This research explores the potential of wind energy as a sustainable solution to environmental challenges by focussing on the deployment of vertical axis wind turbines in confined spaces

such as urban rooftops

ship decks

gas stations

etc. It aims to identify the optimal configuration for a cluster of three rotors by varying geometric parameters

assessing the omnidirectionality of the rotors

and using an artificial neural network as an optimization tool. The methodology involves 2D unsteady Reynolds-averaged Navier–Stokes computational fluid dynamics simulations followed by building and training an artificial neural network. The results indicate that the best cluster configuration

achieving a power coefficient of 0.2517

consists of a horizontal distance of 0.08D

vertical distance of 0.1D

and wind direction of 232°. The accuracy of the neural network is demonstrated by a mere 1.206% discrepancy with the computational analysis

thus highlighting its potential as a time-saving tool in the optimization process. Additionally

the average power coefficient across all wind directions was found to be 0.194

thus confirming the strong omnidirectional characteristics of the cluster. The study also notes a significant speed-up region between the rotors

which creates a low-pressure area

enhancing the power coefficient of the downstream rotor and demonstrating the advantages of this pressure-based turbine configuration. This comprehensive study establishes the efficacy of neural networks in optimizing wind turbine clusters

thus significantly expediting their design and development process.