Presentations

Abstract

Wind turbine power performance is heavily influenced by atmospheric conditions such as turbulence intensity, wind shear, upflow, and veer across the rotor plane. While this can lead to several percent of site-specific performance losses, validation of these effects remains limited. Prior efforts by the Power Curve Working Group have established that IEC 61400-12-1 adjustments for rotor equivalent wind speed (REWS) and hub-height turbulence intensity do not consistently reduce performance variation across atmospheric conditions. As a result, a common industry practice to model site-specific power performance losses relies on the use of power deviation matrices derived from power curve tests. These approaches rely on limited and proprietary data sets and may not generalize outside test conditions, particularly for larger turbines, which are more sensitive to these effects yet less represented in existing datasets. In the present work, we evaluate physics-based adjustments to the IEC 61400-12-1 methodologies to account for turbulence intensity, wind shear, upflow, and veer. The approach accounts for rotor averaging of turbulence intensity and adjusts the REWS calculation to better represent the expected energy a turbine can recover compared to the ideal REWS. We perform a wide-scale field validation across 11 sites and 334 turbines, including 2 offshore sites, processing 1.55 million freestream and normal operation SCADA timestamps. The datasets are normalized by rated power and wind speed to enable comparison across turbines of different sizes. The validation assesses predictive modeling of power residuals, defined as the deviation between a 10-min data point and the mean power curve trend. To evaluate model performance, we analyze the correlation between measured and simulated residuals aggregated by turbulence intensity, REWS adjustments, and turbine operation regimes. Simulated multi-height wind speed and turbulence intensity are derived from numerical weather predictions calibrated using pre-construction wind measurements. After reducing the data, we observe power residual variations from -10% to +10% of rated power, with trends depending on wind speed and turbulence intensity that are consistent with power performance tests. The model predicts these variations with a correlation of 0.78 and virtually no bias (below 0.15% of rated power). These results confirm the model's ability to capture turbine-level performance variations driven by atmospheric conditions, with applications to pre-construction energy yield assessment and operational performance monitoring.