|
|
Motivations & Objectives:
It is recognized that conventional
sources of energy such as fossil fuel may be a major cause f global warming and
several other pollutions. Recent polls show that customers are willing to pay more for energy from renewable sources
such as wind than they would from conventional sources. Partly because of
this reason, the development
of low-cost, technologically advanced wind turbines becomes a high priority for
the U.S. wind energy industry faced with growing energy demands. Eligibility for federal tax credit for construction of new wind turbines
nowadays has
made cost of generating wind energy more competitive with the conventional
source of energy compared with the past. Nevertheless research is still
required in order to improve extreme and fatigue load predictions used in wind turbine design purpose
since the today's design load cases are merely deterministic intended to provide a
conservative envelope for all possible loads likely to be experienced which may
be over-estimated.
The key objective of this project
is then to arrive at an appropriate strategy to predict extreme and fatigue
loads of wind turbines by studying the correlations between inflow wind
characteristics (that may be described by mean wind speed or coherence structure
of wind turbulence components) and structural responses of wind turbines (such
as blade bending moments) based upon available field measurements as well as
simulation models. By combining statistical analyses of the data on inflow and
structure response, correlation studies, simulation of wind turbine aero-elastic
response, and extrapolation of limited-duration loads, we may arrive at
techniques for deriving site-specific design load spectra for wind turbines sited at complex terrain/topography with
site-specific turbulence and inflow wind characteristics, instead of using
merely deterministic design loads used in today's design standards.
Approaches:
-
Correlation studies using
simple parameters: Correlation studies of the inflow and
structural responses for establishing long-term load spectra are based
mainly on (i) the measured data of a tested 65 kW, horizontal-axis,
three-blade tested Micon 65/13M wind turbine provided by
Sandia National Laboratories through the Long-Term Inflow and Structural
Test (LIST) program and (ii) simulation models for
the wind turbine aero-elastic response developed at
National Wind Technology Center (NWTC). A simple bi-variate linear regression analysis is employed to correlate the
inflow parameters with the response parameters. The primary
inflow parameters are taken as the 10-minute horizontal mean wind speed and
turbulence intensity at hub height level of the LIST tested turbine whereas
the response parameters are the 10-minute extreme flapwise (out-of-plan) and
edgewise (in-plan) bending moment at blade roots. The results may be used to
predict long-term load spectra given two simple site-specific inflow wind
characteristics. However, by considering only single-point statistics of
inflow in this simple approach, coherence structure of the inflow and the
rotation sampling effect due to rotating rotor frame are excluded. (see some
preliminary results here)
|
-
Correlation studies using
coherence
structures
of inflow:
It is recognized that there should be certain inflow parameters that
can better predict the load/response than the mean wind and turbulence
at a single point do. These parameters could be
derived from spatial nature of the inflow. Because of this motivation, we intend to
scrutinize in details of the coherence structure of the wind and
investigate how it could be statistically related to the load/response
of the wind turbine based on measured and simulated data. By studying this we might be
able to determine additional descriptors that could better explain the
variability of the responses. Several approaches can be applied to
detect the coherence structure of the inflow and response including
|
|
|
1. Fourier analysis (power spectral density
function & coherence function)
|
|
|
|
2
Proper Orthogonal Decomposition (POD) |
|
|
3 Wavelets Decomposition
|
|
|
(see some preliminary results
here) |
-
Reliability
analysis for wind turbine design:
Probabilistic models of the inflow and short-term conditional response
parameters derived from the correlation studies of the measured and
simulated data
are employed to establish nominal design loads against ultimate limit states
for use in a conventional load-and-resistance-factor-design (LRFD) format. For a prescribed return period or failure probability, the nominal design
load is derived based upon an inverse first-order reliability method
(inverse FORM). The method permits inclusion or exclusion of randomness in
the gross wind environment as well as in the extreme response given wind
condition, and consequently can uncouple environment variables from
structural response which may significantly simplifies the design procedure. To uncouple environment variables from structural response, the load model
however does not retain the complete probabilistic description of the
response. Instead we may need to ignore the uncertainty of the response
variable, using its appropriate “fractile level” conditional on the random
inflow variables. The simplified model may still be appropriately used if "fractile levels"
of the response are derived based on the use of omission sensitivity factors
(Madsen, 1988) that account for variability in the response term. Local
gradients of the limit state function are determined for calculating the
omission sensitivity factors and derive the appropriate fractile of the
response. This can lead to reasonably accurate design loads while retaining
the advantages of the simplified model (namely, that inflow and response are
uncoupled). Besides, the importance of each random variable to the load
modeling for wind turbine design can be investigated by including the
uncertainty of that random variable and then compare with the design load
from the case where the complete probabilistic model is considered. Finally,
the error of design load predictions due to the linearization of the limit
state function is considered by making second-order corrections using
local curvatures of the limit state surface at the design point. (see some
preliminary results here
|
Back to Top |
|
