CFD and ANN surrogate modelling for fixing angle optimisation of a Darrieus vertical axis wind turbine

Global decarbonisation requires efficient and reliable renewable energy systems, with Vertical Axis Wind Turbines [Formula: see text] offering advantages in turbulent and urban environments. This study examines the effect of blade Fixing Angle [Formula: see text] on the aerodynamic performance of a three-bladed Darrieus [Formula: see text] using Unsteady Reynolds-Averaged Navier–Stokes[Formula: see text] based Computational Fluid Dynamics [Formula: see text] simulations coupled with an Artificial Neural Network [Formula: see text] surrogate model.A parametric analysis was carried out for seven [Formula: see text] values across a Tip-Speed Ratio [Formula: see text] range of [Formula: see text]. Results show that a small positive [Formula: see text] (toe-out) improves efficiency by delaying stall onset and extending upwind torque production. The optimum case [Formula: see text] achieved a [Formula: see text] increase in power coefficient [Formula: see text] compared to the baseline [Formula: see text]. In contrast, negative FA reduced performance due to earlier separation and greater downwind losses.The [Formula: see text] surrogate, trained on [Formula: see text] data, effectively captured the nonlinear dependence of [Formula: see text] on [Formula: see text] and [Formula: see text], achieving [Formula: see text] and a Mean Absolute Error [Formula: see text] below [Formula: see text]. Once validated, the [Formula: see text] enabled rapid optimisation of the design space, locating the global optimum with negligible computational cost.Flow-field and force-coefficient analyses further reveal that positive [Formula: see text] modifies the incidence-angle distribution, mitigates torque ripple, and suppresses dynamic stall, thereby explaining the observed performance enhancement. Overall, the integration of [Formula: see text] and [Formula: see text] provides both physical insight and computational efficiency, establishing a solid foundation for future multi-objective optimisation of [Formula: see text] considering parameters such as [Formula: see text], torque ripple, and self-starting capability.