Computational Study of Utility-Scale Marine Hydrokinetic Turbines Using Large-Eddy Simulation and Artificial Intelligence
Click Here for Virtual Registration
EVENT DESCRIPTION
Tidal and riverine flows are a predictable source of renewable energy, yet the potential of marine hydrokinetic (MHK) turbines in natural waterways is strongly site-specific, governed by local topography and bathymetry. This dissertation combines device-scale simulations, field-scale large eddy simulation (LES), and machine learning (ML) to assess the energy potential of a real tidal site, the Piscataqua River in Portsmouth Harbor, New Hampshire.
First, LES coupled with bed morphodynamics were performed to investigate the impact of debris accumulation and sediment transport on a utility-scale MHK turbine under rigid- and live-bed conditions, using resolved logs and a novel mid-fidelity debris model. Denser accumulations divert more flow beneath the rotor, reducing the momentum reaching the lower blades and increasing power fluctuations. Under live bed conditions, the spinning rotor amplifies scour around the tower, while debris accumulation controls the sediment dynamics: the scour depth, deposition height, and size of the sand waves migrating along the wake edges all increase with the debris blockage ratio. Morphodynamic processes also accelerate wake recovery, slightly increasing the mean power while substantially increasing its fluctuations.
Second, to overcome the limited resolution of regional models, LES of a 6 km reach of the river were performed under ebb and flood tides using a high-resolution digital map. The simulations were verified through a grid-sensitivity study, tested with four turbulent inflow boundary conditions, and validated against ADCP measurements. The validated LES were then used to (i) identify potential tidal sites based on the predicted power density, (ii) downscale the study area to a 1 km array-scale domain, (iii) assess the performance of turbine arrays at the selected site, and (iv) examine Reynold's independence in terms of power production, enabling future techno-economic analysis.The results showed that the mechanical turbulence generated by the complex bathymetry and hydraulic structures dominates the flow field, overriding both the inflow turbulence and the turbine wakes. The flow is ebb-dominant: downstream turbines lost about 20% of the leading-row power under ebb tide and up to 70% under flood tide, and a 5 m rotor diameter emerged as the most effective configuration for the site.
Third, a novel ML surrogate model trained on instantaneous LES data filtered using proper orthogonal decomposition (POD) was developed to predict mean flow fields downstream of turbine arrays at a fraction of the cost of LES. The POD filtering suppresses the dominant background fluctuations while preserving information relevant to turbine wakes, lowering the mean-squared error by about 80% compared to the same network trained on unfiltered data. Together, these contributions show that tidal energy assessment can achieve both physical fidelity and practical efficiency, bringing it closer to the reliability required for commercial deployment of MHK turbines in natural waterways.
AGENDA
1700 – Social Hour and Arrivals
1800 – Dinner
1900 – Technical Session
SPEAKER
Meriç Aksen, PhD candidate at Stony Brook University
Meric Aksen received his Bachelor of Science from İzmir Institute of Technology and his Master of Science from Middle East Technical University in Ankara, Turkey, where he focused on the computational modeling of landslide-earthquake-tsunami wave interactions. He is currently a Ph.D. candidate in the Civil Engineering Department at Stony Brook University, graduating in December 2026, and a research assistant in the Energy, Environmental & Biological Fluid Dynamics Laboratory of Dr. Ali Khosronejad. As part of the Atlantic Marine Energy Center, a U.S. Department of Energy–funded center that brings together four universities to advance marine energy along the Atlantic coast, Meric led the first high-resolution tidal energy assessment of the Piscataqua River in New Hampshire and developed a machine-learning approach that makes the design of turbine layouts far faster than conventional simulations. He also studies how submerged debris and sediment transport affect the power production and foundations of marine hydrokinetic (MHK) turbines. Meric has published nine peer-reviewed journal articles.
Click Here for Virtual Registration
#HybridEvent