Simulation of daily total wind energy using a time-series model

The focus of this research is the development, evaluation, and implementation of a mathematical/statistical time-series model to simulate daily total wind energy in the central United States. By modeling the annual cycle, sub-annual cycles, autocorrelation, and random component of the time series, a concise representation of daily total wind energy is provided. The model can be used to create scenarios and probability distributions for wind energy that are conditional on time of year and persistence (autocorrelation). As an example, probabilistic forecasts are generated to determine the probability of exceeding the 80th percentile wind-energy value for two days of the year, based on both a high and low wind-energy value for the previews day. These forecasts indicate that seasonality is a much more important factor in determining a particular day's wind-energy value (and probability distribution) than is persistence in the central United States. Scenarios of climatic change were simulated by altering the amount of persistence in the time series. Changes in persistence had a much larger effect on wind-energy values for a particular day than for longer periods such as a month. Effects of altering persistence were much more pronounced on the day of the year associated with the highest annual cycle value when a low wind-energy value occurred on the previous day, and on the day of the year associated with the lowest annual cycle value when a high wind-energy value occurred on the previous day (i.e., persistence of unusual events).