Methods used to estimate power curves for offshore wind energy systems focus on wind speed without considering the influence of marine and other environmental factors on power generation output and reliability. A machine learning model was combined with a physics-based simulator by Rutgers University researchers to overcome these power prediction limitations.

The analysis applied Gaussian Processes (GPs), a nonparametric statistical learning approach, and the  Rotation and movement axes, and dimensions of the 15 MW reference wind turbine in an offshore environment. Source: Behzad Golparvar et al. Rotation and movement axes, and dimensions of the 15 MW reference wind turbine in an offshore environment. Source: Behzad Golparvar et al.OpenFAST simulator to meteorological data collected from buoys deployed off the New Jersey coast near three future U.S. offshore wind projects.

The research confirmed that in addition to wind speed, air density and relative wind direction are essential in predicting the mean offshore power output. In predicting the design power curve of a 15 MW turbine, the GP-based models showed significant improvements over existing power curve estimation methods.

The results, published in the journal Applied Energy, also underscore the impact of wave-related variables on the second-order properties of offshore wind power and can provide valuable information for power grid integration and storage decisions.

To contact the author of this article, email shimmelstein@globalspec.com