| The flow of coherent turbulent structures into a wind turbine is associated with vibrational blade excitation. Successful forecasting of such turbulent events for control system input would increase the lifetime of turbine components. The coherence of these features suggests description by model reduction. To this end, an array of pressure transducers was deployed on the ground at Reese Technology Center in Lubbock, Texas, and the pressure fluctuations were recorded over nearly two diurnal cycles. A program for computation of the dynamic mode decomposition was developed with special consideration for the case of a non-stationary, nonlinear system. A simulated surface-pressure perturbation was first decomposed, to inform the interpretation of experimental data. Several sets of surface-pressure data were decomposed for various meteorological conditions. The resulting dynamic modes and eigenvalues describe the spatial and temporal coherence of local features in the atmospheric boundary layer. In each case, modes were identified that can be associated with wave-like pressure fluctuations that propagate either at convective or acoustic speeds. |
