Effectively Using Multifidelity Optimization for Wind Turbine Design

Research output: Contribution to journalArticlepeer-review

8 Scopus Citations

Abstract

Wind turbines are complex multidisciplinary systems that are challenging to design because of the tightly coupled interactions between different subsystems. Computational modeling attempts to resolve these couplings so we can efficiently explore new wind turbine systems early in the design process. Low-fidelity models are computationally efficient but make assumptions and simplifications that limit the accuracy of design studies, whereas high-fidelity models capture more of the actual physics but with increased computational cost. This paper details the use of multifidelity methods for optimizing wind turbine designs by using information from both low- and high-fidelity models to find an optimal solution at reduced cost. Specifically, a trust-region approach is used with a novel corrective function built from a nonlinear surrogate model. We find that for a diverse set of design problems - with examples given in rotor blade geometry design, wind turbine controller design, and wind power plant layout optimization - the multifidelity method finds the optimal design using 38%-58% of the computational cost of the high-fidelity-only optimization. The success of the multifidelity method in disparate applications suggests that it could be more broadly applied to other wind energy or otherwise generic applications.

Original languageAmerican English
Pages (from-to)991-1006
Number of pages16
JournalWind Energy Science
Volume7
Issue number3
DOIs
StatePublished - 2022

Bibliographical note

See NREL/JA-5000-79836 for article as published in Wind Energy Science Discussions

NREL Publication Number

  • NREL/JA-5000-83409

Keywords

  • computational modeling
  • multidisciplinary design optimization
  • multifidelity optimization
  • wind turbine design

Fingerprint

Dive into the research topics of 'Effectively Using Multifidelity Optimization for Wind Turbine Design'. Together they form a unique fingerprint.

Cite this