Existing tools for sewer network modelling are accurate but too slow for a range of modern applications such as optimisation or uncertainty analysis. Reduced complexity sewer network models have been developed as a response to this, however, current applications are slow to set up and still require high-fidelity models to be run for calibration. In this study, we compare and develop graph partitioning techniques to automatically group sections of sewer network into semi-distributed compartments. These compartments can then be simulated without calibration in the integrated modelling framework, CityWat-SemiDistributed (CWSD), which has been developed for application to sewer network modelling as part of this study. We find that combining graph partitioning with CWSD can produce accurate simulations 100-1,000x more quickly than existing high-fidelity modelling. We compare a range of graph partitioning techniques to enable users to specify the level of spatial aggregation of the partitioned network, also enabling them to preserve key locations for simulation. We test the impact of temporal resolution, finding that accurate simulations can be produced with timesteps up to one hour. Our experiments show a log-log relationship between temporal/spatial resolution and simulation time, which would enable a user to pre-specify the speed and accuracy needed for their application. We expect that the speed and flexibility of the approach presented in this work may facilitate a variety of novel applications of sewer network models ranging from continuous simulations for long-term planning to spatial optimisation of network design.