eFHMM: Event-Based Factorial Hidden Markov Model for Real-Time Load Disaggregation
Lei Yan, Wei Tian, Jiayu Han, Zuyi Li
Abstract
This letter proposes an event-based Factorial Hidden Markov model (eFHMM) instead of time-based FHMM for multiple appliances in a household and integrates the transient signatures into mathematical formula, which is the first of its kind. It decreases computational complexity to be linear to the number of events by performing inference only when events occur and increases accuracy by utilizing transient signatures extracted from event detection based on high-resolution data, thus ensuring real-time accurate load disaggregation. Tests on high-resolution LIFTED dataset of 50Hz demonstrate that eFHMM outperforms other state-of-the-art FHMM variants in both computational time and accuracy in load disaggregation.