CFD Investigation of the IEA Offshore 15 MW Reference Wind Turbine Performance in Full Scale: A Temporal Discretization Analysis

Abstract

Abstract A significant increase in wind turbine size was noticed in the past decade due to the expansion of the wind energy market to offshore sites, where steadier and stronger wind resources are available. This continuous rising in the wind turbines size leads to several design and manufacturing challenges. From the design perspective, the prediction of the flow behavior and wind turbine performance is a difficult task, since the use of reduced-scale experimental testing may not lead to a proper representation of the physics in the full scale. In this context, numerical simulations in full scale became a valuable tool to assist the design process and performance analysis of the wind turbine. This paper presents a Computational Fluid Dynamics (CFD) methodology to perform blade-resolved simulations of the IEA 15 MW Offshore Reference Wind Turbine to predict its performance. In this regard, a temporal discretization investigation is performed by testing three different values of Courant-Friedrichs-Lewy (CFL) number. The rotor performance is assessed when the turbine operates in optimal wind-power conversion efficiency, for a wind speed of 10 m/s at hub height. The results are compared against those obtained using the blade element momentum theory implemented in OpenFAST, in terms of power production, generated thrust, and distribution of forces along the blade span. The adequacy of each CFL number is assessed considering the computational cost and accuracy of the results in capturing the physics of the flow. Amongst the CFL numbers investigated, the results obtained with CFL = 1 and 2 presented similar behavior and satisfactory accuracy, while those obtained with CFL = 4 presented unsatisfactory results. Detailed information of the flow field in the wake internal gradient region and the flow structures that detach along the blade span were noticed using CFL = 1, while a significant reduction in the computational demand was achieved with CFL = 2. Based on the results presented, we concluded that for the cases being investigated in this paper, a good balance between accuracy and computational demand can be achieved using a CFL number equal to 2.