A multi-project validation of wake loss estimates at onshore wind farms
10 Sep 19
5:00 PM
-
6:00 PM
Wake loss estimation is a key input to pre-construction energy estimates and remains a significant source of their uncertainty. Prior validation studies of wake models typically focus on single turbines, lines of turbines, or at most, a few projects, usually offshore. In this study, we carry out a validation study of ten projects at onshore locations, a sample size large enough to give a sense of the statistical error distribution for project wake loss estimates. The wake loss estimation method that we are validating is a time series approach, in which wakes are calculated every time step from numerical weather prediction (NWP) output. The wake loss estimation uses the Larsen single-turbine wake model. The modeled wake estimate is compared to an observed wake estimate derived from SCADA data. The use of onshore projects presents a special challenge for separating wake effects from terrain and other effects. The same NWP model that is used to drive the wake model can be used to separate out these other sources of variability from the observed wakes. Finally, the observed ad modeled wake estimates are compared at the 10 projects. The key result is that Vaisala’s wake loss estimate is biased low (wake estimates are too small), but the bias-adjusted RMSE is 18% of the wake loss, which is at or below typical industry assumptions for uncertainty on wake losses.