Abstract
Aeroelastic wind turbine simulations are extensively used in the turbine design process to estimate structural fatigue loading. While it is assumed that more stochastic realisations (seeds) and higher resolution simulations will converge to the ‘true’ result faster, a detailed study has not been published to support how many seeds and what resolution is required to achieve an arbitrary level of convergence. In this study, we perform a comprehensive study on the convergence of turbine fatigue statistics as a function of turbulence grid resolution, number of seeds, and the type of turbulence scaling. The study is performed using the HAWC2 aeroelastic code in conjunction with Mann turbulence fields of varying resolution. Key findings are that over 21 seeds are required to ensure a standard error of the mean of less than 5% for fatigues statistics, and a turbulence resolution of at least 2048×64×64 (in the longitudinal, lateral and vertical, respectively) is required to ensure statistical convergence of a 10 minute simulation on one turbine. The presented study provides a valuable insight into the uncertainties involved in turbine fatigue load calculations, as well as recommendations on how to mitigate them.
Original language | English |
---|---|
Title of host publication | Turbine Technology; Artificial Intelligence, Control and Monitoring |
Number of pages | 10 |
Publisher | IOP Publishing |
Publication date | 2022 |
Article number | 032049 |
DOIs | |
Publication status | Published - 2022 |
Event | The Science of Making Torque from Wind 2022 - Delft, Netherlands Duration: 1 Jun 2022 → 3 Jun 2022 Conference number: 9 https://www.torque2022.eu/ |
Conference
Conference | The Science of Making Torque from Wind 2022 |
---|---|
Number | 9 |
Country/Territory | Netherlands |
City | Delft |
Period | 01/06/2022 → 03/06/2022 |
Internet address |
Series | Journal of Physics: Conference Series |
---|---|
Number | 3 |
Volume | 2265 |
ISSN | 1742-6596 |