Oregon State University School of Mechanical, Industrial & Manufacturing Engineering

A System for Eagle Detection, Deterrent and Collision-Detection for Wind Turbines

Roberto Albertani, Matthew Johnston, Sinisa Todorovic, OSU Manuela Huso, Todd Katzner, USGS

National Wind Coordinating Collaborative Webinar: Upcoming Research on Eagle Impact Minimization Technologies Supported by the U.S. Department of Energy

May 19th, 2017

2

Outline

• Motivations

• Eagle detection

• Eagle deterrent

• Blade impact detection

• Acknowledgements

3

Motivations

• Support the growth of domestic wind energy and protecting

wildlife

• Simple and affordable system for eagle detection and deterrent

• Autonomous blade strike detection and species recognition for

deterrent validation, certification and site assessment

4

Wireless Network Topology

• Bluetooth low energy (BLE) for short range communication provide maximum battery life for on-blade module

• Nacelle-mounted module acts as range extender, communicating with ground-based computer over long-range RF wireless channel (Symphony Link @ 915 MHz ISM band)

5

Synchronized Sensors• Three nodes controlled by central computer

• Eagle detection • Eagle deterrent • Blade impact detection and specie recognition

Blade Optical Node Wireless Micro Camera

Remote Transmission of “Event” Data

Central Processing

Blade Vibration Node

Eagle Detection

Eagle Deterrent

18

cameras needed. We tested this camera location using a GoPro camera mounted to the blade of the GE turbine at MCC (Figure 3.2.8). Additional testing is needed to determine proper focal length, resolution, and frame rates for blade mounted cameras, data and power integration, as well as camera placement on blades (e.g., center of chord, leading or trailing edge, windward, leeward, or both sides), but we feel this is the most promising direction for the next generation impact system detection design (see Chapter 4. Next Generation System Design & Commercialization).

Figure 3.2.8. Views from a GoPro camera mounted at the root of the blade with sky and ground background on the GE turbine at MCC. Placement of cameras on turbine blade is a likely solution to provide the resolution needed for target identification while minimizing the number of cameras needed to cover the entire rotor swept area.

6

Eagle Detection

• Single 3600 FOV camera for eagle detection, standard structure-from-

motion framework for eagle trajectory estimate

• Off the shelf components, affordable, reliable

• Limited number of frames required, low resolution

• Easy installation, standard interface with computers

• Deep neural network algorithm

7

Computer Vision for Eagle Detection in Videostim

e

time

video    frames

deep  convolutional    neural  network    

applied  to  each  frame  extracts  local  features

Long-­‐Short  Term  Memory  network    extracts  long-­‐range  features  across  all  frames

Preliminary  Results

20  videos 30  videos 18  videos

Each  video  ≈  200  frames  with  320x240  pix.

network falcon  +  seagull eagle

CNN 93.3% 69.2%

CNN  +  LSTM 100% 84.6%

video  classification  accuracy

8

Eagle Deterrent

• Wind dancer: simple, affordable, reliable off the shelf

• Cluster activated by detection system

• Operation in sequence or random order

• Fail-safe activation for false positive detection not a problem

for energy production or turbine operations

Features:

¾ Ag News.

¾ Berry Organization News

¾ Focus on Food Safety

¾ On the Organic Side…

¾ Focus on Pest Management

¾ Articles

¾ Upcoming Events

Highlights:

New CCE Associate Director 3

WPBR Biology, Ecology, and Management 17

Pollination of Highbush Blueberries 21

Pollinators and Pesticide Sprays 27

Volume 13, Number 4 May 30, 2014

Air Dancers as a Potential Bird Deterrent in Blueberries

Heidi M. Henrichs, Paul D. Curtis, Jay R. Boulanger, Cornell University Department of Natural Resources

As part of a USDA-SCRI study, our research team has spent the last two years examining bird damage to fruit crops in New York State, as well as Michigan, Oregon, and the Pacific Northwest. We examined many different aspects of bird damage including:

1) bird species causing damage and their behavior;

2) spatial distribution of damage within a plot (edge vs. interior);

3) effect of the surrounding landscape;

4) grower opinions; and

5) economic costs.

The main goal of this project is to identify cost-effective, efficient, and environmentally-friendly ways to deter birds from eating cherries, blueberries, apples, and wine grapes.

In 2013, we pilot tested several different techniques in New York State, including bird distress callers, hawk kites, and “air dancers” (inflatable, flexible fabric,

colorful “people”, powered by a fan to move around; see photo above).

In this article, we focus on our results from the Blue Crop blueberry trials and assessments. However, more information on the full suite of fruits can be found in an upcoming issue of the New York Fruit Quarterly.

Air dancer in blueberries; photo courtesy of Heidi Henrichs

9

Impact Detection: Previous Research• Funded by US DoE Golden Field Office • Event-based recording of video data

• Selected number of frames before and after event • Central data processing on PC in turbine nacelle

Accelerometers

Contact Microphones

Camera (downward orientation)

Bioacoustic Microphone

Recording of simulated impact on one blade

Tests performed with WWG-0600 CART3 turbine at National Wind Technology Center (NWTC) at NREL

10

Vibrations Node for Blade Impact Detection• Integrated IMU for position and blade velocity

• On-board signal processing for real-time event detection

• Cross-correlation of sensor signals removes noise and improves SNR

• Sensor fusion to decrease missed detection rate and false alarm rate

11

Acknowledgements

Current funding:

• US Department of Energy, Golden Field Office

Team: • US DoE-NREL National Wind Technology Center

• North American Wind Research and Training Center, Mesalands CC, NM

• External Advisory Board:

• Wind Energy industry

• Dr. Rob Suryan, Associate Professor - Senior Research, Department of

Fisheries and Wildlife, Oregon State University

• Doug Warrick (Associate Professor, Department of Integrative Biology,

Oregon State University

12 Questions ?

Oregon State University School of Mechanical, Industrial & Manufacturing Engineering

A System for Eagle Detection, Deterrent and Collision-Detection for Wind Turbines

Roberto Albertani (roberto.albertani@oregonstate.edu)

18

cameras needed. We tested this camera location using a GoPro camera mounted to the blade of the GE turbine at MCC (Figure 3.2.8). Additional testing is needed to determine proper focal length, resolution, and frame rates for blade mounted cameras, data and power integration, as well as camera placement on blades (e.g., center of chord, leading or trailing edge, windward, leeward, or both sides), but we feel this is the most promising direction for the next generation impact system detection design (see Chapter 4. Next Generation System Design & Commercialization).

Figure 3.2.8. Views from a GoPro camera mounted at the root of the blade with sky and ground background on the GE turbine at MCC. Placement of cameras on turbine blade is a likely solution to provide the resolution needed for target identification while minimizing the number of cameras needed to cover the entire rotor swept area.