Several studies have indicated a potential to exploit the thermal inertia of individual residential buildings for demand response purposes using model predictive control and time-varying prices. However, studies that investigate the response obtained from applying these techniques to larger groups of buildings, and how this response affects the aggregated load profile, are needed. To enable such analysis, this paper presents a modelling methodology that enables bottom-up modelling of large groups of residential buildings using data from public building registers, weather measurements, and hourly smart-meter consumption data. The methodology is based on describing district heating consumption using a modified version of the building energy model described in ISO 13790 in combination with a model of the domestic hot water consumption, both of which are calibrated in a Bayesian statistical framework. To evaluate the performance of the methodology, it was used to establish models of 159 single-family houses within a residential neighbourhood located in the city of Aarhus, Denmark. The obtained bottom-up model of the neighbourhood was capable of predicting the aggregated district heating consumption in a previously unseen validation period with high accuracy: CVRMSE of 5.58% and NMBE of - 1.39%. The model was then used to investigate the effectiveness of a simple price-based DR scheme with the objective of reducing fluctuations in district heating consumption caused by domestic hot water consumption peaks. The outcome of this investigation illustrates the usefulness of the modelling methodology for urban-scale analysis on demand response.