Tidal stream power generation offers the prospect of predictable, low-CO2 power at a number of locations around the UK and the world. Previous assessments of tidal energy resources have taken the form of desk studies based on simplified navigational data. Where numerical model data has been used it has been at too low a resolution to capture high velocity tidal flows constrained by coastal topography. Analytical solutions for maximum energy extraction in simple tidal channels have been produced, but they have not been extended to more complex open-boundary cases such as flow around headlands and islands. There is therefore a role for site-specific numerical modelling, which when validated, offers the twin advantages of a high-resolution picture of the resource and allowing simulation of momentum extraction within the model to take place. In order to parameterize the sub-grid-scale momentum extraction in such models, a new analytical model of the velocity reduction in a large array of tidal turbines has been derived. The model extends previous models of large wind turbine arrays and uses analogies with flow through submerged vegetation. It provides an equivalent added drag coefficient suitable for use in a 2-D coastal numerical model. A numerical model of the flows in the region of the Portland Bill headland has been produced, forced by tidal elevations at the free boundary. A site selection exercise was carried out for the Portland Bill location and an area of around 12 km2 was identified as having a high potential for development using mean cubed speed found through tidal analysis of model results without energy extraction. A large tidal stream generator array has also been simulated within the Portland Bill model—linked to the new model for momentum extraction—and was found to have a significant effect on the tidal parameters in the locality. This was the first time that a large tidal array has been simulated in a realistic coastal domain of large extent, with a parameterization that takes into account the interaction of the turbines with the rough-wall flow in the natural state. Results predict that there is a region downstream of the array extending approximately 5–10 km around the simulated tidal stream turbine array in which the tidal stream ellipse major axis is reduced by at least 5%. In the area of momentum extraction the principal semi-diurnal tidal stream ellipse major axis length was reduced by 10–15%.