Trend based periodicity detection for load curve data
Zhihui Guo, Wenyuan Li, Adriel Lau, Tito Inga-Rojas, Ke Wang · 2013
The authors propose a novel periodicity detection for load curve data that is trend based, therefore, noise resilient. This method models key information in load curve data by a sequence of peaks and valleys extracted from a smoothing curve, and extends Dynamic Time Warping technique to discover repeating subsequences of such shapes while allowing variations due to background noises. Our experimental results show that it is able to detect periodicities more accurately than existing algorithms.