No abstract

Upscaling wind turbines has resulted in levelised cost of energy (LCoE) reductions. However, larger turbine diameters pose significant design challenges, often with conflicting requirements. For example, non-linear dynamics of aeroelastic tailored blades must be accurately predicted whilst, for the sake of efficient gradient-based design, it is also desirable to simplify the numerical definition of such blades—keeping design variables (DVs) to a minimum. This work presents and validates two features of the ATOM code (Aeroelastic Turbine Optimisation Methods), developed at the University of Bristol, that enable accurate and efficient modelling of large-scale wind turbine blades. Both an efficient parameterisation method and high-order beam elements illustrate the capacity for increasing the speed of gradient evaluations whilst accurately predicting blade dynamics—either by reducing DVs or simulation time. As a preliminary validation, aero-servo-elastic simulations from ATOM and an industry-standard software—DNV GL Bladed—are compared against field measurements gathered from an existing 7 MW turbine.

This paper presents, through the structural design of a 20 MW wind turbine blade, a selection of novel analysis and optimisation methods for wind turbines. These methods are integrated in the software—Aeroelastic Turbine Optimisation Methods (ATOM). A key feature is the novel, computationally-efficient piecewise linear model for running rapid design load case simulations (up to 16 times speed-up over conventional methods). Further, a comprehensive set of realistic design constraints is also proposed to ensure structural feasibility and aeroelastic stability. To demonstrate these methods, a sequential gradient-based optimisation process is employed, relying on the globally convergent method of moving asymptotes (GCMMA). The process begins with an aerodynamic optimisation to generate twist distributions, followed by an iterative loop during which load envelope updates and ‘frozen-load’ blade structural optimisations are performed independently. Aeroelastic loads are therefore considered, but are not directly optimised for. The present study investigates the structural design of a 20 MW wind turbine blade with hybrid carbon-glass spar caps. The optimised 122 m blade is found to have a mass of 83,622 kg, which decreases to 81,396 kg (-2.66%) with the addition of sweep. The GCMMA is found to converge successfully at each structural optimisation step. By contrast, the iterative loop is observed to oscillate, albeit within small bounds. Finally, results suggest that convergence of multi-step optimisation methods for aeroelastic blade design may not be guaranteed if design variables inducing aeroelastic couplings are considered.