Disaggregating high-resolution gas metering data using pattern recognition

Abstract

Growing concern about the scale and extent of the gap between predicted and actual energy performance of new and retrofitted UK homes has led to a surge in the development of new tools and technologies trying to address the problem. A vital aspect of this work is to improve ease and accuracy of measuring in-use performance to better understand the extent of the gap and diagnose its causes. Existing approaches range from low cost but basic assessments allowing very limited diagnosis, to intensively instrumented experiments that provide detail but are expensive and highly disruptive, typically requiring the installation of specialist monitoring equipment and often vacating the house for several days. A key challenge in reducing the cost and difficulty of complex methods in occupied houses is to disaggregate space heating energy from that used for other uses without installing specialist monitoring equipment. This paper presents a low cost, non-invasive approach for doing so for a typical occupied UK home where space heating, hot water and cooking are provided by gas. The method, using dynamic pattern matching of total gas consumption measurements, typical of those provided by a smart meter, was tested by applying it to two occupied houses in the UK. The findings revealed that this method was successful in detecting heating patterns in the data and filtering out coinciding use.