Detecting Activities of Daily Living with Smart Meters

Abstract

Smart meters provide us new information to visualize, analyze, and optimize the energy consumption of buildings, to enable demand-response optimizations, and to identify the usage of appliances. They also can be used to help older people to stay longer independent in their homes by detecting their activity and their behavior models to ensure their healthy level. This paper reflects methods that can be used to analyze smart meter data to monitor human behavior in single apartments. Two approaches are explained in detail. The Semi-Markov-Model (SMM) is used to train and detect individual habits by analyzing the SMM to find unique structures representing habits. A distribution of the most possible executed activity (PADL) will be calculated to allow an evaluation of the currently executed activity (ADL) of the inhabitant. The second approach introduces an impulse based method that also allows the detection of ADLs and focuses on temporal analysis of parallel ADLs. Both methods are based on smart meter events describing which home appliance was switched. Thus, this paper will also give an overview of popular strategies to detect switching events on electricity consumption data.