Inferring Swiss residential occupancy patterns through smart meter analysis

Abstract

The need for energy efficiency strategies in buildings necessitates a deeper understanding of occupancy patterns, given their major impact on energy demand. Traditional building energy models often rely on standardized schedules, leading to discrepancies between predicted and actual consumption. This issue is particularly relevant for residential buildings, where occupancy varies based on household composition, culture, and geography. In Switzerland, although the Swiss Energy Strategy 2050 calls for refined household occupancy insights, research in this area remains relatively limited. This study addresses this gap by proposing a data-driven methodology to infer domestic occupancy profiles from smart meter data. Nearly 5000 anonymized smart meters from 2022, sampled at 15-minute intervals, were analyzed. Residential data was identified based on annual electricity consumption thresholds and classified by household size using time-series k-means clustering. A rule-based algorithm inferred deterministic occupancy states, generating schedules for over 1600 households. Validation against the TimeUse+ dataset, comprising four weeks of activity logs from 1300 participants, demonstrated good alignment and confirmed the method’s validity. The results contribute to more accurate building energy simulations and inform the development of more effective, human-centered energy-saving strategies at the national scale.