An Energy Disaggregation Approach Based on Deep Neural Network and Wavelet Transform

Abstract

Energy disaggregation allows identifying individual consumption of different appliances using only the aggregated signal measured from a single point. This work proposes a neural network trained with wavelets reduced data to perform energy disaggregation. Besides the disaggregation, usually a binary answer by identifying the appliance activation moment, we are interested in estimating the appliance’s consumption value. We consider the U.K.-DALE dataset to perform our experiments, containing data from different appliances of five houses from England. Using our strategy, compared with another well-established work, we achieved improvements per appliance of 11.4% (estimated accuracy) in the disaggregation process and 27.8% ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$F_1$</tex-math></inline-formula> -score) in the appliance’s consumption value. Our main contribution was to identify satisfactorily that the coefficients of approximation of the wavelet transform are enough to estimate the individual consumption of household appliances.