Presentations

Abstract

Wind farm blockage causes a deceleration of the wind in front of a wind farm, along with the tendency of the wind to move around and over the farm. Wind speed deceleration is easily identifiable in SCADA data and high-fidelity simulations, such as RANS and LES. However, identifying wind direction changes due to blockage is far less straightforward. Wind direction measurements, such as those from wind vanes, require accurate northing calibration. Without this, identifying wind direction changes due to global wind farm blockage becomes significantly more complicated. Furthermore, even with proper calibration, analyzing a single wind direction to detect these changes is insufficient. This effect should instead be identifiable across the entire wind rose. In this study, the effect of global wind farm blockage on the wind direction in the wind farm is analyzed using SCADA data. The case study wind farm consists of over 4 GW of wind turbines, and the SCADA analysis is performed on at least 2 GW of these turbines. First, northing calibration is performed for each wind turbine on one year of SCADA data to eliminate directional bias. This calibration uses energy ratios of the five closest wind turbines, following the FLASC framework developed by NREL. Next, the freestream wind direction is calculated as the circular mean of all turbines for each 10-minute window. The data is grouped into eight bins of 45-degree intervals based on the freestream wind direction for wind speeds between 6 and 12 m/s. For each turbine in each bin, the circular mean of the wind direction is computed. Results show that wind direction changes due to blockage can be identified across all wind directions. Differences between freestream wind direction and the wind direction at the edges of the farm can reach up to 6 degrees, with absolute differences between opposing edges of the farm reaching as much as 12 degrees. For each bin, wind direction changes due to blockage are analyzed as a function of downstream distance (parallel to the freestream wind direction) and the distance perpendicular to the freestream wind direction. Both Pearson and Spearman correlation coefficients are calculated for each bin. Results indicate that the correlation with downstream distance is negligible. However, the correlation with perpendicular distance is significant, with coefficients consistently lower than -0.5. This suggests a clear trend in wind direction changes as a function of distance perpendicular to the freestream wind direction. For example, for southerly winds, turbines on the west side of the wind farm are observed to have wind directions closer to north-northwest, while turbines on the east have wind directions closer to north-northeast. Analytical wake models do not account for this effect. To address this, the source code of FLORIS is modified to incorporate wind direction changes due to blockage, similar to the heterogeneous inflow library in FLORIS. By adjusting the background flow, the heterogeneous wind direction due to blockage has effectively been incorporated, as observed in the SCADA data.