Appliance load disaggregation based on maximum a-posterior particle filter (MAP-PF)

A fine-grained and real-time energy data provided by the smart meters is one of the main facilitators for achieving improved energy efficiency in future smart grids. Further disaggregating the information provided by smart meters into energy consumption of individual appliances, often referred as appliance load disaggregation (ALD), can provide valuable information. In this paper, we propose a non-intrusive ALD based on the maximum a-posterior particle filter (MAP-PF). The proposed non-intrusive ALD allows to identify state of an appliance by utilizing the probabilistic appliance models and the aggregated power of the household, measured by a single smart meter. A combination of multi-state appliances modeled by hidden Markov models and the corresponding aggregated power is modeled by a factorial hidden Markov model (FHMM). The FHMM models the household as a finite state machine with discrete states. The proposed MAP-PF based ALD uses the principles of particle filter to infer the maximum probable state of the household and subsequently, the state of each appliance at a given time under the Viterbi framework. Furthermore, we evaluated the proposed method using a simulated and well known real-world data sets in comparison with the PF based ALD. In our simulation results, the proposed method achieves similar or higher estimation accuracy with significantly less computational complexity compared to the PF based ALD method.