Showcasing the Applications of Smart Meter Open Data

Smart meter open data is becoming widely available due to the mass installation of smart meters. Consumer data analytic methods are available in literature since several years ago, but they seldom specify the steps taken in data treatment, nor do they highlight the particularities that can arise due to specificities in each dataset. In this paper we showcase data treatment and pattern analysis for three datasets from Belgium, the United Kingdom, and Austria. We demonstrate how to analyse the quality of the raw input data. Each set has different levels of quality in terms of missing data and zero values which are ambiguous. We then show that the literature pattern recognition studies either decompose the data into days, without reference to a specific customer; or into representative customers, using significantly less data of each set to eliminate missing values. In the per day analysis, our results find particularities in each dataset pointing towards installed PV capacity and heat pumps. In the per customer analysis we find that abnormal consumers can create a cluster onto themselves, and we can also observe the seasonal evolution of consumption. Finally, we highlight the state of the art in terms of outlier identification and load disaggregation possibilities explored in literature