Probabilistic Maximum-Value Wind Prediction for Offshore Environments

• Activity Number: 612
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Risk Analysis
• Author(s): Andrea Staid*+ and Pierre Pinson and Seth Guikema
• Companies: Johns Hopkins University and Technical University of Denmark and Johns Hopkins University
• Keywords: wind forecasting ; maximum winds ; offshore risk

Abstract

High wind speeds pose a great risk to structures and operations in offshore environments, and there is a need to accurately predict periods of dangerously high winds. In contrast to many wind forecasting models, which focus on average wind speeds in a given time period, we present statistical models and training methods to predict the distribution of maximum winds. With a response variable of maximum wind speed in a three-hour interval, we assess the performance of linear models, GAMs, and MARS models using meteorological covariates from ECMWF forecasts. We apply these to actual data for a site in the North Sea. The models are trained to predict the mean value of maximum wind speed, and the residuals from model training are used to develop the full probabilistic distribution around the mean. The models outperform traditional baseline methods, and we show their predictive accuracy across lead times and different training methodologies. These simple statistical models, which provide knowledge of the maximum wind speed, may allow for more informed decisions regarding wind turbine operations, planned maintenance, and power grid scheduling for improved safety and reliability.