Modelling district heating demand: A synthetic dataset for two residential neighbourhoods

The extensive artificial datasets developed in this study capture the energy demands of two districts and, with reasonable constraints, emulate monitoring campaigns typically conducted on-site in inhabited houses. Generated datasets are the following, one 1) representing low-performing building stock from before the deployment of the Energy Performance of Buildings Directive (EPBD) (2006), and the other 2) reflecting high-performing stock constructed after 2006. The buildings in the datasets were simulated representing single-family homes, typical of neighbourhoods in Flanders, Belgium. The datasets were generated using Dymola and the IDEAS package integrated with TEASER. Each simulated house differs in geometry, size, envelope characteristics, occupancy patterns, and installed gas heating systems. Envelope characteristics for older houses were derived from Energy Performance Certificate (EPC) data, and categorized into four construction periods, while newer houses were based on Energy Performance of Buildings (EPB) reports, both evaluated in Flanders. The datasets feature only heavy-weight houses in terraced, semi-detached or detached configurations, modelled as either single- or two-zone buildings, depending on their number of storeys. The simulations incorporate a natural infiltration model and a stochastic occupant behaviour model that sets the temperature requirements and accounts for heat gains from occupants and appliances. Due to the computational effort of large-scale simulations, heating system performance was postprocessed using a data-driven approach assuming gas-fired heating systems. Six heating system configurations were allocated, including condensing and non-condensing boilers, each combined with one of three domestic hot water (DHW) setups: no integrated DHW, direct DHW, and DHW with a storage tank. For all system designs, production efficiency varied with the load ratio. Urban-scale simulations were carried out at 10-minute frequency using weather data for Heverlee, Belgium, from the year 2016. The primary objective of this dataset was to support the development of statistical tools for characterizing the heat transfer coefficient of building envelopes. However, the generated artificial datasets offer a wide range of typically hard-to-measure inputs, making them ideal for evaluating the impact of simulated components on the overall energy balance. While the initial focus was on the behaviour of individual buildings, these datasets are also valuable for urban-scale analyses, particularly in energy planning contexts.