Open cellular convection (OCC) over, for example, the North Sea is often observed in connection with cold-air outbreaks. It is accompanied by large temporal and spatial variability in wind speed, which affects offshore wind energy in the area. This study uses the global Model for Prediction Across Scales (MPAS), with regional mesh refinement down to convection-permitting scales of 2 km, to simulate an OCC episode in the North Sea, with a focus on wind-speed variability. Modelled data are combined with wind speeds retrieved from satellite data and in situ measurements to investigate the spatial and temporal variability of offshore wind speeds under OCC conditions from a synoptic to mesoscale perspective, and to examine the model’s ability to represent the OCC structures and wind-speed variability. The model can simulate realistic OCC structures and mesoscale wind-speed variability within the limits set by the effective model resolution. Under OCC conditions, significant differences from climatological conditions are found in the spatial wind-speed power spectrum and in 10-min wind-speed step changes. The very high horizontal mesh-cell spacing in the refinement region of 2 km, and the focus on OCC wind-speed variability, makes this the first investigation of this kind using the MPAS modelling framework with mesh refinement.