Using a regime switching Markov model to enhance AI for offshore wind predictions to assist in siting and maintaining turbines.

Authors: Adam Brawn, Shuang Tang

SUNY Campus: SUNY Poly

Presentation Type: Poster

Location: UU 108

Presentation #: 86

Timeslot: Session C 1:45-2:45 PM

Abstract: This study explores the use of a Markov regime switching model to accurately predict wind behavior, with particular focus on offshore wind data. Traditionally, wind behavior has been modeled using distributions such as the Weibull or normal distribution to estimate wind velocity frequency. However, these distributions are often inaccurate in predicting occurrences of both low-speed and high-speed winds. By utilizing different logarithmic distributions, the model can more accurately reflect the full spectrum of wind speeds, leading to improved prediction accuracy and better model verification. This study aims to investigate various methods for determining and verifying the use of a Markov chain model for predicting wind velocity. Data from offshore buoys in the northeastern region, which record wind velocity, will be used to train and test the model for accuracy. By applying real-world wind data, we can determine the appropriate regression coefficients and estimate the state-switching probabilities for the Markov model. The Markov model will consist of at least two states, corresponding to a change in wind speed, allowing for a dynamic representation of wind behavior. Once the best-fit parameters are identified, artificial intelligence (AI) can be employed to continually predict wind behavior. Improved wind prediction models will not only lead to better placement of turbines but also contribute to the overall efficiency and cost-effectiveness of wind energy production. As wind energy becomes more efficient, it has the potential to become a leading energy source, significantly reducing the environmental impact of power generation.