A data-driven framework for comprehensive identification of operational wind turbines under uncertainty

Abstract

In order to capture the characteristic short- and long-term variability of Wind Turbine (WT) systems, it is crucial to incorporate the uncertainties related to various sources within mathematical prediction models. A data-driven framework able to link the temporal variability characterizing the system with the random evolution of environmental and operational parameters affecting the system is applied on long-term data collected from a real operating WT structure located in Dortmund, Germany. Particular focus placed us on the configuration of the input variable set, namely the evaluation of Polynomial Chaos Expansion (PCE) model estimates for the case of blind and reasoned source selection. The overviewed framework leads to effective reduction of the input set dimension facilitating the implementation of the proposed approach in an automated fashion. The developed data-driven tool proves robust in quantifying the uncertainty linked to the evolution of the structural dynamics throughout the structure’s operational envelope.