The benefits of decentralised energy systems can be realised through the optimal siting of distributed energy systems and the design of highly interlinked district heating networks within existing electrical and gas networks. The problem is often formulated as a Mixed Integer Linear Programming (MILP) problem. MILP formulations are efficient and reliable, however the computational burden increases drastically with the number of integer variables, making detailed optimisation infeasible at large urban scales. To tackle complex problems at large scale the development of an efficient and robust simplification method is required. This paper presents an aggregation schema to facilitate the optimisation of urban energy systems at city scale. Currently, spatial and/or temporal aggregation are commonly employed when modelling energy systems at spatio-temporal resolutions from plant scheduling up to national scenarios. This paper argues for solving different scales separately using a bottom-up approach, while keeping track of the error made by reducing the resolution when moving from building to urban scale. Novel modelling formulations and optimisation techniques are presented. They enable drastic reduction of the computational time (by up to a factor of 100) required to find an optimal solution in reasonable time without sacrificing the quality of the results (no more than 1% loss in accuracy). A density-based clustering algorithm enables intelligent division of a large city-scale problem into sub-optimisation problems by creating clusters of different density. In each cluster, the trade-off between centralised and decentralised energy systems and the associated district heating network design is evaluated. A solution is selected based on a local optimisation of the network costs. Demand profiles of each building are assigned appropriately, then at an upper level the energy optimisation problem is solved considering the network losses at lower levels. This method enables large-scale modelling of urban energy systems while taking into account building-scale levels of detail. The clustering method enables assessment of the potential of district heating networks on city scale based on building characteristics and available urban energy systems.