Date of Award

Spring 2019

Project Type

Dissertation

Program or Major

Mechanical Engineering

Degree Name

Doctor of Philosophy

First Advisor

Martin M Wosnik

Second Advisor

Thomas C Lippmann

Third Advisor

Chris White

Abstract

Large offshore wind turbine array power plants will soon be appearing along the United States' eastern seaboard. With recent advances in technology, turbines have been growing in diameter and arrays have been growing in scale, but not without technical challenges. There exist knowledge gaps in the fluid dynamics that govern the interaction of the incoming atmospheric boundary layer with the wakes of the turbines and the sequential wake-wake interactions that result in an unavoidable decreased bulk power production compared to predicted capacity from the array.

Wind farm experiments were conducted with scale model wind turbines in a high Reynolds number boundary layer wind tunnel. The studies were conducted in the University of New Hampshire Flow Physics Facility which is the world's largest flow physics quality turbulent boundary layer wind tunnel, with test section dimensions of 6 m wide, 2.7 m tall and 72 m long. The long fetch of the facility offers unique opportunities to study the downstream evolution of the wake of single wind turbines, and the flow through model wind turbine arrays over long distances.

Two different types of model turbines were built for these studies at a 1:500 scale based on the National Renewable Energy Laboratory 5 MW offshore reference turbine. Nine 0.25 meter diameter rotating model turbines, and 95 drag matched porous disks of equal diameter were constructed. The two models were shown to have similar enough wake characteristics that they were then used to build up an experimental array. Several experimental campaigns were carried out and selected results are presented here.

An experimental campaign using an array of porous disks placed in atmospheric boundary layer flow was carried with spacings of 8 diameters in the stream wise direction and 4 diameters in the span wise direction. Far downstream within a wind farm it is proposed that the flow through the farm reaches a fully developed state where the flow field becomes similar from one row to the next. It is suggested that the wind turbine array acts as a sparse displaced roughness, which creates an internal layer whose origin (in the wall-normal direction) remains fixed in space, while the turbulent boundary layer it was placed in continues to grow. To within experimental uncertainty, a fully developed wind turbine array boundary layer condition is observed in the mean velocity, for defined inlet conditions and spacings, from row 12 on.

A careful consideration of experimental uncertainty is discussed due to the large physical scale of the wind tunnel and open-to-atmosphere nature. An expanded uncertainty analysis using the Taylor series method is executed to predict uncertainty for the system of interest in the mean flow. This expanded uncertainty prediction was confirmed by a Monte Carlo simulation. A workable compromise between data acquisition time and uncertainty was used in the experiments, mitigating changing initial conditions due to exposure to atmospheric conditions and temperature changes. This experimental array is sufficiently large to converge on a statistically stationary state in the mean to within 95% confidence level.

Another campaign was carried out with the combination of porous disks and rotating model turbines to study the phenomenon of wake meandering, a dynamic non-periodic shift in the wake over time caused by the atmospheric wake interaction. High temporal resolution velocity time series were obtained at high enough frequencies to resolve oscillatory trends behind individual and coalescing wakes of turbines in boundary layer flow. For single turbine models, incipient wake meandering frequencies decay with downstream distance, along with peak spectral energies and eventually return to those of the incoming turbulent boundary layer flow. However, the meandering also presents itself in the large experimental array, and far downstream the peak meandering frequency is dominated by the turbine spacing, indicating a type of resonance of the array itself.

Porous disk turbine models are the experimental equivalent of numerical actuator disks, therefore, in addition to insights gained into the flow physics of turbine arrays, the publicly available data set is expected to be useful for numerical model validation.

Share

COinS