This article studied a year of generation data from the Thames Estuary wind farm cluster for 2014, with particular interest given to a ramping event which occurred on the 3rd of November of that year. This consisted of a ramp-up event of 67% Capacity Factor (CF) in 2h 45 min. Followed closely by a 73% CF ramp-down event, which took 1h 45min.
High resolution forecast data, both deterministic and ensemble, from the UK Met Office was analysed on this day to assess the ability of these forecast systems to predict such extreme ramping events.
Deteministic (UKV) forecasts exhibited a consistent phase error of about 2 hours for all lead times, even for a forecast initialised 6 hours before the event. Choosing the fastest wind speed within 10km of each turbine produced a more accurate representation of the amplitude of the ramp compared to simply using the nearest model grid-point.
Ensemble (MOGREPS-UK) forecasts did exhibit more variability, though a phase error was again evident (a 24 member ensemble using the combination of 03Z and 09Z forecasts on 03/11/14 had 12.5% of its members producing a ramp of more than 40% CF within a 2 hour window of the observed ramp).
Thoughts for our research:
- The use of fastest wind speed within an arbitrary spatial window, as opposed to the nearest model grid point, is worth keeping in mind for future forecast analysis.