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

Area of Science:

• Renewable Energy Engineering
• Artificial Intelligence in Engineering
• Computational Fluid Dynamics

Background:

• Savonius wind turbines (SWTs) show promise for urban renewable power generation.
• Traditional design methods (experimental, CFD) face limitations in optimizing SWT efficiency under complex urban wind conditions.
• Artificial intelligence (AI) and machine learning (ML) offer new avenues for design optimization.

Purpose of the Study:

• To investigate the efficacy of artificial neural network (ANN)-based virtual clones for determining Savonius wind turbine (SWT) performance.
• To assess if ANN virtual clones can achieve optimal performance with reduced time and resources compared to traditional methods.
• To develop and validate an ANN-based virtual clone model for SWT performance prediction.

Main Methods:

• Development of an artificial neural network (ANN)-based virtual clone model.
• Validation of the ANN model using both computational fluid dynamics (CFD) and experimental datasets.
• Performance comparison against traditional simulation and combined ANN + Genetic Algorithm (GA) metamodel approaches.

Main Results:

• The ANN-based virtual clone model achieved over 98% fidelity when validated with experimental data.
• The proposed model generated results five times faster than the existing simulation (ANN + GA metamodel) method.
• The model successfully identified optimal operating points for enhancing turbine performance.

Conclusions:

• ANN-based virtual clones offer a significantly faster and more resource-efficient alternative to traditional methods for SWT performance determination.
• The developed ANN model demonstrates high accuracy and potential for optimizing urban wind turbine designs.
• This AI-driven approach accelerates the design cycle for renewable energy technologies in built-up areas.