Real-time and low cost energy disaggregation of coarse meter data

A novel real-time non-intrusive appliance load monitoring algorithm is introduced. By applying finite state machine training, dynamic appliance disaggregation, rule-based filtering and discrete Fourier analysis of coarse data, several improvements are achieved: 1) reduced appliance model training complexity as compared to existing algorithms, and 2) novel and augmented detection stage. The application of this algorithm on real world data demonstrates that efficiency does not have to be compromised by the relatively lower complexity.