Innovative Savonius rotors evolved by genetic algorithm based on 2D-DCT encoding

Abstract

The Savonius rotor is mainly known by its design simplicity, low cost, easy installation, good starting ability, relatively low operating speed and independent to wind directions. Over the last few decades, although many investigations around the world have reported performance gains of the Savonius rotor, none of the available techniques is capable of automatic generating innovative Savonius rotor geometries other than predefined models such as semicircular or its variations. This study aims to investigate the geometry optimization ability of evolution algorithms based on two-dimensional discrete cosine transformation (2D-DCT) through numerical simulations. By considering the cross-sectional area of a Savonius rotor as a picture, 2D-DCT has been employed to compress and encode the cross-sectional geometry into genomes which will be evolved by Genetic Algorithms to maximize their average torque coefficient. According to numerical simulation, the proposed technique improved the efficiency by 13.77% at tip speed ratio (TSR) 1.0 and 21.11% at TSR 0.0838, which discovered impressive and thought promoting innovative geometries automatically.