The development of offshore wind energy, especially the steps towards deep water and/or higher density wind farms, revives the prospects of vertical axis wind turbines (VAWTs). Because VAWTs may reduce the cost of floating structures, there is a potential to lower energy costs. However, VAWTs are often assumed to be less efficient and less reliable due to a lack of understanding of their complex aerodynamics. This research is motivated by the fact that the performance of isolated turbines is no longer the most important factor, but rather performance at the wind farm level. The objective is to comprehend the possible performance of VAWTs in a wind farm. This dissertation advances the knowledge of wind farm aerodynamics of VAWTs mainly in four aspects: a) It demonstrates the relationship between the rotor loading and wake deflection/deformation, indicating directions for simplified modelling of VAWT wake control; b) It identifies vital characteristics of a VAWT wake, confirming the positive effects of wake deflection on wake recovery and interaction; c) It presents high fidelity experimental data on the wakes and wake interactions of VAWTs placed upwind and downwind, and validates some cutting-edge models with the data; d) it demonstrates the potential of increased power performance of VAWT arrays by controlling the VAWT flow fields. In pursuit of these advances, the dissertation identifies and tackles a series of research topics. The first is on the simplified wake models. The state-of-the-art VAWT wake models are mostly transposed from that for HAWTs, based on the planar actuator disc model. However, the effects of the actuator discs’ shape, specifically the aspect ratio of rectangular ones (corresponding to VAWTs with various height-to-width ratios), on the wake recovery are not considered. We propose the effective mixing diameter D∗ to normalise the shape effects on the wake velocity recovery based on momentum conservation. D∗ is validated through particle image velocimetry (PIV) experiments and Reynolds averaged Navier-Stokes (RANS) simulations, and it outperforms the existing scaling lengths in the literature. The dissertation further questions the validity of planar actuators as surrogates of VAWTs. It compares the three-dimensional wakes of an actuator disc and a lab-scale VAWT using robotic volumetric PIV. The comparison reveals substantial differences in the vortex systems, pointing out the limitations of planar actuators in reproducing VAWT wakes, especially when the wakes are deflected. The results indicate that surrogates for VAWTs should be three-dimensional, coinciding with the swept areas of blades. Based on the three-dimensional actuator cylinder model and a simplified formulation of the vorticity transport equation, we demonstrate the underlying physics of the generation of the streamwise vortex system, highlighting the effect of different load distributions on the wake convection and mixing. We propose four idealised force distributions resulting in different vortex systems and wake topologies; The proposed model is validated qualitatively with stereoscopic PIV measurements on a lab-scale VAWT. We quantify the faster wake recovery consequent from the wake deflection using the experimental data. Furthermore, the wake interaction of two VAWTs placed upwind and downwind is investigated experimentally via PIV and load measurements. The upwind VAWT with positively pitched blades deflects the wake significantly, improving the inflow condition of the downwind VAWT, and thus increasing the overall extraction of the streamwise momentum. With the high-quality experimental data, we validate the state-of-the-art analytical wake models and simulations for VAWTs and identify their validity ranges. Two analytical wake models (the Jensen model and the Bastankhah-Porte-Agel model), five wake superposition models (four algebraic models and one momentum-conservation based model) and an unsteady Reynolds averaged Navier-Stokes (URANS) simulation with VAWTs represented by the actuator line model are compared in both isolated and interaction scenarios. Based on the validated URANS simulation, we explore the wake deflection effects on the enhancement of wind power extraction for two up-scaled VAWTs placed in tandem. The blades of these large H-type VAWTs operate in a high Reynolds number (chordbased, Rec ≈ 1×107), which ensures a high stall angle; The tip speed ratio is set to a relatively high value (4.5) to avoid severe dynamic stalls. And thus, the simulated VAWTs are optimised for engineering operations and perform better than the lab-scale model introduced earlier. Combinations where each turbine operates in three different fixed pitch angles (-10◦, 0◦, 10◦) resulting in different wake deflections are compared. With wake deflections, the overall power coefficient is increased by up to 45%for a tested configuration, which also depends on the inter-turbine distances. Most interestingly, when the turbine blades are pitched in the same direction, the vorticity system in the wake is enhanced and thus yields a flying formation effect for a VAWT array. Furthermore, wakes of three inline VAWTs are scrutinised, focusing on the wake interactions, floor effects and momentum recovery. For all the cases the three VAWTs’ blades are pitched in the same direction following the so-called flying formation scheme. The vertical flux of momentum is notably enhanced by the VAWT array with positive blade pitches even with the floor present, which is vital to the overall increment of power extraction in a wind farm operating in the atmospheric boundary layer. The overall power extraction is increased by 35% compared to the array with zero blade pitches; More importantly, the downwind VAWTs increase their performances by 113%-154%. The latter indicates the tremendous potential of large wind farms consisting of VAWTs employing blade pitching.