A Knowledge Based Approach for Disaggregation of Low Frequency Consumption Data

Smart electric meters that are capable of measuring and transmitting the power consumed by a household are being installed across the world. The low frequency average consumption data generated by these smart meters are now available widely for application development. A key enabler for the applications is the disaggregation of the consumption data into its constituent components, which can be useful to a wide spectrum of target audience from end user to utilities. Disaggregation of active power data sampled once in a few seconds has been studied extensively in the literature. However, there are only limited attempts to use the once in a few minutes consumption data for disaggregation. This paper discusses some preliminary results obtained using a knowledge based approach for disaggregating the low frequency average consumption data into consumption classes. The approach utilizes the spatial and temporal characteristics extracted from the data to estimate class consumption. The initial results fromthis approach are promising and have inspired to develop a larger probabilistic framework, which is also described in this paper. This proposed framework will be useful in generalizing and scaling the disaggregation algorithms across data of different sampling rates.