Limited Area Forecasting and Statistical Modelling for Wind Energy Scheduling
Abstract
The worldwide deployment of wind energy has continually accelerated during the last few decades, and the implications of day ahead predictability for this highly uctuating renewable energy source for feasible wind power integration in electrical grids are mul-tifaceted. This thesis concerns forecast accuracy for operational wind power scheduling. Numerical weather prediction history and scales of atmospheric motion are summarised, followed by a literature review of limited area wind speed forecasting. Hereafter, the original contribution to research on the topic is outlined. The quality control of wind farm data used as forecast reference is described in detail, and a preliminary limited area forecasting study illustrates the aggravation of issues related to numerical orography representation and accurate reference coordinates at ne weather model resolutions. For the o shore and coastal sites studied limited area forecasting is found to deteriorate wind speed prediction accuracy, while inland results exhibit a steady forecast performance increase with weather model resolution. Temporal smoothing of wind speed forecasts is shown to improve wind power forecast performance by up to almost 1 %, and the explanatory value for wind power forecasting of six di erent prognostic and diagnostic weather model variables modelled semi-parametrically is found to di er depending on the local terrain. In terms of wind speed ramp predictability, the study nds consistent improvement for better resolved forecasts, and indications of wind speed uc-tuation phase-drift with weather model integration time are countenanced, which in part explains the faster decline in limited area forecast performance with leadtime, relative to global model forecasts. The limited area forecasting study is rounded o with a demonstration of the feasibility of forecasted wind speed variability for predicting wind power uncertainty. Finally, a statistical postprocessing framework for numerical wind speed forecasts is developed and evaluated, and the proposed methodology made possible the discovery of the lifted index weather model diagnostic as containing systematic corrective potential for wind speed forecasts generated by the weather model studied.
Info
Thesis PhD, 2015
UN SDG Classification
DK Main Research Area
Science/Technology
