Using pattern recognition to identify habitual behavior in residential electricity consumption

Recognizing habitual behavior and providing feedback in context are key to empower individuals to take control over residential electricity consumption. Yet, it is a challenge to change habitual behavior, embedded in everyday routines. This paper intends to discover whether habitual behavior can be identified by pattern recognition techniques. The data source is an experiment similar to a utility led advanced metering infrastructure implementation.The analysis discovers: (1) persistent daily routines and (2) patterns of consumption or baselines typical of specific weather and daily conditions. Approximately 80% of household electricity use can be explained within these two patterns, with several applicable “profiles” for this population, including: unoccupied baseline, hot working days, temperate working days, cold working days, and cold weekend days.The proposed methodology demonstrates that it is possible to use pattern recognition methodologies to recognize habitual electricity consumption behavior given the intrinsic characteristics of the family. This approach could be useful to improve small scale forecast, and as a mechanism to enable the provision of tailor-made information to the families.

► We use residential electricity consumption data sampled every 15 min.
► Pattern recognition methodologies to put in evidence habitual behavior.
► Principal components analysis highlights recurrent routines in majority of days.
► Cluster eigenvector coefficients highlights weather-dependent consumption patterns.