Wind Turbine Yaw Misalignment Detection Based on Simulated Data

This work presents an analysis and selection of variables applied to a new data set produced using the OpenFAST simulation framework, based on the NREL 5-MW baseline turbine model, focusing on the identification of yaw misalignment conditions in wind turbines. A total of 50 variables were simulated, with scenarios including normal operating conditions at different wind speed and turbulence levels, as well as the simulation of 8 different yaw misalignment conditions. Furthermore, the contributions of this work include data visualization in a reduced-dimensional space and analysis of features most relevant to the detection of misalignment. Experiments demonstrated that the simulated variables are relevant and deserve attention in sensor-based data collection processes, where a relevance ranking indicates promising paths for the identification and classification of yaw misalignment conditions, both through statistical analyzes and through the application of machine learning methods.