This study presents a novel model for predicting wind turbine power output at a high temporal resolution in wind farms using a hybrid graph neural network (GNN) and long short-term memory (LSTM) architecture. By modeling the wind farm as a graph, the model captures both spatial and temporal dynamics, outperforming traditional power curve methods. Integrated with a normal behavior model (NBM) framework, the model effectively identifies and analyzes power loss events.

In this study, we assessed the total quantity of microplastics emitted by wind turbines currently operating in the Dutch North Sea. The estimates of microplastics currently emitted from offshore wind turbines in the Netherlands account for a very small portion of the total microplastics released offshore in the Netherlands, specifically less than 1 ‰.

A machine learning model is developed using lidar stations around US coasts to extrapolate wind speed profiles up to the hub heights of wind turbines from surface wind speeds. Independent validation shows that our model vastly outperforms traditional methods for vertical wind extrapolation. We produce a new long-term gridded dataset of wind speed profiles from 20 to 200 m at 0.25° and 6-hourly resolution from 1987 to the present by applying this model to the National Oceanic and Atmospheric Administration (NOAA)/National Centers for Environmental Information (NCEI) Blended Seawinds product.

Numerical models, used to assess loads on floating offshore wind turbines, require many input parameters to describe air and water conditions, system properties, and load calculations. All parameters have some possible range, due to uncertainty and/or variations with time. The selected values can have important effects on the uncertainty in the resulting loads. This work identifies the input parameters that have the most impact on ultimate and fatigue loads for extreme storm load cases.

This work employs deep neural networks to identify damage in the mooring system of a floating offshore wind turbine using measurements from the platform response. We account for the effect of uncertainty caused by the existence of multiple solutions using a Gaussian mixture model to describe the damage condition estimates. The results reveal the capability of the methodology to discover the uncertainty in the assessment, which increases as the instrumentation system becomes more limited.