The power big data load decomposition technique utilizing electric meter data is the key factors in the novel power system, especially NILM, because the detailed physical behavior and mathematical model of each load components is the basis of load modelling and demand response management. In this paper, a NILM algorithm based on noise filtering and Conditional Random Field model (CRF) is proposed. The data is firstly preprocessed by two filters with first-order lag filter and glitch filter, and then each load in the household is modeled by CRF. Finally, the effect of signal processing on model learning is verified by evaluating the results before and after filtering.

non-intrusive load monitoring;load disaggregate;conditional random field;glitch filter