Date of Award

Spring 1-1-2018

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

First Advisor

Julie K. Lundquist

Second Advisor

John J. Cassano

Third Advisor

Peter E. Hamlington

Fourth Advisor

Caroline Draxl

Fifth Advisor

Perdo Jiménez

Abstract

This dissertation explores the interactions between the atmosphere and wind turbines from numerous perspectives. The work presented here outlines three subjects: the characterization of wind-turbine wakes in the evening, the evaluation of simulated wind-power productions in a numerical weather prediction model, and the attempt to systematically quantify wind-speed (WS) variability over decades.

After introducing the background of wind-energy meteorology, the first part of this dissertation discusses the evolution of wind-turbine wakes during the evening transition. In observations as well as simulations from the Weather Research and Forecasting (WRF) model, turbine wakes, namely the in downwind WS reduction and turbulence enhancement, become more prominent in the evening. Hence, the power generations of downwind turbines decrease when the atmosphere changes from unstable to stable.

The second section of this dissertation focuses on validating the power-production predictions of the wind farm parameterization (WFP) scheme in the WRF model. Using the WFP with fine (~12 m) vertical grid resolution leads to the most accurate power simulations. Compared to the actual power generations, the WFP tends to underestimate power in stable conditions with high winds and low turbulence. Overall, the accuracy of the WRF model in WS prediction dictates the skill of the WFP in simulating wind power.

The third topic of this dissertation explores optimal methods to assess the variability of WS and energy production. Among the 27 methods tested, the Robust Coefficient of Variation (RCoV), as a normalized, statistically robust and resistant spread metric, yields the strongest correlation in connecting the variations between monthly mean WS and monthly net energy generation. By comparison to a long data record from a reanalysis product, the RCoV also requires 6 years of WS data to effectively quantify the long-term variability of a location.

Finally, this dissertation ends with a remark on the importance of correctly using the WRF WFP and statistics. Future work includes improving the power curve and applying the variability metrics in evaluating financial risk.

Share

COinS