Forecasting the Usage of Home Appliances with Denoised Signal Patterns



As the Internet of Things becomes reality and software appears in every device and in every building, we are interested to reach a level of ambient intelligence that promotes the understanding of our home real usage. Thus the need for forecasting electrical home consumptions to help changing human behavior towards energy saving for a sustainable environment. For this purpose, in this paper we propose a Non-Linear AutoRegressive with eXogenous inputs (NARX) forecasting consumption model of an electrical aggregated signal.
Support Vector Regression (SVR), which exhibits good generalization due to its regularization capability, was successfully tested on onsite real data measurements.
The method yields good to excellent results when analyzing the Mean Squared Error (MSE) and Squared Correlation Coefficient (SCC) on the household collected test data.


Energy Consumption Forecasting, Support Vector Regression


Energy Consumption Forecasting


19th edition of the Portuguese Conference on Pattern Recognition. 2013

Cited by

No citations found