Non-intrusive load monitoring plays an important role in obtaining electrical power information, which is conducive to strengthening demand-side management, guiding users to use electricity, and realizing energy conservation and consumption reduction. In terms of the variety of electrical appliances and the changeable working state, the means-shift clustering algorithm is utilized to construct load power templates. Based on the load power characteristics, the state probability factors are introduced as a new feature into the fitness function, which can be solved by the genetic algorithm. Furthermore, abnormal state correction is performed on the disaggregation results. Simulation results show that the proposed method can be well applied to non-invasive load identification.

non-intrusive load monitoring, load power templates, state probability factors, genetic algorithm