Incorporating Geostrophic Wind Information for Improved Space-Time Short-Term Wind Speed Forecasting

• Activity Number: 81
• Type: Contributed
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Xinxin Zhu*+ and Marc G. Genton and Kenneth Bowman
• Companies: Texas A&M University and KAUST and Dept of ATMO, Texas A&M University
• Keywords: Geostrophic wind ; Lasso ; Space-time statistical model ; Wind energy ; Wind speed forecasting

Abstract

Accurate short-term wind speed forecasting is highly needed for the rapid development of wind energy. However, it is a very challenging problem. Statistical space-time models have been considered to be quite advanced for short forecast lead times (Zhu and Genton, 2012). Hering and Genton (2010) proposed so-called TDD (trigonometric direction diurnal space-time) model to generalize and improve the RSTD (regime-switching space-time) model by including trigonometric functions of wind direction as predictors. In the atmospheric physics point of view, besides wind direction, pressure and temperature are also closely related with wind and should be included into models to improve prediction accuracy. However, directly incorporating pressure and temperature into the TDD model is found not working well. In this article, we introduce a new variable, geostrophic wind that not only extracts information in pressure and temperature but also has physical interpretability, into the TDD model and improved prediction accuracy is achieved based on West Texas data. Additionally, simpler but more efficient methods are proposed to fit the diurnal pattern in wind and better forecasts are obtained.