A building clustering approach for urban energy simulations

Abstract

Within the context of amongst others urban energy planning and energy system design, urban and district energy simulations have gained interest to quantify the energy use of existing districts. To reduce calculation time and in absence of adequate detailed building level data, urban energy simulations often deploy reductive modelling approaches based on a limited set of buildings, or archetype buildings. This may lead to significant modelling errors when the archetype buildings are not tailored to the studied location. This paper explores a building clustering approach that harvests available local building information, e.g. geospatial data, to generate a tailored set of archetype buildings. Focussed on simulating the annual heat demand or peak heat demand, this paper evaluates if clustering on building properties can be an alternative to clustering on the energy key performance indicators of interest to define the tailored archetypes. As consumption data on a building level is often not available, such an approach would eliminate the need to simulate the energy use for all buildings. The results show that indeed clustering on the properties is a viable alternative with robust results for both annual energy use and peak energy demand and a comparable accuracy compared to clustering on the targeted performance indicators.