A simulation-based procedure for reliability anaylsis of wind turbines
MetadataShow full item record
Deterministic procedures for wind turbine design in use today provide a conservative envelope for structural loads likely to be experienced. This research addresses an alternative approach based on a probabilistic framework that integrates the use of simulation techniques and reliability-based load extrapolation for establishing loads associated with any specified reliability level. Most commonly ten-minute simulations of inflow turbulence and wind turbine response are needed to empirically derive “short-term” extreme response statistics. Reliability-based approaches then extrapolate such statistics to “long-term” extreme loads for exposures associated with a typical desired turbine service life on the order of 20-50 years. While the short- and long-term load statistics require complex modeling for full field characterization of the inflow turbulence and the turbine response, computations of this nature can be expensive often involving extensive simulations. The alternative procedure proposed involves efficient low-dimensional in- flow representations using Proper Orthogonal Decomposition (POD) to address the short-term problem as well as an inverse reliability framework for the longterm problem. In the short-term problem, the POD techniques employed yield simpler representations of the inflow field that preserve spatial patterns of turbulence energy distribution over the rotor plane and thus give rise to realistic, though approximate, turbine response statistics. Usefully, POD techniques also provide physical insights into site-specific spatial coherence structures of the inflow stochastic field. For the long-term problem, an Inverse First-order Reliability Method (Inverse FORM) is employed that allows rapid searches of critical combinations of inflow parameters that drive turbine loads. A considerably smaller number of short-term response simulations is then needed at each search point to arrive at the design load in question. The overall probabilistic design framework that results from merging these two procedures is then efficient, accurate, and can be conveniently applied in practice for the design of wind turbines. In this dissertation, first, the benefits of each of the two procedures—POD and Inverse FORM—to the short-term and long-term problems are addressed separately. Following this, illustrations of this integrated approach on utilityscale wind turbines of different types are presented to demonstrate its efficiency and accuracy in deriving design loads.