Genetic Algorithm Optimisation of Actively Deforming Vertical Axis Wind Turbine Blade Profiles

Some studies to optimise the turbine blade profile to mitigate or overcome the dynamic stall issues that plague Vertical Axis Wind Turbines (VAWTs) have shown promise, however, they do not consider how the new optimised dynamic turbine blade profile could be replicated in an actual turbine, and typically utilise prescribed uninformed deformation profiles for the turbine blade. This thesis uses a novel approach by creating a dynamically changing turbine blade profile that alters the blade profile’s camber based on azimuthal position, but the profile is derived from using a genetic algorithm optimisation process. The core body of work is in developing a genetic algorithm to optimise the blade camber magnitude at specific positions within the turbine’s rotation, significantly improving overall and instantaneous power generation. Three techniques were tested for predicting the torque of each candidate camber profile in the GA optimisation process: using XFOIL to quickly predict torque at the desired azimuthal positions, using a transient rotating VAWT CFD simulation to model real-time blade deformation, and using an Artificial Neural Network Surrogate to predict torque based on turbine tip speed ratio, windspeed and azimuthal angle be. These three methods were integrated into a bespoke genetic algorithm optimiser, to find the optimal turbine blade profiles through it’s rotation. The thesis also explores variations in the genetic algorithm’s optimisation parameters, such as the number of optimisation positions, bounds of camber adjustment, and variations in the aerofoil configurations. All three techniques integrated into the GA showed significantly strong improvements in turbine power generation and reduced blade loading. The XFOIL GA showed the lowest improvement in mean moment at 59.1%, the CFD-In-The-Loop provided an uplift of 150% and the ML optimiser had an impressive 164% improvement!