The efficient selection of a wind turbine at a given site is presently limited by the developer's knowledge of what turbines are available on the market, and their inability to test and compare available turbine designs before investing. Poor turbine selection results in a financially sub-optimal investment. In this paper we develop an approach to optimize theoretical wind turbine designs based on Blade Element Momentum theory, multiple Evolutionary Computing algorithms, and a realistic cost model. The physical model was verified and tested using raw, real-world data from a met mast and two Enercon E-44 turbines installed at Burfell, Iceland. The selection method applied in this paper identified an optimum theoretical wind turbine design for Burfell which decreases the Levelized Cost of Energy by 10.4% when compared to the existing E-44 turbines, manually selected by a trial and error approach. The power curve of the theoretically optimal wind turbine was then used as a search parameter in a set of real turbines, to determine the optimum real turbine model for Burfell. The use of this real turbine would decrease the Levelized Cost of Energy by 8% when compared to the existing Enercon E-44 turbines. The approach used in this study also produced better results than previous trial-and-error based studies of turbine selection at Burfell reported in the literature. (C) 2014 Elsevier Ltd. All rights reserved.