Efficiency-Driven Supervised Learning Regressors in Power Modeling and Optimization of Vertical Axis Wind Turbines

Abstract

Listen

Translate article

Abstract Of all the clean energy sources, wind power stands as the most widely available and employed form. Besides the Horizontal Axis Wind Turbines (HAWT), Vertical Axis Wind Turbines (VAWT) are attracting significant attention. This study focuses on optimizing a 12-kW lift-based 3-blade VAWT by introducing an optimal diffuser to enhance performance using two optimization procedures. To do that, a nonsymmetric diffuser in the shape of NACA4405 is introduced to the VAWT which is simulated using Computational Fluid Dynamics (CFD) and validated with experimental data. To optimize the power coefficient (Cp) of the VAWT, adjustments will be made to the position and orientation of both the upper and lower walls of diffuser. Two optimization methods were employed: one involves direct optimization, where a Genetic Algorithm (GA) is integrated with CFD. The second method utilizes Machine Learning (ML) models, such as Gaussian Process Regression (GPR), Support Vector Regression (SVR), and AdaBoost, fitted to the data set extracted from CFD. These ML models then replace CFD and are used in optimization. Results show that optimal diffuser can contribute to 27% increase in Cp. When employing GPR and AdaBoost, the optimal Cp values closely match those from direct optimization while optimization using ML models reduces computational costs by nearly 78% fewer simulation runs.