Nonintrusive appliance load monitoring (NIALM) for energy control in residential buildings

Significant saving in energy consumption can be achieved by better energy management and control in residential buildings. More advanced control approaches require real-time information on the appliances in use. A straightforward method to obtain this information uses a power sensor attached to each household appliance of interest. Unfortunately, this method is costly and requires significant installation and maintenance efforts. A more sophisticated way to obtain appliance-specific data is by disaggregation of total power consumption data acquired at the main breaker level. Such nonintrusive appliance load monitoring (NIALM) uses a single point of power measurement (e.g., of the electric feed for the whole house), combined with special signal processing techniques. Our paper presents overviews and test results for two novel NIALM approaches we are developing and testing. The first system uses an inexpensive off-the-shelf sensor (TED - The Energy Detective) to obtain real power data at a 1 Hz sampling rate. A statistical algorithm, optimally implementing historical data on power draws and power consumption patterns, has been developed for this system. The algorithm is capable of semi-automatic detection of major household appliances and outperforms the known NIALM algorithms in terms of detection accuracy. The second system relies on a custombuilt sensor capable of sampling voltage and current at a rate of up to 100 kHz. At this sampling rate, voltage and current waveforms and their features are used to identify and track a much broader range of household appliances.