Abstract

Transitioning to clean energy is necessary to meet the climate targets of the Paris Agreement. Accelerating decarbonisation requires improving energy efficiency and making large-scale green energy investments, inter alia in residential homes. Household energy behaviours and investment decisions are mostly suboptimal as individuals often face significant psychological barriers and are subjected to cognitive biases. Consequently, one-size-fits-all interventions, that are aimed at fostering green energy behaviours, lead to information overload and rebound effects, thereby being inefficient. A growing proposition in behavioural sciences is to personalise the delivery of behavioural interventions (BIs) to facilitate the uptake of energy-efficient behaviours. This is typically done, for example, by tailoring different BIs to individuals to overcome individual biases in the adoption of green appliances and renovations. Nonetheless, there is no clear know-how to use different statistical methods to tailor BIs. While researchers rely on various techniques to customise BIs for specific groups, this segmentation process lacks coherence overall. In this paper, we systematically review and sort the literature on statistical classification and clustering models, including machine learning methods, that have been used to optimise BIs for improving residential energy efficiency. Our review provides a holistic overview of these different methods, along with recommendations for practitioners to use them. It further highlights the role that machine learning algorithms can play in automating BIs, for example, by using sophisticated data analysis and pattern recognition to identify intricate relationships between decision-making factors. These insights can lead to highly optimised personalised strategies for increased energy efficiency.