Abstracted/indexed

Often in wind power forecasting the mean wind speed is forecasted at a plant, converted to power, and multiplied by the number of turbines to predict the plant's generating capacity. This methodology ignores the variability among turbines caused by localized weather, terrain, and array orientation. We show that the wind farm mean wind speed approach for power conversion is impacted by Jensen's inequality, quantify the differences, and show machine learning can overcome these differences.

We investigate large wind speed fluctuations from a 10-year period at the Danish coastal site Høvsøre. The most extreme fluctuations are not turbulent but due to larger-scale weather phenomena. We find how these fluctuations impact wind turbines using simulations. The results are then compared to an extreme turbulence model described in the wind turbine safety standards, and it is found that the loads on the different turbine components are not the same as what the standard describes.

In this paper, detailed inflow information extracted from measurements is used to improve the accuracy of simulated wind turbine fatigue loads. Inflow information from nearby met masts is utilised as well as information from a blade-mounted flow sensor in combination with a method to compensate for the disturbance to the flow caused by the presence of the wind turbine.