Upp í vindinn - 01.05.2017, Blaðsíða 49

Harnessing of wind energy using wind turbines is an essential part of developing a
sustainable energy grid for the future. Some of the countries that are leading in wind en-
ergy development, such as USA and China, are very seismically active and have known
earthquake faults running along a large section of their borders. With growing interest
in utilizing this renewable energy, it is inevitable that wind farms in some countries are
installed close to earthquake sources. In Iceland, for example, the National Power Com-
pany, Landsvirkjun, is planning to build a wind farm in the Búrfell area, which lies close
to potential earthquake faults in the South Iceland Seismic Zone. In the near-fault area,
ground motion is often affected by forward directivity effects. Such ground motions are
known to severely affect tall and flexible structures (see, for example, [7]). Wind turbine
towers are slender and tall structures and are more flexible than common buildings. It
is therefore expected that near-fault ground motions would affect wind turbine towers
in a different way than they would affect the low-rise apartment buildings that are tra-
ditionally built in that area. Existing guidelines for wind turbine design [6,10] mostly
rely upon building design codes (for example, [9]) which do not account for forward
directivity effects. Response spectral shapes specified in design codes are derived mostly
from far-fault ground-motion data and underestimate the effect of velocity pulses com-
monly observed in near-fault ground motion. The dominant period of near-fault veloc-
ity pulse is proportional to earthquake size, with increasing pulse period, the ground
motion becomes more critical to structures with long fundamental period of vibration.
The fundamental period of a typical wind turbine can be close to the pulse period of a 7
Mw earthquake, meaning that large seismic demands can be expected.
The model being used in this study is the one described by [2]. The turbine is a
conventional three-bladed, upwind, variable-speed, with 5-MW rated power. The focus
is on the tower structure; hence the modelling of the nacelle and rotor is simplified as
lumped masses. The base of the tower is considered as ffxed, assuming the structure is
anchored to the engineering bedrock. Tlie tower itself is a steel circular hollow-section
with a diameter and thickness which decreases linearly along the height. A finite ele-
ment model of the tower is created by using linear elastic beam-column elements (100
elements in the tower). Ihe damping ratio used is 1% of critical, which is the recom-
mended value used for most standards [6]. The translational and rotational head mass
applied at the tip along with other relevant parameters of the model are listed in [4].
Undamped natural frequencies obtained from eigenvalue analysis of the finite el-
ement model are presented in Table 1 along with the effective modal mass. The funda-
mental modes are displayed both for side-to-side (SS) and for-aft (FA) motion in Fig, 1.
The effective modal mass for the perpendicular directions are computed separately. For
verification, the frequencies obtained by Bir and Jonkman [1] using the BModes and
ADAMS software are presented in the corresponding columns in Table 1. The dynamic
analysis presented here considers ground shaking in one horizontal direction only: due
to the symmetry of the simplified structured, and since pulse like ground motions are
significantly stronger in one direction, this simplification is justified.
As seen from Table 1 the effective modal mass is nearly the same in both directions,
and the 2 first modes account almost 80% of effective modal mass in both cases. Subse-
quently linear elastic dynamic analysis is performed using time-histories and response
spectral methods. The near-fault ground motion data used in this study is a subset of
data described in [7] For a site close to tectonic plate boundaries or known earthquake
faults, a consideration to near fault effects could be incorporated by use of near fault re-
sponse spectra. For this study the near fault response spectra [7], which will henceforth
be referred to as RR2011, is used and its effectiveness evaluated against the simulated
time history, ground motion response spectra obtained using the SRSS combination
rule (GM SRSS), and the Eurocode 8 response spectra (EC8) scaled by the PGA of each
ground motion record.