You are on page 1of 10

Journal of Environmental Management 241 (2019) 397–406

Contents lists available at ScienceDirect

Journal of Environmental Management


journal homepage: www.elsevier.com/locate/jenvman

Research article

An analysis of enhanced tree trimming effectiveness on reducing power T


outages
Jason R. Parent∗, Thomas H. Meyer, John C. Volin, Robert T. Fahey, Chandi Witharana
Department of Natural Resources and the Environment, University of Connecticut, 1376 Storrs Road, U-4087, Storrs, CT, 06269-4087, USA

ARTICLE INFO ABSTRACT

Keywords: We evaluated the effectiveness of an enhanced tree trimming (ETT) program for its ability to reduce tree-related
Tree trimming power outages, and thereby improve resilience, on an electric utility distribution system during storm events.
Vegetation management Evaluations encompassed thirteen years of trimming (i.e. 2005–2017) data and were performed for both
Resilience backbone and lateral utility lines. Backbones included all three phase lines between a substation and a faultable
Power distribution
device whereas all other lines were considered laterals. The study site spanned the entire state of Connecticut,
Outage
where the dominant vegetation is temperate deciduous forest. We controlled for variations in weather, tree
Electric utility
cover, and wire type, by pairing ETT-treated zones with nearby untreated zones. ETT-treated conductors had
storm outage rates that were 0.07–0.36 outages/km/year lower than untreated conductors or 35–180% lower
than the service-area's average annual outage rate for untreated conductors. ETT-treatment was associated with
lower outage rates for “minor” outage types (i.e., blown fuse, tripped recloser, etc.) but the treatment effect was
not statistically significant for “major” outage types (damaged poles or wires). System-wide ETT application, for
the approximately 27,000 km of conductors in the study area, was predicted to reduce annual storm-related
outages by an average of 81–104 and 318–759 outages/year for backbone and lateral lines, respectively. Our
study provided a robust empirical evaluation of ETT and also proposes a geospatial methodology that controls
for variations in weather and environment.

1. Introduction is near the conductors. In many areas, maturing roadside forests receive
little or no silvicultural management which promotes the development
Tree failures contribute to large numbers of power outages during of hazard trees with poor structure and health. Utility tree trimming
storm events in the forested regions of the world (Campbell, 2012). programs aim to prevent tree growth into conductors as well as mitigate
Simpson and Bossuyt (1996) reported that tree failures accounted for risks from hazard trees. In the past, the goal of these programs has
40% of preventable outages in their Massachusetts, USA study area. generally been to improve system reliability which is the ability of the
Eversource Energy outage data for Connecticut, between 2005 and grid to deliver uninterrupted power during relatively normal operating
2017, shows that more than 80% of outages during storms are due to conditions (Vugrin et al., 2017). However, grid resilience has recently
tree failures (Redding, personal comm.). In the northeastern U.S., emerged as a strategic objective for utilities where “resilience” is the
Winter Storm Alfred and Hurricane Irene (2011) caused massive da- ability of the grid to withstand and recover rapidly from disruptions
mages that affected 1.2 million customers, in Connecticut alone, and (Obama, 2013). This new strategic objective has likely been motivated
took 9–12 days to fully restore (McGee et al., 2012). The North by the massive disruptions caused by storms in recent years and pre-
American Ice Storm of 1998 affected 4 million customers in the dictions that extreme weather events may become more common under
northeastern U.S. and Canada and took several weeks to fully restore future climate change scenarios (Meehl et al., 2007).
(Abley, 1998). For each of these storm events, trees were reported as In the northeastern USA, tree trimming programs are generally si-
the leading cause outages. milar among the states (Hansen, 2011) and equivalent programs are
Electric utilities with overhead infrastructure in forested regions likely used in other forested areas throughout the world. These strate-
have extensive vegetation management programs that attempt to re- gies include: standard maintenance trimming (SMT), enhanced tree
duce the frequency of tree-related outages by removing vegetation that trimming (ETT), and hazard tree removal. Although the names and


Corresponding author. Department of Natural Resources and the Environment, University of Connecticut, 1376 Storrs Road, Storrs, CT, 06269-4087, USA.
E-mail addresses: jason.parent@uconn.edu (J.R. Parent), thomas.meyer@uconn.edu (T.H. Meyer), john.volin@uconn.edu (J.C. Volin),
robert.fahey@uconn.edu (R.T. Fahey), chandi.witharana@uconn.edu (C. Witharana).

https://doi.org/10.1016/j.jenvman.2019.04.027
Received 8 November 2018; Received in revised form 8 April 2019; Accepted 8 April 2019
Available online 24 April 2019
0301-4797/ © 2019 Elsevier Ltd. All rights reserved.
J.R. Parent, et al. Journal of Environmental Management 241 (2019) 397–406

specifications of these programs may differ among utilities, they can “Major” storm conditions were defined using IEEE 1366 methodology
generally be described as follows. The SMT occurs on a regular basis, and “non-major” storm conditions included all remaining events that
typically a 4–5 year cycle, and requires the removal of vegetation occurred during “non-blue sky” days. The study concluded that ETT
within a zone around the primary conductors. As an example, Ever- reduced outages by 40–50% during non-major storms and 35–40%
source Energy's SMT specifications require that all vegetation be during major storms. Although Burley (2012) provides compelling
cleared from a zone 15 feet above, 8 feet to the sides, and 10 feet below evidence for positive effect of ETT, the report does not describe the
the primary conductors (Goodson et al., 2008). The ETT specification methodology so it is unclear how or whether the study accounted for
typically requires the removal of all vegetation, from ground to sky geographic variation in weather and forest cover conditions within the
(Goodson et al., 2008), within the “utility protection zone” which, in Eversource Energy Connecticut service area. Furthermore, the evalua-
Connecticut, is defined as the area extending eight feet horizontally tion was based primarily on comparing the performance of backbone
from the outermost electrical conductors and vertically from ground to lines, which had received ETT, to lateral lines which had not received
sky (PURA 2017). Hazard tree removal requires the removal of trees ETT. By Eversource's definition, backbones included all three phase
that are near utility infrastructure and have poor health or structure. lines between a substation and a “faultable” device whereas all other
Across North America, vegetation management is regarded overall lines were considered laterals. Backbone lines are far more critical to
as one of the largest recurring expenses associated with maintaining the power grid than lateral lines so they are constructed to a more ro-
overhead utility infrastructure (Novembri, 2004). Thus, vegetation bust specification and receive priority treatment; thus, it is unclear to
management, for electric utilities, has been the subject of much scien- what degree the differences in performance can be attributed to the ETT
tific research which has largely focused on optimizing the application of treatments.
vegetation management to maximize its benefit on outage reductions as A substantial limitation of many previous studies on tree trimming
well as minimize costs. This research had three main focus areas. The is that they did not account for temporal or geographic variation in
first focus area was to create models using environmental variables to storm events. Guikema et al. (2006) and Radmer et al. (2002) used only
predict locations of increased vegetation risk (Dokic and Kezunovic, “blue sky” outages that occurred during fair weather conditions.
2018, Poulos and Camp 2011; Poulos and Camp 2010) and explore the Simpson and Bossuyt (1996) included storm-related outages but their
use of remote sensing technology to regularly monitor vegetation methods do not appear to have controlled for the differing storm events
conditions (Ahmad et al., 2013). A second focus area was to identify the in the pre- and post-treatment time periods. Louit et al. (2009) assumed
mechanisms by which trees cause outages to help define what should be that weather conditions were similar from one year to the next in their
the priorities of vegetation management in terms of improving relia- study. Dokic and Kezunovic (2018) did account for weather in their
bility (Qi et al., 2013, Appelt and Goodfellow, 2004). Finally, a third model; however, their observed effects of tree trimming were based on
focus area was to develop guidelines and optimization models for simulation results rather than directly on empirical data. In the north-
scheduling and implementing vegetation management (Dokic and east and in many other parts of the U.S., the highly variable occurrence
Kezunovic, 2018; Louit et al., 2009, Nowak and Ballard, 2005; of storms, for a given geographic area, makes it critical to use a
Goodfellow, 2000). methodology that accounts for year-to-year weather variation.
Despite the well-developed literature on vegetation management, Our study adds to the current literature by comparing the effec-
we found relatively few studies that empirically evaluated the specific tiveness of two specific types of tree trimming programs (ETT versus
tree trimming programs employed to reduce power outages on elec- SMT), in the context of system resilience, while controlling for con-
trical distribution systems. This scarcity of studies is likely due to the founding factors such as weather and forest cover characteristics. Our
challenges for researchers in acquiring the necessary data from utilities. specific objective is to quantify any added benefits of ETT over SMT in
Simpson and Bossuyt (1996) found that hazardous tree removal and terms of reducing outage occurrences in an electric distribution system
“storm proof” trimming reduced outage rates by 20–30% for an electric during all types of storm events. Because SMT trimming is applied
utility in Massachusetts. Guikema et al. (2006) found evidence that regularly throughout our study area, our evaluation of ETT is, by ne-
increasing trim frequency results in measureable decreases in outages cessity, relative to SMT. We consider both minor (i.e. blown fuse,
on an electric distribution system in the southeastern U.S.; based on broken limb) and major (i.e. broken pole or damaged conductor) outage
statistical models, they predicted that decreasing the trim cycle from events associated with storms during the years 2005–2017. Rigorous
five years to four years, across their study utility system, would have evaluation of ETT will help inform electric utilities’ decision-making on
resulted in 0.25 fewer outages/circuit/year during their 43 month study directing investments toward the most effective solutions for improving
period. Radmer et al. (2002) used statistical models to predict tree-re- power system resilience. Furthermore, the geospatial-based metho-
lated outages but the models had relatively poor fits, which were at- dology that we used to evaluate ETT, in this study, will allow for
tributed to the limited size of the dataset. Louit et al. (2009) found a controlled empirical evaluations of other vegetation management and
delayed effect of tree trimming with outage rates reaching a minimum 4 resilience programs.
years after the treatment which they speculated was due to the treat-
ment increasing the wind exposure of trees that were sheltered prior to 2. Methods
the treatment. In their study area, which was anonymous, the most cost-
effective interval between trimmings was 7 years if utilities consider We evaluated ETT by comparing outage rates between: 1) backbone
only the costs of the repairs versus costs of trimming. Dokic and conductors, which have received ETT, and lateral conductors that have
Kezunovic (2018) showed that, for their Texas study area, an optimized not received ETT; and 2) lateral conductors that have received ETT and
tree trimming schedule could reduce outage risk by 33% over a 3 month lateral conductors that have not received ETT. From Burley (2012), we
period without increasing costs. assume that all backbone conductors have received ETT in the past and
The costs of implementing ETT is much greater than standard have since been maintained to the ETT standard (Goodson et al., 2008).
trimming and thus widespread implementation of such “ground-to-sky”
specifications needs to be justified by superior performance. We have 2.1. Study area
found no scientific literature that specifically studied the effectiveness
of ETT on reducing outages in an electric utility distribution system; The study area for this analysis consisted of the Eversource Energy
however, Eversource Energy did provide us with their internal perfor- service area in Connecticut which includes roughly 27,000 km of dis-
mance review of ETT (Burley, 2012). Burley (2012) evaluated the effect tribution lines and covers approximately 11,660 km2 (Fig. 1). Con-
of ETT in reducing tree-related storm outages during both major and necticut is the 4th most densely populated state in the USA (based on
non-major storm conditions that occurred between 2008 and 2011. U.S. Census Bureau 2015 estimates) with tree canopy covering nearly

398
J.R. Parent, et al. Journal of Environmental Management 241 (2019) 397–406

0 2,000 km
³ 0 50 km

Fig. 1. The study area included the Eversource Energy service area in Connecticut (gray) which covers an area of approximately 11,660 km2.

Table 1 We used the US National Land Cover Database’s 2011 Tree Canopy
Numbers of outage events during study period (2005–2017). (cartographic) dataset (NLCD 2011) to estimate the average forest
Note that “major” refers to outages with broken poles and cover within the device exposure zones used in our analysis. This data
down wires and “minor” refers to outages with tripped fuses set provides estimates of percent tree cover based on Landsat satellite
or devices or limbs on wires. imagery, within 30 m × 30 m grid cells, across the United States. .
Type of outage Number of Outages

All outages 175,145 2.3. Relative outage rates


Tree-related 107,020
Storm-related 65,561 The number of outage occurrences associated with each device ex-
Major severity 7564
posure zone was summed for each year from 2005 to 2017. Outages
Minor severity 54,250
were only included if they were tree-related, associated with a primary
isolating device, and occurred during the following weather conditions:
65% of the state (Nowak and Greenfield, 2008). Many of the roads and blizzard, hurricane, ice storm, rain, snow, thunderstorm, tornado, and
utility corridors in the state are lined by trees and forests. Connecticut's high winds. Note that we included weather events that do not ne-
forests are temperate deciduous and dominated by oak/hickory cessarily meet the requirements of storm events as defined by IEEE
(Quercus spp. and Carya spp.) and other northern hardwood forest types 1366 or other guidelines. These tree-related storm-weather conditions
(Wharton et al., 2004). The topography is generally hilly with eleva- included 65,561 outage events and accounted for 37% of the total
tions ranging from sea level in the south to nearly 750 m in the outages that occurred during the 2005–2017 time period (Table 1). The
northwest. outage causes in utility databases are often questionable; however,
given that most storm-related outages in our study area are caused by
trees, we expect Eversource's outage database to be sufficiently reliable
2.2. Data for our study.
The substantial variations in year-to-year storm events in the
Eversource Energy provided the following datasets for this study: an northeastern U.S. make it difficult to assess temporal trends in outage
outage database (2005–2017), ETT trimmed areas (2009–2017), and frequency for a given device exposure zone. To account for these var-
isolating device exposure zones. The outage database provided detailed iations, we matched each ETT-treatment zone with one or more nearby
information for each outage including: the date of occurrence, the as- (< 5 km separation) control zones that had not received ETT during the
sociated isolating device, the cause, the weather condition, and the 2009–2017 time period – the close proximity made it likely that the
action taken to repair the outage. The database included outages from a control zone(s) experienced similar weather events to their matching
variety of storm events ranging from small-scale thunderstorms to treatment zone. When we used lateral ETT-treatment zones, we paired
broad-scale storms including Hurricane Sandy (2012), Hurricane Irene each treatment zone with a single control zone. However, due to the
(2011), and Winter Storm Alfred (2011). The ETT trimmed areas were much longer lengths of backbone zones, we paired backbone treatment
represented by polygon shapefiles that allowed identification of con- zones with a group of one or more control zones that had, altogether, a
ductor segments along which ETT was performed in a given year from similar length to the treatment zone. We considered these control zones
2009 to 2017. The device exposure zones, in a polyline shapefile to represent the background outage rates for a given area in the absence
format, indicated the conductor segments that were associated with a of ETT treatment. The following rules were used to match the control
given isolating device; the dataset also included attribute information zones to ETT-treatment zones:
about the zone including the dominant wire type of the conductors and
whether the zone was associated with backbone or lateral conductors. • No history of ETT between 2009 and 2017
All zones were associated with the primary distribution system. We • Nearest zone(s) within 5 km
excluded from analysis the secondary distribution system which di- • Total length of zone(s) was within ± 20% of the treatment zone
rectly connects the primary system to customers via service lines - the length
goals of Eversource's vegetation management programs do not include • Average percent tree cover within ± 20% of the treatment zone
protecting this secondary system. Isolating devices include fuses and percent tree cover
reclosers that open a circuit during an outage event. All outage events • Underground conductors comprise < 25% of the zone
are registered to a specific isolating device which is associated with a • Same wire type as treatment zone (i.e., covered vs. uncovered wire)
specific device exposure zone. Thus, the device exposure zone is the • Not paired with any other treatment zone (if a control for lateral
finest scale at which an outage event location is known; therefore, we ETT-treated zones)
used device exposure zones as the units on which we conducted our
analyses. Zone lengths range from < 0.1 to 25 km with an average These selection rules help ensure that a given treatment zone had
length of 0.5 km. Approximately 50% of zones had lengths < 0.25 km similar environmental and structural characteristics and experienced
which accounted for 10% of the total length of the network. similar weather events as their matching control zone(s). The average

399
J.R. Parent, et al. Journal of Environmental Management 241 (2019) 397–406

percent tree cover for all zones was determined using the NLCD (2011) “major” and “minor” outages separately in our calculations of outage
Tree Canopy dataset. The conductor line segments associated with each rates. We determined outage severity based on the action needed to
zone were buffered by 30 m and the values of the Tree Canopy dataset repair it. An outage was considered “major” if it required either a pole
grid cells, within the buffers, were averaged to give the average percent or conductor to be repaired or replaced; an outage was “minor” if it
tree cover. We ensured that the wire cover type matched between required a fuse to be replaced, a device to be reset, or a limb to be
treatment and control zones because the wire covering is intended to cleared from the conductor.
reduce tree-related outages. ETT-treatment zones were excluded from It is important to note that any observed difference in outage rates
further analysis if a suitable set of control zones could not be found. between treated and untreated zones would pertain only to areas that
Suitable control zone matches were found for approximately 28% of are affected by storm damage in a given period of time. The majority of
backbone zones and 46% of lateral zones that had received ETT. lateral and backbone zones are not affected by storms in a given year.
Outage rates were calculated for each treatment zone as well as for Thus, we roughly extrapolated our findings to all lateral and backbone
each control zone or group of control zones. Rates were calculated in lines in the network by multiplying the estimated impact of the treat-
terms of outages per km of conductor. Relative outage rates were cal- ment by the percent of lines expected to be affected by outages in a
culated as follows: given year. This extrapolation assumes that all backbone or lateral
zones are treated with ETT and, therefore, it estimates the maximum
relative outage rate = treatment zone outage rate potential impact of ETT.
control zone outage rate

As such, positive relative outage rates indicate that the treatment 2.5. Statistical analyses
zone performed worse than the control zone and negative rates indicate
that the treatment zone performed better than the control zone. We used ArcGIS Desktop 10.5 and Python 2.7 with the SciPy module
Treatment zones were excluded from analysis if they had lengths less to conduct all geospatial and statistical analyses in this study. We in-
than 250 m. The reduced exposure of these shorter zones made them cluded both parametric and non-parametric statistical analyses to avoid
more likely to have no outages before or after treatment. In addition, any potential subjectivity in evaluating the normality of the data.
these short zones would have contributed to an increase in large out- For the pairs of treated and untreated lateral lines, we used two
liers in the analysis. sample t-tests and Wilcoxon rank-sum tests to compare differences in
means and medians, respectively, between the pre- and post-treatment
2.4. Calculating outage rates relative outage rates. Significant changes in the mean relative outage
rates between the pre-treatment and post-treatment periods can be at-
Outage rates for the ETT-treated lateral zones and their control tributed to the ETT treatment effect.
zones were calculated for the 3-year periods before and after ETT was For the pairs of treated backbone and untreated lateral line groups,
applied to the treatment zone; outages occurring during the year of ETT we used 1 sample t-tests and Wilcoxon signed-rank tests to determine if
treatment were excluded because it was not possible to tell when the the mean and median relative outage rates, respectively, were sig-
treatment occurred relative to the outages for that year. Outage rates nificantly different from zero. Relative outage rates that are sig-
for the pre- and post-treatment periods were calculated as: nificantly different from zero may be due to an effect of treatment al-
though other factors could be involved. A more robust evaluation based
outage count in 3 year period on comparisons between pre- and post-treatment time periods was not
outage rate (lateral analysis ) =
length of zone (km ) possible, in this case, because no pre-treatment outage data were
available for the backbone lines.
The pre- and post-treatment comparisons were used to check for any
biases in the ETT-treated zones toward higher or lower outage rates in
comparison to background outage rates. Differences found between the 3. Results
pre- and post-treatment outage rates can be attributed to the ETT
treatment. Eversource adopted more aggressive ETT specifications in The total length of lines trimmed with ETT, each year, ranged from
2011, for lateral conductors, which required all vegetation to be re- 273 km to 1084 km with annual trimming increasing sharply in 2012
moved from ground-to-sky within 8 lateral feet of the conductors; (Fig. 2). More than 175,000 outages occurred between 2005 and 2017
previously, the specification only required vegetation to be removed with tree failures accounting for slightly more than 107,000 outages.
within 8 lateral feet to a height 20 feet above the conductors (Goodson Approximately 65,500 tree-related outages occurred during “storm”
et al., 2008). Thus, for ETT-treated lateral zones, we considered ETT conditions and 7564 of those outages were considered “major” damage,
treatments prior to 2012 separately from treatments applied during or based on our definition (Table 1).
after 2012. Relative failure rates were excluded from analysis, for either
the pre- and post-treatment periods, if both the treatment and control 1200
zones had zero outages during that time period because these data
Length of trimmed lines (km)

points provided no information on the effect of ETT.


We did not have outage data corresponding to pre-treatment time
800
periods for the backbone zones; thus, we considered only the post-
treatment period in the part of our analysis in which backbone zones
were used to represent the ETT-treatment. Outage rates for the back-
bone zones and their corresponding lateral control zones were calcu- 400
lated as:
outage count in 2005 2017 period
outage rate (backbone analysis ) = 0
total length of zone (s ) (km)
2009 2010 2011 2012 2013 2014 2015 2016
Outages are not equal in terms of the time and resources needed to Year
repair the outage. Outages involving broken poles and downed con-
ductors are much more costly and hazardous than outages that occur
without major physical damage to the system. Thus, we consider Fig. 2. Total lengths of conductors trimmed with ETT between 2009 and 2016.

400
J.R. Parent, et al. Journal of Environmental Management 241 (2019) 397–406

Table 2
Statewide tree-related outage rates, during storm conditions, by zone type (i.e. backbone, untreated lateral, ETT-treated lateral). Note that
outage rates are not controlled for variations in weather, tree cover, wire type, etc. Includes all outage severities (i.e. major and minor).
Zone Total length Total outages Outages/km/12 yrs Outages/km/year
(km) (2005–2017) (2005–2017)

Backbone 7615 6323 0.8 0.06


Untreated lateral 12,002 30,805 2.6 0.20
Treated lateral (2009–2011) 1080 3639 3.4 0.26
Treated lateral (2012–2014) 2842 5504 1.9 0.15

Table 3 rate of 0.15 outages/km/year and lateral lines that had never been
Statistics for lateral and backbone zones affected by at least one outage during treated with ETT had outage rates of 0.20 outages/km/year. Note that
the 2005–2017 study period; all outage types are included. these outage rates do not control for variation in weather, forest cover,
or wire type - we present them here only to provide a context for
Total Total length Percent of total length affected by outages
number (km) per year evaluating the magnitude of differences in outages rates that were ob-
served between ETT-treatment and control zones.
Mean Min Max The study area included 46,334 lateral zones and 3147 backbone
zones with total lateral and backbone zone lengths of 18,477 km and
laterals 46,334 18,478 13.3 6.9 33.7
backbones 3147 8127 14.3 6.6 49.6 8127 km, respectively (Table 3). During the 2005–2017 time period, the
mean annual percentage of total line length affected by at least one outage
was 13.3% and 14.3% for the lateral and backbone zones, respectively.
System-wide, the backbone lines had the lowest overall outage rates
during the 2005–2017 study period with 0.06 outages/km/year 3.1. Lateral zones
(Table 2). Lateral lines that had been treated with ETT between 2009
and 2011 had the highest overall outage rate during the 2005–2017 The ETT-treatment resulted in significantly lower outage rates in
study period with 0.26 outages/km/year. During the study period, lateral lines for both the 2009–2011 and 2012–2014 time periods, when
lateral lines treated with ETT between 2012 and 2014 had an outage considering all outage types (Table 4, Figs. 3 and 4). For 2009–2011,

Table 4
Comparison of mean and median relative outage rates (outages/km/3years) for pre- and post-treatment analysis of ETT-treated lateral zones. Negative rates indicate
better performance in ETT-treated zones.
Time period Mean (95% CI) Median (95% CI) n p (t-test) p (Wilcox)

2009–2011 (all outages) Pre 0.70 (0.48–0.93) 0.81 (0.64–0.96) 537 < 0.001 < 0.001
Post 0.06 (−0.14–0.27) 0.04 (−0.13–0.18) 655
2012–2014 (all outages) Pre 0.20 (0.08–0.33) 0.06 (0.00–0.19) 1660 < 0.001 < 0.001
Post −0.44 (−0.59–−0.30) −0.76 (−0.83–−0.58) 580
2012–2014 (major outages) Pre 0.03 (−0.17–0.23) −0.03 (−0.52–0.53) 204 0.138 0.107
Post −0.33 (−0.76–0.10) −0.75 (−0.85–−0.33) 46
2012–2014 (minor outages) Pre 0.21 (0.09–0.33) 0.07 (0.00–0.21) 1623 < 0.001 < 0.001
Post −0.46 (−0.61–−0.31) −0.75 (−0.82–−0.59) 528

Fig. 3. Box-plots showing pre- and post-treatment


relative failure rates (outages/km/3years) for ETT-
treated and untreated lateral zones. Subsets include
2009–2011 with all outage types and 2012–2014
with all, major, and minor outage types. Negative
values indicate better performance in the treatment
zone. Stars indicate mean relative outage rates.

401
J.R. Parent, et al. Journal of Environmental Management 241 (2019) 397–406

Fig. 4. Histograms showing pre- and post-treatment relative failure rates (outages/km/3years) for ETT-treated and untreated lateral zones. Subsets include
2009–2011 with all outage types and 2012–2014 with all, major, and minor outage types. Negative values indicate better performance in the treatment zone.

outages/km/year which is 35–180% of the mean annual system-wide


100
outage rate of untreated lateral lines (0.2 outages/km/year) during the
2005–2017 period. The change in mean relative outage rates, for the
75 63 67 64 2012–2014 trim period, was 0.13–0.31 outages/km/year (Table 5)
which is 65–155% of the mean annual system-wide outage rate of un-
Percent

49 48 50 48
50 treated lateral lines. Based on the 2012–2014 trim period, the overall
37
outage reduction extrapolated across all laterals is 0.017–0.041
25 outages/km/year (Table 5) which equates to an average of 318–759
fewer outages per year (Table 5).
The ETT-treatment effect, in the 2012–2014 time period, was pre-
0 sent for both major and minor outage types (Figs. 6 and 7). For major
2009-2011 2012-2014 2012-2014 2012-2014 outage types (i.e. damaged poles or wires), the mean and median re-
(all) (all) (major) (minor) lative failure rates dropped from 0.03 to −0.33 and −0.03 to −0.75
outages/km/3years, respectively; however, these changes were not
Pre Post statistically significant (p > 0.10). The treatment zones performed
better than the control zones 50% and 67% of the time in the pre- and
Fig. 5. Percent of occurrences in the pre- and post-treatment time period when post-treatment time periods, respectively (Fig. 4). For minor outage
ETT-treated lateral zones had better performance than the untreated control types (i.e. no physical damage to system), the mean and median relative
zones. failure rates dropped significantly (p < 0.001) from 0.21 to −0.46 and
0.07 to −0.75 outages/km/3years, respectively. The treatment zones
the mean and median relative outage rates dropped from 0.70 to 0.06 performed better than the control zones 48% and 64% of the time in the
and 0.81 to 0.04 outages/km/3years, respectively, between the 3 year pre- and post-treatment time periods, respectively (Fig. 4).
pre- and post-treatment periods; the improved performance for ETT-
treated zones was statistically significant (p < 0.001). The treatment 3.2. Backbone zones
zones performed better than the control zones 37% and 49% of the time
in the pre- and post-treatment time periods, respectively (Fig. 5). For Backbone zones performed significantly better than untreated lat-
2012–2014, the mean and median relative outage rates also dropped eral control zones for all outage types (Table 6; Figs. 6 and 7). For the
significantly (p < 0.001) from 0.20 to −0.44 and 0.06 to −0.76 2005–2017 period, the mean relative outage rates were −0.99, −0.10,
outages/km/3years, respectively. The treatment zones performed and −0.94 outages/km/12years for all, major, and minor outage types,
better than the control zones 48% and 63% of the time in the pre- and respectively. Median relative outage rates were −0.97, −0.23, and
post-treatment time periods, respectively (Fig. 5). The change in mean −0.94 outages/km/12years for all, major, and minor outage types,
relative outage rates, for the 2009–2011 trim period, was 0.07–0.36 respectively. Relative outage rates were significantly less than zero

402
J.R. Parent, et al. Journal of Environmental Management 241 (2019) 397–406

Table 5
System-wide extrapolation of outage reductions (based on all outage types for the 2012–2014 sample).
Mean sample outage rate reduction System-wide outage rate reduction System-wide reduction in outages
(outages/km/year) (outages/km/year)

Mean Min Max Mean Min Max

laterals 0.13–0.31 0.017–0.041 0.009–0.021 0.044–0.104 318–759 165–393 809–1930


backbones 0.07–0.09 0.010–0.013 0.005–0.006 0.035–0.045 81–104 38–48 282–363

untreated lateral lines (0.2 outages/km/year) during the 2005–2017


period. The mean overall outage reduction extrapolated across all
backbones was 0.010–0.013 outages/km/year which equates to an
average of 81–104 fewer outages per year (Table 5).

4. Discussion

Our study demonstrated empirically that the ETT treatment has an


observable effect on reducing power outage rates during stormy
weather conditions. To our knowledge, this is the first study in the
scientific literature to provide an evaluation of ETT, and, more gen-
erally evaluate tree trimming for storm-related outages with empirical
data. The ETT treatment significantly reduced relative outage rates such
that treated zones had fewer outages, on average, than untreated con-
trol zones. Overall, the ETT treatment significantly reduced outage
rates, in both lateral and backbone zones as compared to untreated
control zones. It should be noted that although these “untreated” con-
trol zones had not received ETT, they had received SMT multiple times
during the study period; thus our estimates of these reductions in
outage frequency are in addition to the effects of the SMT program.
Fig. 6. Box-plots showing post-treatment relative failure rates (outages/km/ The Eversource Energy Connecticut service area is in the eastern
12years) for backbone and untreated lateral zones. Subsets include 2005–2017 temperate forest region which covers nearly all of the eastern half of the
with all, major, and minor outage types. Negative values indicate better per-
United States (EPA 2006). The forests in Connecticut have similar
formance in the treatment zone. Stars indicate mean relative outage rates.
species compositions and land use histories as the forests throughout
much of the eastern region of the U.S. and the entire region is suscep-
(p < 0.005) in all cases – indicating that backbone zones had better tible to damaging weather events. Therefore, it is reasonable to expect
performance than the control zones. Backbone zones performed better that ETT could have similar positive impacts on reducing tree-related
than control zones 74%, 63%, and 78% of the time for all, major, and outages throughout the region for utilities with overhead infrastructure.
minor outage types, respectively (Fig. 8). The mean relative outage The transferability of our study results to regions with very different
rates, for all outage types, was 0.07–0.09 outages/km/year (Table 5) forest compositions and histories is more speculative and empirical
which was 35–45% of the mean annual system-wide outage rate of assessments focused on these regions are warranted.

Fig. 7. Histogram showing post-treatment relative failure rates (outages/km/12years) for backbone and untreated lateral zones. Subsets include 2005–2017 with all,
major, and minor outage types. Negative values indicate better performance in the treatment zone.

403
J.R. Parent, et al. Journal of Environmental Management 241 (2019) 397–406

Table 6
Comparison of mean and median relative outage rates (outages/km/12years) for backbone zones compared with untreated lateral zones. Negative rates indicate
better performance in backbone zones.
Time period Mean (95% CI) Median (95% CI) n p (t-test) p (Wilcox)

2005–2017 (all outages) −0.99 (−1.12–−0.85) −0.97 (−1.18–−0.85) 660 0.000 0.000
2005–2017 (major outages) −0.10 (−0.16–−0.03) −0.23 (−0.26–−0.16) 425 0.004 0.000
2005–2017 (minor outages) −0.94 (−1.06–−0.82) −0.94 (−1.07–−0.78) 650 0.000 0.000

when a dataset with a sufficient time span becomes available.


100
Our estimates of system-wide impacts on outage reductions are
78 much greater for lateral lines than for backbone lines even though ETT-
74
75 treated laterals and backbones had similarly better performance than
63 untreated laterals. This discrepancy can be attributed to the fact that
Percent

lateral lines dominate the system in terms of total length. In our study
50 area, lateral lines comprise nearly 70% of the total length of lines in the
utility network, therefore, a reduction in outage rates for laterals will
have a greater overall impact than a similar reduction in outage rates
25
for backbones. However, it is important to note that outages occurring
on backbone lines typically have much greater impact, in terms of
0 customers affected, than outages occurring on lateral lines; so in terms
of overall impact, a smaller decrease in total outages for the backbone
all major minor may be more valuable than a substantially larger decrease in outages
for the laterals.
Fig. 8. Percent of occurrences when backbone zones had better performance Eversource's change to a more aggressive “ground-to-sky” ETT
than the untreated lateral control zones for all, major, and minor outage se- specification, in late 2011, had no discernible effect on the performance
verities. of the treated lateral zones, because the magnitude of the treatment
effect was similar in both the 2009–2011 and 2012–2014 time periods.
Our study found a 35–45% reduction in annual storm-related outage For lateral zones treated with ETT in the 2009–2011 period, the treated
rates for backbone lines, in storm-damaged areas, when compared to zones tended to have poorer performance than the control zones in the
untreated laterals lines; this result is consistent with Eversource's in- pre-treatment period; however, in the post-treatment period, ETT ap-
ternal performance review which attributed a 35–40% reduction in pears to have improved performance in these zones such that they had
outage rates to ETT for backbone lines during major storms (Burley, similar outage rates to the control zones. For lateral zones trimmed in
2012). These reductions are also similar those found by Simpson and the 2012–2014 period, the pre-treatment performances tended to be
Bossuyt (1996) with regard to the impact of hazard tree removal and similar between treatment and control zones; and ETT-treatment re-
storm-proof trimming programs in their study area. Guikema et al. sulted in improved performance in the post-treatment period when
(2006) predicted a 0.005 outages/km/year reduction in outage fre- compared to control zones. Eversource initiated the ETT program for
quency, across the entire utility network in their study, in response to the lateral lines in 2009 so the poor pre-treatment performance of zones
decreasing time between trim cycles from five years to four years. Al- trimmed during the 2009–2011 period may be due to Eversource
though similar in magnitude, our study found substantially larger re- prioritizing zones for ETT based on outage history to ensure that the
ductions in outage rates for ETT-treatment on both lateral and back- most failure prone zones received treatment first. By 2012, many of the
bone zones (0.010–0.041 outages/km/year). Although Guikema et al. most problematic zones may have received treatment; thus, the pre-
(2006) does not explicitly state the type of trimming that was eval- treatment failure rates for treated zones were more similar to the
uated, their description seems to correspond to our definition of SMT. control zones. In addition, the major storms in 2011 caused massive
Thus, the larger reduction in outages that we found for ETT might vegetation removals in untreated zones. It seems likely that these
suggest that the application of ETT has a greater benefit than increasing storms removed many of the trees that would have failed during future
the frequency of SMT by one year. However, the differences may also be storms and thus helped to equalize outage rates across all zones.
due, at least in part, to the methods used to extrapolate outage rates The positive performance of ETT during inclement weather is par-
across the entire system and to differences in the study areas. Dokic and ticularly beneficial because it reduces outage frequencies during storms.
Kezunovic (2018) predicted a 33% decrease in outage risk by adopting The effect of ETT was evident despite the occurrence of three major
a more targeted tree trimming program that prioritizes high risk areas. storms (Hurricane Irene, Winter Storm Alfred, Hurricane Sandy) during
Their predictions are similar to what we observed for ETT although it is the study period (in 2011 and 2012) which inflicted massive damages
unclear whether the tree trimming program in their study was to utility infrastructure in Connecticut and throughout the region. The
equivalent to ETT or SMT. high frequencies of outages during storms can affect large numbers of
Louit et al. (2009) observed a declining trend in outages during the customers as well as result in longer post-storm recovery times; thus,
2–3 years following the tree trimming treatment which was presumably storm-related outages tend to have more severe negative impacts on
due to the forest edges becoming acclimated to the treatment dis- public well-being than outages during blue-sky conditions (Campbell,
turbance. Despite focusing on only the 3 years immediately following 2012). A reduction in outage frequencies would help improve post-
ETT treatment, our study showed a significant effect of ETT on reducing storm recovery time thus increasing overall system resiliency.
outages. Thus, it is possible that the effect of ETT would have be larger Intuitively, we expected that ETT treatment would be most effective
if we had excluded the initial post-treatment acclimation period. It also at reducing outages of minor severity rather than for outages of major
seems possible that outage rates during the acclimation period could be severity. Failure of overhead branches, which are completely removed
temporarily elevated above the pre-treatment rates as a result of ex- by ETT, seems more likely to cause isolating devices to trip without any
posing previously sheltered trees to edge environments. It would be actual physical damage to the conductors or poles. On the other hand,
worthwhile to compare the short-term and longer-term impacts of ETT ETT may do little to prevent whole tree failures if many are from hazard

404
J.R. Parent, et al. Journal of Environmental Management 241 (2019) 397–406

trees outside of the utility protection zone – whole tree failures would treatment and that these same zones would also receive high priorities
be more likely to result in damage to poles and conductors. We did not during restoration events. Thus, we might expect ETT-treated zones to
find a statistically significant effect of ETT treatment on reducing major have shorter duration outages simply because they are higher priority
outage types for lateral lines because of the large confidence intervals zones. In our study, we do not attempt to quantify or model the impact
around the post-treatment mean and median relative outage rates of ETT on standardized indices because we felt such attempts could be
which may have been due to the small sample size (n = 42). Although misleading and complicate the generalization of ETT's impacts.
insufficient to affect the mean or median, the treatment did seem to
increase the frequency with which treated lateral zones performed 5. Conclusions
better than untreated zones – the magnitude of this change was similar
to that observed for minor outages. In addition, backbone zones did Our study evaluated the effectiveness of enhanced tree trimming
have significantly fewer major outages than untreated lateral zones (ETT) on reducing outage rates during storm weather conditions in the
which may be due, in part, to the ETT treatment. However, the lower temperate deciduous forests of Connecticut, USA. Our analyses in-
rate of major outages on the backbones may also be due to their more cluded both lateral and backbone type zones and controlled for
robust construction when compared to lateral lines. The statistically weather, tree cover, and wire type. We found significant effects of ETT
inconclusive findings for ETT's effect on major outage types also sug- in reducing rates of minor outage types (i.e., blown fuse, tripped re-
gests that hazard trees outside of the utility protection zone, which closer, etc.). These effects were not statistically significant for major
cannot be mitigated by ETT, may cause much of the physical damage to outage types (broken poles or down wires) for lateral conductors, which
the grid. may be due, in part, to the relatively small sample sizes of these events.
Backbone zones performed significantly better than untreated lat- On the other hand, treatment did improve the frequency with which
eral zones over the 2005–2017 period regardless of outage type; ETT treated zones out-performed untreated zones for both major and minor
treatment in the early 2000s likely accounts for much of this improved outage types. Backbone conductors, most of which received ETT prior
performance. However, while our analysis controlled for weather, to the study period, experienced fewer major and minor type outages
forest cover, and wire type, there are likely differences between back- than untreated lateral lines. The better performance of backbones,
bone zones and lateral zones for which we could not account. Backbone when compared to laterals, may also include structural differences in
zones are the most important components of any utility circuit and combination with ETT. Our study provided a rigorous evaluation of ETT
consequently receive preferential treatment for hazard tree removal as well as proposed a geospatial-based method that controls for weather
and resilience improvements, which could account for some of the and other environmental conditions in evaluations of utility vegetation
improved performance when compared with lateral zones. The trees management and resilience programs.
and forest edges along backbone zones have been maintained to the
ETT specification for 15 or more years; thus, it is possible that the Conflict of interest
trimmed trees and forest edges have adapted (i.e. become more wind
firm) to repeated trimming over time. When compared to laterals, This project was funded by Eversource Energy CT.
backbone zones tend to have taller and larger diameter poles which
carry multiple conductors, also with larger diameters, which may be Acknowledgements
more resistant to the major outage types as defined in this study.
Finally, the recloser devices used on backbone zones can reduce the This research was funded by Eversource Energy Connecticut
occurrence of minor outages by allowing the circuit to close after an through the Eversource Energy Center at the University of Connecticut,
initial temporary power disruption such as when a branch falls across USA. All utility data for this research were provided by Eversource
the conductors and subsequently falls to the ground. During the study Energy Connecticut. The authors are grateful for the data and support of
period, Eversource had not yet been using “fuse-saving” devices on Eversource Energy Connecticut and the Eversource Energy Center. We
lateral lines which can help avoid outages from temporary power dis- are also grateful to the reviewers for their helpful comments.
ruptions.
The geospatial-based methods, developed in this study, can provide References
robust evaluations of utility and other systems that are strongly influ-
enced by weather and other environmental conditions, such as tree Abley, M., 1998. The Ice Storm: an Historic Record in Photographs of January 1998. M&S,
cover. Locations that are in close proximity are likely to experience the Toronto, pp. 192 ISBN 0-7710-6100-5.
Ahmad, J., Malik, A.S., Xia, L., Ashikin, N., 2013. Vegetation Encroachment monitoring
same weather events during a given period of time; thus pairing for transmission lines right-of-ways: a survey. Electr. Power Syst. Res. 95, 339–352.
treatment zones with nearby control zones helps control for year-to- https://doi.org/10.1016/j.epsr.2012.07.015.
year weather variations. Similarly, it is important to match treatment Appelt, P.J., Goodfellow, J.W., 2004. Research on how trees cause interruptions – ap-
plications to vegetation management. In: IEEE Rural Electric Power Conference.
and control zones with similar tree cover and exposure (i.e. length) for Scottsdale, Arizona. https://doi.org/10.1109/REPCON.2004.1307064. 2004.
evaluations of tree trimming since these factors can greatly affect Burley, W., 2012. Distribution Circuit Component Reliability Performance Review.
outages regardless of the vegetation management history. Northeast Utilities, pp. 4.
Campbell, R.J., 2012. Weather-related power outages and electric system resilience.
Standardized resilience indices do not yet exist and procedures for
Congressional Research Service Available at. http://docs.jalite-group.com/data/
evaluating resiliency tend to be developed on a case-by-case basis to docs/pdf/R42696.pdf, Accessed date: 25 January 2019.
account for the specific objectives of the utility (Vugrin et al., 2017). Dokic, T., Kezunovic, M., 2018. Predictive risk management for dynamic trimming
schedules for distribution networks. IEEE Transactions on Smart Grid. https://doi.
Standardized reliability indices such as the System Average Interrup-
org/10.1109/TSG.2018.2868457. In press.
tion Duration Index (SAIDI), the System Average Interruption Fre- Environmental Protection Agency (EPA), United States, 2006. Ecoregions of North
quency Index (SAIFI), and the Customer Average Interruption Fre- America. Available online at: ftp://newftp.epa.gov/EPADataCommons/ORD/
quency Index (CAIFI) are routinely used to evaluate grid reliability. Ecoregions/cec_na/NA_LEVEL_I.pdf, Accessed date: 28 January 2019.
Goodfellow, J., 2000. Applying reliability centered maintenance (RCM) to overhead
These indices are typically calculated as annual averages that exclude electric utility distribution systems. In: IEEE Power Engineering Society Summer
storm-related outages and usually are not sufficient to provide useful Meeting, https://doi.org/10.1109/PESS.2000.867647.
information on system resilience (Vugrin et al., 2017). Furthermore, Goodson, D.A., Carey, A.C., Cavanagh, J.B., Layton, C.W., Redding, S., 2008. Northeast
Utilities Specifications for Local Distribution Line Clearance Tree Work and Brush
reliability indices are affected by the crews and resources available for Control. Northeast Utilities, pp. 12.
restoration, the distribution of customers, restoration priorities, and Guikema et al. 2006Guikema, S.D., Davidson, R. a., Liu, H., 2006. Statistical models of the
other factors that may be pertain to a particular event or utility. For effects of tree trimming on power system outages. IEEE Trans. Power Deliv. 21 (3),
1549–1557. http://doi.org/10.1109/TPWRD.2005.860238.
example, we would expect that high priority zones are targeted for ETT

405
J.R. Parent, et al. Journal of Environmental Management 241 (2019) 397–406

Hansen, L.R., 2011. Utility Tree Trimming in Other States. OLR Research Report 2011-R- CT, Accessed date: 2 July 2018.
0459. Accessed online at. https://www.cga.ct.gov/2011/rpt/2011-R-0459.htm, Obama, B., 2013. Presidential Policy Directive 21: Critical Infrastructure Security and
Accessed date: 19 February 2019. Resilience. White House Press Release Retrieved from. https://obamawhitehouse.
Louit, D., Pascual, R., Banjevic, D., 2009. Optimal interval for major maintenance actions archives.gov/the-press-office/2013/02/12/presidential-policy-directive-critical-
in electricity distribution networks. Int. J. Electr. Power Energy Syst. 31 (7–8), infrastructure-security-and-resil.
396–401. https://doi.org/10.1016/j.ijepes.2009.03.021. Poulos, H.M., Camp, A.E., 2010. Decision support for mitigating the risk of tree induced
McGee, J., Skiff, J., Carozza, P., Edelstein, T., Hoffman, L., Jackson, S., McGrath, R., transmission line failure in utility rights-of-way. Environ. Manag. 45 (2), 217–226.
Osten, C., 2012. Report of the Two Storm Panel: Presented to Governor Dannel P. Poulos, H.M., Camp, A.E., 2011. Mapping threats to power line corridors for Connecticut
Malloy. Accessed online, at. http://www.governor.ct.gov/malloy/lib/malloy/two_ rights-of-way management. Environ. Manag. 47 (2), 230–238.
storm_panel_final_report.pdf, Accessed date: 18 January 2018. Public Utility Regulatory Authority (PURA), 2017. PURA Regulations (Title 16 – Public
Meehl, G.A., Stocker, T.F., Collins, W.D., Friedlingstein, P., Gaye, A.T., Gregory, J.M., Service Companies). Connecticut Department of Energy and Environmental
Kitoh, A., Knutti, R., Murphy, J.M., Noda, A., Raper, S.C.B., Watterson, I.G., Weaver, Protection Ch. 238, sec. 16-234. Available online at. https://www.cga.ct.gov/
A.J., Zhao, Z.-C., 2007. Global climate projections. In: Solomon, S., Qin, D., Manning, current/pub/chap_283.htm#sec_16-234, Accessed date: 24 January 2018.
M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L. (Eds.), Climate Qi, Junjian, Mei, Shengwei, Liu, Feng, 2013. Blackout model considering slow process.
Change 2007: the Physical Science Basis. Contribution of Working Group I to the IEEE Trans. Power Syst. 28 (3), 3274–3282.
Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Radmer, D.T., Kuntz, P.A., Christie, R.D., Venkata, S.S., Fletcher, R.H., 2002. Predicting
Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. vegetation-related failure rates for overhead distribution feeders. IEEE Trans. Power
National Land Cover Dataset (NLCD), 2011. U.S. Forest service tree canopy (cartographic) Deliv. 17 (4), 1170–1175 Oct. 2002.
dataset. In: Multi-Resolution Characteristics Consortium, Available at. https://www. Simpson and Bossuyt. 1996Simpson, P., Bossuyt, R. Van, 1996. Tree-caused electric
mrlc.gov/nlcd11_data.php, Accessed date: 22 December 2017. outages. J. Arboric. 22 (3), 117–121. Retrieved from. http://www.utilityarborist.
Novembri, R., 2004. Utility Vegetation Management Final Report. CN Utility Consulting, org/images/Articles/Tree Caused.pdf.
LLC, pp. 62. at. https://www.ferc.gov/industries/electric/indus-act/reliability/ Vugrin, E., Castillo, A., Silva-Monroy, C., 2017. Resilience Metrics for the Electric Power
blackout/uvm-final-report.pdf, Accessed date: 18 January 2018. System: A Performance-Based Approach. Sandia National Laboratories Report.
Nowak, C.A., Ballard, B.D., 2005. A framework for applying integrated vegetation man- SAND2017-1493. pp. 49. Retrieved from. https://prod.sandia.gov/techlib-noauth/
agement on rights-of-way. J. Arboric. 31 (1), 28–37. access-control.cgi/2017/171493.pdf.
Nowak, D.J., Greenfield, E.J., 2008. Urban and Community Forests of New England. Wharton, E.H., Widmann, R.H., Alerich, C.L., Barnett, C.J., Lister, A.J., Lister, T.W.,
United States Department of Agriculture Forest Service. General Technical Report Smith, D., Borman, F., 2004. The Forest of Connecticut. USDA Forest Service
NRS-28. Available online at. https://www.nrs.fs.fed.us/data/urban/state/?state= Resource Bulletin NE-160.

406

You might also like