A Bayesian approach for site-specific wind rose prediction

Abstract

Feasibility of wind-farm projects emphasizes the need for a timely evaluation of the site-specific wind potential (i.e., the electrical power production from wind sources) because, unfortunately, it is usually hampered by the need for long-term anemometric sampling. In contrast, short-term (i.e., less than one year’s worth of) data may not contain enough information if collected when the wind is not blowing from the prevailing direction(s). From a technological point of view, wind turbine performances drops down when they work one in trail of another, then wind direction plays a strategic role since it determines the optimal wind-farm layout design. By means of a real-case study, a Bayesian approach is proposed and shown to be capable of enhancing the annual wind rose prediction at the given candidate site by integrating the short-term sample data with both historical information (from a neighboring survey station) and expert opinion. Predictive wind direction and speed distributions are obtained marginally at first. Then, the predictive wind rose is derived by non-parametric modeling of the dependency between wind speed and direction based on the short-term data collected at the candidate site.