Using Nacelle Mounted Lidars and Machine Learning Model & Validate Turbine Performance in Complex Flow
11 Sep 18
4:00 PM
-
5:00 PM
Tracks:
Data Science and Digitalization, Poster Presentation
The quantification of turbine power performance in complex terrain has historically been hampered by significant sources of uncertainty. These include site calibration uncertainty, upwind mast measurements only at discrete points in the flow field, and an overall lack of visibility of the full complexity of wind flow across the rotor. In addition, if no mast is available, nacelle anemometry can be unreliable. Nacelle lidar allows for a detailed view of wind flow and turbine response, which is particularly important to inform our understanding of performance as it relates to pre-construction yield assessments. Natural Power uses advanced data analytics to evaluate nacelle mounted lidar data, including application of machine learning algorithms which are utilized to create a multi-parameter power model based on all available inputs. Broadly, as measurements improve, our approach to gain useful insight must rely on increasingly complex data science techniques, which are explored here.