric utilities' strategies. Even though load flow and related modeling and associated analysis is today's industry adopted practice, there are many electric utilities that lack such models for their distribution feeders and can benefit from a balance that entails modeling a strategically defined portion of their systems. Additionally, there are niche studies that do not require running individual models of every single distribution feeder and can also rely on sample analysis and subsequent extrapolation. This paper presents a clustering method to derive representative samples of distribution feeders for common distribution planning studies. This work was driven by the needs of a Caribbean electric utility that operates about 1,400 distribution feeders concentrated in an island, but only about 3% of these feeders have a certified load flow model.