New data framework illuminates uncertainties in offshore wind farm conditions

Wind speeds and wave heights can have a major effect on offshore wind farms. But because they are correlated, their combined significance for wind farm designs couldn’t be factored in until now. CWI researcher Anne Eggels developed methods which take the effect of such correlations or dependencies into account. Today, she will publicly defend her thesis at the University of Amsterdam.

Publication date: 06-11-2019

Wind speeds and wave heights can have a major effect on offshore wind farms. But because they are correlated, their combined significance for wind farm designs can’t be factored in until now. CWI researcher Anne Eggels developed methods which take the effect of such correlations or dependencies into account. Today, she will publicly defend her thesis Uncertainty quantification with dependent input data at the University of Amsterdam.

When designing offshore wind farms, many circumstances must be taken into account. The trick is to reduce the available data or these conditions as much data as possible, so you can take full advantage of the computational power of simulating software. The question is what data should be processed and what data can be discarded. For dependent or correlated data, this is a well-studied problem. But for independent data, efficient methods are currently not available

Occurring together
CWI researcher Anne Eggels developed methods which take the effect of dependent data into account. An example of dependent data is wind speed and wave height. “Both large wind speeds and large wave heights can impact the structure of offshore wind parks”, says Eggels. “Since these conditions are dependent, they often occur together, and this needs to be taken into account in the structural design.”

Detecting dependencies
Another topic Eggels studied is how to detect dependencies in datasets. “When the dependencies are known, one can also get more insight in the structure of the data,” says Eggels. “For example, wind speeds measured at different heights above the ground are strongly dependent. The same goes for air density, pressure and temperature. In a dataset, these variables form two groups which are independent of each other, but strongly dependent within each group.”

Unseen data
Next to these two topics, Eggels studied other aspects of uncertainty quantification, such as emulation of unseen data, sensitivity analysis and adaptive methods to select samples – all with a focus on dependent data.

Wide range of applications
The methods Eggels developed are not specifically designed for offshore wind farms. Eggels: “They can be employed in a wide range of applications, for instance calculating the strength of storm surge barriers. The only requirement is that you have both a simulation model and input data for that model.”

Eggels performed her research within CWI’s Scientific Computing group, supervised by Daan Crommelin (CWI) and Barry Koren (Eindhoven University of Technology). Her research was part of the research programme EUROS (Excellence in Uncertainty Reduction of Offshore wind Systems), which aimed to greatly reduce the costs of offshore wind energy. The programme focused on reducing uncertainties in the design, construction, and logistics of offshore wind farms.