| Popis: |
The volatility of wind power feed-in threatens system security. Statistical minute-scale power forecasts are used to support grid balancing but fail during ramp events. The aim of the thesis was the development of probabilistic minute-scale power forecasts for offshore wind farms based on lidar measurements and turbine operational data as a physical-based alternative. Firstly, a method based on dual-Doppler radar data was extended to the needs of single long-range lidar scans. Next, uncertainty related to wind speed height extrapolation and atmospheric stability was characterized. The forecast availability and skill for wake-impacted turbines were enhanced by combing lidar and operational data. Finally, the forecasts of wind turbines were aggregated to a probabilistic wind farm power forecast using copulas. Our work proved the potential of observer-based forecasts and gave valuable insight into the dependence of forecast skill on atmospheric conditions and the measurement set-up. |
