Optimization Method Analysis for Tidal Turbine Blade Design

Abstract

As demand for green energy increases, Minesto AB produces marine technology for renewable tidal energy solutions. Underwater turbines are used to harness the energy of ocean currents caused by tidal movements. To maximize the electricity generation, the design of the turbine is crucial. The purpose of this thesis project is to study the optimization problem related to a turbine model developed by Minesto AB. The turbine performance for different blade geometries was estimated by calculating the power and thrust coefficients (C_P and C_T ) based on results from computational fluid dynamics (CFD) simulations in the software OpenFOAM. The former coefficient relates the generated to the total available power, and the latter the thrust force to the water flow momentum, and the primary focus was to maximize C_P and the ratio C_P/C_T . There were 18 parameters to be optimized, related to the geometry of the turbine, including chord lengths, and angles defining the bend and rotation of the blade. The optimization methods tested and analysed were Nelder-Mead, NSGA-II, simulated annealing, and Bayesian optimization with Gaussian processes. The methods were implemented primarily in Python with different settings and operators, and combinations of the methods were also tested. Initially, the optimization was focused on exploring the parameter space, and the maximum power coefficient was found to be C_P = 0.3902. The final goal was to minimize the number of CFD simulations needed to find the optimum, by identifying and implementing the most effective algorithm. The project found that, for a constant parameter space, Bayesian optimization is the most efficient individual method, and reliably produces results close to the overall maximum C_P within about 260 simulations, or about 30 h. However, combining it with NSGA-II and Nelder-Mead can be advantageous.