In accordance with the guidance of the West Virginia Public Service Commission (WVPSC) and with recommendations from the Pinnacle Technical Advisory Committee, which included members from the WVPSC, U.S. Fish and Wildlife Service (USFWS), West Virginia Division of Natural Resources (WVDNR), and NRG Energy, we initiated a study in July 2015 to test alternative wind turbine operational strategies to reduce bat fatalities at the Pinnacle Wind, LLC (PWF), Mineral County, West Virginia. Our primary objective was to test the effectiveness of a novel operational minimization strategy to reduce bat fatalities at Pinnacle Wind Farm, LLC (PWF). A secondary objective was to examine potential mechanisms that effect fatality risk to bats.
We randomly selected 15 of the 23 turbines at the PWF for the experiment to evaluate 3 operational minimization strategies. We used a completely randomized block design and treatments were randomly assigned to turbines each night of the experiment, with the night when treatments were applied as the experimental unit. We conducted daily fatality searches between 15 July and 30 September 2015, which represents the expected peak fatality period of bats for this region. The following treatments involved the decision framework to initiate turbine start-up and included:
Treatment A: increased the wind speed requirement to initiate turbine start-up from 3.0 m/s and fully feathered blades until wind speed reached 5.0 m/s based on a 10-minute rolling average as measured at a nearby meteorological (met) tower anemometer at 76 m above ground level (agl). Turbine blades were fully feathered until wind speeds reached 5 m/s. This treatment is currently the standard operating procedure at the PWF from 15 July–30 September,
Treatment B: increased the wind speed requirement to initiate turbine start-up from 3.0 m/s and fully feathered blades until wind speed reached 5.0 m/s based on a 20-minute rolling average measured at the same meteorological tower as Treatment A, and
Treatment C: increased the wind speed requirement to initiate turbine start-up from 3.0 m/s and fully feathered until wind speed reached 5.0 m/s based on a 20-minute rolling average as measured from anemometers on individual turbines at 80 m agl. To reduce the effects of the turbine blades on the wind speed measured downwind on the nacelle mounted anemometer, proprietary calculations were implemented to determine the “free-stream wind speed”.
Decisions to shut-down operations, or curtail turbines, were all based on a 10-minute rolling average of wind speed <5.0 m/s as measured at the met tower for Treatments A and B and the individual turbine for Treatment C. Thus, the shutdown/start-up decision framework for Treatment A was symmetrical (10-minute average to shut-down and start-up), whereas Treatments B and C were asymmetrical (10-minute average to shut-down, 20-minute average to start-up), but with wind speed measurements based on the met tower for Treatment A and B and the individual turbine Treatment C.
During standardized searches, we found 57 fresh bat carcasses, representing 5 different species, including 31 eastern red bats (Lasiurus borealis), 11 hoary bats (Lasiurus cinereus), 10 big brown bats (Eptesicus fuscus), 4 silver-haired bats (Lasionycteris noctivagans), and 1 tri-colored bat (Perimyotis subflavus). No Myotis species were found. We found 17 bat fatalities associated with 4 Treatment A, 12 under Treatment B, and 23 under Treatment C. We removed 5 carcasses prior to analysis because they were associated with nights that experienced treatment implementation error.
We used two methods, Poisson regression and estimated fatality, to evaluate the 3 operational minimization strategies based on fresh bat fatalities. These two methods generally supported each other, although estimated fatality, corrected for detection bias, was the only one that showed a significant difference and only between Treatments B and C. The best Poisson Regression model explaining the number of bat fatalities found under turbines only included turbine differences, but the models with mean hours on and treatment were within 1 Deviance Information Criterion (DIC) unit. In general, Treatment C had higher bat fatalities, significantly higher than Treatment B, and turbines in that treatment were operational for significantly longer periods compared to Treatments A and B. The turbine anemometer had an average 1.03 m/s higher wind speed value compared to the met tower, which likely caused turbines under Treatment C to start-up earlier and shut-down later increasing the operating time. Operating time was not significant and therefore was not solely determined to be the reason for higher bat fatalities based on our Poisson regression models.
As a secondary objective, we examined potential mechanisms that influenced fatality only on nights when turbines were operating regardless of treatment. The best logistic regression mixed model of bat fatalities found per hour the turbine was spinning included number of stops. However, stops/starts and starts were within one AIC unit. This suggested that bats may be at risk during operational transitions (i.e. during turbine start-up or shut-down), specifically the probability of finding a fatality increased significantly with an increasing number of stops. Alternatively, since all treatments were based on a wind speed of 5 m/s it is difficult to separate risk to bats when turbine operations were in transition compared to risk at relatively low wind speeds (e.g. ~ 5 m/s), which might influence changes in turbine operation.
The results of this study suggest that fewer bat fatalities occurred when turbine operations were based on the meteorological tower (Treatment A and B) rather than the individual turbine (Treatment C). This is likely associated with the amount of time turbines were in operation each night, which was longer for Treatment C, although our models suggested other factors may also influence fatality. Furthermore, extending the decision time, from 10 minutes (Treatment A) to 20 minutes (Treatment B), to begin operating turbines when wind speeds exceed the cut-in speed, also may reduce fatalities by reducing the number of transitions (i.e., turbine start-ups and shut-downs). Minimizing the number of start-ups/shut-downs may assist in reducing wear-andtear on turbines and, at least in this study, may also reduce the power loss related to this reduction strategy. Thus, Treatment B represents a decision framework with fewer fatalities, significantly fewer than Treatment C, and compared to Treatment A had less wear-and-tear on turbines (i.e. start-up and shut-downs) with no additional loss in power, and may be the most cost effective option of the 3 treatments studied in this experiment.
The relationship between turbine transitions and bat fatalities is unclear and additional research is needed at other wind energy facilities to better understand bat/wind turbine interactions during start-ups and shut-downs. Until more data are gathered, implementing strategies that limit operational transition of turbines at low wind speeds, such as extending the average decision time period (e.g., from 10 to 20 minutes) to inform turbine operation, may further reduce bat fatalities at the same cut-in speed. Moreover, limiting the number of times 5 turbines start-up and shut-down may reduce turbine wear-and-tear and power loss, which provide benefits for wind facility operators. Future research across a variety of facilities, turbine types and species should consider comparing differences between a longer decision framework and higher cut-in speeds or combine different decision frameworks with additional weather variables to assess the most cost-effective strategy to reduce bat fatalities at wind turbines.