Online detection of power events using the end-user communication interface of residential smart meters.

Even though smart meters are a reality in several countries, the interaction that end-users have with these systems is still relatively low. Being an easily accessible source of information, smart meters can be used to effectively promote changes in energy habits and increase the acceptance of different energy programmes. However, a timely, privacy-friendly, and understandable feedback is required to increase the engagement of users with their energy consumption. From the different types of feedback strategies, close-to-real time metrics on per-appliance usage has proven to have significant impact. However, to do this with low-frequency measurements, as those captured by smart meters, several challenges are faced. One of them is the accurate detection of change points in the aggregate signal that help identify the different appliances. This paper aims to implement and evaluate and a series of algorithms for online detection of power events using a live stream of data, collected locally from the household smart meter. These event detection techniques can then serve as the starting point for subsequent clustering methodologies, which in turn can help identify major energy consumers and opportunities for demand response actions.