Vertical-Axis Wind Turbine Design Using Surrogate-assisted Optimization with Physical Experiments In-loop

Abstract

Most of the existing wind power comes from the use of traditional Horizontal-Axis Wind Turbines (HAWT), which typically require installation over vast countryside areas (in wind farms) due to considerations such as wake effects and noise. Recently, there has been an increasing interest to explore an alternative solution - Vertical-Axis Wind Turbines (VAWT) - that may be more suitable for compact urban areas. In this paper, we conduct design optimization of twin-blade VAWTs by evaluating the candidate designs through direct small-scale prototyping and physical experiments in-loop. The problem is a practical example of expensive optimization where the number of evaluations affordable are severely limited. In addition to the conventional single-objective form (maximize rotational speed), we also solve a multi-objective version of the problem (maximize rotational speed and minimize mass). For conducting optimization, we leverage surrogate-assisted approaches that make use of both predicted mean and uncertainties in modeling for an efficient exploration of the design space. The experiments demonstrate that the approach is able to generate non-intuitive designs that are competitive or better than the baseline (classic twin-blade Savonius design) within a small evaluation budget. The study also strengthens the case for applying this approach for design optimization in general.