|
Activity Number:
|
99
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
| Abstract - #300453 |
|
Title:
|
Wind Forecasting Models and Loss Function
|
|
Author(s):
|
Amanda S. Hering*+ and Marc G. Genton
|
|
Companies:
|
Texas A&M University and University of Geneva
|
|
Address:
|
3143 TAMU , College Station, TX, 77843-3143,
|
|
Keywords:
|
Directional data ; Loss functions ; Skew-t distribution
|
|
Abstract:
|
Making accurate predictions of wind speed reduces the risk that electrical grids face after accepting wind energy as a source. Two models for making short-term forecasts are presented that handle the wind vector as either Polar or Cartesian coordinates. Modeling the wind vector with speed and direction is more intuitive but also more difficult since direction is a circular variable. Transforming these into the Cartesian components is commonly done, and a regression model with a skew-t error distribution can provide a flexible fit to the data. The quality of the predictions from these models can be more realistically assessed with a loss function that depends upon the "power curve" that describes the nonlinear relationship between speed and power. The proposed loss function yields more insight into the true worth of each model's predictions.
|
