A New Event-detection Method Based on Composite Windows in NILM for Industrial Settings

A wide variety of industrial loads lead to the difficulties of non-intrusive load monitoring method for industrial settings, and the transient states of industrial electric loads vary as well. Accurate and timely detection of status changes of loads is the cornerstone of the entire algorithm. A new event-detection method based on composite-window analysis in non-intrusive load monitoring for industrial settings is presented in this paper. The method is compared with cumulative-sum method and the results indicate that the new method can detect transient state on multiple time scales with an average F1-measure equal to 82.9%, better than 62.3% of the state-of-art on industrial loads dataset.