Abstract
Renewable energy sources receive increasingly more attention as the world seeks cheap, reliable and less CO2-intensive ways of producing electricity. Wind is recognized as an important source of renewable energy and reaches worldwide an installed capacity of 732 GW.
However, being intermittent, wind power may imply difficulties to balance supply and demand, triggering curtailments and financial losses. In this context, accurate wind supply forecasts become imperative.
The present research merges market experience with complex statistical and applied probability techniques in order to improve our understanding of wind power forecasting. We gradually approach the problem by studying wind speeds with a daily resolution, then with an hourly granularity and finally, to a refinement of supply forecasting. In practice, we employ spatio-temporal modeling methods to predict wind speeds in the sites that are most relevant to the wind generation. At a later stage, we develop a grid-wide production model encompassing the onsite wind speed forecasts. We perform this study on the wind farms of Southern California power price area.
The main contributions of this research are (1) a novel technique of downscaling and forecasting wind speeds at wind farm locations and (2) a market-wide power production model that considers real market constraints.