A novel use of the K-mean clustering technique for operational efficiency in electricity distribution

Abstract

Observed data shows that extreme temperature is the main driver of electricity demand in locations of the world where seasonal variations in the weather are a yearly occurrence. Although utility planners and management are aware of this fact, setting expectations and load projections during extreme temperature events have been a challenge. Extreme temperatures in the winter, usually have a lot of variations with very infrequent occurrences for certain events. Despite these uncertainties, it is imperative to design the electrical systems to be ready for all uncertain events that may come up. This brings about the consideration of economical, safe, and efficient alternatives to shave off or manage the peak demand during extreme temperature conditions. Modeling the behavior of temperature and demand becomes very important for decision-making in this regard. We show that the first step needed to develop such a model of temperature and demand is to split days into groups that share similar temperature trends. The K-mean clustering technique is used in a novel way to do this. The paper gives insight into some interesting behavioral trends and patterns observed in historical temperature and demand data and sets a stage for effective forecasting of electric demand in the industry.