A non-event based approach for non-intrusive load monitoring

This chapter investigates the use of factorial hidden Markov models (FHMMs) to identify the most likely sequences of appliance states that correspond to the time series of aggregated power measurements. It discusses the probabilistic framework for modelling and estimation of hidden appliance. The chapter discusses the model definition and provides an overview of learning and inference methods for an FHMM. To initialize the load disaggregation model, the initial state and transition probability must be specified. The chapter provides the detail of different models that were considered for hidden appliance state estimation. Hidden Markov models have been widely used to model stochastic processes and are also well suited to model a combination of independent processes. Moreover, empirical evaluations suggest that non‐event based and event based approaches are competitive in performance for recognizing individual appliance operations