Title: Wind Ramp Prediction.

The number of wind turbines and wind farms in the Pacific Northwest has increased dramatically in the past six years, which represents a significant amount of electrical generation capacity connected to the public electric grid. However, the variable nature of wind sometimes introduces excessive power, or conversely shortages, in power delivery from the wind farm possibly leading to grid instability in the region. Knowing the short-term wind profile for a wind farm would allow system operators to better schedule generation resources yielding better grid stability. This thesis presents a method for predicting the power output of a Pacific Northwest Wind Farm by using data collected from wind anemometers located at the wind farm and from off-site meteorological stations. An auto-regressive moving average model (ARMA) with wind velocity inputs from off-site meteorological stations along with current and past wind velocities from the wind farm was used to predict wind velocity changes up to two hours in advance. The predicted wind velocities were then used to compute the future wind farm power output. A fuzzy logic inference system (FLIS) was used to detect and classify wind power ramps. The FLIS provides outputs indicating the degree of membership of power ramps from 10 to 50 % of the nameplate rating of the wind farm. Wind Power Ramp prediction capability will allow system operators better management of the grid and reserve generation resources