Constrained optimization of offshore wind turbine positions under uncertain wind conditions with correlated data

Abstract

Offshore wind power production is subject to substantial uncertainty due to variability in wind conditions. Planning of new wind parks should take these uncertainties into account by means of stochastic modeling and uncertainty quantification. Wind speed and wind direction exhibit dependence that needs to be properly modeled to yield reliable uncertainty estimates. In particular, the dependence is strong when data are averaged over a short time-horizon, e.g., on a monthly rather than an annual basis. In this work we introduce a stochastic model for wind speed and wind direction using Rosenblatt transformation and an empirical model. This allows efficient numerical quadrature that replaces thousands of Monte Carlos samples by a few tens of model evaluations at quadrature points in parametric space. A robust design formulation for maximizing power production honoring dependent data is proposed and demonstrated on data from the North Sea site Sørlige Nordsjø II.