Journal of the Illuminating Engineering Society

ISSN: 0099-4480 (Print) (Online) Journal homepage: http://www.tandfonline.com/loi/uzie20

Occupancy and Time-Based Lighting Controls in

Open Offices

Judith Jennings, Nesrin Colak & Francis Rubinstein

To cite this article: Judith Jennings, Nesrin Colak & Francis Rubinstein (2002) Occupancy and
Time-Based Lighting Controls in Open Offices, Journal of the Illuminating Engineering Society,
31:2, 86-100, DOI: 10.1080/00994480.2002.10748395

To link to this article: http://dx.doi.org/10.1080/00994480.2002.10748395

Published online: 16 Sep 2013.

Submit your article to this journal

Article views: 25

View related articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=uzie20

Download by: [University of Tennessee, Knoxville] Date: 07 June 2016, At: 11:20
 86

Occupancy and Time-Based Lighting Controls in

O p e n Offices
Judith Jennings, Nesrin Colak, and Francis Rubinstein

Lighting controls are making their way into increasing On weekdays, it is the custodians' job to turn out the
numbers of U. S. buildings, but the potential for appli­ lights when they finish their rounds. On weekends, the
cation of these technologies is still far from being fully custodians do not clean the offices and therefore do not
achieved. A key element for decision-makers in choosing turn off the lights. The weekend analysis compares the
lighting control technologies is evidence that these tech­ fraction of lights-off events due to the schedule versus
nologies will be effective in reducing rising energy costs. the wall switch. In this building, the lights are turned on
This long-term study adds to the small but growing body only rarely on weekends, either by employees putting in
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

of literature describing the potential savings from occu­ overtime or when work is being carried out by the build­
pant sensors and time-scheduling controls in large open ing crew or contractors after hours. Figure 1 shows the
office areas. layout of the entire testbed. The un-shaded areas are the
Occupancy sensing has been shown to be an effective subject of this paper.
means of reducing lighting energy use in private
offices. ' ' Time scheduling can also save significant Culling the data
energy in similar large spaces.4'5 In the lighting testbed For the weekday analysis, only days with more than
on the General Services Administration floors of the eight hours of occupied time were considered. This elim­
Phillip Burton Federal Building in San Francisco, we inated only a couple of days in two or three zones, most­
examined lighting energy use in large open spaces with ly due to faulty data collection. Holidays were also omit­
either occupant sensors or time scheduling controls ver­ ted. On weekends, there are no normal working hours,
sus wall switching alone. We discuss the operation of so the 8-hour criterion was not applied.
both control systems in the context of one working office We had an unusually large number of problems in this
building, and describe the impacts of user behavior on testbed due to inexperienced installers and complex
the potential for energy savings. data collection needs. In addition to inconsistently wired
switches, occasionally a problem occurred in a lighting
Description of the site control panel that necessitated bypassing the control
Floors 3, 4 and 5 of the Phillip Burton Federal functions until the problem was solved. In bypass mode,
Building are devoted to the advanced lighting testbed. the lights are forced on and remain on until the problem
The testbed encompasses many different technologies in is corrected and the bypass is removed. We omitted such
three types of office space. This work describes the time- days from the analysis, both for weekdays and weekends.
scheduled switching and occupant-sensing technologies The final average count of weekdays included in the
implemented in non-private open office areas in the analysis in floor 3 was 90 in 1998 and 217 in 1999. For
third and fifth floors of the testbed. weekends on floor 3, an average of 37 days were includ­
Overhead lighting consists of 2 ft by 4 ft 3-lamp T8 fluo­ ed in 1998 and 1996 in 1999. For floor 4, the analysis
rescent parabolic troffers on an 8 ft grid over the entire included an average of 186 weekdays and 85 weekend
ceiling, with a few 2 ft by 2 ft 2-lamp fixtures. Under-shelf days. On floor 5 the data for 1998 and 1999 were ana­
task lighting is available in most of the cubicles. After re- lyzed together, with an average of 317 weekdays and 129
lamping and initial lamp burn-in at the start of the testbed, weekend days. The data sets for both years contain data
the average full-light illuminance in open areas was about from each season, though more data were missing or
700 Ix, with a minimum of 231 and a maximum of 992. The unusable in 1998.
wide range reflects the lack of coordination between furni­ Several entire areas were dropped from the data set
ture and lighting layout. Before the installation of controls, for two reasons. Some had remained unoccupied during
bi-level switching was in use throughout the building, one major renovation for long periods, and in a few others
switch operating the inboard lamp in each 3-lamp fixture the sensors malfunctioned and did not turn out the
and the other operating the outboard lamps. lights. In others, incorrect wiring caused data to be unus­
able. This left us with valid data on the third floor for all
Author's affiliation: Building Technologies Program, Environmental the large zones but only 12 of the original small zones,
Energy Technologies Division, Ernest Orlando Lawrence Berkeley
and on the fifth floor for two of the original six large
National Laboratory, University of California, Berkeley, CA
zones and three of the original four small zones. On
Summer 2002 JOURNAL of the Illuminating Engineering Society
 87
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

IrNTERO* - NOT SHCtUDEO IN M S RtfER

SAN FRANCISCO FEDERAL BUILDING JOPHN &SYUT - NOT INCLUDED !N THtS fftPER
GO.DEN GATE CA
NOIINCtUOEOWTESfBEO

CONFERENCE ROOM - NOT INCIUDED IH1SST6ED

Figure 1—Floor plan of the test-bed showing large open areas.

floor 4, fewer changes occurred and we only lost one a low-voltage relay rather than directly controlling a
small zone that turned out to be a corridor. switch leg. The new electronic switches allowed all wall
switched "on" and "off events to be recorded.
Methods On the fifth floor, occupant sensors were installed in
The testbed was subdivided into zones, each with a each zone. All used ultrasonic technology, and were set
separate kWh meter. Energy data were collected in the with a time delay of 15-20 minutes as in previous work.1
form of accumulated pulses every 15 minutes. In the two The sensors were selected to cover the area of each zone
floors (3 and 5) where controls were installed, the bi- as accurately as possible. When the wall switch was in the
level switches were replaced by a single electronic switch "off position the occupant sensors did not turn the
that controlled both loads. This switch appeared identi­ lights on, but every occupant sensor "on" or "off event
cal to a standard wall switch except it actually controlled was recorded regardless of the actual state of the lights.
JOURNAL of the Illuminating Engineering Society Summer 2002
 88

On the third floor during 1998, there were State Description Wall Switch Occupant Overhead Light
no additional controls in the open areas Sensor State
besides the switches. Load (on or off status 0 Vacant, wall switch off Off Off Off
of inboard and outboard lamps) and rea­ 1 Occupied, wall switch off Off On Off
son (switch, time schedule, operator, etc.) 2 Vacant, wall switch on On Off Off
data were also collected for each zone, and 3 Occupied, wall switch on On On On
served as a means of checking any anom­ Table 1—State definitions
alies in the switch and occupant sensor
data.
In November-December 1998, a time schedule was results to estimate the daily average "on" time, the aver­
added to shut off the lights automatically at two-hour age morning "on" event, and the average evening "off'
intervals on evenings and weekends, but no changes event. The bi-level switching reduced die accuracy of our
were made to the building's standard practice of allow­ calculated data to +-15 minutes. We were also able to
ing the lights to be turned on in the morning by the first determine the average number of times lights were left
person who desires them.1
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

on overnight on this floor both on weekdays and week­

Although dimming ballasts were installed on the third ends, for comparison with switching behavior on the
and fifth floors, in this paper only switching data are ana­ other two floors.
lyzed, and any dimming methods that affect energy con­ On the third floor in 1998, the time schedule had not
sumption are ignored. However, the zones near windows been implemented, so we were able to include the
are smaller and arranged in separately switched rows, so switching data from that year in the analysis of uncon­
they are analyzed separately from the larger multi-row trolled areas.
interior zones just on the basis of their size. The two types
of zones are called "large" and "small" throughout the Occupant sensors
rest of the paper, where "small" zones have 20 or fewer fix­ Using wall switch and occupant sensor state data and
tures and "large" zones have more than 20 fixtures for var­ a computational method based on previous work, we
ious reasons. Most of the small zones were narrow adja­ determined the total time the lights were in each of four
cent rows with very low or shared occupancy (each desk states defined in Table 1, for each zone in the fifth floor.
in more than one narrow zone), and so are not entirely Note that states 1 and 3 include the time of the occupant
independent, although they have separate switches and sensor delay. The lights are off in all states but State 3.
can be operated separately. Two of the small zones on the Noting that wall switches are regularly turned off in
fifth floor were adjacent and shared an occupant sensor, uncontrolled areas of this building, we determined that a
but had separate wall switches right next to each other. reasonable calculation of savings from occupant sensors
These were almost always operated simultaneously on could only be carried out widi die assumption diat in die
weekdays but were often operated separately on week­ baseline (without sensors) the lights would have been
ends. In addition, the small zones have a larger ratio of turned off manually as noted in die site description above.
corridor to office space than the large zones, and so are
likely to see more casual passersby. The two large zones Calculating savings from occupant sensors
on the fifth floor shared a wall switch but had separate We split the days into two types: days when the wall
occupant sensors that operated the zones differendy, so switch is turned off by midnight (Day Type 1), and days
we analyzed them separately. Because of the high level of with lights left on at midnight, either at the end (23:59)
diversity of our data in this "real life" office environment or the beginning (0:01) of the day or both (Day Type 2).
we do not attempt to analyze these differences in this The split allowed us to examine how use of wall switches
paper. For the reader's information, we noted the size dif­ varied between controlled and uncontrolled areas. We
ferences and presented the data for large and small zones performed die calculations below for each day type to
separately, though grouped visually, in die tables. determine the boundaries of savings from die occupant
sensor in the controlled area. We then applied a factor to
Uncontrolled areas the two values based on the ratio of Day Type 1 to Day
The entire fourth (reference) floor had no controls Type 2 in the third and fourth floor uncontrolled areas to
but the building standard bi-level switching. No switch­ arrive at an estimated nominal occupant sensor savings.
ing data were available on this floor, so die switching The total wall switch "on" time between the first and last
times were calculated from the energy data and the events of the day serve as a "moving baseline" for each day,
known installed wattage in each zone. We used the as calculated below. This equation can be used to simulate
die "on" time assuming no other controls are present:
i Personal communication with building staff

Summer 2002 JOURNAL of the Illuminating Engineering Society

 89

The Moving Baseline At the end of 1998, the time scheduling system was
implemented. Each zone in the time-scheduled area was
Hours with Hours in Hours in given a four-digit code. These codes were supplied to the
•+■
wall switch alone State 2 State 3 occupants according to their location, along with the
telephone override phone number, in a letter from the
building manager's office describing the operation of
We can also simulate the effect of using an occupant the scheduling system with instructions for overriding
sensor alone (without a wall switch): the schedule using their telephone keypad. Stickers on
the phone jacks allow occupants who move to a different
area to locate the access code in their new space.

H
Hours with
Hours in Hours in
occupant sensor Operation - Five minutes before the lights are sched­
State I '+• State 3
alone uled to be turned off, the lighting control system causes
the lights to blink in the scheduled zone(s). If an occu­
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

pant wishes to keep the lights on, he or she may either:

We define the nominal savings from occupant sensors Pick up the phone and dial the telephone override num­
as the time in State 2. In all other states, the lights are ber and punch in the appropriate 4-digit code and the
either off due to the state of the wall switch, or on number of additional hours of light desired (The lights
because they are apparently needed." We can calculate will shut off after the requested time), or walk over to the
the percentage savings as follows: wall switch after the lights turn off, turn it off manually
and then on again (The lights will stay on until the next
scheduled sweep).
Time in State 2 The original time schedule flashed the warning lights
Nominal %
Savingsfrom every hour from 6 pm until 11 pm, and at 2-hour inter­
occupant sensors Hours with vals thereafter until 5 am. However, the occupants of the
wall switch alone area thought the custodian was turning out the lights
while they were still in their cubicles, whereupon they
would begin to call out to her to turn the lights back on.
where the denominator is the "moving baseline" In order to make his or her life a bit more comfortable,
defined above. we changed the schedule to shut the lights off at two-
hour intervals beginning at 7 pm and from midnight to
Finally, we calculated the energy that would have been 4 am. On weekends, the schedule warns the occupants
wasted by an occupant sensor in the absence of wall and turns out the lights every two hours over the whole
switches, by examining the occupant sensor intervals of 24-hour period.
exactly the duration of the occupant sensor time delay
and comparing it with the wall switch "on" and "off' Calculating savingsfrom time scheduling
times. For all longer periods the conservative assumption In the open areas on the time-scheduled third floor
is that the lights were desired, because the occupant we were able to use a modified "moving baseline" tech­
stayed in the room long enough to do more than just nique by a careful analysis of the switching events both by
pass through. the time schedule and by occupants or custodians. Our
analysis differentiated the first evening "off' event after
Time scheduling with telephone override the normal (8:00 am to 5:00 pm) working day, because
Time scheduling is the practice of turning off the most workers had left before that event. Also, we noted
lights in a building automatically at specified hours after the first "off" was usually due to the schedule while sub­
the normal working day. The time scheduling system sequent "off events tended to be due to the wall switch.
used in the testbed is similar to systems used in other On some days, the lights were turned on one, two or
buildings. In some time scheduling systems the lights are (rarely) more times after hours, by individuals coming in
also turned on automatically at the beginning of the work after hours. On other days, the lights stayed on quite late
day, but in this office this was considered an unnecessary because late-night workers overrode every scheduled
step that had the potential to waste energy (there are no event. For each zone-day in the time-scheduled zones we
automatic-on schedules anywhere else in the building). selected all intermediate "off' periods that began with a
time-scheduled event, and summed them to obtain the
ii This statement ignores the occupant sensor time delay. time savings for that day due to the schedule. Any "off'
periods initiated by a manual (wall switch) event were
JOURNAL of the Illuminating Engineering Society Summer 2002
 90

Weekday Weekend
Zone Total Average Average Average Days Day Type 2 Total Days with Overall Day Type 2
not counted as savings. The "moving days total on first on first off with days any lights total on
baseline" for each day was the time event event 2nd on on

between die first "on" and the last "off'

(count) (counQ (*) (hours) (count)(%)
4-1 195 17.11 5.24 22.35 0 0 0.00 85 0 nVa 0 0.00
4-2 195 17.13 5.22 22.35 0 0 0.00
event of the day regardless of reason. 4-3 183 17.08 5.27 22.34 1 0 0.00
85
85
2
3
40.35
1.97
0 0.00
0 0.00
Final "off' events by the time schedule 4-4 183 16.92 5.26 22.26 2 0 0.00 85 2 0.79 0 0.00
4-5 189 16.32 5.79 22.27 2 2 1.06 85 2 2.00 2 2.35
were not counted as savings. It is likely 8?
— 4-6 187 14.90 5.99 20.65 32 4 2.14 85 9 6.35 1 1.18
that a few of the final "off events by the 4-7 185 16.18 5.64 22.16 1 7 3.78 85 23 19.43 8 9.41
4-8 184 15.27 5.94 21.30 2 13 7.07 85 6 5.76 5 5.88
schedule signaled as much as an entire 4-9 181 16.59 5.57 22.17 0 6 3.31 85 16 13.99 5 5.88
lg. Total 40 32 63 21
night's worth of energy saved, but we 4-10 180 16.09 5.31 21.40 0 1 0.56
_ 85 4 12.71 1 1,18
did not count these savings because P 4-11
4-12
189
183
15.90
14.53
5.84
5.93
21.74
20.69
3 2 1.06 85 4 3.46 3 3.53
2 6 3.28 85 3 2.64 2 2.35
lights are rarely left on all night in the sra. Total 5 9 11 6
uncontrolled areas of this building.
Table 2— Fourth floor analysis (decimal hours)
Results
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

Third and fourth floor "no controls" results 1998. No distinction was made between data for 1998
The data in Table 2 illustrate the switching behavior in and 1999. Because we collected occupancy data inde­
the fourth floor with no added controls. On the third pendently of switch data, we were able to examine occu­
floor, the time schedule was not implemented until the pancy patterns even when the lights were off (States 0
end of 1998, and the data from 1998 in Table 7 (see and 1). State 1 includes time when occupants chose to
Time Scheduling Analysis) are used with the fourth floor work without overhead lights, as well as time when the
data to provide a reference for switching behavior in occupant sensor was tripped on by a passerby (see False
uncontrolled areas. On Analysis below).
Table 2 shows the lights are turned off very regularly Figure 2111 illustrates an example switching profile for
at about the same time, and fairly late, practically every two large adjacent zones with occupant sensors. In each,
weeknight in this floor, probably by the custodian. Of all the upper trace shows the state of the wall switch (on or
weekday zone-days on the fourth floor, the lights were off), and the middle trace shows the state of the occu­
left on all night (Day Type 2) only 32 out of 1682 (1.9 pant sensor. The lower trace shows the actual measured
percent) total days in large zones, and 9 out of 552 (1.6 energy data for the zone.
percent) in small zones. In four of the 12 zones the lights Using the state definition given in Methods, we were
were never left on all night throughout the year. We esti­ able to calculate the average hours in each state for each
mate that each night the lights were left on accounted zone over the entire 24-hour period. The results of these
for about 10 additional lighting hours. In only one zone calculations are summarized separately for weekdays and
was there a significant number of second "on" events. weekend days as percentages in Table 3 for "Day Type 1"
The average first "on" event of the day was 5.58 (5:35) and "Day Type 2." We found that Day Type 2 is consider­
and the average first "off' was 21.8 (21:48). On this floor, ably larger in the area controlled by occupant sensors
there are only two states: state 0 (off) and state 3 (on). than in the uncontrolled areas, presumably because occu­
On weekends, custodians are not on duty. The sum of pants have come to trust the sensors to turn off the lights.
the weekend "Days with any lights on" column in Table 2 Weekdays (presented as daily average hours) show
is 63 for large zones, 11 for small. Of these zone-days, 21 quite regular hours in the different states with average
(33 percent) were Day Type 2 for large zones and 6 (54 standard deviation < o" > between 2 and 3, while on week­
percent) were Day Type 2 for small zones. The Day Type 2 ends the range of < a > is between 30 and 40. For week­
percentage of all weekend days is much smaller as seen in end days when lights were turned on, the total number
the last column of Table 2. By similar calculation in the of wall switch "on" hours varied from 0.25 to 24, but on
time-scheduled area (see Table 7 for weekends later in this more dian half the total weekend days the lights were not
paper), lights were left on overnight on 12 of the total 29 turned on at all (days with State 3=0). Because of these
(41 percent) weekend zone-days on which lights were huge weekend variations we presented the weekend data
used in 1998, and in small zones 31 out of 97 (32 percent). as total hours over the analysis period rather than daily
average hours.
Occupant sensors
The control system in this part of the testbed was
placed in operation in May 1996 and remained the same
iii The precision of our graphics program makes the 15-minute intervals
throughout the analysis period, so the occupants were appear more variable in width than they actually are.
fully accustomed to the way the controls operated by
Summer 2002 JOURNAL of the Illuminating Engineering Society
 91

Zone 5-4 01/04/99

From the state data we calculated the aver­
age percentage of "on" hours for the occupant
l__ sensor (states 1+3), the wall switch (states 2+3),
i runx. and both wall switch and occupant sensor
(state 3, lights actually on) for both day types,
on weekdays and weekend days. Day type 2 con­
stituted 13.2 to 16.2 percent of the total week­
days, and 31.7 to 46.7 percent of total weekend
days. Table 4 gives the results of the separate
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time of Day (hr) analyses, allowing us to characterize die range
of possible savings as a function of switch
behavior in the baseline. For these zones, the
lower bound to the estimate of savings over the
Zone 5-5 01/04/99
period studied is the savings calculation for Day
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

Type 1, representing a case where wall switches

-|_ are always used to turn off lights at night. An
i2Ln_Jirui r i n i i n * mi upper bound is the savings calculation for Day
Type 2, where the wall switches are left on for
I either the late night or early morning hours or
both. The actual savings we found are between
these values. Assuming that in the baseline
without occupant sensors the ratio of Day Type
1 2 3 4 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time of Day (hr)
2 to Day Type 1 would be as it is in the uncon­
trolled spaces on floors 3 and 4, we can use this
ratio to calculate an "adjusted wall switch on
time" to use as the baseline. This new baseline
Figure 2—Wall switch and occupant sensor activity and concurrent ener­
gy use for two large adjacent zones.
is a means to adjust for possible effects of occu­
pant sensors on wall switching behavior. The
savings estimate for occupant sensors in the last
column of Table 4 is relative to this baseline.
There is a dramatic difference between
WEEKDAYS (Daily Average Decimal Hours) weekday and weekend percentage savings cal­
Day Type 1(1) Day Type 2 (2) (3) culated in this way. On weekdays, the occu­
Number Number
Zone StateO State 1 State 2 State 3
of days
State 0 State 1 State 2 State 3 <a> pancy patterns are very regular, while on week­
of days
5-4 282 7.4 1.9 1.1 13.5 43 3.0 0.9 5.8 14.3 2.04 ends there may be no occupants, or the room
1 5-5 282 7.5 1.9 1.1 13.5 43 2.9 0.9 5.7 14.4 2.14 may be occupied for many hours. The small
5-1 255 2.9 1.1 12.5 48 2.8 1.2 5.7 14.4 2.49 percentage savings on weekdays are over a
i 5-2
5-3
254
278
" 2.9
1.2
1.1
2.0
12.5
12.6
49
51
2.8
3.3
1.1
0.6
5.7
6.9
14.3
13.2
2.55
2.59
moderately large estimated baseline usage
wEEKEHDS (Tot al Decimal Hours) ("Adjusted WS On" column in Table 4), while
on 5-4 97 1936.4 370.6 1.6 19.5 43 3.0 0.9 5.8 14.3 33.9 the large weekend percentage savings are over
5-5 97 1990.2 316.8 3.7 17.4 43 2.9 0.9 5.7 14.4 34.7 a much smaller baseline usage. Despite the
5-1 69 1468.3 120.1 27.5 40.0 53 432.0 45.8 600.1 194.0 30.3 smaller percentage savings on weekdays, the
i 5-2 65 1392.4 105.7 23.0 38.7 57 489.1 52.4 623.4 203.0 30.8 actual weekday and weekend savings are quite
5-3 77 1762.9 84.2 0.5 0.5 41 205.6 3.6 679.3 95.6 40.8
similar, because of the larger number of total
Table 3—State data hours the lights were used on weekdays. The
weekend percentage applies only to the wall
(1) Lights are off at beginning and end of day switch "on" time on weekend days shown by zone in
(2) Lights are on at either beginning or end of day or both
(3) Average standard deviation Table 4. The average actual "on" time for those days is
very small, varying between 0.02 and 1.37 percent for Day
Type 1 and between 3.38 and 6.93 percent for day type 2,
while the WS On time varies a great deal between Day
Type 1 and Day Type 2.
Applying the appropriate estimated nominal savings
to the actual percent wall switch "on" for all days (includ-
JOURNAL of the Illuminating Engineering Society Summer 2002
 92

WEEKDAYS (% of daily hours, averaged over all days)

Da »Typel Da ^Type2 Adj. Day Overall
Without a wall switch, there would be no
Nominal Nominal Estimated state 1 as we have defined it. Therefore, to
Adjusted
Zone
OS On WSOn Actual Savings OS On WSOn Actual Savings
WSOn
Nominal determine how much energy would have
(%) (*) On(%) Lower (*) (*) On(%) Upper Savings been wasted in the absence of a wall switch
Bound(%) Bound(%) (%) (%)
5-4 64.2 61.1 59.7
by these false ons, we can simply add up all
56.3 7.8 63.3 84.0 28.9 61.7 8.6
! 5-5 64.2 61.0 56.2 7.8 63.8 84.0 60.2 28.4 61.6 8.6 occupant-sensor 15-20 minute periods in
5-1 64.1 56.8 52.1 8.3 64.7 83.4 59.9 28.2 57.4 8.9 state 1. Longer periods in state 1 indicate
1 5-2 63.9 56.5 51.9 8.2 64.4 83.5 59.6 28.6 57.1 8.8 that someone probably was working in the
5-3 57.4 60.9 52.5 13.8 57.3 83.8 55.0 34.4 61.4 14.4
zone, and presumably wanted the lights
WEEKENDS (% of total hours, all days)
off. Notice in the figure that there appears
8, 5-4 11.45 0.62 0.57 7.4 6.59 25.84 5.90 77.2 i 1.34 46.1
5-5 9.81 0.62 0.51 17.4 5.58 25.84 5.10 80.3 1 1.34 52.3 to be a period after hours when an occu­
5-1 5.47 2.31 1.37 40.7 8.19 27.12 6.63 75.6 2.66 45.8 pant worked without overhead lights, per­
i 5-2
5-3
4.93 2.11
2.99 0.03 0.02
1.32 37.3
52.6
8.72 28.22 6.93
3.50 27.36 3.38
75.4
87.7
2.48
0.43
43.5
84.9
haps using task lights or the light from a
computer monitor (see Figure 2, at *).
Most nights the wall switch was turned off,
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

Table 4—Occupant sensor results

so false ons did not occur. In large office
areas with multiple occupants, false ons in
State 3 are rare during normal office hours, but common
after hours until the wall switch is turned off, or in Day
Zone Savings for all days (%)
Type 2. On weekends, the OS On value in Table 4 is quite
Large large in comparison to the Actual On time for Day Type
5-4 9.0 1. This includes false ons, but is also an indication that
5-5 9.0 some occupants may work for significant periods on
Small weekends without overhead lights.
5-1 9.6 False ons are analyzed in Table 5. Subtracting these
5-2 9.4 false ons from State 1 gives the time when occupants are
5-3 14.6 working in the area but have chosen not to turn on the
lights ("Time Occ. Worked w/o lights" columns in Table
Table 4.1—Occupant sensor total nominal savings (all days) 5). The time in this condition would also have con­
tributed to wasted energy if wall switches had been omit­
ted or were difficult to find.
ing both weekends and weekdays), we can calculate a
Day Type 1
total savings. The results of this calculation are given in Zone 15-20 min 15-20 min OS+WS Time wasted if no Time occupant
Table 4.1. OS interval ON interval wall SW worked w/o lights
The savings in the upper and lower bound columns in Large
Table 4 illustrate the role of wall switches in conjunction 5-4 1.35 0.27 1.08 0.82
5-5 1.29 0.26 1.03 0.87
with occupant sensors. With no wall switches at all, the Small
lights would have been on for the "OS On" time on aver­ 5-1 0.81 0.25 0.56 2.31
5-2 0.80 0.25 0.55 2.32
age each day. When the occupant sensor alone showed 0.85 0.38
5-3 0.47 0.69
more "on" time than the wall switch alone, either there
were enough passersby to account for the extra hours, or Table 5—Occupant sensor false on analysis (daily average decimal
there were occupants working with the lights out after hours)
hours. In the test-bed at 450 Golden Gate, the sensor
delays were chosen to minimize occupant complaints. Time scheduling results
To test the intended before/after analysis of time
"False ON" analysis scheduling we did an initial data analysis to compare
If wall switches are left on, occupant sensors will turn average daily on hours from a month before (August
on the lights whenever a person passes by, whether or not 1998) and a month after (August 1999) the application
they are needed. Events triggered by passersby ("false of the control technique in the same experimental areas.
ons") will cause the occupant sensor to be on for almost The averages for these months are presented in Table 6.
exactly its delay time. Such short intervals are seen in the We noted that in August 1999 there were 8.6 percent
occupant sensor data in both state 1 and state 3, and in fewer "on" hours than in August 1998, for both large and
the middle (OS) traces in Figure 2. small zones. At first glance, this seemed a reasonable esti­
mate of savings.
Summer 2002 JOURNAL of the Illuminating Engineering Society
 93

Aug-98 Aug-99 difference On 12/17/99, the lights were turned off by the time
average of large zones 14.32 13.08 8.60% schedule at 19:00, and then turned on again by an occu­
average of small zones 13.95 12.75 8.60%
pant coming in to work after hours at about 19:20. This
Table 6—Preliminary comparison: August 1998 - August 1999
time the occupant left the lights on, and the schedule
turned them off again at 21:00. In the last example, the
time schedule saved energy in the period between the
Next, we did the same simple analysis for the entire first "off event and the arrival of the late-evening work­
year in 1998 and 1999. Figure 3 and Table 7 indicate that er (the second "on" event).
the differences in average "on" hours are much higher in An analysis of switching patterns for the two years is
some zones in 1999 than in 1998, and vice-versa in oth­ presented in Table 7. Note that the average first "off
ers, which we guessed might be caused by something value in 1998 is always later than in 1999, except in two
other than the implementation of time scheduling. small adjacent zones where it was the same in both years.
Delving deeper into the occupants' switching behavior Subsequent "on" and "off events are quite different
enabled us to evaluate the effects of time scheduling and between the two years. As an example, in zone 3-4 in
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

other influences in the different zones. We concluded 1998, there were few "on" events after the lights were first
that differences in occupants' schedules over time turned off at an average of 20.3 hours (20:15). However,
swamped the differences caused by the control treat­ in 1999 returning occupants chose to turn on the lights
ment in several zones. To overcome these problems we a second time 136 out of 233 total days in 1999, at about
developed the "moving baseline" method for this control 20.2 (20:12) on average, and yet a third time on 13 of
technique described in Methods. those days.lv The average last off time (by any means) is
20.4 hours (20:24).
The nominal savings from time scheduling for
this example are calculated as described in
Methods. Of the 233 first "off events, 97 percent
ir were due to the time schedule in 1999, but only
136 were followed by a second "on". For the 136
E second "off events, 60 percent were due to the
I-

as schedule, and 13 were followed by a third "on."

IE? * lii *^ ijit Only one (8 percent) of the third "off events was

3 *f «<» (p r- Vf
I 1 !_!
*- (M n v
due to the schedule. The rest were due to the wall
switch. Since savings are counted when a time-
<o th m <o « « A « <h
scheduled "off" event is followed by any "on"
event, the total average savings calculated from
the event data for this zone were 0.73 hours per
day, or 170 hours over the 233 days studied, for a
Zone nominal average savings of 5.1 percent shown in
Table 8. Keeping in mind that most of the small
Figure 3—Percent difference in total "on" time for scheduled zones 1998-99. zones are not strictly independent, the average
nominal savings calculated in this way for all the
zones is 4.8 percent in both large and small zones,
Figure 4 illustrates four example days of switching with a range from 0.7 percent to 6.3 percent.
data in 1999 for a zone with time scheduling. On Continuing our example, in 1999 there were 97 days
1/5/99, the lights were turned off by the time schedule with no second "on" time. Of these, the lights were
at the first scheduled "off time (19:00). On 4/27/99, turned off by the schedule on 85 days at 19.00 (19:00),
the lights were turned off at the wall switch either by the and on 8 days at 21.00 (21:00). The remaining 4 were
occupants or by the custodian before 19:00. On turned off by wall switches between 22.3 and 22.42 (22:18
12/6/99, the time schedule tried to turn the lights off and 22:25), indicating the schedule had been overridden
twice but an occupant overrode the schedule and kept twice before being turned off at the switch. None of these
the lights on. Upon leaving, the occupant turned the off events were eligible to contribute to savings.
lights off again at about a quarter past 22:00. For these
first three examples, no savings were attributed to the
time scheduling system because all persistent switching iv There were even a very few fourth ON events in some of these zones
was done either by occupants or custodians, or was the that we did not include due to space.
last event of the day.
JOURNAL of the Illuminating Engineering Society Summer 2002
 94

Zone Total Day Average Average Average Average Days with 2 Days 1st off 2nd 3rd
days Type 1st ON 1st OFF Last OFF Total ON or more ON with 3rd bvTS off by off by Although no savings are claimed for
2 time time time hours events ON (%) TS TS
event <%) (%) final "off events by the time schedule, it
Year 98| 99 98 98| 99 98| 99 981 99 98| 99 98| 99 99 99 99 99 is probable that some savings are due to
Large these events, and even likely that on a
3-1 96 235 4 6.23 6.15 19.3 19.0 19.3 19.2 13.4 12.9 1 28 2 87% 54% 0%
3-2 88 184 0 6.40 6.34 19.8 19.1 19.9 20.0 13.5 13.1 3 80 6 96% 70% 0% few of these nights the lights would oth­
3-3 87 233 10 5.96 6.36 20.3 19.1 20.4 20.4 14.5 13.7 6 137 12 97% 61% 8% erwise have been left on all night (see
3-4 88 233 5 5.96 6.36 20.3 19.1 20.3 20.4 14.4 13.7 5 136 13 97% 60% 8%
3-5 94 225 8 5.99 6.14 22.2 19.3 22.4 21.7 16.5 14.6 4 189 77 86% 43% 6% Results, Third and Fourth Floor "No
3-6 95 225 6 6.00 6.08 21.2 19.3 21.3 21.5 15.5 14.5 3 181 61 92% 67% 7% Controls" Results).
3-7 91 215 0 6.49 6.45 21.4 19.5 21.4 21.7 14.9 14.2 2 170 67 92% 71% 7%
Small
3-8 96 233 4 6.34 6.19 19.3 19.0 19.3 19.2 13.3 12.8 1 28 2 85% 61% 0%
3-9 88 183 0 6.38 6.34 19.8 19.1 19.9 19.9 13.5 13.1 3 84 6 95% 71% 0%
Weekends - On weekends in this build­
3
zl2j 88 184 0 6.40 6.34 19.8 19.1 19.9 20.0 13.5 13.1 3 85 8 95% 72% 0% ing, the total savings from the time sched­
3-11 88 183 0 6.40 6.34 19.8 19.0 19.8 20.0 13.5 13,0 3 84 7 95% 70% 0%
3-12 87 233 4 5.96 6.36 20.3 19.1 20.3 20.3 14.2 13,6 2 128 11 97% 59% 18%
ule were small, in part because there are
3-13 87 233 4 5.96 6.36 20.3 19,1 20.3 20.3 14.2 13.7 2 135 14 97% 59%J 29% few total weekend days when lights were
3-14 87 233 4 5.96 6.36 20.3 19.1 20.3 20,3 14.2 13.7 2 135 12 97% 60% 17%
3-15 94 225 3 6.01 6.07 20.1 19.3 20.2 21.3 14.1 14.4 5 175 53 87% 59% 8% turned on. There were no custodians on
weekends, and we noticed that occupants
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

3-16 94 225 3 6.02 6.14 19.8 19.3 20.0 21.6 13.9 14,6 5 191 71 88% 66% 8%
3-17 90 219 0 6.50 6.45 20.9 19.3 21.1 21.5 14.5 14.0 4 179 64 91% 66% 6%
3-18 91 216 0 6.49 6.45 19.5 19.5 19.7 21.9 13.1 14.3 6 178 84 90%" 69% 10% left the lights on all night in several of the
3-19 91 215 0 6.50 6.45 19.5 19.5 19.7 21.8 13.1 14.3 4 175 82 91% 65% 12% zones in 1998 (see Table 7, Weekends, and
Table 7—Time scheduling analysis weekdays (decimal hours)
3rd and 4th Floor "No Controls" section).
In 1999 the schedule always turned off the
lights. For large zones, multiplying the
1998 fraction of days with lights left on (12/29) by
Four Sample Days For A Zone With Time-Scheduttng the number of 1999 days with any lights on (78), we
can speculate that the schedule may have saved a
•" J, rs |
night's worth of energy on 32 days in 1999, and sim­
1/5/99

ws
1 ws
1 ilarly on 98 (or (31/97) *306) days in small zones.
OFF J, J,
4/27/99 On this floor, we do not know how often peo­
ws1 1
ple may have worked on weekends without lights
OFF
12/6/99
f ™f lTsiws
1 1 1 because we have no occupancy data. The savings
ws
J, TS I , ] , WS
12/17/99
1
r
1
potential is highly dependent on whether or not
u workers use the wall switches on weekends, either
to turn the lights on or off.
8 8 8 8 8 8 8 8 8
O 1- M O T- CJ O .- CM
First Morning "On" Event - Another illustration
of the pitfalls of the simple before/after analysis is
the variability of the average first morning "on"
Figure 4—Four sample days for a zone with time scheduling. event over time, which in some zones differed by
as much as 23 minutes between 1998 and 1999 as
seen in Figure 5. The first morning "on" event is
Considering all large zones, we found that although unrelated to the operation of the time-scheduling sys­
92 percent of first "off events were due to the time tem, but affects the total "on" time during the day. The
schedule, only 52 percent of all subsequent "off' events first "on" event of the day varied over time according to
(including those that are the last "off' of the day) are due occupant habits or changes in personnel. The percent­
to the schedule. Overall, the time schedule was responsi­ age savings are affected by the first morning "on" event
ble for 76 percent of all "off' events in the time-sched­ because of the definition of the moving baseline.
uled zones.
Note also that there is a trend toward more late night Discussion
activity in the zones toward the bottom of the table. The The potential for savings from switching controls such
larger zone numbers are on the east end of the building, as occupant sensors and time schedules depends on the
where we noticed that people seemed to stay late during type of space and the behavior of occupants of the space,
a certain period of 1999, possibly due to a group in that whether cubicle occupants or custodians. Such controls
area working on a big project deadline or a change in can save a great deal of energy in areas that are fre­
employees. In the west wing of the building there were quently unoccupied and where the lights are left on.
fewer late night events. However, they have very little effect during normal work­
ing hours or in areas where occupants are diligent in
Summer 2002 JOURNAL of the Illuminating Engineering Society
 95

Zone Nominal Moving Nominal turning out lights when they leave rooms, and they can
Savings Baseline % savings cause some waste if their presence causes occupants
(hours/day) (last off- NOT to use die wall switches. Weekends are quite differ­
first on) ent from weekdays: open areas can become more like
private offices widi a single occupant, and the savings
Large potential depends perhaps even more on occupant
schedules and behavior.
3-1 0.09 13.02 0.7% The usefulness of diese controls in large spaces is in
3-2 0.53 13.61 3.9% ensuring that lights are switched off when not needed,
3-3 0.72 14.41 5.0% and the achievable savings depends heavily on die
3-4 0.73 14.40 5.1% method of calculation of the baseline. If the last occu­
3-5 1.03 15.72 6.6% pant to leave always turns out the lights, no switching
control system, occupant sensing or time scheduling
3-6 0.94 15.47 6.1% included, can save any energy. But in most buildings,
3-7 0.95 15.18 6.3% some people forget, and control systems act as valuable
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

Small insurance against wasted energy.

3-8 0.10 12.92 0.8% Large open office areas are very different in character
3-9 0.56 13.64 4.1% than small private offices, in which occupant sensors can
save very significant amounts of energy right in the mid­
3-10 0.56 13.64 4.1% dle of the day for occupants who frequently leave their
3-11 0.56 13.61 4.1% offices. In larger spaces during normal working hours,
3-12 0.68 14.33 4.8% there is no clear ownership of the area, and the space is
3-13 0.73 14.39 5.1% rarely unoccupied. Occupant sensors "see" a diverse pop­
3-14 0.73 14.39 5.1% ulation, and can only turn off the lights when every one
3-15 0.83 15.28 5.4% of the occupants in an area is gone. Similarly, time sched­
ules are overridden for the entire area controlled by a
3-16 0.96 15.59 6.1% single switch when only one person is left requesting that
3-17 0.90 15.01 6.0% the lights remain on. After hours, multiple occupants
3-18 0.97 15.36 6.3% potentially unseen behind cubicle walls may cause reluc­
3-19 0.96 15.30 6.3% tance to turn off wall switches at the time when most peo­
ple are leaving. These characteristics limit the ability of
Table 8—Time schedule savings estimate these control methods to save energy to a degree which
depends in part on the number of occupants and the
diversity of dieir schedules, and in part on the area con­
trolled by a single switch. But, during normally
7 — — unoccupied hours with individual people working
overtime, large spaces operate much more like pri­
vate offices. Workers who stay late are likely to be
aware that they are the only person in a large office

fl
es

■
J ■
■J
■ 1
L
■
L k b ■ i:::-i
area, and they appear much more likely to turn off
the light switches by hand.
While the percentage savings from controls in
H"
c
SS

S2S

S
I 1I I 1I 1
I I ■1 I l 1
LM_Jt .■-_■_-■__■__■_
3-1 3-2 3-3 3-4 3-5 3-e 3-r
large spaces may seem unimpressive, the larger size
of the space means tfiat the actual savings potential
from a single occupant sensor or time-scheduled
switch is greater. For example, in a private office
widi two 3-lamp light fixtures, the potential savings
for an hour of unoccupied time with an occupant
sensor set at 15 minute time delay would be 3/4 hr
Zoo*
* 180 W = 135 W/h, or 0.135 kWh. In a large area
with 36 3-lamp fixtures, the potential for savings
Figure 5—Illustration of changes in first morning "on" event over test from the same unoccupied hour is 3/4 hr * 3240 W
period. = 2430 W/h, or 2.43 kWh. Though in a small private
office you may achieve an average of 26 percent sav-
JOURNAL of the Illuminating Engineering Society Summer 2002
 96

ings from occupant sensors on a weekday, the result for On weekends, the percentage savings from occupant
each sensor is a small number. In the large office, a 10 sensors can appear much larger than on weekdays, if wall
percent savings can have a much more substantial bene­ switches are not turned off. However, the actual total sav­
fit for the relatively small cost of installing the sensor orings depend on how often occupants work on weekends
implementing a time schedule. However, to achieve the and whether or not they are conscious of the need to flip
greatest energy savings it is advisable to use more smaller the switch when they leave. We also noted that workers
controlled zones as long as it is practically and economi­ put in significant hours without turning on the lights on
cally justifiable. weekends, probably using the task lights that are avail­
In a working office with uncontrollable variables, it is able in all cubicles in these open areas. Task lights also
important to set up the data analysis very carefully. permit lower general lighting levels, and we recommend
Before-after analysis can be deceptive where changes that they be included in all office lighting designs.
occur over time in office personnel or in project activi­ Comparing these results with earlier estimates of sav­
ties. Close examination of the data can help refine and ings in private offices,1 it is clear that greater energy sav­
increase the accuracy of the analysis method. ings are achievable where occupant sensors are designed
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

to control smaller areas, albeit at greater installation and

Occupant sensors - Our results indicate that occupant commissioning cost. A narrower field of view from the
sensors save an average 10 percent of lighting energy in occupant sensor would eliminate a portion of the false
open offices in this installation, compared to a moving on events relative to the size of the area covered.
baseline with light switches alone. Savings from occupant
sensors are limited by a number of factors. "False ons" Time Scheduling - Times scheduling controls can save a
(inadvertent activation of occupant sensors due to great deal of energy if compared to a baseline where light
passersby) caused some energy waste that would not have switches were not regularly turned off. On the other hand,
occurred with time scheduling, but the occupant sensors if someone diligendy turns off the lights off every night at
also turned off lights soon after the last occupant left 6 pm, adding a time schedule to turn the lights off at 7 pm
rather than waiting until a scheduled time. False ons are would not produce any savings at all, and might cause ener­
included in the total "on" hours on a given day, because gy waste by encouraging people not to use the switches. In
they are part of the typical operation of occupant sen­ this building, the time schedule consisted of multiple
sors. Their effect varies with the programmed time delay sweeps at 2-hour intervals with wall switches as overrides.
in the occupant sensors. Emergency egress lighting For comparison, a building with a single sweep off at 11 pm
should be adequate for incidental traffic, and occupant and another sweep on in the morning, with no wall switch­
sensors that are adjusted not to "see" corridors will have es available, would doubtless reveal quite a different value
a lower incidence of false ons. for savings depending on how the baseline was chosen.
Another limit to the potential savings of occupant sen­ For days with at least a second "on" time, we have
sors is the increased likelihood of occupants relying on inferred that the occupant who was intending to come
the sensors to turn off the lights. If the switch is off at back would not have turned out the lights, based on the
night, passersby will not trigger occupant sensors and large fraction of first "off" events due to time schedule in
cause a "false on" to occur. this example zone and the scant number of second "on"
Occupant sensor time delays can be shorter than events in 1998. The first "off" event after normal working
those in this study, thereby diminishing the effect of false hours appears to be different in nature from subsequent
ons. However, without careful installation and commis­ "off events. The data suggest that people who stay late
sioning short delay times have the potential to cause are more inclined to take the initiative to turn off lights,
occupant dissatisfaction. This is a matter of choice for probably because they are likely to know that they are the
building operations personnel. In open offices, the occu­ last person in the space.
pant sensor time delay has little effect on savings during Lights in time-scheduled zones are not typically
normal working hours, but it can be significant after turned on by passersby if emergency egress lighting is
hours. The total effect depends on the amount of traffic properly designed, so energy is not normally wasted in
near enough to the controlled area to cause the sensor this way. However, since the automatic switching takes
to trip, and can vary widely from day to day. We did not place only at preset hours, energy may be wasted
modify our calculation to include the error due to occu­ between the time the last occupant leaves and the time-
pant sensor time delays. A potential improvement to the scheduled switching if the wall switch is not used or is not
state of the art for occupant sensors might be to permit available. The amount of such potential waste can be
reduced time delays on evenings and weekends, thought adjusted by increasing the frequency of scheduled
this would add to the complexity of commissioning. switching events, but this must be balanced against
potential annoyance of the occupants.
Summer 2002 JOURNAL of the Illuminating Engineering Society
 97

Wall switches - We have observed occupants in this all "first off' events in these areas, and 52 percent of all
building will turn off lights, particularly when they are subsequent "off' events, and 76 percent of all off events
likely to be the last person left in a particular area. Wall overall, suggesting that workers alone in a space are more
switches are a very important part of any lighting control likely to take responsibility to turn off the lights.
system, because they can allow lights to be turned off pos­ Occupant behavior has a marked effect on the savings
itively. If there is no automatic "on" control in the morn­ attainable by lighting controls and the appropriateness
ing, wall switches allow the lights to stay off until they are of one control technique vs. another. No savings can be
needed. As always their effect depends greatly on how attributed to controls whenever a wall switch is used to
they are used. In an office where people pass through turn off the lights in any of these zones. However the
often on their way to another area or while making secu­ presence of wall switches which positively turn out the
rity rounds, occupant sensors can waste a great deal of lights gives occupants control over their lighting envi­
energy if the wall switches are not available to override ronment, and allows them to save additional energy over
them or are hidden. Wall switches that are all in the same that which could be saved by automatic controls.
place are likely to be turned on or off all together, partic­ Occupants who assiduously turned out their light switch­
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

ularly if they are remote from the lights they control, but es in areas controlled by either technology were respon­
this effect is reduced after normal working hours. Data sible for considerable energy savings, none of which was
from this test-bed showed that occupant sensors without attributable to the control technologies.
wall switches would have saved no more than wall switch­ The decision to choose one particular control system
es alone in the areas in which they were installed. We over another should be based on an analysis of manage­
believe all commercial office buildings should have acces­ ment practices and expected occupant behavior (includ­
sible manual switches capable of overriding the occupant ing custodial or service personnel) in the area to be con­
sensors even if they are not required by code, regardless trolled. Both control systems considered in this study
of whether automatic switching controls such as occupant save energy consistently, though how much they save
sensors or time scheduling are used. depends on how they are commissioned and coordinat­
ed with the needs of the users of the space, and on the
Conclusion baseline condition. We propose that, in addition to other
In a long-term test of functioning office spaces with lighting controls, building managers encourage occu­
diverse occupants, we determined that both occupant pants to turn off the lights when they leave.
sensors and time scheduled controls have the potential Our findings contrast with previous results for private
to save significant energy, but the savings that can be offices in which only a single occupant is present and
attributed to controls depend greatly on occupant "ownership" of the lighted space is more personal. In pri­
habits, custodial practices and the baseline of the build­ vate offices, the occupant usually turns out the light
ing in question. We found both methods saved energy switch when leaving at night, knowing that no one else is
consistently, but that the presence of lighting controls in the space, and the occupant sensor can save consider­
affects occupant switching behavior. Since our moving able energy in the middle of the day. The diversity of
baseline method ignores potential savings from the last occupant schedules means the lights are almost always
"off' event in any day, our nominal savings results are on in large zones during the normal workday regardless
conservative. of the lighting control technique.
Occupant sensors in general turned the lights off well
before the wall switches were turned off, saving a nomi­ Acknowledgment
nal average of 10 percent of lighting energy in the zones This work was supported by the General Services
they controlled. On weekends, the percentage savings Administration, Pacific Rim Region, and by the Pacific,
from occupant sensors was much larger at 47 percent - a Gas & Electric Co., and by the Assistant Secretary for
number highly dependent on occupancy patterns, and Energy Efficiency and Renewable Energy, Office of
applicable only to a small number of weekend hours. In Federal Energy Management Programs and the Office of
the absence of wall switches that avoided occupant sen­ Building Technology, State and Community Programs of
sor "false ons" by positively turning off the lights, savings the U.S. Department of Energy under Contract No. DE-
from occupant sensors in the areas studied would have AC03-76SF00098. The authors would like to thank Bob
been negligible. Clear for the incisive clarity of his technical advice,
A conservative nominal average savings of 5 percent Dennis DiBartolomeo for development of software to
was calculated in the time-scheduled areas with a period support the calculations, and Ann Nguyen for assistance
of 2 hours between scheduled "off' events, compared with data and graphics.
with wall switches alone operated by custodians and occu­
pants. Time scheduling was responsible for 92 percent of
JOURNAL of the Illuminating Engineering Society Summer 2002
 98

References zone, or one per workstation, or something else? Where

1. Jennings, J., Rubinstein, R, et al. 2000. Comparison were they located? There would be great benefit in the
of Control Options in Private Offices in an Advanced floor plans showing the location of workstation partitions
Lighting Controls Test-bed. J of the IESNA. 29 (No. 2). pp. and the location of occupancy sensors.
39-60. These quibbles aside, their series of reports on this
2. Maniccia, D., Rutledge, W., et al. 1999. Occupant building are welcome because they address the perfor­
Use of Manual Lighting Control in Private Offices. Jofthe mance of lighting controls as they are installed in real
IESNA. 28 (No. 2). pp. 42-56. buildings, and not theoretical performance in a con­
3. Pigg, S., Eilers, M., et al. 1994. Behavioral Aspects of trolled laboratory setting. If we wish to increase the pene­
Lighting and Occupancy Sensors in Private Offices: A tration of lighting controls in the marketplace we must
Case Study of a University Office Building. ACEEE 1994 focus on increasing the as-installed field performance.
Summer Study on Energy Efficiency in Buildings. (No. 8). pp. The real performance might be regrettably lower than the
161-170. lab performance but, in my opinion, we do ourselves a dis­
4. Rubinstein, R, Verderber, R., et al. 1985. service if we overstate potential savings only to disappoint
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

Determination of the Optimum Sector Size for the building manager later. Such a strategy is unlikely to
Automatic Lighting Controls. / of the IESNA. 15 (No. 1). promote the long-term adoption of energy-saving lighting
pp. 3-20. controls. I would appreciate the authors' comments.
5. Rubinstein, R and Karayel, M. 1984. The Measured Another important lesson for the controls industry
Energy Savings From Two Lighting Control Strategies. can be drawn from an easily overlooked sentence early in
IEEE Transactions on Industry Applications. 1A-20 (No. 5). the paper: "We had an unusually large number of prob­
lems in this test-bed due to inexperienced installers ...".
Discussions This is not the only anecdotal evidence that poor instal­
Anyone who has undertaken a field study related to lation and commissioning is seriously impeding the abil­
building system performance knows just how difficult ity of lighting control systems to deliver the promised sav­
dealing with real buildings and their occupants can be. I ings. I believe if the industry fails to address this issue
can only marvel at the fortitude of the LBNL researchers market penetration will be compromised. Again, I would
who have carried out the studies at 450 Golden Gate on appreciate the authors' comments.
such a large scale. This report is particularly welcome in Finally, a comment on the reported energy savings.
that it addresses open-plan spaces. These spaces are per­ They are substantial from a conservation point of view,
haps the most common office environment in North but I suspect inadequate to justify to a building manager
America, yet tfiey are under-represented in lighting con­ the cost of installation at prevailing energy prices at most
trols research compared to individual offices because they locations in North America. I believe to encourage the
are much harder to deal with. Indeed, many refuse to adoption of lighting controls in general we need to fur­
consider controls for the open-plan because of the impact ther explore the benefits beyond energy savings. For
of one occupant's actions on their colleagues. Yet if we example, a number of recent studies have suggested that
ignore these spaces we give away a large potential market. offering die occupant some form of individual control
As the reader will quickly realize, the complexity of may result in improved environmental satisfaction
field studies, and the reduced ability for the researcher (Newsham and Veitch, 2000; Boyce et al., 2000; Jennings
to control conditions, makes analysis of lighting control et al., 2000; Carter et al., 1999 and Maniccia et al.,
impacts difficult. While I have no reason to doubt the 1999) .a_e Can the authors offer any insights from their
credibility of the calculated savings - and I am pleased to vast experience on the issue of non-energy benefits?
see them err on the conservative side - their method, as
presented in the paper, smacks a little of "Black Magic". References
It is difficult for the reader to keep track of all the a. Newsham, G.R. and Veitch, J.A. 2000. A New
assumptions and corrections made in reaching the final method of deriving illuminance recommendations for
savings estimates. I recommend that in any future publi­ VDT offices." Proceedings ofIESNA Conference. Washington,
cation, a fully worked example for one space be present­ D.C. pp. 211-217.
ed with every step in the savings calculation clearly b. Boyce, P R , Eklund, N.H. and Simpson, S.N. 2000.
explained. Individual Lighting Control: Task Performance, Mood,
Another area where more information would be very and Illuminance, f of the IESNA. 29(No. 1). pp. 131-142.
valuable is the description of the control systems in the c.Jennings, J.D., Rubinstein, F.M., DiBartolomeo, D.
open-plan spaces. I assume that in the large zones there and Blanc, S.L. 2000. Comparison Of Control Options In
were many fixtures for many occupants on a single cir­ Private Offices. / of the IESNA. 29 (No. 2). pp 39-60.
cuit. How many occupancy sensors were there? One per d. Carter, D.J. Slater, A.I. and Moore, T. 1999. A Study
Summer 2002 JOURNAL of the Illuminating Engineering Society
 99

Of Occupier Controlled Lighting Systems. Proceedings of availability may provide greater reductions in response to
the 24th Session ofCIE. Warsaw, pp. 108-110. use. From a behavioural perspective, these may also be
e. Maniccia, D., Rutledge, B., Rea, M.S. and Morrow, superior in that the user has some control over the oper­
W. 1999. Occupant Use of Manual Lighting Controls in ation of the system. The reduction in on-time for private
Private Offices. JofthelESNA. 28 (No.2). pp. 42-56. offices seems low relative to the findings by other author-
sa,b who reported private perimeter offices they studied
Guy Newsham to be vacant 20 to 70 percent of the time. Changes in
Institute for Research in Construction, technology and the spread in results suggests a need for
National Research Council Canada continuing field studies of control performance.

I congratulate the authors for their extensive study References

and detailed analysis on die use of time-based and occu­ a. Opdal, K. and Brekke, B. 1995. Energy Saving in
pancy-based lighting controls in open spaces. This valu­ Daylighting by Utilization of Daylight. Proceedings of Right
able data will help building managers and lighting prac­ Light Three: The Third European Conference on Energy-
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

titioners alike to better understand what are the typical Efficient Lighting. Vol. 1 - Published papers. Newcastle-
pitfalls to avoid as well as the opportunities to capitalize upon-Tyne, UK- Northern Electric, pp. 67-74.
when using these common types of controls. I was some­ b. John, R.W. and Salvidge, A.C. 1986. The BRE Low-
what surprised to see such low levels of energy saving Energy Office-Five Years On Garston. Watford, UK Building
from occupancy sensors, as our own field data collection Research Establishment, Department of the
for similar situations has often found much higher val­ Environment.
ues, closer to 30 percent and not 10 percent as found in
this study. The reasons for the difference might be found James Love
in the number and type of sensors used and their loca­ Dept of Mechanial Engineering
tion, their calibration and the furniture type and layout. University of Calgary, Canada
I would appreciate the authors providing detailed
description of these factors. I am also interested in know­
ing if the authors considered restricting, by way of appro­ Author's response
priate masking devices, the typically large area coverage
this type of occupancy sensors usually are designed for, in To Guy Newsham
order to more closely match the actual site occupancy My thanks to Guy Newsham for his comments. He
conditions. Obviously, occupancy sensors used at the rate brings up a number of excellent points that I would be
of one/occupant would reduce lighting energy use at a happy to discuss further, and I will try to demystify the
much higher rate, albeit at a prohibitive financial cost. "black magic."
Would the authors estimate the recommended density of The key concept in the paper is the moving baseline,
occupancy sensors they would suggest for this type of which arises out of a need to account for the habits and
facility to maximize energy savings with a typically accept­ practices of individual occcupants and groups of occupants
able return on investment? in the spaces under consideration. Using a fixed baseline
or a before/after baseline can give false results due to vari­
Ivan Pasini, M. Eng., P. Eng., FIES ability between spaces and between time periods in the
Public Works Government Services Canada, same space. In my revision of this article for this publica­
Ottawa, ON, Canada tion, I have attempted to clarify the methods for both the
occupant sensing and the time-scheduling analyses.
I fully agree with the comment that it is a mistake to
Controls present great opportunities to enhance both overstate the performance of energy-saving lighting con­
user satisfaction and occupant comfort. Independent, rig­ trols. Building owners and managers are more likely to
orous testing of the type reported here is essential to make use of energy-saving technologies if they feel they
establish reliable savings estimates and guidelines for can trust engineers' savings predictions. Inflated claims
application. With the type of occupancy control can cause these useful technologies to be discredited.
described in this paper, it appears efficiency in ambient I particulary appreciate the point about the installers.
lighting design is crucial in large open plan areas due to In our previous work we have taken pains to point out
the low fraction of off switching that can be achieved. The the importance of proper installation and commission­
advent of commercial systems that allow occupant control ing of control devices.a Commissioning problems are a
of individual luminaires in illumination preferences in major reason for systems failing to produce savings, and
addition to control based on occupancy and daylight it is critically important for owners and facility managers
JOURNAL of the Illuminating Engineering Society Summer 2002
 100

to hire contractors who can ensure that commissioning gies are difficult to justify in terms of economics alone
is done correctly. Measurement and verification are until the cost comes down.
often neglected in energy conservation projects, but they Our paper covered large open office areas, not private
are critical elements of demonstrating the functioning offices, so our results cannot be compared directly to pri­
and effectiveness of the energy conservation measures, vate office results. In previous work on this test-bed (see
and can help provide an incentive to installers to com­ Reference 1 in the paper), we found that some of the pri­
mission the project correctly. vate offices we studied were indeed unoccupied as much
In regard to the last comment about benefits beyond as 70 percent of the time, but others were occupied
energy savings, I would like to point out that the energy almost constantly. In that paper we found an overall aver­
savings in this building seem low in part because the age savings from occupant sensors of 26 percent when
baseline itself is quite low. The paper states clearly that averaged over all offices studied for the seven-month
this is a well-operated building where the lights are very period we analyzed.
rarely left on for long unoccupied hours. I'm not sure In contrast, in this study we are looking stricdy at large
that simple occupant sensors and time schedules in open spaces widi multiple occupants, corridors adjacent, occu­
Downloaded by [University of Tennessee, Knoxville] at 11:20 07 June 2016

offices offer many other benefits, unless people prefer to pied most of the time at least by someone during every
have the lights turn on automatically when they walk in. normal working day. It is rare for a whole open office
But as I've stated, you have to leave the wall switch on for area to be empty during working hours.
this to happen, and there is significant waste associated
with doing this in an open office space.
As Guy points out, it has been shown that giving indi­
viduals a degree of control over their lighting environ­
ment is perceived as a benefit. In these open spaces, we
have seen evidence of occupants choosing to work with­
out lights, and also to make use of bi-level switching in
areas where it was available. I believe it is likely that they
would make greater use of more individualized controls
were they to be made available. The least we can do is to
provide wall switches, as well as the benefits of other
types of control such as individualized dimming and bi-
level switching.

To Ivan Pasini
My thanks to Ivan Pasini for the questions. I've
answered some of them in my responses above. In each of
the "large" zones, we controlled from 25 to 35 fixtures with
a single ceiling-mounted occupant sensor. This is a pretty
large number, and we think it would be far better to con­
trol smaller zones with more sensors, though it would cost
more to implement. In small zones, we had from 10 to 20
fixtures per sensor. One problem is not fully overcome by
multiple sensors in an open area is the problem of passers-
by. A person walking through a large area will trip every
sensor he or she passes if the wall switch is on.
We did not consider masking devices, but relied on
the manufacturers' coverage areas in laying out the sen­
sors with as little overlap as possible.
Regrettably, we were unable to do a return on invest­
ment calculation on these sensors because of limited
resources.

To Jim Love
We appreciate Professor Love's comments. The
thorniest obstacle to individual control of illumination is
the present cost of controllable ballasts. Such technolo­
Summer 2002 JOURNAL of the Illuminating Engineering Society