Appliance Load Disaggregation based on Bayesian Sequence Estimation using Importance Sampling

Appliance load disaggregation (ALD): Disaggregating the total power demand of an household into power demands of individual appliances can provide valuable information to consumers. In this paper, a non-intrusive ALD method based on Bayesian sequence estimation (BSE) is presented. Given a sequence of measured aggregated power demands of the house-hold, the proposed method estimates the maximum probable state sequence of each appliance in the household by utilizing the Viterbi framework. In addition, this paper proposes three importance sampling (IS) methods namely uniform, transition and weighted transition methods to reduce the complexity of the proposed BSE based ALD by state reduction. Furthermore, the proposed method has a backtracing mechanism to smooth the estimated state sequence based on the future measurements. The proposed method is compared with the particle filter (PF) based ALD using two well known real-world data sets. The simulation results show that the proposed method achieves higher estimation accuracy than the PF based ALD method depending on the chosen IS method.