Abstract
We present a proof of concept of wind turbine wake identification and characterization using a region-based convolutional neural network (CNN) applied to lidar arc scan images taken at a wind farm in complex terrain. We show that the CNN successfully identifies and characterizes wakes in scans with varying resolutions and geometries, and can capture wake characteristics in spatially heterogeneous fields resulting from data quality control procedures and complex background flow fields. The geometry, spatial extent and locations of wakes and wake fragments exhibit close accord with results from visual inspection. The model exhibits a 95% success rate in identifying wakes when they are present in scans and characterizing their shape. To test model robustness to varying image quality, we reduced the scan density to half the original resolution through down-sampling range gates. This causes a reduction in skill, yet 92% of wakes are still successfully identified. When grouping scans by meteorological conditions and utilizing the CNN for wake characterization under full and half resolution, wake characteristics are consistent with a priori expectations for wake behavior in different inflow and stability conditions.
Original language | American English |
---|---|
Article number | 4438 |
Number of pages | 20 |
Journal | Remote Sensing |
Volume | 13 |
Issue number | 21 |
DOIs | |
State | Published - 2021 |
Bibliographical note
Publisher Copyright:© 2021 by the authors. Licensee MDPI, Basel, Switzerland.
NREL Publication Number
- NREL/JA-5000-80496
Keywords
- Complex terrain
- Convolutional neural network
- Image processing
- Lidar
- Wake characterization
- Wind turbine wakes