(735bm) Machine Learning Predictive Modelling of Velocity and Turbulent Kinetic Energy of a Bladeless Wind Turbine Implementation in an Urban Environment

The installation of wind turbines to all possible extents has contributed to meeting the ever-increasing energy demand. Technologically developed areas have a high potential for wind energy, including the rooftops of high-rise buildings, railway tracks, the region between or around multi-story buildings, and city roads. Harnessing wind energy from these areas is quite challenging since it has a dramatic nature, being chaotic and turbulent on urban surfaces. A modelling paradigm is developed to assist predictive models of two turbulence models; SST and RNG, by using generated data from computational fluid dynamics (CFD) simulations. The key components of the current study involve four machine learning techniques validated to estimate the velocity and turbulence kinetic energy of a wind pattern in an urban environment: support vector regression, artificial neural networks (ANN), random forests, and extreme gradient boosting (XGBoost). To create training and test datasets for the four machine learning algorithms, a series of high-fidelity numerical simulations for a bladeless wind turbine implementation in an urban environment were performed; The capability of this bladeless wind turbine can harness winds from any direction. The ML models' anticipated velocity and turbulence kinetic energy are also compared to key existing analytical velocity and turbulence kinetic energy models.

The machine learning methods estimate velocity and turbulence kinetic energy in a manner comparable to CFD simulations. The findings show that machine learning-based systems can estimate velocity more accurately than traditional methods. The broader vision is that a comparison study of these turbulence models was established against machine learning techniques to reveal the form of the models' discrepancy and improve the predictions of data-driven turbulence models not just for wind patterns but for more general flows.