Evaluating the determinants of household electricity consumption using cluster analysis

Abstract

Identifying the determinants of household electricity use is a key element in facilitating the efficient use of energy. Even more so, segmenting households into well-resolved and characterised groups makes it possible to explore electricity use trends at more disaggregated levels, revealing consumption patterns and reduction opportunities for different consumer groups. Considering such groups, the drivers and implications of consumption trends can be better understood, bringing new insights into electricity use and offering opportunities to target policies and interventions that represent the needs of population sub-groups. For this reason, the aim of this research is to develop distinct household typologies using a k-means cluster analysis method. This was developed using questionnaire data of 310 households collected in a locality in Botswana. A feature selection procedure that maximises the silhouette was also developed to select the variables with the most significant clustering tendency. The analysis resulted in four distinct groups that are distinguishable by dwelling type, tenure, the number of rooms, the number of bedrooms, annualised electricity consumption and the number of appliances. The clusters identification enhanced the understanding of the fundamental factors underlying electricity consumption characteristics of different household segments. With this known, it is possible to identify those groups that offer the greatest energy saving potentials, thus providing insights for targeted demand-side management (DSM) and other possible strategies aimed at efficient energy use by customers. • A k-means clustering method and a feature selection method that maximises the silhouettes was developed. • Survey data from 310 households were utilised. • Four distinct groups that offer opportunities for different energy-saving strategies were discovered. • Socio-economic factors were not significant in the discovery of clusters. • The outcome of clusters provide insights that help support DSM strategies.