Synergistic ANN-GA-CFD framework for high-performance Savonius wind turbine optimization with experimental validation

Area of Science:

• Renewable Energy Engineering
• Computational Fluid Dynamics
• Machine Learning Applications

Background:

• Existing Savonius wind turbine (SWT) optimization studies are limited by small datasets.
• This restricts comprehensive exploration of the SWT design space.

Purpose of the Study:

• To develop a comprehensive dataset for SWT optimization.
• To create an iterative optimization framework integrating artificial neural networks (ANN), genetic algorithms (GA), and computational fluid dynamics (CFD).

Main Methods:

• Constructed a multisource dataset covering key geometric parameters and operating conditions.
• Employed CFD simulations to enrich the dataset and fill data gaps.
• Developed ANN surrogate models for straight and twisted SWTs.

Main Results:

• Achieved high-accuracy ANN models (correlation coefficients up to 0.98).
• Identified optimal designs with maximum power coefficients of 0.1856 (straight) and 0.1927 (twisted).
• Quantified the influence of design parameters and operating conditions on SWT performance via sensitivity analysis.

Conclusions:

• The ANN-GA-CFD framework accurately predicts SWT performance.
• Experimental validation confirmed the model's predictions.
• The study provides a robust method for optimizing Savonius wind turbine designs.