A clustering-based analysis of heat demand profiles for energy storage and demand flexibility operations

Large scale increases in electrification of domestic heat, combined with increasing intermittent renewable supply, are motivating the implementation and exploration of various flexibility approaches (time of use tariffs, direct load control, energy storage etc.) to deal with resulting network constraint issues. Understanding of the dynamics of household demand are important to the operation and design of these flexibility measures. Current understanding of heat demand patterns has been heavily influenced by expected seasonal and aggregated daily profiles. However, these aggregations of heating demand are not necessarily useful for understanding how storage and control technology can enact flexibility at the scale of households and neighbourhoods. Variation in demand matters. Using heat-pump demand data from 418 households, as part of the Renewable Heat Premium Payment trial in the UK, clustering and property classification techniques were combined to explore demand variability in the sample population. Previous studies on clustering of heat demand patterns have offered little explanatory power as to why demand clusters exist. To overcome this limitation, clustering is combined with meta-data on the 418 households as an exploration of the ability of property type, property age, heat pump type, number of bedrooms and emitter type to inform on clustered features of demand. The clustering took 45 different attributes into account, using k-means for exploration of seasonality and agglomerative hierarchical clustering for grouping profiles based on the selected attributes. This paper presents key attribute-based heat demand patterns with interpretations that could help system planners develop better tariff plans, and thermal storage control algorithms.