Power Appliance Disaggregation Framework Via Hybrid Hidden Markov Model

Abstract

The recent proliferation of the Internet of Things (IoT) has provided consumers unprecedented connectivity and access to many of their devices via mobile applications and smart home energy management systems. Although many platforms are available for consumers to remotely fine-tune their energy consumption patterns, awareness is still lacking of specific details about their past consumption trends. When contextual data is presented regarding specific appliances that were active in the past along with associated costs, consumers will be incentivized to make power consumption decisions that result in increased cost-savings and energy conservation. In this paper, we propose a hybrid classification system based on the Hidden Markov Model (HMM) and k-Nearest Neighbours (KNN) algorithms for classifying and disaggregating power consumption data of individual households in a non-intrusive manner. We also apply the Pareto's 80/20 Principle for accurate identification of appliances that draw significant power and contribute to majority of energy costs.