Abstract
This document represents deliverable MD1.4and describes the application of EdF’s open-source CFD solver Code_Saturne to large-eddy simulations (LES) of a full-scale tidal-stream turbine (TST). All simulations were performed with realistic approach-flow mean-velocity profiles. The objective of the work was to supplement this with realistic approach-flow turbulence and to extract frequency spectra of the fluctuating loads.
LES was conducted with both zero inlet turbulence and with a fluctuating velocity field synthesised from the stress profiles and length scales generated by a fully-developed channel-flow simulation. Turbulence profiles were further scaled to represent field measurements more accurately. The simulations follow on from Reynolds-averaged Navier-Stokes (RANS) computations reported earlier in MD1.5, but allow a more comprehensive evaluation of fluctuating loads on the turbine.
The main achievements are:
- full LES(up to 17.6M cells) of the 1MW Alstom turbine at fixed rotation rate, including effects of turbulence and mean-flow velocity shear with realistic mean and turbulent velocity profiles from the EMEC test site;
- comparison of turbine loads, including time variation and spectra of blade bending moments, with measurements from the 1 MW Alstom turbine during operation at EMEC;
- Simulations with three different turbulence characteristics and with two different approach-flow speeds to inform changes in behaviour during tidal cycles.
The primary findings of this report are as follows.
- LES is substantially more sensitive to spatial discretisation and gradient-reconstruction algorithms than RANS; in particular, it places stringent requirements on mesh quality (notably skewness), careful meshing near sliding interfaces (to ensure that ghost nodes lie wherever possible within the set of cells abutting the interface), better gradient reconstruction(to compute more accurate advective fluxes), centred differencing(for accuracy and to conserve energy) and smaller time steps(for stability).
- Mean power coefficients predicted by LES with inlet turbulence compare favourably with those measured at the EMEC test site, provided that the hub-height reference velocity is that immediately upstream of the turbine (say, one diameter upstream) and not on the inlet plane; this is because of flow development upstream of the turbine associated with the inlet turbulence prescribed by synthetic eddy modelling (SEM).
- If significant inlet turbulence levels are prescribed, velocity-profile development occurs between inlet plane and turbine. It is unclear whether this is an intrinsic fault of SEM (which can only provide statistically-correct first and second moments of a fluctuating velocity field) or the fact that statistics were supplied from a channel-flow simulation at much lower Reynolds number(the computing requirements for higher-Reynolds-number flow at comparable resolution are impractical).
- Intra-cycle variations of blade loading occur because of approach-flow velocity shear, interaction with the support tower, blade-generated turbulence and approach-flow turbulence. Only an eddy-resolving method such as LES provides any practical means of modelling the last two of these.
- Inflow turbulence makes little difference to mean blade bending moments or mean power, but greatly increases fluctuations in load and the influence of the support tower during a rotation.
- Power coefficients and blade bending moments (mean, phase-averaged and spectra) have been compared with field data. Realistic inflow turbulence is necessary to reproduce the spectrum of fluctuations in loads that are observed in experimental data. The vertical profiles of Reynolds stresses and length scales necessary to use SEM come from detailed channel-flow calculations(at a lower Reynolds number). We have investigated further scaling of these to match the Reynolds stresses and streamwise length scales measured at the EMEC site.
As this is the final Report of the series, a summary of the work undertaken earlier in the project is given as an Appendix