Time-Series Based Surrogate Model For Wind Farm Performance Prediction

Abstract

Aeroelastic codes are state-of-the-art simulation tools in both industry and academia for the modelling of wind turbine loads and power output. Although these codes are widely used for the analysis of individual turbines, they are in general computationally too expensive for the calculation of all turbines within a wind farm. Engineering models that are computationally cheaper but also provide a lower fidelity are therefore typically used for wind farm power performance predictions. In this paper, an alternative approach to simulate wind farm performance is presented: the use of a data-driven surrogate model that is trained on time series that were generated by the in-house aeroelastic tool BHawC. This surrogate model provides results with potentially higher fidelity than more simplistic engineering models but is computationally much cheaper than BHawC simulations.