Content uploaded by Robert M Suryan
Author content
All content in this area was uploaded by Robert M Suryan on Sep 16, 2016
Content may be subject to copyright.
DE-EE0005363
A Synchronized Sensor Array for Remote Monitoring of Avian
and Bat Interactions with Offshore Renewable Energy Facilities
Final Report of Results
Principal Investigators
Dr. Robert Suryan, Associate Professor – Senior Research, Department of
Fisheries and Wildlife, Hatfield Marine Science Center, Oregon State
University
Dr. Roberto Albertani, Associate Professor, School of Mechanical, Industrial,
and Manufacturing Engineering, Oregon State University
Dr. Brian Polagye, Assistant Professor, Department of Mechanical
Engineering, Northwest National Marine Renewable Energy Center,
University of Washington
Additional Investigators
Trevor Harrison, PhD candidate, University of Washington
Jeremy Flowers, MS student, Oregon State University
William Beattie, PhD candidate, Oregon State University
Congcong Hu, PhD student, Oregon State University
Brian Hrywnak, MS student, Oregon State University
July 15, 2016
i
Executive Summary
Wind energy production in the U.S. is projected to increase to 35% of our nation’s energy by 2050.
This substantial increase in the U.S. is only a portion of the global wind industry growth, as many
countries strive to reduce greenhouse gas emissions. A major environmental concern and potential
market barrier for expansion of wind energy is bird and bat mortality from impacts with turbine blades,
towers, and nacelles. Carcass surveys are the standard protocol for quantifying mortality at onshore sites.
This method is imperfect, however, due to survey frequency at remote sites, removal of carcasses by
scavengers between surveys, searcher efficiency, and other biases as well as delays of days to weeks or
more in obtaining information on collision events. Furthermore, carcass surveys are not feasible at
offshore wind energy sites. Near-real-time detection and quantification of interaction rates is possible at
both onshore and offshore wind facilities using an onboard, integrated sensor package with data
transmitted to central processing centers.
We developed and experimentally tested an array of sensors that continuously monitors for
interactions (including impacts) of birds and bats with wind turbines. The synchronized array includes
three sensor nodes: 1) vibration (accelerometers and contact microphones), 2) optical (visual and infrared
spectrum cameras), and 3) bioacoustics (acoustic and ultrasonic microphones). Accelerometers and
contact acoustic microphones are placed at the root of each blade to detect impact vibrations and sound
waves propagating through the structure. On-board data processing algorithms using wavelet analysis
detect impact signals exceeding background vibration. Stereo-visual and infrared cameras were placed on
the nacelle to allow target tracking, distance, and size calculations. On-board image processing and target
detection algorithms identify moving targets within the camera field of view. Bioacoustic recorders
monitor vocalizations and echolocations to aid in identifying organisms involved in interactions. Data
from all sensors are temporarily stored in ring (i.e., circular) buffers with a duration varying by sensor
type. Detection of target presence or impact by any of the sensors can trigger the archiving of data from
all buffers for transmission to a central data processing center for evaluation and post-processing. This
mitigates the risk of “data mortgages” posed by continual recording and minimizes personnel time
required to manually review event data.
We first conducted individual component tests at laboratories and field sites in Corvallis and
Newport, Oregon, and Seattle and Sequim, Washington. We conducted additional component tests on
research wind turbines at the North American Wind Research and Training Center, Mesalands
Community College (MCC; General Electric 1.5 MW turbine), New Mexico, and the National Wind
Technology Center, National Renewable Energy Laboratory (NREL; Controls Advanced Research
Turbines 3 [CART 3] 600 kW Westinghouse turbine), Colorado. We conducted fully integrated system
tests at NREL in October 2014 and April 2015. We used only research wind turbines so that we could
conduct controlled, experimentally generated impacts using empty and water-filled tennis balls shot from
a compressed air cannon on the ground. The ~57 - 140 g tennis balls (depending on water content) were
at the upper mass range for bats, but lower mass range for marine birds. Therefore, the ability to detect
collisions of most seabirds is likely greater than our experiments demonstrate, but possibly lower for
some bats depending on the background signal of a given turbine. Vibration data demonstrated that
background signals of operating turbines varied markedly among the CART 3 under normal operation
(greatest), GE (moderate), and CART 3 during idle rotation (generator not engaged; least). In total, we
measured 63 experimental blade impacts on the two research turbines. Impaction detection was
dependent on background signals, position of impact on the blade (a tip strike resulted in the strongest
ii
impact signal), and impact kinetics (velocity of ball and whether the ball struck the surface of the blade or
the leading edge of the blade struck the ball). Overall detection percentage ranged from 100% for the
“quietest” conditions (CART 3 idle rotation), down to 35% for the noisiest (CART 3 normal operation).
Impact signals were detected from sensors on more than one blade (i.e., blades other than the blade
struck) 50% - 75% of the time. Stereo imaging provided valuable metrics, but increased data processing
and equipment cost. Given the cost of cameras with sufficient resolution for target identification, we
suggest mounting cameras directly on the blades to continuously view the entire rotor swept area with the
fewest number of cameras. Bioacoustic microphones provide taxonomic identification, as well as
information on ambient noise levels. They also assist in identifying environmental conditions such as hail
storms, high winds, thunder, lightning, etc., that may contribute to a collision or a false positive detection.
We demonstrated a proof of concept for an integrated sensor array to detect and identify bird and bat
collisions with wind turbines. The next phase of research and development for this system will
miniaturize and integrate sensors from all three nodes into a single wireless package that can be attached
directly to the blade. This next generation system would use all “smart” sensors capable of onboard data
processing to drastically reduce data streams and processing time on a central computer. A provisional
patent for the blade mounted system was submitted by Oregon State University and recorded by the U.S.
Patent and Trademark Office (application no. 62313028). Eventually, technology and industry advances
will allow this low cost monitoring system to be designed into materials during manufacturing so that all
turbines could be monitored with either a subset or full suite of sensors. As standard equipment on all
commercial turbines, the sensor suite would allow the industry to effectively monitor whether individual
turbines were causing mortalities or not and under what circumstances. It would also provide real-time
evaluation of mechanical and structural integrity of a turbine via vibration, image, and acoustic data
streams, thereby permitting modifications in operation to limit environmental or mechanical damage.
iii
Table of Contents
Executive Summary ............................................................................................................................... i
1. Introduction ................................................................................................................................... 1
1.1. Background ........................................................................................................................... 1
1.2. Need ...................................................................................................................................... 1
1.3. Multi-sensor Array ................................................................................................................ 2
2. Study Design and System Description .......................................................................................... 3
2.1. Vibration Node ...................................................................................................................... 3
2.2. Optical Node ......................................................................................................................... 4
2.3. Bioacoustic Node .................................................................................................................. 5
2.4. System Integration and Data Acquisition ............................................................................. 6
2.5. Impact Event Detection and System Trigger ........................................................................ 6
2.6. Experimental Impacts ........................................................................................................... 7
2.7. Project Timeline .................................................................................................................... 7
3. Results ........................................................................................................................................... 9
3.1. Vibration Node ...................................................................................................................... 9
3.2. Optical Node ....................................................................................................................... 13
3.3. Bioacoustic Node ................................................................................................................ 19
3.4. System Integration, System Triggering, and Event Detection ............................................ 20
4. Next Generation System Design & Commercialization ............................................................. 22
4.1. Vision Statement ................................................................................................................. 22
4.2. System Description ............................................................................................................. 22
4.3. Application .......................................................................................................................... 22
4.4. Commercialization Plan ...................................................................................................... 23
5. Summary of project accomplishments ........................................................................................ 26
5.1. Patents ................................................................................................................................. 26
5.2. Publications & Other Products ............................................................................................ 26
5.3. Presentations ....................................................................................................................... 26
5.4. Media .................................................................................................................................. 27
6. Acknowledgements ..................................................................................................................... 28
7. Literature Cited ........................................................................................................................... 29
1
1. Introduction
1.1. Background
Wind energy production in the U.S. is projected to increase to 35% of our nation’s energy by 2050
(Dept of Energy 2015). Offshore wind is expected to play a significant role in reaching targets (Musial
and Ram 2010). The installation of between 50 and 90 GW of offshore wind capacity will require an
investment of over $200 billion for construction, operation, and infrastructure development (Musial and
Ram 2010). Concerns for environmental impacts, however, especially bird and bat collisions, can add to
these costs in the form of construction delays, mitigation, project permitting, etc.
Similar to onshore wind facilities, offshore wind has the potential to affect avian populations through
reduction in habitat, disruption of migratory pathways, and injury/mortality through collision (Allison et
al. 2008). For many wind facilities, direct impact through collisions is of principal concern. Even low
levels of collision mortality are of concern for endangered or protected species. Unfortunately, unless
collision and recovery rates are sufficiently high, standard carcass surveys below turbines are unlikely to
produce mortality estimates with confidence intervals narrow enough to effectively inform management
decisions (e.g., Huso et al. 2015). In the marine environment, standard carcass surveys are not feasible,
therefore post-installation impact assessment is problematic. A turbine-mounted detection system with
data transmitted back to shore for post-processing is an efficient strategy for long-term assessment of bird
and bat casualties in offshore wind energy installations.
1.2. Need
A compact, integrated monitoring system capable of directly observing injury/mortality events 24
hours per day is required to validate site-specific risk models for offshore wind – and could be equally
valuable for land-based turbines. Such a system must be relatively simple to deploy on operating turbines
and minimize requirements for manual review of data. Because the consequences of injury/mortality
depend on the relative significance of the species (i.e., consequences are greater for threatened or
endangered species), the system must not only identify injury/mortality events, but also the affected
species. Finally, because mass collision events can occur during periods of low visibility (Desholm et al.
2006), the integrated system must be able to operate over a broad range of meteorological conditions and
ambient light levels. An integrated impact detection system would also be necessary to validate the
efficacy of deterrent or operational measures that are employed to mitigate impact rates.
Several attempts have been made to develop an automatic detection system for avian and bat
collision with wind turbines including vibration or acoustic sensing devices and visual‐ or infrared‐
spectrum cameras (Desholm et al. 2006, Wiggelinkhuizen et al. 2006, Evans 2012). Desholm et al.’s 2006
Thermal Animal Detection System (TADS) identified detections and taxonomic classification through
wing beat analysis and animal size, however, the system required manual review of imagery collected on
a pre‐determined duty cycle. Wiggelinkhuizen et al.’s 2006 WT‐Bird system used vibration sensors to
trigger the visual cameras, thereby recording only imagery of greatest interest. Each of the various
components (e.g., vibration vs. acoustic sensors, infrared vs. visual cameras) have benefits and drawbacks
and no single one is capable of providing all of the information needed to detect impacts and identify the
species involved. While previous projects have effectively used multiple sensors, no one project has
attempted to integrate all sensors into an automated collision detection and identification system. In this
study, we developed and experimentally tested a multi-sensor array for collision detection and
identification.
2
1.3. Multi-sensor Array
Our multi-sensor array consists of three primary sensor nodes communicating with a central
controller and data acquisition system (Fig. 1.3.1). The vibration node, composed of accelerometers and
contact microphones, provides monitoring for collisions. The optical node, composed of visual and
thermal infrared (IR) cameras configured for stereo imagery, provides information for taxonomic
classification, as well as information on animal presence and near misses. The bioacoustics node,
composed of audio and ultrasonic microphones, provides detection of bird and bat calls/echolocation to
aid in species identification, while also providing ambient acoustic information to assess other factors
affecting a collision event or otherwise. Temporal coverage of all sensors is continuous, while spatial
coverage ranges from omnidirectional (e.g., bioacoustics) to narrow fields of view (e.g., IR camera).
Wireless connectivity and on‐board battery power for the vibration node allowed sensors to be installed
on existing turbines with minimal impact. Both visual and IR cameras were included, as they offer
complementary capabilities. Visual cameras provide the most comprehensive taxonomic information and
are relatively inexpensive, but are limited to daytime use and favorable meteorological conditions. In
contrast, IR cameras are effective in a broader range of environmental conditions and provide higher
contrast imagery for target detection, but suffer from lower resolution and higher cost. the three
individual types of nodes communicate through a central computer in the nacelle of the turbine. Data
from all sensors are stored in individual ring buffers until detection of a collision event, at which point the
data surrounding the event is saved. Currently the system is triggered by the vibration node, however,
automated event detection could be programmed into all sensors so that each sensor could trigger an event
recording. Event data are transmitted back to a central processing location for manual review, thereby
limiting the amount of personnel time needed to review sensor data.
Figure 1.3.1. Diagram of a multi-sensor array for detecting and identifying impacts on a wind
turbine showing the three main nodes controlled by a central computer and collision event data
transmitted to a remote center for evaluation.
Optical Node
Stereo Infrared & Visual
Spectrum Cameras
Remote Transmission of
“Event” Data
Central Processing
Onboard Turbine
Bioacoustics Node
Acoustic and
Ultrasonic Microphones
Vibration Node
Accelerometers &
Contact Microphones
3
2. Study Design and System Description
All system components were initially tested in a laboratory
setting before tests on operating turbines. For operational tests,
we used two wind research turbines. The use of research turbines
allowed us to install devices and have full control of turbine
operation, including start-up and shut down, to conduct controlled,
artificial impact experiments using tennis balls. We used a 1.5
MW General Electric turbine operated by Mesalands Community
College (MCC) at the North American Wind Research and
Training Center in Tucumcari, New Mexico, and a 600kW CART
3 Westinghouse turbine operated by the National Renewable
Energy Laboratory (NREL) at the National Wind Technology
Center in Boulder, Colorado.
2.1. Vibration Node
The vibration node consisted of a paired wireless (UHF)
contact microphone and accelerometer positioned at the root of
each blade on the turbine. The contact microphone (Sun-
Mechatronics USK-40 w/ UZ-10 UHF receiver) was 35mm
diameter, 35mm height, weighed 31.7 g, and was powered by an
external 3.0V DC cell (Fig. 2.1.1). We used a National
Instruments (NI) USB-4431 DAQ to convert the microphone
analogue signal to digital. The 3-axis accelerometer (LORD
MicroStrain G-Link LXRS with 104-LXRS base station) was
45mm L x 60mm W x 20mm H, weighed 40.8 g, and was
powered by external 3.6V DC cell (Fig. 2.1.2). The accelerometer
length was oriented with the length of the blade. The paired
vibration sensors were attached on the outside of the blade near
the root using double-sided tape on the contact surface and black
Gorilla® tape along the edges of the boxes to provide an extra firm
hold. The double sided tape was commercial 3m plastic tape
without a foam core (3m part# 927 or F9460PC were both used).
A heat gun was used to heat the blade at the application location
to increase tackiness of the double sided tape when attaching to
the blade. During tests at MCC in 2013 and NREL in 2014, the
sensors were attached on the outside with access from on top of
the nacelle (MCC) or using a boom lift from below (NREL 2014;
Fig. 2.1.3). During tests at NREL in 2015, the vibration sensors
were also placed at the blade roots, but inside the shroud over the
hub that allowed easy access for installation from the nacelle (Fig.
2.1.3). Wireless receivers for the contact microphones and
Figure 2.1.1. Wireless contact
microphone.
Figure 2.1.2. Wireless
accelerometer.
Figure 2.1.3. Placement of the
vibration sensors – (top) external
or (bottom) inside of the shroud
over the hub.
Fig. 2.1.4. Receivers for wireless
contact microphones and
accelerometers were placed
inside the nacelle.
4
accelerometers were placed inside the nacelle near the hub (Fig.
2.1.4).
Sampling rates:
Contact microphones and accelerometers were programmed to
sample at 512 Hz and 1000 Hz, respectively, using a custom built
LabView graphical user interface. Recording frequency was
determined by the desired temporal resolution of impact
identification and the capacity of the central computer to record and
store desired quantities of data.
2.2. Optical Node
While IR cameras provide excellent contrast for target
detection, they have limited resolution in comparison to visual
spectrum cameras. The generation of uncooled microbolometer-
based IR cameras at the time of this study had a maximum
resolution of 640x480 (0.3 Megapixel – Mpx) in comparison to
visual spectrum cameras approaching 10 Mpx resolution.
Consequently, in selecting a camera lens there is a significant
trade-off between the percentage of the rotor swept area being
imaged and the ability to detect a target (e.g., a “fish eye” lens with
a wide field of view may have insufficient resolution to detect
targets over the intended operating range). We evaluated three
types of cameras for the optical node: 1) thermal infrared (FLIR
A655sc), 2) visual spectrum standard (Allied Vision Technology
Manta 201‐C), and 3) visual spectrum “smart” cameras with
onboard processing capabilities (Ximea Currera‐RL50C). The
stereo-infrared and stereo-visual cameras were enclosed in a
weather-resistant housing (Fig. 2.2.1) and were integrated into the
sensor array via a LabView visual interface. Dimensions of the
housing was 0.5m W x 0.2m H x 0.4m D and the cameras and
housing combined weighed 12 kg. The optical node was secured
within a pan-and-tilt frame (Fig. 2.2.2) that was attached to the
railing on the CART 3 turbine at NREL (Fig. 2.2.3). The pan-and-
tilt frame was 0.9m W x 0.8m H, and 0.4m D. The combined frame
with drive motors weighed 50 kg, but could be broken down into
smaller 15 kg modules for transport up the tower and onto the
nacelle. The worm gears and frame were sized to withstand gusts
of up to 140 mph with the system in an adverse orientation. The
pan-and-tilt frame was controlled using a LabView virtual
instrument (VI) on the central computer.
The optical node required 120 V AC power, transformed to 12 V DC. In operation, the optical node
itself draws < 100 W of power. The pan-and-tilt motors require 120 V AC and can draw up to 500 W of
Figure 2.2.1 Stereo-infrared and
stereo-visual cameras enclosed
in a weather-resistant housing.
Figure 2.2.2 Pan-and-tilt frame
for optics node.
Figure 2.2.3. Pan-and-tilt frame
attached to railing on nacelle of
Cart 3 turbine at NREL.
Fig. 2.2.4. Smart cameras with
onboard image processing
ability.
Power distribution
FLIR Infrared Cameras
MANTA Visual
Camera 1
MANTA Visual
Camera 2
Ethernet
Switch
5
power. In addition to power supply, the optical node required an Ethernet connection to the central
computer.
The smart cameras (Fig. 2.2.4) operated independently, not in stereo. These cameras were used on
an experimental basis with the anticipation that they would prove valuable in a subsequent design of a
miniaturized system. For the initial design, however, we preferred that all sensors be under control of the
central computer that was running the sensor integration and triggering software.
Sampling rates and field of view:
The infrared cameras had a resolution of 640 x 640 and sampled at 12 frames per second (fps) with a
15° field of view lens. The stereo visual Manta cameras had a resolution of
1624 x 1234 and sampled at 6 fps with a 52o field of view lens. The Ximea
smart cameras had a resolution of 2650 x 1920 and sampled at 15 fps with
a 19.4 o or 25.7o field of view lens.
2.3. Bioacoustic Node
The bioacoustics node consisted of one microphone for bird calls and
ambient noise, a second ultrasonic microphone for bat echolocation calls.
Ultrasonic microphones are sensitive to sounds greater than 20 kHz, which
is typical of bat echolocation calls. We initially used an stand alone
acoustic recorder that had self contained power, microphone, analogue to
digital converter, filter, and data storage (Wildlife Acoustics SM3 Bat/Bird;
Fig. 2.3.1). This recorder was used during our initial, tests at NREL,
however at that time, the bioacoustics were not integrated into the array.
During our final tests at NREL, we used a completely redesigned
bioacoustic node that was fully integrated with other sensors in the array
(Fig. 2.3.2). The components of the integrated bioacoustics node included:
1) a general purpose electronic piezoelectric microphone for audio signals
powered by a 4 mA 1-Channel CCP Power Module with A-weighting filter
(G.R.A.S. Sound and Vibration A/S); 2) a Knowles FG Electret ultrasound
microphone for ultrasonic signals (Avisoft Bioacoustics); and 3) a National
Instruments NI-9223 DAQ analog to digital converter. Although we
designed the system to record both frequency ranges, we used only the
acoustic microphone during impact testing given that all tests were
conducted during daylight hours, negating potential to record bat
echolocation calls. Microphones were mounted both inside and outside the
nacelle during tests on the CART 3 turbine at NREL.
Sampling rates:
The four channel NI-9223 DAQ can sample at 1 MHz channel-1. The
sampling frequency for recordings was approximately 5-10X the
frequencies of the anticipated vocalizations or echolocation. Sampling
frequency was 200 kHz for the acoustic microphone and 500 kHz for the
ultrasonic microphone.
Figure 2.3.1.
Independent
bioacoustic recorder.
Figure 2.3.2. Integrated
bioacoustics node
showing the
microphone (top),
analogue to digital
converter (middle), and
microphone power
module with filter
(bottom).
6
2.4. System Integration and Data Acquisition
We used LabView system design software operating on the
central computer for individual component control and data
acquisition (Fig. 2.4.1). We created a custom LabView VI to
control instruments from each sensor node, including sampling
frequency and duration. These individual sensor node VIs were
embedded within an overarching system control VI that triggered
data collection from all sensors and stored the data in a
standardized format on the central computer separately for each
triggering event.
The central computer controlling the system was placed in
the nacelle and controlled by a second computer at ground level
during operational testing via an Ethernet connection and remote
desktop. Data from each sensor streamed continuously to the
central computer in the nacelle during system operation. The
volume of data from all sensors, especially the optical node, was
too large and time consuming to archive data continuously. To
manage this volume of data, we used a ring buffer architecture to
save data from impact events. Without a triggering event, the
ring buffer was continually overwritten. Once an event recording was triggered, however, data streams
from all sensors were written to disk for a specified period of time before and after the event (Fig. 2.4.2).
For our test purposes, we chose 20 second buffers for all sensor nodes, with the impact event nominally
centered in the buffer. In a non-experimental setting, additional extended recording time for the
bioacoustics could capture vocalizations or echolocations from the species of bird or bat involved in a
collision, provide a relative assessment of how many individuals were nearby, and assess ambient
environmental conditions that may have led to the collision or a possibly false positive detection (e.g.,
hail, lightning, etc.). The software allowed individual control of the buffer size for each node, therefore
the user could determine the amount of data to be stored for each sensor before and after an event trigger.
2.5. Impact Event Detection and System Trigger
We tested several event detection and system triggering mechanisms ranging in complexity.
Automated detection algorithms tested included continuous wavelet transformation (Flowers et al. 2014,
Flowers 2015). These more complex algorithms were tested during retrospective analyses of impact
event data collected during lab or field experimentation. Both approaches showed promise, but require
further development to function in a fully automated, operational setting. We did use a simple automated
threshold trigger during static integrated system tests on a non-operational turbine (CART 3) at NREL.
This automated threshold triggering test verified the full functionality of the integrated system software
and provided a proof of concept for inclusion of more complex real-time detection algorithms during
future developments. We used a manual trigger implemented through the system control VI software for
all operational tests of the integrated system.
Figure 2.4.1. Example window
of custom built LabView virtual
instrument user interface.
Figure 2.4.2. Ring buffer
architecture for multi-sensor
data recording, event detection,
and data storage.
Oldest data
20 sec buffer
Newest data
Event Detector
Data Streams
7
2.6. Experimental Impacts
We used a compressed air
cannon capable of shooting tennis
balls to create experimental impacts
on research turbines (Fig. 2.6.1).
The propellant source for the canon
was an air compressor with a pre-
pressurized air tank. The regulator
for the air cannon was set to 115 psi
(the max pressure for the electric
valve). The launcher was controlled
using an off/on toggle safety switch
and a momentary switch. The air
cannon is operated by holding down
the momentary switch to charge the
air cannon’s air tank and then
releasing the momentary switch to
open the valve and launch the tennis
ball. Switches controlling the electric
pneumatic valve were powered by
two 12V DC batteries connected in
series (24V DC). The cannon is
barreled to shoot a standard tennis
ball that weighted ~57 g empty (used
at MCC & NREL) or ~140 g filled
with water (MCC only). Empty
tennis balls were similar in mass to
the largest bats and smallest seabirds
(Fig. 2.6.2). During some tests at
NREL, we soaked tennis balls in
warm water to provide a stronger
thermal and residual impact signal
for the IR cameras.
2.7. Project Timeline
The first three years of the project primarily involved refining system design and individual
component tests (Table 2.7.1). In designing the system, we relied on considerable input on many aspects
of the system and performance requirements from advisory panel members representing industry and
agency perspectives, as well as science and biological expertise. Several key modifications resulting from
advisory panel input include: 1) stereo camera configuration to calculate size and distance to object, 2)
integration of the bioacoustics node into the sensor array because of its potential contribution to impact
detection (Evans 2012), and 3) importance of identifying individual species of birds or bats interacting
with the turbine, therefore requiring high resolution imagery. These considerations led to extra design,
Figure 2.6.1. Compressed air cannon used to create experimental
impacts on research turbines.
Figure 2.6.2. Mass of an empty (i.e., not water filled) tennis ball
used in experimental wind turbine impact tests relative to large bats
and small to large seabirds
Utah Dept. of Natural Resources
Puffin
Pelican
Shorebird
Storm-Petrel
Albatross
Hoary Bat
10 g …….………..800 g………………. 6,000 g
© John Avise
Western
Gull
8
laboratory, and field testing (stereo cameras and integrated bioacoustics) and sensor mounting
configuration on the turbine. To ensure success of these efforts, we made three rather than the one
planned visits for testing at NREL and added a stereo camera test in collaboration with Pacific Northwest
National Laboratory scientists in Sequim, WA. We required a 1-year extension, for a total of 4-years, to
complete all of the project tasks (Table 2.7.1).
Table 2.7.1. Timeline for major tasks of the project. Total project period was October 1, 2011-September
30, 2015.
Component
Task
Vibration
Optic
Bioacoustic
Location
Performance Period
(Budget Periods 1-4)
Advisory panel
meetings
Newport &
Corvallis
1st meeting in BP1
2nd meeting in BP3
Individual component
testing and calibration
X
X
X
Seattle, Corvallis,
Newport
BP1-BP4
Target identification
#1&2
X
Newport
BP2, BP3
Test site assessment
X
X
X
NREL, NWTC
BP2
Collision event
detection (individual
components, non-
integrated)
X
MCC, NAWRTC
BP3
Target identification
(stereo)
X
Sequim
BP3
Collision event
detection (individual
components, non-
integrated)
X
X
X
NREL, NWTC
BP3
Collision event
detection (fully
integrated system)
X
X
X
NREL, NWTC
BP4
9
3. Results
3.1. Vibration Node
Laboratory Testing
Initially the vibration node was tested in the laboratory
using a mechanical shaker (Fig. 3.1.1). With the
accelerometer on a bar attached to the shaker, gently tapping
the bar with a screw driver served as an impact signal. The
accelerometer signal from this test is a 10hz sine wave (57
mVrms) combined with white Gaussian noise (80 mVrms).
This results in a very noisy signal with periodic
characteristics, much like the actual wind turbine data we
received (Fig. 3.1.2). Once low pass and high pass filters of
discrete wavelet transformations (Coiflets 5) are applied, a
clear impact signal is evident (Fig. 3.1.2). The high pass
filters out the white noise (which was a bit more dampened by
the bar than expected) and most of the periodic signal. The
filtering and detection algorithm produced good results on
artificial and actual wind turbine data with impact signals,
however, additional refinement is necessary for operation in
real-time.
Archived turbine operation data
We obtained accelerometer data from turbines operated
by Floating Power Plant A/S and NREL to provide
background operational signals in which we could insert
simulated impact signals to test detection algorithms. Data
from Floating Power Plant were collected at too low of a
frequency and, unfortunately, could not be used for these
purposes. Some data from the NREL CART 3 turbine,
however, were collected at 400 Hz and, therefore, beneficial
for continued laboratory testing and calibrating of the impact
detection algorithm.
Tests on Turbine
We conducted operational tests in 2013 at MCC and in 2014 and 2015 at NREL. A summary of data
collected during these tests is provided in Tables 3.1.1 and 3.1.2.
Figure 3.1.1. Accelerometer
mounted to an aluminum bar
atop a shaker. An input/output
module controls the shaker and
receives data from the
accelerometer.
Figure 3.1.2. (top) Raw
accelerometer data from shaker.
(lower left) Data after low pass
filtering and (lower right)
impact signal after a combined
low then high pass filtering
using discrete wavelet
transformations.
10
Table 3.1.1 Tests performed on the MCC GE turbine with only vibration sensors mounted on the turbine
(Ximea smart cameras collected some image data from the ground), 9-13 December 2013. Sample sizes
(# of recordings) are shown for each sensor node (n = 630 recordings from individual sensors). All
recordings were using a manual trigger.
Sensor
Description
Vibration
Turbine start-up, shut-down, and engaging generator
sequences
22
Turbine Normal Operation, single ball shot from air
cannon on ground and impacting a blade
17
Turbine Normal Operation, single ball shot from air
cannon on ground and impacting the tower
11
Turbine Normal Operation, single ball shot from air
cannon on ground and impacting the tower and a blade
6
Total =
Total blade impacts =
55
23
Table 3.1.2. Tests performed on the NREL CART 3 turbine with all sensor components, 22 - 23 October
2014 and 13 - 17 April 2015. Sample sizes (# of recordings) are shown for each sensor node (n = 2,252
recordings from individual sensors – 6 vibration sensors, 4 optical, 1 bioacoustic). Due to varying wind
conditions and greater background noise of the older CART 3 turbine, we ran the turbine under several
modes. Recordings are summarized among the different operational modes.
Sensor
Description
Vibration
Optic
Bioacoustic
Total
Turbine Normal Operation1, start-up and shut-down
sequences
13
1
9
23
Turbine Idle Rotation2, start-up and shut-down sequences
2
0
2
4
Stationary blade, single ball hand thrown from the
nacelle
21
19
2
42
Stationary blade, multiple balls (2-4) hand thrown from
the nacelle
3
3
-
6
Stationary blade, single ball shot from air cannon on
ground (manual trigger)
2
20
21
43
Stationary blade, single ball shot from air cannon
(automatic trigger)
4
3
4
11
Turbine Normal Operation1, single ball shot from air
cannon on ground (manual trigger)
6
6
6
18
Turbine Idle Rotation2, single ball shot from air cannon
on ground (manual trigger)
4
4
4
12
Total =
55
56
48
159
Total blade impacts (stationary) =
30
45
27
Total blade impacts (operational) =
10
10
10
1The generator is engaged during normal operation.
11
2The wind speed is too slow to engage the generator during idle rotation. Operating the CART 3 in this
state was beneficial to our tests because the background noise of the turbine was much quieter than the
intermediate noise from the GE turbine at MCC and the most noise from the CART 3 during normal
operation.
Static Impact Tests
We initially tested sensor operation and
recorded impact signals of a tennis ball
thrown by hand from atop the nacelle and
impacting a stationary blade. The 3 axes
accelerometers showed the strongest signal in
“Z” or out of plane axis of the blade (Fig.
3.1.3). The paired contact microphone and
accelerometer both showed a clear signal of
the impact (Fig. 3.1.4). Repeated blade
impact signals are distinctive from three
tennis balls sequentially thrown from the
nacelle (Fig. 3.1.5).
Dynamic Impact Tests
We first obtained background levels of
turbine operation by recording sequences of
turbine start-up, shut down, and generator
engagement for both the GE and Cart 3
turbines (e.g., Fig. 3.1.6). This along with
prior CART 3 operational data obtained from
NREL provided vital information on
characteristic periodic and non-periodic
vibration signals that an impact detection
algorithm must account for. Furthermore,
Figure 3.1.6. Accelerometer data from a
single blade (GE) during turbine startup
(top) and shut down (bottom).
Figure 3.1.3. Data from the three axes of one
accelerometer on a single, stationary blade (CART 3)
showing the impact of a single tennis ball hand
thrown from on top of the nacelle. The strongest
signal is in the Z axis (bottom).
Figure 3.1.4. Data from a contact microphone (top)
and accelerometer (bottom) on a single, stationary
blade (CART 3) showing the impact of a single
tennis ball hand thrown from on top of the nacelle.
The impact signals are slightly offset because the
time recording of the two sensors were not perfectly
synchronized.
Figure 3.1.5. Data from one accelerometer on a
single stationary blade (CART 3) showing three
sequential impacts from three tennis balls thrown
from the nacelle.
12
these measurements clearly demonstrated striking differences in background noise levels between the GE
and CART 3 turbines and the different operational modes of the CART 3. It was clear that the smaller,
but older CART 3 had larger background vibration signals through which to discern impact signals (Fig.
3.1.7). The greater background vibrations of the CART 3 could potentially mask impact signals of less
kinetic energy that were detectable on the larger GE (MCC). Indeed, we observed that fewer impacts
could be visually detected in the accelerometer data from the CART 3 at NREL than what we had
expected based on results from impact tests on the GE at MCC. Due to low wind occurrence during our
final test at NREL, we also operated the CART 3 turbine in idle rotation (generator not engaged). The
CART 3 turbine operating in idle rotation had much lower background signals compared to both the
CART 3 and GE during normal operation (Fig. 3.1.7). Low background signal of the CART 3 during idle
rotation is in part because the generator was not engaged, therefore less load on the gearbox, but also
because the blades were rotating at a lower speed than both the CART 3 or the GE under normal
operation - as indicated by the reduced frequency of the sine wave. Fortunately, the two operational
modes of the CART 3 turbine plus the normal operational mode of the GE provided three background
signal scenarios for us to test impact signal detection.
In total, we measured 23 blade impacts on the GE turbine under normal operation, six blade impacts
on the Cart 3 during normal operation, and 4 blade impacts on the CART 3 during idle rotation (Tables
3.1.1, 3.1.2, Fig. 3.1.8). Impaction detection is dependent on background signals, position on the blade
section, and impact kinetics. Overall detection percentage ranged from 100% for the “quietest”
conditions (CART 3 idle rotation), down to 35% for the noisiest (CART 3 normal operation; Fig. 3.1.8).
Figure 3.1.7. Increasing levels of background “noise” in out-of-plane (Z) axis accelerometer data
from the CART 3 600 kW turbine during normal operation (generator engaged), the GE 1.5 MW
turbine during normal operation, and the CART 3 during idle rotation (generator not engaged). The
frequency of oscillations indicates the CART 3 under normal operation is rotating at the highest rpm,
followed by the GE, then the CART 3 during idle rotation.
13
Impact signals were detected from sensors on more than one blade (i.e., blades other than the blade
struck) 50% (CART 3 during normal operation) to 75% (GE during normal operation and CART 3 during
idle rotation) of the time.
3.2. Optical Node
Laboratory Testing
We first determined lens focal length that would allow identification of target species of birds and
bats that would be expected offshore of North America. For wind turbines deployed on the west coast of
the United States, a signature avian species of regulatory concern is the marbled murrelet
(Brachyramphus marmoratus). These birds are quite small relative to other seabirds (10 cm body size)
and can fly at speeds up to 44 m/s (100 mph) and, therefore, is a good minimum target identification size.
The candidate locations for IR camera are on the turbine nacelle or the tower. Prior experience suggests
that the camera should not be oriented towards the water surface since small targets will be
indistinguishable from sea clutter (i.e., variations in emissivity associated with wave propagation and
breaking). In the following example, three lenses (15o, 25o, 45o field of view) are considered for an IR
camera either: 1) Mounted 20 m above the water line on the tower, oriented towards the sky or 2)
Figure 3.1.8. Percentage of turbine blade impacts from a 57 g tennis ball shot from a
compressed air cannon on the ground that were detected by accelerometers or contact
microphones mounted on blade roots (one accelerometer and one contact microphone on each
blade). Data are from two different turbines with one of the turbines operating in two
different states providing different background signals (accelerometer graphs below x-axis;
see Fig. 3.1.7 for explanation)
% of Impacts detected
n=6
n=4
n=23
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
CART 3 - normal
operation GE - normal
operation CART 3 - idle
rotation
Accelerometer
14
Mounted on top of the nacelle and oriented towards the sky. Neglecting lens distortion, at a given
distance (D) from the camera, the width (L) of the field of view is given as a function of the lens angle (θ)
by the trigonometric relation
2
tan2
DL
.
The size of a pixel at distance D is then given as L/Rx, where Rx is the horizontal resolution (i.e., 640
pixels). The number of pixels spanning a marbled murrelet can then be readily calculated from the
murrelet’s body size.
For the example, we consider the case of Siemens SWT-6.0-154 (6 MW) offshore wind turbine,
representative of the scale of an offshore turbine making use of the monitoring system. The turbine blades
for the SWT-6.0-154 are 75 m long and the hub is 4 m in diameter. Hub height is selected in a site-
specific manner and here we assume a hub height of 116 m (consistent with the hub height of the
prototype SWT-6.0-154). Target resolutions for each lens and camera orientation combination (6 total)
are shown in Figure 3.2.1. Target detection is unlikely to be possible if there are less than 3-4 pixels
spanning a target.
These results suggest that only an IR camera with a narrow field of view (e.g., 15o lens) would allow
for target detection over the length of a turbine blade. Coverage with this narrow angle lens, however, is
restricted to a small portion of the rotor swept area, requiring multiple cameras to achieve sufficient
coverage. Conversely, a 45o lens could cover the majority of the rotor swept area. In this configuration,
detection of a murrelet is unlikely at appreciable distances, but detection of larger seabirds is feasible.
A second metric for target detection is the number of frames containing a target. If a murrelet is
assumed to be traveling in a straight line in a horizontal plane, then the number of captured frames is
given by
Figure 3.2.1. Trade-off between percentage of rotor imaged and spanwise target resolution for
a 10 cm marbled murrelet (i.e., potential for detection) for three lens fields of view and two camera
positions (upper – top of nacelle; lower – 20 m above sea surface). Dashed black line indicates
extent of rotor sweep for Siemens SWT-6.0-154. Solid black line indicates tower height.
15
Fx
uL
target
where utarget is the target speed (m/s), F is the camera frame rate (Hz), and L is as previously defined.
The FLIR A655sc can record full resolution frames at up to 50 Hz. Marbled murrelets are likely to fly at
speeds between 17 and 44 m/s (40-100 mph). Figure shows the number of frames that would be captured
under this idealized scenario for two lens angles. Due to the wider field of view, the 45o lens is able to
capture many more frames containing a target than the 15o lens at equivalent distances and target speeds.
These results suggest that high-speed target detection using a 15o lens may be challenging at distances
less than 10 m from the camera (i.e., the target will be present in relatively few frames). This is balanced
against a high number of pixels per target at close range, which will facilitate detection and identification.
In summary, analyses suggest that if marbled murrelets are a species of concern, then target detection
will be best achieved by a narrow angle lens (i.e., 15o). In order to achieve significant rotor and tower
coverage this will, however, require on the order of ten cameras – or twenty for stereo imaging – per
turbine. For locations where larger species are likely to be of greatest concern (e.g., albatross in mid-outer
continental shelf deployments), a wider angle lens may be effective and preferred.
Early in the project, we pursued stereo imaging techniques upon recommendation of our advisory
panel. Stereo imaging allowed us to determine target size, speed, and position, but also increased the
complexity of the optical node. Stereo imaging required camera calibration (Fig. 3.2.3) and software
permitting target identification and tracking. We were successful in using stereo imaging to measure size
and distance of bats (Fig. 3.2.4) and birds (Fig. 3.2.5) during field trials in Sequim Washington.
Figure 3.2.2. Number of frames captured as a function of target speed and distance from camera for
two lenses.
16
Experimental Turbine Tests
We completed 18 tests using single tennis balls soaked
in warm water to provide a thermal signature for IR cameras
while in flight and during impact (Table 3.1.2). The warm
water left a strong thermal signature on the blade at the point
of impact (Fig. 3.2.6). The point of impact was less evident,
but still detectable with the visual cameras (Fig. 3.2.6). We
collected IR and visual imagery from 20 impacts on a
stationary blade with a single tennis ball shot from the
ground using a compressed air cannon (Table 3.1.2). In
addition to inherent differences in IR and visual images, we
were able to document the effect of different camera
resolution and frame rates in identifying the object and
capturing the object near the moment of impact. The IR
cameras were higher frame rates (12 fps) than the visual
cameras (6 fps), allowing us to capture imagery near the
moment of impact for IR, but not visual cameras (Fig. 3.2.7).
For the purposes of this project, we recommend using 15°
field of view, similar to the IR camera, and > 12 fps. The
lower resolution of the IR camera (640 X 480) was
somewhat compensated by the longer focal length (narrower
field of viewer) to help identify the object (Fig. 3.1.7). The
thermal signature of the warmed tennis ball and water left on
the blade at the point of impact was also evident on the
moving blade in the IR image (Fig. 3.2.7).
Figure 3.2.3. Stereo camera
calibration chart.
Figure. 3.2.4. Target tracking (50 fps)
and distance calculation of a bat. Two
bats are in image. Species was either
silver-haired, California myotis, or
little brown – all were present in audio
recordings during image capture.
Figure. 3.2.5. Target tracking (12 fps)
and distance calculation of a Brewer’s
blackbird.
17
During fully integrated system tests at NREL, tennis balls were fired from the upwind side of the
blades due to safety constraints. With cameras mounted from the top of the nacelle looking at the
downwind side of the blade, few images captured a tennis ball traveling through the air or striking the
blade, due to the blade itself visually obstructing the event. This experimental difficulty demonstrates the
significance of camera placement in target identification. An alternative camera location is to mount
cameras directly on the blade, near the hub looking outward. Each blade is always in full view, and this
configuration allows shorter imaging distances, higher target resolution, and minimizes the number of
Figure 3.2.6. Infrared (left) and visual (right) image of a tennis ball soaked in warm water
bouncing off a stationary blade. IR and visual cameras are mounted side-by-side (Fig. 2.2.1).
The IR image (left) was taken with a resolution of 640 X 480 at 12 fps and 15° field of view
lens. The visual image (right) was taken with a resolution of 1624 X 1234 camera at 6 fps
and 52° field of view lens.
Figure 3.2.7. (Left) IR camera showing impact to a stationary blade of a tennis ball fired
from an air cannon on the ground. (Right) Visual camera showing the ball falling away after
the same impact event. The IR image (left) is taken at 12 fps, allowing us to capture the
moment of impact The visual image (right) is taken at 6 fps and did not allow us to capture
the moment of impact.
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.
19
3.3. Bioacoustic Node
Since it was highly unlikely that we would
record bird or bat vocalizations or echolocations
during our short-duration experimental tests, we
used audio recordings of turbine operation and
impact sounds to demonstrate successful operation
of the bioacoustics node. These results highlighted
the additional value of the bioacoustics node to
provide ambient acoustic information that could
help identify an impact signal or possible reasons
for false positive impact signals from the vibration
node, such as rain, large hail stones, lightning, etc.
Static Tests
We collected acoustic data during initial
impact tests on a stationary blade. Tests included
tennis balls hand thrown from the nacelle, as well
as those shot from the compressed air cannon on
the ground. During static tests the acoustic signal
of the compressed air cannon and the impact of the
tennis ball were both detectable and about 1 sec
apart in the acoustic spectrogram with the
microphone located inside the nacelle (Fig. 3.3.1).
Operational Tests
We collected acoustic data of turbine
operation and impacts from inside and outside the
nacelle. The spectrogram from the inside of the CART 3 turbine during startup and shutdown with the
generator engaged demonstrated the intense background noise (Fig. 3.3.2) that could mask some impact
and bioacoustics signals. Indeed, during impact tests the signal of a tennis ball impacting the blade was
not readily detectable with the microphone located inside the nacelle (Fig.3.3.3). With the microphone
mounted outside of the nacelle, however, an impact signal was detected (Fig. 3.3.3).
Figure 3.3.1. Spectrogram from an acoustic
recorder located inside the nacelle of the CART
3 turbine during a stationary blade impact test.
The left signal is the compressed air cannon
being fired from the ground and the right signal
is the impact of the tennis ball on the blade.
Figure 3.3.2. A spectrogram from an acoustic
recorder located inside the nacelle of the CART
3 turbine during an 82 second startup and
shutdown sequence.
20
3.4. System Integration, System Triggering, and Event Detection
We used National Instruments LabVIEW to write custom programs to operate instruments within
each node and to integrate all nodes to record through a single trigger interface (Fig. 3.4.1).
Figure 3.3.3. (top) Spectrogram from an acoustic recorder located inside the nacelle during normal
operation of the CART 3 600 kW turbine. A blade impact of a tennis ball shot from the compressed
air cannon on the ground was masked by the background operational noise of the turbine. (bottom)
Spectrogram from an acoustic recorder located outside the nacelle showing a blade impact of a tennis
ball shot from the compressed air cannon on the ground during idle turbine rotation.
Cart 3 Idle Operation
microphone OUTSIDE nacelle with impact
Accelerometer
Impact acoustic signal detected
21
Manual Triggering
The LabVIEW VI was initially programmed for manual triggering during impact testing. As soon as
an impact was observed, manual triggering saved vibration, optical, and bioacoustic node data from the
ring buffer for 10 seconds before and after triggering (i.e., total buffer length of 20 seconds). These data
were later post-processed to assess impact detection and magnitude of signals. We used manual
triggering for all dynamic tests with the turbine in operation.
Automated Triggering
As noted above, wavelet analysis showed promise for automated event detection. The preliminary
results validate the planned implementation of a wavelet-based event detection algorithm. Using both
artificially generated vibrations data and real vibrations measurements, it was possible to find the time of
occurrence of a simulated impact at high-level of kinetic energy. Low-energy impact detection requires
further investigation. Additional experimental data from laboratory and field tests will be critical for
system tuning, implementation and validation (Flowers 2015). Furthermore, we continue to evaluate
other event detection algorithms. Use of automated event detection and triggering in real time is feasible,
but requires additional development beyond our project.
As a proof of concept for automated triggering, however, we did use a threshold filter on the
accelerometer data stream during static blade impact tests, therefore not requiring removal of background
signals. After selecting an appropriate threshold, the automatic trigger on static blades was highly
successful, with all four impacts successfully triggering the system (Table 3.1.2). This test demonstrated
Figure 3.4.1. Schematic diagram of individual LabVIEW programs and communication links
between nodes of the sensor array.
User
Trigger
(#1)
(#2)
22
that it is feasible to trigger the sensor array through single impact events sensed by the vibration node,
however, a more advanced triggering algorithm will need to be developed.
4. Next Generation System Design & Commercialization
4.1. Vision Statement
Wind energy production in the U.S. is projected to increase to 35% of our nation’s energy by 2050.
This substantial increase in the U.S. is only a portion of the global wind industry growth, as many
countries strive to reduce greenhouse gas emissions. A major environmental concern and potential
market barrier for expansion of wind energy is bird and bat mortality from collisions with turbines.
However, there are no commercially-available, real-time, impact detection system to document mortality
events, inform impact risk assessment, or verify whether impact deterrents are working.
Our blade-mounted multi-sensor array can detect impacts and identify the impact source, including
small birds and bats, in real-time on operating, commercial scale turbines. Data streams to wind facility
operators could allow immediate assessment of impact events. Technology and industry advances over
time will allow this low-cost monitoring system to be designed into materials during manufacturing so
that all turbines could be monitored with either a subset or full suite of sensors. As standard equipment
on all commercial turbines, the industry could effectively monitor whether individual turbines were
causing mortalities or not and under what circumstances, as well as evaluating mechanical and structural
integrity of a turbine via real-time vibration, image, and acoustic data streams – permitting modification
or shut-down to limit environmental or mechanical damage.
4.2. System Description
We envision a commercially viable
system will have all vibration, optical, and
acoustic sensors integrated in a single unit to
attach to the blade of a turbine (Fig. 4.2.1).
The sensor unit will be powered by batteries
that are charged by solar panels and kinetic
blade motion. Smart sensors will permit
near-real-time onboard processing of data,
thereby reducing the computational burden
of the central computer and improving
overall system efficiency. Integrated sensor
units can be attached to currently operating
turbines, but ultimately we anticipate that
sensors will be built into blade design and
manufacturing.
4.3. Application
Once installed, the accelerometers and contact microphones could be used to build a vibration
“signature library” for individual turbines, each signature being unique yet temporally variable, which a
learning algorithm could adapt to. Over time the sensor array and event detection algorithms could
become increasingly more informed to detect subtle impacts beyond standard operational signals.
Furthermore, continuous monitoring of the turbine by the sensor array will provide real-time assessment
Figure 4.2.1. Integrated sensor array to be mounted on
turbine blade. Ultimately, we anticipate that sensors
will be built into blade design and manufacturing.
23
of the structural integrity of the turbine. Turbine operators could also use the vibration, image, and
acoustic data streams to detect irregularities, environmental conditions, or unforeseen events that may
affect power output, or lead to potential failure before the next scheduled turbine maintenance.
There are numerous ways in which the sensor array could be scaled to meet commercial needs.
Initially, the array could be deployed on several turbines at a small-scale evaluation or demonstration site.
At a commercial-scale facility, full sensor arrays could be deployed on a subsample of the turbines in
strategic, but statistically relevant, locations. The remainder of the turbines could be instrumented with
only the vibration node, allowing detection of impacts and structural monitoring of the turbine, but no
image or audio data to help identify causes of impacts of operational irregularities of turbine.
4.4. Commercialization Plan
Commercialization of the system could require roughly four major steps (Table 4.4.1). 1) Obtain
intellectual property protection for the concept and system design; 2) Obtain additional funding and seek
capital investment and partnering with an engineering firm to refine current system design. Conduct
additional experimental tests, and extended test deployments on commercially operating turbine in a high
impact area; 3) Scale-up system design for commercial deployment; 4) Identify commercialization
pathway and creation of division or new company that builds end-user software products or provides fee-
based system set-up and life-cycle monitoring.
Table 4.4.1.
Action
Description
Date Completed
Step 1: Provisional patent
Oregon State University filed a United States
provisional patent application (No. 62313028) for the
blade-mounted, integrated sensor array
March 2016
Step 2: Potential Partners
and funding
Oregon:
OSU Advantage Accelerator to analyze market,
identify customers, examine supply chain to
identify the right commercialization partners
University Venture Development Funds to
improve prototype
Funding from Oregon BEST (Oregon Built
Environment & Sustainable Technologies Center,
Inc.)
Other States:
California Energy Commission
New York State Energy Research & Development
Federal:
Department of Energy
Private:
Renewable Energy Systems Americas, Inc.
NextEra Energy Resources
Before March
2017
TBD
TBD
TBD
Step 3: System design for
commercial scale use
Possibly license to an existing company or creation of
a new venture (decision based on information
gathered and results from above tasks)
After March
2017
Step 4: Commercialization
pathway
Use information gathered from above tasks to develop
a commercialization pathway.
TBD
OFFICE FOR COMMERCIALIZATION AND CORPORATE DEVELOPMENT
J , .
Available
Tec hnol ogy :
Environment
& Sensing
Advantage – Impact
Please Contact:
David Dickson
IP & Licensing Manager
541.737.3450
David.dickson@oregonstate.edu
Technology Ref. # OSU-15-21
oregonstate.edu/research/occd
Wind Turbine Sensor Unit for
Monitoring of Avian & Bat Collisions
Wind turbines are a substantial and growing source of renewable electricity. However, the collision of
endangered bird and bat species with turbines poses a serious barrier to turbine deployment, both offshore
and on land. Standard carcass counting land survey methods are fraught with uncertainty and error, and
this method is impossible for offshore turbines. An integrated multi-senor system, capable of providing
temporal and spatial coverage of collision events has been developed to monitor collision events to address
this need. The invention enables the environmental impacts of wind turbines to be remotely monitored and
ensure the benefits of renewable power generation are not outweighed by mortality of protected species.
TECHNOLOGY DESCRIPTION
An integrated multi-sensor detection
package with event-based data collection
has been developed using accelerometers,
microphones, and a cameras coupled with
integrated signal processing. The sensors
are installed directly on wind turbine
blades and on-board processors wirelessly
transmit data to the central controller
and data acquisition system immediately
after a collision. The accelerometers and
contact microphones provide continuous
temporal coverage for collision detection.
The visual and IR cameras and
bioacoustic recorders provide
taxonomic classification. Wireless
connectivity, low power consumption,
and small size allow these sensors to be
installed on existing turbines with
minimal impact and easily integrated
into new turbines.
STATUS
Provisional patent application filed.
Seeking commercial development
partner, available for licensing.
Applications
•Bird and bat collision monitoring
•Off-shore and terrestrial turbines
•Species classification
Features & Benefits
•Low-cost
•Retrofittable on existing turbines
•Easily integrated into new turbines
•Temporal and spatial coverage
J , .
Advantage – Impact
OFFICE FOR COMMERCIALIZATION AND CORPORATE DEVELOPMENT
About the Principal Investigator
DR. ROBERTO ALBERTANI
Dr. Robert Albertani is an associate professor of mechanical engineering at Oregon State
University. Dr. Albertani received his MS in aeronautical engineering from the Polytechnic
University of Milan, Italy, and a PhD in mechanical & aerospace engineering from the University of
Florida.
Dr. Albertani’s research interests include aerodynamics and stress analysis of flexible structures,
environmental impact of wind energy, high-performance sailboat testing techniques, fiber
composites technology, micro air vehicles, and biological flight mechanics.
e OCCD supports research development and commercialization of University intellectual property. Focusing on the protection and transfer
of intellectual property through license, confidentiality and material transfer agreements, the OCCD is the bridge between researchers and
commercial entities. From Oregon-based startups to large international companies, the OCCD facilitates OSU research to impact the world.
Visit oregonstate.technologypublisher.com to view technologies available for commercialization.
oregonstate.edu/research/occd
26
5. Summary of project accomplishments
5.1. Patents
Oregon State University filed a United States provisional patent application (No. 62313028) for the
blade-mounted, integrated sensor array, March 2016
5.2. Publications & Other Products
Proceedings
Flowers, J., R. Albertani, T. Harrison, B. Polagye, R. Suryan. 2014. Design and initial component
tests of an integrated avian and bat collision detection system for offshore wind turbines.
Proceedings of the 2nd Annual Marine Energy Technology Symposium, Seattle, WA, April 15-
18, 2014. http://vtechworks.lib.vt.edu/bitstream/handle/10919/49197/65-
Flowers.pdf?sequence=1
Thesis
Flowers, J. M. 2015. Design and testing of an integrated wildlife-wind turbine interactions detection
system. M.S. thesis. Oregon State University, Corvallis, Oregon.
Book chapter
Henkel, S.K., R.M. Suryan, B.A. Largerquist. 2014. Marine renewable energy and environmental
interactions: Baseline assessments of seabirds, marine mammals, sea turtles, and benthic
communities on the Oregon shelf. In Marine Renewable Energy Technology and Environmental
Interactions. M. A. Shields and A.I.L. Payne (Eds.). DOI 10.1007/978-94-017-8002-5. Springer,
Dordrecht.
5.3. Presentations
Invited
Suryan, R.M. 2015. Assessing potential marine bird impacts from offshore energy development.
Oregon State University, Corvallis, Oregon
Suryan, R.M. 2014. A synchronized sensor array for remote monitoring of avian and bat interactions
with offshore renewable energy facilities. Department of Energy, wind power peer review.
Arlington, Virginia.
Suryan, R.M., R. Albertani, B. Polagye. 2012. A sensor array for remote monitoring of avian and bat
interactions with wind turbines. Northwest National Marine Renewable Energy Center annual
meeting. Corvallis, Oregon.
Scientific conference
Suryan, R.M., R. Albertani, B. Polagye, J. Flowers, T. Harrison, C. Hu, W. Beattie. 2015. Design
and Development of an Integrated Avian and Bat Collision Detection System for Wind
Turbines. North American Wind Energy Academy, Blacksburg, Virginia
Suryan, R.M., R. Albertani, B. Polagye, J. Flowers, T. Harrison. 2014. Near real-time detection of
avian and bat interactions with wind turbines. National Wind Wildlife Research Meeting X,
Broomfield, Colorado.
27
Harrison, T. B. Polagye, R.M. Suryan. 2014. Remote monitoring of birds and bats using visual and
infrared stereo imagery. National Wind Wildlife Research Meeting X, Broomfield, Colorado.
Flowers, J., R. Albertani, B. Polagye, R. Suryan, T. Harrison. 2014. Remote monitoring of avian and
bat interactions with offshore wind energy facilities. 2nd Annual Marine Energy Technology
Symposium, Seattle, Washington.
Suryan, R.M., R. Albertani, B. Polagye, J. Flowers, T. Harrison. 2014. A synchronized sensor array
for remote monitoring of avian and bat interactions with offshore wind turbines. Ocean Sciences
Meeting, Honolulu, Hawaii.
Suryan, R.M., R. Albertani, B. Polagye. 2012. A synchronized sensor array for remote monitoring of
avian and bat interactions with offshore renewable energy facilities. Pacific Seabird Group
Annual Meeting, Turtle Bay, Hawaii.
Public
Suryan, R.M. 2014. Seabirds and wind energy. Newport Intermediate School 6th grade science
classes, Newport, Oregon
Suryan, R.M. 2014. Seabirds and Marine Renewable Energy. Renewable Energy Challenge for high
school students, Oregon Sea Grant, Hatfield Marine Science Center, Newport Oregon.
Suryan, R.M. 2014. Seabirds and marine renewable energy off the Oregon coast. Yaquina Birders
and Naturalists. Newport, Oregon
Suryan, R.M. 2012. Lost at sea: Monitoring the effects of wind energy devices on seabird mortality.
Oregon Sea Grant Career Day, Hatfield Marine Science Center, Oregon State University.
5.4. Media
TV News
KEZI News, Eugene: Offshore Wind Turbines May Be Dangerous for Birds, Bats (2012)
http://kezi.com/news/local/239071
Radio
OPB: OSU Gets $600K To Study Wind Power's Effect On Birds, Bats (2012)
http://news.opb.org/article/osu-gets-600k-study-wind-powers-affect-birds-bats/
28
Print/Web
Oregon State University http://oregonstate.edu/ua/ncs/archives/2012/feb/researchers-eye-system-
monitoring-offshore-wind-energy-impacts-seabirds-bats
Hatfield Marine Science Center Currents (newsletter)
http://blogs.oregonstate.edu/currents/2012/02/14/researchers-to-develop-system-for-monitoring-
wind-energy-impacts-on-seabirds-bats/
Portland Tribune: OSU to study impact of offshore wind turbines on birds (2012)
http://portlandtribune.com/sustainable/story.php?story_id=132924528122429800
Daily Astorian: OSU Gets $600K To Study Wind Power's Effect On Birds, Bats (2012)
http://www.dailyastorian.com/news/northwest/osu-gets-k-to-study-wind-power-s-affect-
on/article_30030d0f-1fe8-578d-866f-07cb442fc7f1.html
The Chronicle, Lewis County, WA: OSU Gets $600K To Study Wind Power's Effect On Birds, Bats
(2012) http://www.chronline.com/news/northwest/article_2abbbc30-2436-5c57-9351-
8cc2a0b67df4.html
Sustainable Business Oregon: OSU center gets grant to study offshore wind impact on birds (2012)
http://www.sustainablebusinessoregon.com/articles/2012/02/osu-center-gets-grant-to-study.html
The Daily Barometer, Oregon State University Student Media: OSU pursues project on ocean-based
wind turbines (2012) http://oregonstate.edu/dept/student_affairs/studentmedia/osu-pursues-
project-ocean-based-wind-turbines
Quay Country Sun: Oregon students, faculty using Mesalands turbines for research
http://www.qcsunonline.com/2013/12/10/oregon-students-faculty-using-mesalands-turbines-for-
research/
6. Acknowledgements
We are greatly indebted to our collaborators at the National Renewable Energy Laboratory’s
National Wind Technology Center and the Mesaland Community College’s North American Wind
Research and Training Center without who’s help we could not have conducted this study. We also thank
all of the science, management, and industry representatives on our advisory panel (Table 6.1) who
willingly shared their knowledge, encouragement, and provided valuable feedback on system
components, regulatory needs, and design for installation on commercial scale turbines. We thank
Department of Energy program managers and grant administrators for their guidance and support of this
project. The project was funded by the U.S. Department of Energy under Funding Opportunity
Announcement Number: DE-FOA-0000414, U.S. Offshore Wind: Removing Market Barriers, Oregon
State University, and the University of Washington. In-kind matching contributions were provided by
29
Bat Conservation International, Floating Power Plant A/S, and Leidos Maritime Solutions. The
Northwest National Marine Renewable Energy Center was instrumental in project development.
Table 6.1 Names and affiliations of advisory panel
Name
Dr. Judd Howell
Dr. Sharon Kramer
Dr. Jon Plissner
Ms. Manuela Huso
Dr. Mike Lawson
Ms. Karin Sinclair
Mr. Lee Jay Fingersh
Dr. Cris Hein
Dr. Rebecca Holberton
Mr. Anders Køhler
Mr. Craig DeBlanko
Mr. Jerry Roppe
Mr. Kevin Banister
Mr. Mark Waller
Mr. Richard Williams
Ms. Roberta Swift
Ms. Laura Todd
Affiliation
H.T. Harvey and Associates
H.T. Harvey and Associates
ABR Environmental Research and Services, inc.
U.S. Geological Survey, For. and Range Eco. Sci. Ctr.
National Renewable Energy Laboratory
National Renewable Energy Laboratory
National Renewable Energy Laboratory
Bat Conservation International
University of Maine
Floating Power Plant AS
Coastal Community Action Project
Iberdrola Renewables
Principal Power Inc.
Bridgeworks Capital
Leidos Maritime Solutions
U.S. Fish and Wildlife Service
U.S. Fish and Wildlife Service
7. Literature Cited
Allison, T. D., E. Jedrey, and S. Perkins. 2008. Avian issues for offshore wind development. Marine
Technology Society Journal 42:28-38.
Dept of Energy. 2015. Wind Vision: A new era for wind power in the United States. U.S. Department of
Energy.
Desholm, M., A. D. Fox, P. D. L. Beasley, and J. Kahlert. 2006. Remote techniques for counting and
estimating the number of bird-wind turbine collisions at sea: a review. Ibis 148:76-89.
Evans, W. R. 2012. An evaluation of the potential for using acoustic monitoring to remotely assess aerial
vertebrate collisions at industrial wind energy facilities. New York State Energy Research and
Development Authority.
Flowers, J., R. Albertani, T. Harrison, B. Polagye, and R. M. Suryan. 2014. Design and initial component
tests of an integrated avian and bat collision detection system for offshore wind turbines.
Proceedings of the 2nd Marine Energy Technology Symposium, April 15-18, 2014, Seattle, WA
METS 2014.
Flowers, J. M. 2015. Design and testing of an integrated wildlife-wind turbine interactions detection
system. M.S. thesis. Oregon State University, Corvallis, Oregon.
Huso, M. M. P., D. Dalthorp, D. Dail, and L. Madsen. 2014. Estimating wind-turbine-caused bird and bat
fatality when zero carcasses are observed. Ecological Applications 25:1213-1225.
Musial, W., and R. B. 2010. Large‐scale offshore wind power in the United States: Assessment of
opportunities and barriers. Golden, CO.
Wiggelinkhuizen, E. J., S. A. M. Barhorst, L. W. M. M. Rademakers, H. J. Den Boon, and S. Dirksen.
2006. WT-BIRD®, Bird collision monitoring system for multi-megawatt wind turbines. Energy
research Centre of The Netherlands.
