|
This paper shows a method to obtain optimal chord and twist distributions in wind
turbine blades by using genetic algorithms. The distributions are computed to maximize
the mean expected power depending on theWeibull wind distribution at a specific
site because in wind power systems optimization is highly site dependent [1]. This approach
avoids assumptions about optimal attack angle related to the ratio between the
lift to drag coefficients.
Genetic algorithms are global optimizers that have a wide trade-off between exploration
and explotation on the space problem. The geometry definition of a wind
blade is a problem with many degrees of freedom being suitable to fall in local optima,
which can be surpassed using evolutionary methods. Evolutionary Algorithms
are frequently used as powerful optimization methods. They are stochastic methods
inspired in the natural process of evolution[2, 3], and among their advantages are their
global search due to the management of a population of candidate solutions instead
only one, and also the only requirement of the knowledge of the fitness function value
to perform a evolutionary optimization, without any other consideration such as derivability
or continuity of the function. Many different optimum design problems in
multiple fields of sciences and engineering have been solved outperforming any other
previous results with evolutionary algorithms [4].
To optimize chord and twist distributions, an efficient implementation of the Blade-
Element and Momentum(BEM) theory [1, 5, 6] is used. It is basically a simplified
theory that is used routinely by wind power industry because it provides reasonably
accurate prediction of performance. The BEM theory has shown to give good accuracy
with respect to time cost, and at moderate wind speeds, it has sufficed for blade
geometry optimization.
In the implementation of BEM, the sophistication is dismiss to reduce computational
cost. The time required to evaluate the forces in a typical turbine is in the order
of milliseconds, which allows massive evaluation of trial turbines. The implementation
is validated by comparing power prediction with the experimental data of the
Risø test turbine that is one of six experimental turbines widely tested by the IEA. The
data are contained in the Annex XVIII report [7] and in the public database of rotor
performances at the ECN.
High quality in results is obtained until the stall zone, about wind speed of 13m/s
proximately. Predictions are used to compute the mean power that is the fitness function
in the genetic algorithm. The mean power, which is proportional to the annual
generated energy, is obtained by averaging power predictions with the probability obtained
from the Weibull distribution of the specific site. To obtain the optimal blade,
the upper and lower limits of chord and twist are needed as well as an optional upper
limit of the blade area. |
|