Finite State Machine Household's Appliances Models for Non-intrusive Energy Estimation

Abstract

Non-intrusive loads monitoring NILM is a set of algorithms that aims to leverage smart meter data by extracting more useful information from the smart meter data. NILM involves disaggregation of individual household loads in term of their individual energy consumption. It is considered as low cost alternative to better understand the electrical network and reduce complexity of the management operations. It offers to households monitoring and control possibilities to their everyday energy consumption. This paper contributes toward non-intrusive energy estimation of household's loads through data-driven appliances modelling approach based on finite state machine models that mimic the real operations cycle. First, the models are built based on features extractions and events clustering via dynamic fuzzy clustering. The resulting clusters are further de-noised and processed to reveal accurate appliances operations states. Then finite state machine models are created using transition probability matrix and an optimization approach to extract the operation cycle that best describe real appliance operations. The evaluation of the framework was performed using two public datasets showing its performance to learn appliances models and energy estimation with an average error of 5% to 22%.