Simulation and Optimization in Offshore Wind Turbine Structural Analysis

Abstract

The public interest in renewable energy resources is continuously growing as issues of pollution and shortage of limited resources as coal or oil become evident. One promising technical solution for the extraction of renewable energy is to install wind turbines offshore. Stronger and more steady winds as well as the reduced need for land area are substantial advantages compared to onshore wind turbine installations. However, higher costs for offshore installations as well as operation and maintenance issues are by today limiting factors of offshore developments. Therefore, cost reductions in this field are strongly needed in the future. Investigations in this thesis were focused on simulation and optimization aspects in offshore wind turbine structural analysis. The aim of the work was firstly to contribute to a better understanding of complex time-domain simulations in the research community, and secondly to give insight to several optimization approaches which can be applied to offshore wind turbine structures. Both objectives target at the simplification of simulation tasks, resulting in shorter analysis times and more efficiently designed structures. Analyses performed in this thesis are presented for the example of lattice type sub-structures. In terms of simulation aspects, a large simulation study was performed to determine the influence of input loading variability to structural responses. The variability was found to be an important source of simulation error as both ultimate and fatigue loads might be estimated with an error of up to 34% and 12%, respectively, when following a simulation setup recommended by international standards. Another aspect in the simulation setup is the number of load cases to be simulated. An approach using statistical regression methods was proposed to estimate the total fatigue damage for a set of load cases, by simulating only a few of them. Interestingly, the number of load cases could be reduced from 21 to 3, by still providing a high accuracy with maximum error of 6%. Using the definition of structural weight as the main cost criteria, several optimization studies were performed on different lattice type sub-structures. In total three approaches were developed and/or applied: a local optimization approach, the simultaneous perturbation stochastic approximation, and the genetic algorithm. One result was the identification of fatigue damage as the design driver, another the distribution of loads, resulting in varying member dimensions over the tower height. The local optimization approach was identified as fast and efficient. Stochastic approaches were in comparison more time demanding, but showed some interesting solutions which a human designer would not consider so quickly.