808
Automatic Wind Turbine Detection Using Level-II Data

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 26 January 2011
Automatic Wind Turbine Detection Using Level-II Data
B. L. Cheong, Univ. of Oklahoma, Norman, OK; and R. D. Palmer and S. M. Torres

Poster PDF (11.5 MB)

Wind power is considered a “green” form of electricity production as it is renewable and ecologically friendly. While there are countless benefits from its growth, the negative impacts should not be neglected. One such impact is the contamination induced on radar signals, which are essential for weather forecasting and warning decisions for public safety. In this work, an algorithm for automatic wind turbine detection from Level-II moment data, i.e., reflectivity, Doppler velocity and spectrum width, was developed. The algorithm utilizes a series of consecutive moment maps and a Fuzzy-logic inference system for detection. The impetus for this project is to design and realize a detection technique that requires minimal modifications to the existing WSR-88D infrastructure. That is, we began with a restriction of only utilizing Level-II data. Most of the time, a wind farm within the radar domain can be visually identified from Level-II data by inspecting several consecutive images and looking for stationary features. Most weather features on the radar maps advect and deform but features from ground targets, including wind turbine clutter, would remain at the same locations and that is how they provide us with visual queues for identification. Current operational ground clutter filter, i.e., GMAP (Gaussian Model Adaptive Processing), does not completely filter out wind turbine clutter simply because it is not designed to do so. Even other clutter filters are meant for filtering targets at near-zero velocity. Residual signals from the wind turbine clutter through these ground clutter filters still contaminate meteorological data and it is our primary goal of this work to design, implement and test of an automatic detection technique to identify the residual wind turbine clutter signals using Level-II data. We focus on areas where GMAP has been applied, i.e., areas where the CMD (Clutter Mitigation Decision) flag has been marked. Understanding how human visual systems identify these residual signals of wind turbine clutter from the GMAP, it was believed that by processing several consecutive images at a time, a similar detection scheme, which is also suitable for computer implementation, could be realized. A fundamental algorithm has been developed, which utilizes the CMD flag, six running-temporal textures, and numerical statistics of the moment values, and a fuzzy-logic inference system for the detection. In the paper, a detailed description of the algorithm and some examples from several WSR-88D radars, e.g., KDDC, KDYX, KBUF, etc., will be presented. Of course, moment dataset with the highest temporal resolution of 5-minute would limit the performance of the algorithm and they will be discussed. A more thorough system evaluation is currently underway and the results will be reported.