Open Collections

Psotka, Jillian

University of British Columbia

The accuracy of numerical weather prediction forecasts decreases with lead time due to the propagation of uncertainties and the chaotic nature of weather. Turbine hub-height wind forecasts are skillful only out to ~6 days. The goal of this research is to extend this accuracy horizon. Large synoptic patterns of geopotential heights can be skillful out to >10 days lead time, and many studies have shown that significant correlations exist between these synoptic patterns and near-surface winds. This thesis follows the development of a new method to forecast wind speeds using synoptic patterns. Self-organizing maps are used to cluster past patterns of geopotential heights and their associated wind-speed observations. Forecasts of geopotential height are then matched to the trained map patterns, and the associated hub-height wind distributions are used as probabilistic wind forecasts. This method is evaluated against a climatological forecast as well as hub-height wind forecasts from the Global Ensemble Forecast System. Secondarily, we experiment with averaging over different forecast and observation window lengths from 6 hours to 7 days, to explore trade-offs between forecast skill and temporal resolution. The lead time of skillful forecasts was extended to 10 days during winter and 8 days during spring and summer. Fall forecasts were skillful out to 6-7 days but did not yield significant lead time advantages over traditional forecasting. Shorter time-averaging windows performed better than longer temporal resolutions, suggesting that forecast sharpness may be a limiting factor of the skill of the method. This research contributes to understanding the predictability of wind power at the 1-2 week forecast horizon, which is valuable for optimizing operational power generation and transmission planning.

Text

2024-09-19

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/89264

Atmospheric Science

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/