RANS-based Shape Optimization of Dual-Rotor Wind Turbines using Variable-fidelity Models

摘要

Dual-rotor wind turbines (DRWTs) may be capable of greater power capture when compared to their single-rotor counterparts. The design and analysis of DRWTs may require the use of computational fluid dynamics (CFD) models, among other tools; such models are often computationally expensive. At the same time, numerous model evaluations are typically needed throughout the design process; this, combined with the potentially high cost of the model, may render the overall computational cost of DRWT design optimization to be prohibitive. This paper investigates the use of variable-fidelity CFD models for the design optimization of DRWTs. The high-fidelity CFD model simulates the fluid flow using the Reynolds-Averaged Navier-Stokes equations with a two-equation turbulence model on an axisymmetric mesh. A physics-based surrogate model, used in the optimization process, solves the same CFD model but with a coarser mesh and relaxed convergence criteria. The surrogate is enhanced with limited high-fidelity information, only one evaluation per design iteration, using multi-point output space mapping. The resulting surrogate is fast and yet reliable. Due to the faster simulations and reduced number of high-fidelity model evaluations the optimization process is accelerated significantly. The approach is demonstrated through the design of DRWTs with up to 11 design variables. For cases with two and three design variables, the variable-fidelity method converges up to three times faster than alternative approaches; for the 11-parameter case, it reaches better solutions while requiring 10 to 30 times less computing time.