Volume 11: Sustainable Energy Solutions for Changing the World: Part III

A Hierarchical GMM-based Method For Non-intrusive Load Monitoring (NILM) Jingyu Lu, Mingyue Jiao, Xinmei Yuan*

Abstract

By using non-intrusive load monitoring, energy consumption of individual appliances can be labeled through disaggregating the aggregated consumption of an electrical network by data analytical algorithms. Due to the advantage of low cost and easy installation, and the requirements of smart grid applications, NILM has been widely focused in recent years. However, the accuracy of the NILM can be greatly affected by the difference in power resolution of appliances. In this paper, a two hierarchical Gaussian mixture model-based method is proposed to solve this problem. At the 1st hierarchical level, the aggregated energy consumption signals are disaggregated into high-power appliances and low-power appliances. Consequently, at the 2nd hierarchical level, detailed appliances energy usage behaviors can be estimated with adapted power resolutions, respectively. The pubic dataset– BLUED is used to verify the proposed method. The results show that the proposed method effectively improve the accuracy of NILM, particularly for low-power appliances, compared with conventional Gaussian mixture model method.

Keywords Non-intrusive load monitoring (NILM), unsupervised learning, Gaussian mixture model (GMM), clustering, power resolution.