Global sensitivity analysis of the blade geometry variables on the wind turbine performance

The need for implementing efficient blade designs gains relevance as wind turbine developments require longer blades. The design of blade geometry, traditionally divided in 2D airfoils and spanwise distributions, is usually addressed as an optimization problem. A correct identification of the design variables is crucial to avoid unnecessary computational cost or insufficient exploration of the design space. This paper deals with the identification of the design variables that affect the wind turbine performance. First, the number of design variables for an accurate airfoil representation is resolved. A methodology, based on statistical hypothesis testing applied to the airfoil approximation errors, is presented to assess the accuracy of types of B‐splines. Second, the study is extended to chord and twist distributions besides airfoil geometry with the purpose of assessing the sensitive blade variables in the wind turbine performance. Global sensitivity analysis as multi‐variable linear regressions and variance‐based methods are used. Latin hypercube sampling is applied to generate efficient inputs. MATLAB‐based code is developed to obtain outputs: annual energy production, maximum blade tip deflection, overall sound power level and blade mass. As result of the study, a list of non‐affecting variables is deduced. These variables can be avoided in the optimization without loss of gain in the performance. The method is a powerful tool to analyse in a preliminary phase a design problem involving a high amount of variables and complex physical relations by means of combining different multi‐disciplinar calculation codes and performing statistical treatments. Copyright © 2017 John Wiley & Sons, Ltd.