| Summary: | This paper addresses a major utility and regulator concern of characterizing customer net electricity consumption profiles to realize integrated distribution system planning. This is pivotal in assessing the capability of the power system to accommodate net load variability and its impacts on the grid such as voltage rise, narrowing peak demand duration, and reducing the cost of energy storage. Although the extant literature has focused on load clustering, this paper uses a symbolic aggregate approximation-based (SAX-based) dimensionality- reduction and k-means techniques to cluster net consumption of smart meter data for more than 3500 residential customers in a month at different temporal resolutions. This study proposes the use of cumulative explained variance in the principal component analysis to determine the optimal number of segments and dimensionality of the transformed space during discretization while retaining the data integrity instead of using intuition, as proposed by the extant literature. Also, this paper describes a screening methodology to determine the distribution of high-voltage customers among the resulting clusters of customers with and without on-site solar photovoltaic generation at different time resolutions.
|
|---|
