Predicting wave heights for marine design by prioritizing extreme events in a global model

TL;DR

This paper improves the modeling of extreme wave heights for marine structure design by using a weighted estimation method on the exponentiated Weibull distribution, resulting in better tail fit and more accurate load predictions.

Contribution

It introduces a weighted least squares estimation approach for the exponentiated Weibull distribution to better model extreme wave heights in offshore wind turbine design.

Findings

Better fit at the upper tail of wave height distributions.

Accurate estimation of 1-year return wave height.

Enhanced prediction of extreme loads for marine structures.

Abstract

In the design process of marine structures like offshore wind turbines the long-term distribution of significant wave height needs to be modelled to estimate loads. This is typically done by fitting a translated Weibull distribution to wave data. However, the translated Weibull distribution often fits well at typical values, but poorly at high wave heights such that extreme loads are underestimated. Here, we analyzed wave datasets from six locations suitable for offshore wind turbines. We found that the exponentiated Weibull distribution provides better overall fit to these wave data than the translated Weibull distribution. However, when the exponentiated Weibull distribution was fitted using maximum likelihood estimation, model fit at the upper tail was sometimes still poor. Thus, to ensure good model fit at the tail, we estimated the distribution's parameters by prioritizing observations of high wave height using weighted least squares estimation. Then, the distribution fitted well at the bulks of the six datasets (mean absolute error in the order of 0.1 m) and at the tails (mean absolute error in the order of 0.5 m). The proposed method also estimated the wave height's 1-year return value accurately and could be used to calculate design loads for offshore wind turbines.