Savonius wind turbine blade design and performance evaluation using ANN-based virtual clone: A new approach

Abstract

The drag based Savonius wind turbine (SWT) has shown immense potential for renewable power generation in built-up areas under complex urban wind conditions. While a series of studies have been conducted on improving SWT's efficiency, optimal performance has yet to be achieved using traditional design approaches such as experimental and/or computational fluid dynamics methods. Recently, artificial intelligence and machine learning have been widely used in design optimization. As such, an ANN-based virtual clone can be an alternative to traditional design methods for wind turbine performance determination. Therefore, the main goal of this study is to investigate whether ANN-based virtual clones are capable of determining the performance of SWTs with a shorter timeframe and minimal resources compared to traditional methods. To achieve the objective, an ANN-based virtual clone model is developed. Two sets of data (computational and experimental) are used to validate and determine the efficacy of the proposed ANN-based virtual clone model. Using experimental data, the model's fidelity is over 98%. The proposed model produces results in one-fifth the time of the existing simulation (based on the combined ANN + GA metamodel) method. The model also reveals the location of the dataset's optimized point for augmenting the turbine's performance.