Numerical models of the flow and wakes due to turbines operating within a real-scale offshore wind farm can lead to a prohibitively large computational cost, particularly when considering blade-resolved simulations. With the introduction of turbine parametrisations such as the actuator disk (ADM) or the actuator line (ALM) models this problem has been partially addressed, yet the computational cost associated with these simulations remains high. In this work we present an implementation and validation of an ALM within the mesh-adaptive 3D fluid dynamics solver, Fluidity, under a uRANS-based turbulence modelling approach. A key feature of this implementation is the use of mesh optimization techniques which allow for the automatic refinement or coarsening of the mesh locally according to the resolution needed by the fluid flow solver. The model is first validated against experimental data from wind tunnel tests. Finally, we demonstrate the benefits of mesh-adaptivity by considering flow past the Lillgrund offshore wind farm.