Automatic appliance load signature identification by statistical clustering

By disaggregating load signatures at metering level, we aim to develop new knowledge on load behavior and subsequently devise new applications for customers as well as utilities. The pre-requisite of a meaningful load disaggregation is to have a reliable appliance-based load signature database. This paper presents an automated and statistical approach for load signature identification using clustering methods. One of the key advantages of the clustering method is its capability to identify the major loads accurately. We also investigate the effectiveness on accuracy improvement by pre-filtering the signals prior to cluster-creation and also performing consolidation after using automatic clustering. Actual trial data and results are presented in the paper to demonstrate the performance of the automated identification mechanism.