Using clustering techniques to provide simulation scenarios for the smart grid

Abstract

The objective of this work is to obtain characteristic daily profiles of consumption, wind generation and electricity spot prices, needed to develop assessments of two different options commonly regarded under the smart grid paradigm: residential demand response, and small scale distributed electric energy storage. The approach consists of applying clustering algorithms to historical data, namely using a hierarchical method and a self-organizing neural network, in order to obtain clusters of diagrams representing characteristic daily diagrams of load, wind generation or electricity price. These diagrams are useful not only to analyze different scenarios of combined existence, but also to understand their individual relative importance. This study enabled also the identification of a probable range of variation around an average profile, by defining boundary profiles with the maximum and minimum values of any cluster prototypes.