Lifetime reassessment of offshore wind turbines using meta-models

Abstract

Offshore wind turbines are usually designed for 25 years. After this period, there are various options regarding the further use of the wind turbines. One option is a so-called lifetime extension, i.e., to continue operating even after the 25 years. However, to continue operating a wind turbine after its theoretical end of life, the remaining lifetime must be determined first. This can be done by a so-called lifetime reassessment. Therefore, the actual lifetime is calculated based on the real occurred environmental conditions, e.g., wind speeds. The actual lifetime is then compared with the design case which is usually calculated with conservative assumptions regarding the environmental conditions. Thus, the remaining lifetime of the wind turbine can be determined. However, a lifetime reassessment requires very high computing times due to the large number of simulations. This means that it is not possible to do a lifetime calculation with time-domain simulations. Therefore, generally representative load cases, i.e., representative combinations of environmental parameters, are calculated and the resulting damage of this calculations is then extrapolated to the lifetime using the frequency of occurrence of the combinations of the environmental parameters. However, this method cannot take into account all combinations of environmental parameters that may have occurred, so that this and the extrapolation to the lifetime make the calculation of the lifetime inaccurate. An alternative option that may reduce these inaccuracies is to use a meta-model as a surrogate of the original simulation model. Due to the significantly reduced computing time of the meta-model, all combinations of the environmental parameters can be calculated, although each calculation is only an approximation compared to the original simulation model. Meta-models are already used in wind energy, e.g., in the case of lifetime calculations, and have shown good results, however, they have not been used in this context.

Therefore, in this work, both a lifetime reassessment with the extrapolation method and a lifetime reassessment based on meta-models will be carried out. In order to make a statement about how large the errors are in comparison to an �exact� lifetime reassessment 100,000 additional simulations will be carried out with the original simulation model. These results are then compared with the results from the extrapolation method and the method based on meta-models in order to be able to determine whether an improvement in the accuracy of the recalculated lifetime can be achieved by using meta-models in comparison to the extrapolation method.