Developing and evaluating a probabilistic event detector for non-intrusive load monitoring

Abstract

In this paper we present and evaluate probabilistic event detection algorithm for Non-Intrusive Load Monitoring. Like the other probabilistic event detectors, this algorithm also calculates the likelihood of a power event happening at each sample of the power signal. However, unlike the previous algorithms that threshold or employ voting schemes on the event likelihood, this algorithm employs a maxima/minima (i.e., the extrema) locator algorithm to identify potential power events. The proposed algorithm was evaluated against four public datasets, and its performance was compared to that of other four alternative solutions. The obtained results show that this new algorithm is competitive with the other alternatives in the four datasets. Furthermore, the results also suggest that using an extrema locator instead of a voting scheme, increases the performance of one of the state-of-the art algorithms.