Towards an Improved Understanding of Statistical Extrapolation for Wind Turbine Extreme Loads

Abstract

** † One of the design load cases to be evaluated in the current edition of the design standard for wind turbines as prescribed by the International Electrotechnical Commission (IEC) requires that nominal loads associated with a return period of 50 years be established. This must usually be done by carrying out simulations of the aeroelastic response of the turbine. This design load case is for an ultimate limit state and in order to be able to estimate this rare nominal load, it is necessary to have, as starting point, extreme loads data that are of an adequate quantity and quality to facilitate robust long-term predictions of the load. Practitioners attempting to follow the guidelines set forth in the IEC standard have voiced concerns about aspects of the load extrapolation as expressed therein—for instance, questions have arisen related to the minimum number of ten-minute turbine response simulations that should be performed, about whether only a single (global) maximum load from each simulation should be saved or whether, alternatively, several time-separated (block) maxima may be preferred. Also, while it is clear that not all turbine load types are influenced by each wind speed between cut-in and cut-out to the same degree, no discussion about the desired effort required for each wind speed bin is presented; it is important that simulation effort be focused mainly on wind speed ranges that control the highest load for each load type. The present study attempts to answer all these questions. Using extreme load statistics (global and block maxima) for four different load measures derived from aeroelastic simulations on a 5MW turbine model, we study extreme loads as a function of wind speed (we call these short-term load distributions). For different block sizes (time separations), block maxima are tested for independence and empirical load distributions are compared when global and block maxima are used. We aggregate the short-term load distributions (conditional on wind speed) to the long-term level by integration over all wind speeds. We present convergence criteria which serve to assess whether or not an adequate number of simulations has been performed. Throughout, we highlight the importance of striving for efficiency with regard to the effort expended in running simulations by staying cognizant of important wind speed ranges that are design drivers for each load type. Together, all of the above lead to a proposal that we make for addressing load extrapolation that focuses on efficiency, that spells out how to employ either global or block load maxima, and that provides easy-to-use convergence criteria for deciding on an adequate number of simulations that must be performed before attempting long-term load prediction using extrapolation.