A complementary unsupervised load disaggregation method for residential loads at very low sampling rate data

Abstract

Abstract In this paper, low-resolution smart metering data analysis is applied to breakdown the load consumption for household appliances to facilitate the deployment of residential energy management solutions. A complementary NILM approach for smart meter data with a very low sampling rate is designed to disaggregate the energy consumption data for both household space thermal appliances and small appliances. The load disaggregation approach relies only on the general activity time usage data and public statistical data for appliance consumptions. A novel three-step disaggregation topology is applied to complement NILM problems. The first step is separating white appliances loads using existing NILM algorithms (commercial algorithm of WATT-IS company). The second disaggregation step couples an edge detection technique with a k-means cluster method to detect the ON/OFF event for heating and cooling loads, from the residual aggregated loads. Finally, a novel disaggregation approach using a dynamic fuzzy logic model and a predictive method is applied for the remaining aggregated loads to identify the ON/OFF event occurrence for small appliances. The proposed method is validated using French household dataset with 10\xa0min sample data rates. Then it is applied to disaggregate the load consumption for Portuguese household dataset with 15\xa0min sample data rates.