Design optimization of Savonius hydrokinetic turbine with aid of an artificial neural network model

The Savonius hydrokinetic turbine (SHT), a vertical-axis turbine, efficiently extracts energy from low-speed water currents in rivers, canals, and marine environments. Optimizing the tip speed ratio (TSR), gap ratio (GR), and immersion ratio ( ) enhances its power extraction efficiency. Traditionally, optimization relies on computationally intensive CFD simulations, which are time-consuming. To address this, a machine learning-based approach, specifically an Artificial Neural Network (ANN) model is employed, reducing reliance on extensive CFD simulations while maintaining accuracy. The ANN model was trained using CFD simulations performed in StarCCM+ and then used to predict power coefficients for various design configurations. The CFD simulations are validated against experimental results reported in the literature. The optimal parameters, tip speed ratio (0.7808), gap ratio (-0.0498), and immersion ratio (1.0661) yielded the highest power coefficient (0.2179). This study demonstrates that machine learning accurately predicts turbine performance, reducing reliance on extensive CFD simulations, making turbine optimization more efficient.