Abstract

In this paper, we identify hybrid heating systems on a single residential customer’s premises using smart meter data. A comprehensive methodology is developed at a generic level for residential sector buildings to identify the type of primary and support heating systems. The methodology includes the use of unsupervised and supervised learning algorithms both separately and combined. It is applied to two datasets that vary in size, quality of data, and availability and reliability of background information. The datasets contain hourly electricity consumption profiles of residential customers together with the outdoor temperature. The validation metrics for the developed algorithms are elaborated to provide a probabilistic evaluation of the model. The results show that it is possible to identify the types of both primary and support heating systems in the form of probability of having electric- or non-electric type of heating. The results obtained help estimate the flexibility domain of the residential building sector and thereby generate a high value for the energy system as a whole.