Improving the efficiency of a Savonius wind turbine by designing a set of deflector plates with a metamodel-based optimization approach

Abstract

Savonius wind turbines are the most suitable devices used in urban areas to produce electrical power. This is due to their simplicity, ease of maintenance, and acceptable power output with a low speed and highly variable wind profile. However, their efficiency is low, and the development of optimization tools is necessary to increase the total power output. This work presents a metamodel-based method to optimize the size and shape of a set of deflector plates to reduce the reverse moment of the turbine, using a genetic algorithm combined with an artificial neural network, reducing the computational cost. A parametrization of the deflectors geometry is proposed, and a Computational Fluid Dynamics model was implemented to train and validate the artificial neural network. The method was applied to design the deflectors of an actual 8-blade, 1[kW], 2.5[m] height turbine. Results showed an efficiency increment of 30%, from 0.215, to 0.279 in the turbine with the optimized deflectors. Furthermore, it is capable of producing power at 4[m/s], while the reference design had null power at that point. This methodology demanded 159 h, a substantial reduction of the computational cost of up to 97% in contrast to the classical simulation-based optimization approach.