Data Reduction and Reconstruction of Wind Turbine Wake Employing Data Driven Approaches

Data driven approaches are utilized for optimal sensor placement as well velocity prediction of wind turbine wakes. In this work, several methods investigated suitability in the clustering analysis and predicting time history flow field. The studies start by applying a proper orthogonal decomposition (POD) technique to extract dynamics flow. This is followed evaluations different hyperparameters machine learning algorithms their impacts on accuracy. Two test cases considered: (1) wake cylinder (2) rotating rotor exposed complex conditions. training data both obtained from high fidelity CFD approaches. reveal that combination classification-based algorithm Bi-LSTM sufficient periodic signals, but more advanced required highly near wake. done exploiting set POD modes field reconstruction. A satisfactory accuracy achieved an appropriately chosen horizon networks. results show data-driven can offer alternative conventional