Space-Time Statistics for Wind Power Forecasting

点击翻译按钮获取中文摘要

The technology to harvest electricity from wind energyis now sufficiently advanced to make entire cities poweredby wind a reality. High-quality, short-term forecasts ofwind speed are vital for making this a more reliable energysource. The investigator proposes to study two topics inwind power forecasting, develop relevant methodologies andanalyze their corresponding properties and performances boththeoretically and through Monte Carlo simulations studies.The new methods will be applied to a first-of-a-kind testbedwind dataset from northeastern US. The first topic concernsshrinkage and selection of space-time variables in variousmodels for wind power forecasting. The investigator proposesto combine techniques such as partial least squares and theadaptive lasso with space-time correlation information thatnaturally arises among the predictor variables in the windforecasting models. The second topic concerns realisticforecast evaluations in the context of wind power forecasting.The investigator proposes to develop new loss functions thatare economically relevant for wind power and study theirproperties and their use in building and evaluating forecastingmodels in the wind energy arena.The primary impact of this project is that the new methodsfor wind power forecasting will provide valuable tools toapplied practitioners in wind farming. There is an obviousneed for statisticians to be involved in such importantproblems related to renewable and clean energies for thewell-being of our society. In particular, the research topics address crucial and major challenges for advancing the use of energy from wind.