| dc.description.abstract | Tidal turbines convert kinetic energy from tidal currents into electricity in a similar way as wind
turbines extract energy from wind. The wind energy industry being more developped, tidal turbine
designers use therefore the same methods as their wind turbine counterparts to develop their
power plants. The standard engineering method in use is the Blade Element Momentum (BEM)
theory, which combines the division of the blade in blade elements characterised by their hydrodynamic
properties (lift and drag coefficient), their twist angle and their cord length, and the
momentum theory, based on the assumptions of actuator disc, irrotational and uniform inflow,
among others. However, the differences in the inflow and in the rotor inertia between wind and
tidal turbines questions the validity of this method. For instance, contrary to wind turbulence,
ocean waves have a frequency close to the rotor frequency, leading to significant variations in
the inflow within one rotor revolution. To model loads on tidal turbine blades, one can also call
on two other methods, namely model testing and Computational Fluid Dynamics (CFD) simulations.
These methods are appropriate for design optimisation and/or validation but require
respectively expensive test facilities and time consuming simulations on high performance computers.
Therefore, these methods are less flexible than the quasi-static BEM method, for which
results can be obtained relatively instantaneously for various flow conditions.
To get an idea of the accuracy of the BEM method, a two-bladed tidal turbine rotor was designed
and tested in a towing tank with a regular wave field and at runaway, i.e. when the rotor
is suddenly uncoupled from the grid and rotates freely. These experimental setups were also reproduced
in the BEM code, as well as in CFD simulations, using ANSYS CFX. The thrust force
and shaft torque obtained by these three approaches could then be compared to point out the
need for modeling dynamic effects in the quasi-static BEM method. This was done by analysing
the time series in time domain and comparing statistically the mean values, extreme values and
standard deviations of the wave tests. A spectral analysis was also used to check if all frequency
content appearing in the experimental data was also captured by the BEM, and to what extent.
As a result, the mean values of the wave tests match the BEM mean values, as long a there
is no stall on the blade under wave troughs. However, the spectral analysis showed that the
dynamic (expertimental or CFD) thrust and torque signals may be obtained by high-pass filtering
the quasi-static (BEM) values (low-pass filtering the induction factors), observed through
a phase delay and an amplification of the peaks at high frequencies. This can be explained by
the dynamic behaviour of the wake. Dynamic wake models developped for wind turbines may
be used to correct empirically the BEM codes, with coefficients determined using vortex-based
numerical methods. However, vortex codes are not valid for the turbulent wake state, which
is often observed passed tidal rotors in waves. Therefore, a correction of the BEM method, as
simple as those used for wind turbines, are a priori not adapted for tidal turbines in wave fields.
In addition, the actuator disc simplification, necessary in the BEM method, decreases the influence of blade position on rotor loading under regular waves (showing a non-uniform inflow
due to the decreasing wave velocity profile with depth). Therefore, the spectral peaks observed
at sum frequency (wave+rotor frequency) in the experimental results are attenuated or even not
captured by the BEM. As a conclusion, the smaller, but not unsignificant, load cycles at sum
frequency are not taken into account in a fatigue analysis based on BEM predictions and would
lead to a wrong life time estimation.
Concerning the runaway case, the tests highlighted a critical peak in thrust force, rising extremely
fast after the release of the rotor. This peak does not appear on the BEM results because
it arises from the dynamic properties of the wake. This corroborates with the response of a highpass
filter to a step, as previously observed for the wave cyclic loads. The highly loaded rotor
makes the wake reach the vortex ring state, whose dynamics cannot be captured with available
engineering models. This shows the necessity of an efficient control system braking the rotor in
case of runaway.
Many questions were raised during this PhD project and further research in this field is necessary
to include corrections to the BEM method, so that it takes into account the wake dynamics.
Those corrections should be calibrated using CFD simulations and validated by comparing with
experimental results. | nb_NO |
