Using autoencoder network to implement non-intrusive load monitoring of small and medium business customer

Recently, issues such as the greenhouse gas emission, depletion of fossil fuels, and safe use of nuclear energy led many countries and governments to focus on problems of energy economy. Improving energy efficiency of small- and medium-(S&M) business customers has become a challenge. To monitor the energy usage of appliances used by S&M business customers given the advanced metering infrastructure (AMI) meter's data and to establish a cost-effiective energy system, this study employs a deep learning network, the denoise autoencoder, to propose a nonintrusive load montoring system for data granularity of 1 sample/min. By implementing the system and analyzing actual data acquired from 12 test sites, the feasibility of the proposed system is demonstrated, and future research venues are discussed.