GC086-0002
Optimizing Buoy Locations to Measure Ocean-waves for Wave-energy Converters
Optimizing Buoy Locations to Measure Ocean-waves for Wave-energy Converters
Monday, 14 December 2020
Poster
Abstract:
\noindent
Applying sparse wave measurements, e.g.,~from wave buoys, as boundary conditions that drive regional-scale wave models affords potential to improve wave predictions in the interior of the model domain. Here, we applied an optimization framework to rank-order wave-buoy locations along the model boundary to ensure that their assimilation into model boundary conditions minimized model-data errors inside the model domain at important locations (e.g.,~where wave-energy converters could be located). This optimization was applied to a SWAN wave model, developed for Yakutat, Alaska. Boundary conditions for the SWAN model were from a basin-scale W{\footnotesize{AVE}}W{\footnotesize{ATCH}}\,III model. The western boundary was divided into 65 potential wave-buoy locations rank-ordered to minimize model-data errors in significant wave height, $H_\mathrm s$, and peak period, $T_\mathrm p$, at two wave buoy locations within the model domain.
\\
\\
\noindent
The \underline{par}ameter \underline{e}stimation software, PEST, was used to perform a linear uncertainty analysis based on the Jacobian matrix that described the changes in model outputs with respect to changes in model parameters $J_{i,j}=\frac{\partial o_i}{\partial p_j}$. In this case, the model outputs, $o_i$ were the simulated equivalents of the wave characteristics at the measurement buoys while the model parameters, $p_j$, were $H_\mathrm s$ and $T_\mathrm p$ at each of the 65 boundary segments. To ensure computational tractability, only the midnight $H_\mathrm s$ and $T_\mathrm p$ of the 65 boundary segments were considered parameters for days 25--29 of September (the five full days where buoys data were available) and the rest of each day's data were tied to those measurements. When the reduction in uncertainties to the simulate buoy measurements were normalized by distances between the buoys and boundary segments, there was a notable correlation between boundary segment distance and the point of interest (in this case wave characteristics at each buoy). Of course, wave direction impacts these results and those effects will be discussed.
\\
\\
\footnotesize{SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525}
Applying sparse wave measurements, e.g.,~from wave buoys, as boundary conditions that drive regional-scale wave models affords potential to improve wave predictions in the interior of the model domain. Here, we applied an optimization framework to rank-order wave-buoy locations along the model boundary to ensure that their assimilation into model boundary conditions minimized model-data errors inside the model domain at important locations (e.g.,~where wave-energy converters could be located). This optimization was applied to a SWAN wave model, developed for Yakutat, Alaska. Boundary conditions for the SWAN model were from a basin-scale W{\footnotesize{AVE}}W{\footnotesize{ATCH}}\,III model. The western boundary was divided into 65 potential wave-buoy locations rank-ordered to minimize model-data errors in significant wave height, $H_\mathrm s$, and peak period, $T_\mathrm p$, at two wave buoy locations within the model domain.
\\
\\
\noindent
The \underline{par}ameter \underline{e}stimation software, PEST, was used to perform a linear uncertainty analysis based on the Jacobian matrix that described the changes in model outputs with respect to changes in model parameters $J_{i,j}=\frac{\partial o_i}{\partial p_j}$. In this case, the model outputs, $o_i$ were the simulated equivalents of the wave characteristics at the measurement buoys while the model parameters, $p_j$, were $H_\mathrm s$ and $T_\mathrm p$ at each of the 65 boundary segments. To ensure computational tractability, only the midnight $H_\mathrm s$ and $T_\mathrm p$ of the 65 boundary segments were considered parameters for days 25--29 of September (the five full days where buoys data were available) and the rest of each day's data were tied to those measurements. When the reduction in uncertainties to the simulate buoy measurements were normalized by distances between the buoys and boundary segments, there was a notable correlation between boundary segment distance and the point of interest (in this case wave characteristics at each buoy). Of course, wave direction impacts these results and those effects will be discussed.
\\
\\
\footnotesize{SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525}