Daniel Zwick

Support Structure Optimization for Offshore Wind Turbines With a Genetic Algorithm

This study considers the use of a genetic algorithm for the structural design optimization of support structures for offshore wind turbines. Member diameters, thicknesses and locations of nodes are jointly optimized. Analysis of each design is performed with a complete wind turbine simulation, for a load case in the time domain. Structural assessment is in terms of fatigue damage, evaluated for each joint using the hot-spot stress approach. This defines performance constraints. Designs are optimized with respect to their weight. The approach has been tested with the modified 4-legged UpWind jacket from the OC4 project. The weight is quickly reduced, convergence slows after about 100 iterations, and few changes occur after 250 iterations. Interestingly, the fatigue constraint is not active for any member, and it is the validity of stress concentration factors that determines the best design, which utilizes less than 90 percent of the available fatigue lifetime. These results of the preliminary study using the genetic algorithm demonstrate that automatic optimization of wind turbine support structures is feasible under consideration of the simplified load approach. Even for complex, multi-member structures such as the considered jacket a weight reduction was achieved.Copyright

Simulation-based optimization of lattice support structures for offshore wind energy converters with the simultaneous perturbation algorithm

Support structures for offshore wind turbines are contributing a large part to the total project cost, and a cost saving of a few percent would have considerable impact. At present support structures are designed with simplified methods, e.g., spreadsheet analysis, before more detailed load calculations are performed. Due to the large number of loadcases only a few semimanual design iterations are typically executed. Computer-assisted optimization algorithms could help to further explore design limits and avoid unnecessary conservatism. In this study the simultaneous perturbation stochastic approximation method developed by Spall in the 1990s was assessed with respect to its suitability for support structure optimization. The method depends on a few parameters and an objective function that need to be chosen carefully. In each iteration the structure is evaluated by time-domain analyses, and joint fatigue lifetimes and ultimate strength utilization are computed from stress concentration factors. A pseudo-gradient is determined from only two analysis runs and the design is adjusted in the direction that improves it the most. The algorithm is able to generate considerably improved designs, compared to other methods, in a few hundred iterations, which is demonstrated for the NOWITECH 10 MW reference turbine.

Comparison of different approaches to load calculation for the OWEC Quattropod jacket support structure

Accurate load simulations are necessary in order to design cost-efficient support structures for offshore wind turbines. Due to software limitations and confidentiality issues, support structures are often designed with sequential analyses, where simplified wind turbine and support structure models replace more detailed models. The differences with an integrated analysis are studied here for a commercial OWEC Quattropod. Integrated analysis seems to generally predict less damage than sequential analysis, decreasing by 30-70 percent in two power production cases with small waves. Additionally it was found that using a different realization of the wave forces for the retrieval run in sequential analysis leads to an increase of predicted damage, which can be explained as the effect of applying two independent wave force series at the same time. The midsection of the detailed support structure model used shell elements. Additional analyses for a model with an equivalent beam model of the midsection showed only small differences, mostly overpredicting damage by a few percent. Such models can therefore be used for relatively accurate analysis, if carefully calibrated.