Impact of Temporal Window-size on Pattern Denoising of a Smart Home Electrical Signal



A Smart Home is expected to be energy-efficient. Detailed information for each appliance in the home can influence the consumers behavior. A Non-intrusive load monitoring (NILM) system acquires signals of aggregated consumption of the electrical network in order to extract useful information to proceed to device identification. The successful load disambiguation can be seriously compromised by noise. In this work, we investigate an approach to denoise an electrical signal which combines techniques of Embedding, Wavelet Shrinkage and Diagonal Averaging. In particular, we analyze the influence of the window-size parameter used in the embedding procedure on the denoised results. The computational experience was focused on the whole-house consumption signal and shows that a value near half the length of the signal is the most suitable value for that parameter.


Signal Denoising


Signal Denoising, Non-Intrusive Load Monitoring Systems


17th Portuguese Conference on Pattern Recognition (RecPad), October 2011

Cited by

No citations found