Smart Meter Data Characterization and Clustering for Peak Demand Targeting in Smart Grids

The increasing popularity of smart meters deployed at customer sites provides a vital opportunity for network operators to efficiently support and implement demand response (DR) solutions to consumers. Currently, one focus for DR research is to extract knowledge from the smart meters data using data analytics techniques. Defining correct attributes is vital to access and target customers for DR. In this paper, a novel characterization model is proposed for peak load targeting of consumers. This model specifically describes customers' demand variations over one day period with consumption levels ranging from 0 to 1. A k-medoid clustering algorithm and a dynamic time warping (DTW) distance measure are proposed to cluster the characterized smart meter data. The COP index cluster validation technique is used to derive the optimal number of clusters. The proposed model is applied on the publicly available Irish smart meter data, and results show a well-defined grouping of customers based on their variation and peak load contribution.