Sensitivity analysis of a Small-Scale Darrieus vertical axis wind turbine

Resumo

Demand for alternative and renewable energy sources has grown substantially in recent years as a result,
inter alia, of economic and environmental aspects of conventional sources such as oil and its derivatives. In this
context, wind energy has emerged as attractive renewable source. The growth of wind energy in Brazil and in the
world envision the possibility of developing new technologies and more efficient equipment that meet energy
needs. Thus, the objective of this work is to contribute to the development of Small-Scale Darrieus Vertical Axis
Wind Turbine (VAWT) that can be employed in decentralized electrical generation through computational fluid
dynamics (CFD), operating at low tip speed ratio λ (TSR). A 2D numerical modeling is performed considering an
unsteady-state and turbulent flow. Mesh density analysis is ensured by Grid Convergence Index (GCI)
methodology and the numerical robustness is verified through experimental comparison. The geometric
parameters of the VAWT analyzed are: profile camber (m), camber position (p), aerodynamic profile thickness
(t), profile chord (c) and blade incidence angle (α), the first three being based on the NACA-4 digit
parameterization. These five parameters, associated with the operational parameters, are submitted to a sensitivity
analysis using the Smoothing Spline ANOVA algorithm in order to predict the influence of each one of these on
the power coefficient (Cp) of Darrieus VAWT. With a Design of Experiments (DoE) containing 100 possible
configurations of the Darrieus turbine generated using the Uniform Latin Hypercube Sampling (ULHS) method,
it was possible to evaluate which parameters and interactions of them have the most influence on the power
coefficient. The main finds showed that the contribution index of the angle of incidence (α) on the power
coefficient was the highest among the others, with a contribution of 73%. Another significant contribution is the
combination of angle of incidence parameter (α) and the profile chord (c), presenting an index of 16%. Therefore,
the impact of NACA-4 digit parameterization on the power coefficient is small. Moreover, the sum of the
contribution of the parameters inherent to the parameterization is only 4%.