Zhisong Li, Doctoral Dissertation, University of Cincinnati, August 2012
In the global effort of exploring low-cost renewable energy and reducing greenhouse gas emissions, energy sources in the ocean are receiving more and more attentions. The kinetic energy from ocean currents is enormous and virtually inexhaustible. Among the different kinds of ocean currents, the tidal flow is most predictable and sufficiently rapid for power generation to many well-chosen sites. To convert the stream momentum into electrical power, a device called marine tidal turbine, predominantly horizontal-axis, is used. Working like a windmill underwater, the tidal turbine operates in a unique environment constrained by water-free surfaces and seabeds. The present study will concentrate on the numerical modeling of tidal turbine operations in standalone and array configurations. Dedicated computational fluid dynamics (CFD) models are first created for this particular research application. Based on the solutions obtained from simulations, comparative investigations are then taken to assess the wake characteristics in a turbine farm scenario in order to minimize any unfavorable wake/turbine interactions. The effects of free surface waves and uneven bed terrains are also studied.
To establish the simulations, one steady and one time-transient code are developed using modular programming and parallel processing design. The core solver uses a classic projection method to solve the three-dimensional incompressible flow problem and an efficient σ-coordinate method to model the free surface, both of which are well verified with analytical solutions. An actuator disc model and an actuator line model are numerically implemented for the steady and unsteady codes respectively, validated by experimental data. To include the natural water waves and ambient turbulence unsteadiness, the inflow boundary condition in time-transient code incorporates a fifth-order Stokes wave generator and artificial velocity fluctuations. Foreseeing the rotational movements and anisotropic turbulence from turbine spinning, a second-moment closure or nonlinear eddy viscosity assumption is needed for turbulence modeling. The project adopts an explicit algebraic Reynolds stress model, balancing the demand for solution accuracy and computational economy.
Extensive tests and case running have been performed using the simulation codes with parametric modifications for different evaluation purposes. In scalability tests, both codes can give acceptable speedup up to 30 nodes and the steady state code achieves better performance due to its highly explicit formulations. The steady and unsteady codes are compared as baselines cases and they agree well in predicting wake velocity deficits. In steady state modeling of a single turbine, the study appraises the influence on turbine wake from rotor size, inflow profiles, and three different simple bed terrains. In unsteady modeling, turbine wake under long waves exhibits some velocity superposition behavior. Turbine array simulations first probe the steady state flow features in a number of rotor configurations: side-by-side, transverse, streamwise, co-rotating, and contra-rotating. They are examined for their wake restoration rates and turbulence intensities. Then the time-averaged wake activities in the staggering and tilted line layout are inspected and compared to settle a fluid dynamic preference. Finally an unsteady modeling is carried out for a pair of upstream-downstream and contra-rotating rotors, enabling a dynamic analysis on the turbine interactions.