Sie sind auf Seite 1von 10

Preliminary findings

from California
research project
provide insight into
fan design and

A Fresh Look at

energy savings

Fans

an energy is a large part of mechanical- Public Interest Energy Research (PIER) program.2
system energy usage. In Californias new The projectLarge HVAC Integration3focuses
commercial-building stock for built-up on the air side of built-up VAV systems, including
systems, it accounts for 1 terawatt-hour of electric- fan selection, fan staging, and supply-pressure
controls, with the ultimate goal of
energy usage per year, representing
By MARK HYDEMAN, PE,
producing design guides for airapproximately 50 percent of all
and JEFF STEIN, PE,
side-system design and control.
HVAC energy usage.1 Standard
Taylor Engineering LLC
practices and operation can lead to
The project includes detailed site
designs that use as much as twice
monitoring of five built-up VAV
the energy of optimized designs. This can be attrib- systems. The parameters include fan energy, airflow,
uted in part to the lack of analysis of fan-system fan static pressure, duct pressure, and terminal-unit
performance across the full range of operation and demand. In the process of doing the research, which
the lack of tools to perform such analysis. This will conclude in June, the team developed a new
article reports on a current publicly funded research simulation model of fan-system efficiency as a
project on optimized design techniques for large function of flow and pressure. The model will
variable-air-volume (VAV) systems and control. be presented in a symposium paper during the
Preliminary findings from one of five monitored American Society of Heating, Refrigerating and
sites are presented. These results provide insight Air-Conditioning Engineers (ASHRAEs) 2004
into both methods of design and means to save fan Winter Meeting. Details on the model are available
energy.
on the project Website.3
This article describes our preliminary findings
BACKGROUND
at Site 1, a high-tech office building with data
The authors are part of the team performing the centers located in San Jose, Calif.4 The building is
research under the California Energy Commissions three stories tall, approximately 105,000 sq ft, and

Mark Hydeman, PE, is a principal and Jeff Stein, PE, a senior engineer for Taylor Engineering LLC in Alameda,
Calif., a mechanical consulting firm specializing in the design and commissioning of HVAC and control systems
for commercial buildings. The firm actively participates in the development of building energy codes and
building-science research.
28

May 2003 HPAC Engineering

owneroccupied and has


a central air handler and VAV boxes with
hot-water reheat. The central air handler
consists of two 66-in. plenum fans in
parallel with barometric backdraft
dampers at the fan inlets (Figure 1). Each
of these supply fans was designed to
provide 72,500 cfm at 4-in.-wc static
pressure (145,000 cfm total). The supply
fans have 75-hp motors and variablespeed drives (VSDs). The VSDs are
controlled to maintain a fixed duct
static-pressure setpoint of 1.5 in. wc, as
measured at the bottom of the duct riser.
This air-handling system runs continuously to serve late-night programmers
and server rooms.
Figure 5 (Page 35) presents the measured airflow and static pressure. The
static-pressure measurements, including

t h e this fan, the do not select or surge line


pressure coincides with the line of peak efficiency.7
drop across
the inlet backdraft FINDINGS FROM SITE 1
The new fan model and the monidampers, 5 were taken
from the fan inlet plenum to tored data were used to answer several
the discharge plenum. The red line is questions:
the theoretical sys What fan
tem curve for the
should the engifan, assuming a fixed
neer have selected,
minimum staticand how does it
pressure setpoint of
compare to the one
1.5 in. wc and the
that was selected?
system design static
What is the
pressure.6 Figure 2
optimal staging of
shows the manufacthe two fans, takturers fan curve,
ing into account
while Figure 3 shows
the efficiencies of
the same fan curve
the fans, motors,
in three dimensions,
belts, and VSDs?
with efficiency plot What is the
ted in color on the
effect of resetting
Z-axis as a function
the fan static presof airflow and fan FIGURE 1. Plan view of Site 1 mechanical
sure according to
static pressure. For room.
VAV-box demand?

HPAC Engineering May 2003

29

F A N S

FIGURE 2. The manufacturers fan curve.

It is important to note that fan selection generally applies only to new construction, while staging and reset apply to
both new construction and retrofits. In
theory, one could replace an existing fan,
but this is unlikely to be cost-effective
unless the existing fan is at the end of its
service life. Each of these issues will be
examined in detail in the paragraphs that
follow.

Fan selection. Although they are more


expensive and less efficient than housed
centrifugal fans,8 plenum fans take up
less space and are quieter in the critical
low-frequency bands.9 The engineer for
Site 1 probably selected plenum fans to
save space and provide a hedge for the
acoustical design.
While keeping an eye on sound-power
data and motor size, the engineer likely

66%
52%
Efficiency, percent

54
37%

36

23%

18
0

9%

3.6

rentia
Diffe

l pres

sure

4.8

2.4
1.2
40,000

120,000
80,000
r minute
pe
Cubic feet

160,000

FIGURE 3. Fan efficiency as a function of airflow and static pressure.

30

May 2003 HPAC Engineering

200,000

F A N S

used the manufacturers selection program to pick a fan with a design point
that was efficient, but not too close to the
surge or do-not-select region.
Given that the manufacturer did not
make a plenum fan large enough to
meet the total design flow efficiently,
the engineer decided to use two fans in
parallel. This had the benefit of redundancy (reduced exposure to fan failure)
and improved performance under lowload conditions.10
The engineer probably looked at
something similar to what is shown in
Figure 4 and chose the 66 PL-A because
it provided a good combination of
efficiency, relative cost, acoustics, and
motor size. It is unlikely that any simulation or evaluation of part-load operation
was considered.
Fan selection for Site 1 was evaluated
by simulating a range of potential selections against the monitored fan-load

FIGURE 4. Typical fan-selection software.

profile (total cubic feet per minute and


differential pressure across the fan).
Figure 5 shows the monitored data and
design point. Figure 6 shows that the

Circle 200
32

May 2003 HPAC Engineering

system spends the majority of its time at


very low flows and never came close to the
design condition during the monitoring
period. Figure 6 has the same X-axis scale

Circle 184

F A N S

Differential pressure across fan, inches of water

4.0
Actual data
Perfect system curve (i.e., static-pressure reset)
Design point (145,000 cfm at 4.0 in.)
Likely actual system curve (no static-pressure reset)

3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0

20,000

40,000

60,000
80,000
100,000
Total cubic feet per minute

120,000

140,000

FIGURE 5. Monitored data from Site 1.

as Figure 5; together, they display the frequency of operation for each


region of operation. As Figure 5 shows,
the actual system curve appears to run
through 1.5 in. at 0 cfm. A consequence
of a high fixed static-pressure setpoint is
that the fan operates in the surge region
at low loads (Figure 2). Because they were
oversized, the fans operate in the surge
region more than 60 percent of the time.
As described below, aggressively resetting
the static-pressure setpoint would reduce
or eliminate this problem.
7,000

The system spends many hours at night


between 20,000 and 25,000 cfm

6,000

Records (15-min data)

Figure 7 shows the efficiency of the


Site 1 fan system along the system curve
shown in Figure 5 as a red line. Figure 7
shows that the fan efficiency goes up
and down as cfm changes and as the
system stages from single- to dual-fan
operation. According to our simulations,
the average fan efficiency during the
monitoring period was 57 percent.
Several other fan selections were simulated against the actual measured load.
These included other sizes of plenum
airfoil fans and several sizes of housed air-

5,000
4,000
3,000
2,000
1,000
0
0

20,000

40,000
60,000
80,000
100,000
Airflow bin (cubic feet per minute)

120,000

140,000

FIGURE 6. Histogram of Site 1 cfm.


Circle 192
HPAC Engineering May 2003

35

foil fans. Figure 8 shows simulation


results for the base case and alternate fan
selections. The first two lines in Figure 8
compare the actual fan staging with an
optimal staging using the same fan (see
discussion of optimal staging later in this
article). Figure 8 shows that the 66-in.
plenum fan (66 PL-A) used less energy
than the 49-in. (49 PL-A) fan and about
the same as the 54-in. (54 PL-A), 60-in.
(60 PL-A), and 73-in. (73 PL-A) fans.
It is interesting to note that the annual
energy ranking from the simulation
(Figure 8) does not follow the efficiency
ranking from the manufacturers selection
program (Figure 4). There are several
reasons for this. One has to do with the
valleys and peaks (or sweet spots) in
the efficiency profile of each fan (see, for
example, Figure 7) compared with the
load profile. Different fan systems have
peaks and valleys at different spots.
Figure 8 also shows that housed airfoil
fans (marked A-DWDI) are consistently
more efficient than plenum fans (marked
PL-A). Of course, this is not necessarily a
fair comparison because of the space
requirements and acoustical issues associated with housed fans.
Fan staging. In the simulations, the
most efficient control sequence for staging the fans was employed. As mentioned
previously, the first two lines in Figure 8
compare the actual fan staging with an
optimal staging using the same fan. Based
on the monitored data, the existing
control scheme runs one fan almost all of
the time, with the second fan brought
on for a few hours each week, mostly
late in the afternoon. The optimal staging
routine in the simulation compared the
energy use of one and two fans at each
operating condition, selecting the option
with the lowest energy that could meet
the load based on the installed motor
horsepower and fan curve. For this load
profile, it almost always was better to run
one fan because the majority of the hours
are at very low load conditions.11 Thus,
switching from the base case to the optimal staging with the same fan (66 PL-A)
did not save much energy (Figure 8).
36

Fan efficiency, percent

F A N S

75
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0

Stages from one to two fans operating


(this is the optimal staging point for
this fan system on this system curve)

20,000

40,000

60,000
80,000
Cubic feet per minute

100,000

120,000

140,000

FIGURE 7. Fan efficiency.

In a real control system, staging logic


must be based on a measured input, such
as supply-fan VSD speed. Also, it needs
to have a control differential or other
feature to avoid short-cycling the fans.
The authors sought to develop recommendations for staging the fans based on
speed. Figures 9 and 10 present optimal
staging speeds for this building with and
without supply-pressure reset. In these
figures, the green and blue lines represent

the power consumed by the fans as they


run up and down the system curve. The
red and yellow lines represent the speed
of these fans at each condition. The
dashed lines show the speed at the optimal staging point. For the 66 PL-A fan
system and for a system curve that runs
through 1.5 in. (i.e., fixed static-pressure
setpoint), the optimal point to stage from
one fan to two is when the fan exceeds
about 77-percent speed. Conversely, the

66 PL-A base-case staging

$20,474

66 PL-A optimal staging

$20,206
$21,666

49 PL-A optimal staging

$20,232

54 PL-A optimal staging


60 PL-A optimal staging

$19,713

73 PL-A optimal staging

$19,993

49 A-DWDI optimal staging

$17,424

54 A-DWDI optimal staging

$17,069

60 A-DWDI optimal staging

$16,835
$17,081

66 A-DWDI optimal staging


73 A-DWDI optimal staging

$18,065
$0

FIGURE 8. Simulation results.

May 2003 HPAC Engineering

$5,000

$10,000
$15,000
Annual electricity cost

$20,000

$25,000

F A N S

100
140

90
1.5-in. curveone fan (speed)

80
60

80

50

1.5-in. curvetwo fans (speed)

40
30

1.5-in. curveone fan (KW)

40

1.5-in. curvetwo fans (KW)

20

0
70 0
,0
00
80
,0
00
90
,0
0
10 0
0,
00
11 0
0,
00
12 0
0,
0
13 00
0,
00
14 0
0,
00
0

00

,0

,0

60

00

50

,0

,0

40

30

,0

,0
10

00

0
00

0
0

10
00

20

Cubic feet per minute

FIGURE 9. Optimal staging without static-pressure reset.

not be required). Real implementations


of terminal reset fall somewhere in
between these two extremes. Figure 11
shows the relationship between the
minimum duct static-pressure setpoint
and the optimal point for staging fans.
The Y-axis represents the optimal fanstaging points (up and down), while the
X-axis represents the minimum point
on the control-system curve.
The impact of static-pressure-setpoint
reset on both annual energy use and fan
selection was evaluated. To simulate reset,
a new load profile was developed by
replacing the monitored pressure with
the pressure from the system curve in

0-in. curve
one fan (speed)

140

0-in. curve
two fans (speed)

100
90
80

120
Fan-system KW

70

100

60
80

50
40

60
0-in. curve
one fan (KW)

40
20

30
20
0-in. curve
two fans (KW)

0
70 0
,0
00
80
,0
00
90
,0
0
10 0
0,
00
11 0
0,
00
12 0
0,
0
13 00
0,
00
14 0
0,
00
0

,0

00

10

60

00

,0
50

00

,0
40

00

,0
30

00

,0
20

,0

0
10

optimal point to stage from two fans to


one is when the speed drops below about
63 percent.
The optimal staging point for this fan
system would be very different if the
system curve ran through 0 in. at 0 cfm,
which would be the case if zone-based
reset of the static-pressure setpoint were
working perfectly. Figure 10 shows the
optimal speed to stage from one fan to
two is 55 percent. To stage from two fans
to one, it is about 35 percent.
Static-pressure-setpoint reset. Staticpressure-setpoint reset by box demand is
a control sequence that resets the staticpressure setpoint of the fan VSD based
on the measured demand of the terminal
units. At Site 1, the control system
provides a fixed setpoint of 1.5 in. wc,
regardless of demand. When direct digital
control (DDC) is used to control VAV
zones and the zone controller knows
the damper position (e.g., whether an
analog output or a floating point output
with position feedback is used for damper
control), this pressure setpoint could
be reset from 0 in. wc to 1.5 in. wc as
required to maintain the near-wide-open
position of the most open VAV damper.12
With perfect implementation of reset,
this would provide a system curve that
went from the design condition though
the point of 0 in. wc at 0 cfm (i.e., if no
zone required air, static pressure would

Speed, percent

60

Speed, percent

70

100

20

Fan-system KW

120

Figure 5 (perfect reset line) for the monitored airflow. These reset data were used
to compare the performance of the fans
evaluated in Figure 8. The results are
presented in Figure 12, which shows that
annual fan energy use can be cut by as
much as 50 percent if static-pressure reset
is implemented successfully (compare
Figure 12 with Figure 8). This corroborates the results reported by Hartmand
and others.
Figure 12 also shows that annual
energy ranking now follows the efficiency
ranking shown in Figure 4. This
is because a fan operating on a perfect system-reset curve has constant efficiency.
This is one of the reasons why static-pressure reset saves so much energynot
only is the fan doing less work (maintaining lower static pressures), it is doing it at
a higher efficiency.13
Another advantage of supply-pressuresetpoint reset is that it dramatically
reduces the time that fans operate in
surge. As can be seen in Figure 2, a system
curve that tails off horizontally (like the
actual system curve in Figure 5) will end
up in the surge region more often than
a reset curve (Figure 2) will. A perfect
system curve with reset starting at a point
to the right of surge will never end up in
the surge region.
The results in Figure 12 imply that,
in terms of energy cost, bigger fans are

Cubic feet per minute

FIGURE 10. Optimal staging with static-pressure reset.

HPAC Engineering May 2003

37

F A N S

80
75

Fan speed, percent

70

From one fan to two

65
60
55
50
45
40
From two fans to one

35
30
0.0

0.5
1.0
Duct static-pressure setpoint

1.5

FIGURE 11. Optimal fan-staging point vs. minimum duct static-pressure setpoint.

better for systems with supply-pressuresetpoint reset. Indeed, the approximately


$560 in annual energy savings from
selecting the 73-in. plenum fan rather
than the 66-in. plenum fan pays for
the $1,200 incremental cost increase
(Figure 4) with a simple payback of
less than three years; however, these
results need to be tempered with special
considerations. In addition to the first
cost of the fan, other first costs, including
the impacts on space and the electrical
service, should be considered. Also
considered should be the increased
risk that the fan will operate in surge
should perfect reset not occur. (The most
common cause of less-than-perfect
reset is a zone that is undersized, has
lower-than-design temperature setpoints,
or has consistently high loads, all of
which can result in a consistently high
demand for static pressure, even when
the rest of the system is at low load.)
The bigger the fan, the closer the design
point is to the surge region and the
greater the risk of operating in surge for a
less-than-perfect reset curve.
CONCLUSIONS

design guidelines and tools for fan


selection and system design. Current
simulation and design tools do not
adequately model the part-load performance of the components of a fan system
(i.e., fans, motors, drives, and belts). The
authors research project is addressing
some of these needs.
Clearly, static-pressure reset by zone
demand is a key to fan-system performance. Not only does it greatly reduce

energy use, it can eliminate the noise,


vibration, and instability issues associated
with fans operating in surge. Because
many DDC systems do not provide
feedback on box-damper position,
guidelines for applying this reset across
a range of control-system capabilities
are required.
Traditional fan-selection techniques
yield reasonably good results. Better
results are possible with the use of a
simulation tool to estimate annual fanload profile.
Although, in general, larger fans are
more efficient, bigger is not necessarily
better. The design-point-efficiency rankings in manufacturers software cannot
be extrapolated to annual-energy-cost
rankings, unless static pressure is reset
aggressively.
Optimum fan staging can be determined during design by simulating the
proposed fan system against a handful of
data points on the expected system curve.
The expected system curve and, thus, the
optimum staging point are greatly
affected by the success of static-pressure
reset. Of course, fan staging can be revisited after occupancy, based on the actual
success of static-pressure reset.

Base case (66 PL-A, no reset)


66 PL-A optimal staging

$20,474
$10,314

49 PL-A optimal staging

$16,611
$15,562

54 PL-A optimal staging


60 PL-A optimal staging

$10,806

73 PL-A optimal staging

$9,753

49 A-DWDI optimal staging

$9,720

54 A-DWDI optimal staging

$9,175

60 A-DWDI optimal staging

$8,917

66 A-DWDI optimal staging

$8,685

73 A-DWDI optimal staging

$8,519

It is clear that significant energy


$0
$5,000
$10,000
$15,000
savings can result from careful fan
Annual electricity cost
selection and control. It also is clear that
further work needs to be done to develop FIGURE 12. Simulation results with perfect static-pressure reset.

$20,000

$25,000

HPAC Engineering May 2003

39

F A N S

ACKNOWLEDGEMENTS

FOOTNOTES

The authors wish to acknowledge the


input and work of fellow research-team
members Erik Kolderup and Tianzhen
Hong of Eley Associates, Lynn Qualmann of SBW Consulting Inc., and
Roger Lippman of New Horizon Tech-

1) Data from www.calmac.org.


2) Information on the PIER program
is available at www.energy.ca.gov/pier.
3) Information on the Large HVAC
Integration project is available at www
.newbuildings.org/pier/index.html.

It is clear that significant energy savings can


result from careful fan selection and control.
nologies. They also wish to recognize
the contributions of members of their
technical advisory team. Lastly, they
extend special thanks to building
engineers Ravish Puri and Christopher
Newbury for putting up with the
authors intrusions on their building and
for their significant assistance with the
authors work.

4) The five sites monitored as part of


this research project represent a range of
occupancy types, building sizes, and
climates. Criteria for site selection
are presented in a paper by Kolderup,
Hydeman, Baker, and Qualmann from
the 2002 ACEEE Conference on Energy
Efficiency.a
5) Note that the pressure measured

Circle 154
40

May 2003 HPAC Engineering

was not exactly fan static pressure as


presented in the manufacturers fan curve
(Figure 2) because it included the
backdraft damper. Fan ratings are
described in ANSI/ASHRAE Standard
51-1999/AMCA Standard 210-1999,
Laboratory Methods for Testing Fans for
Aerodynamic Performance Rating.
6) A system curve is a theoretical line
that represents friction losses in a duct
system as a function of airflow. It is
typical to use a quadratic equation (i.e.,
pressure drops as the square of the flow);
however, in practice, this coefficient
is closer to 1.8. The curve in Figure 5
goes to 1.5 in. wc at 0 cfm, as this is the
control point maintained at the staticpressure sensor. With perfect demandbased pressure reset, the curve will go
through 0 in. wc at 0 cfm.
7) Fan surge is a condition of instability
that occurs under high-static-pressure
and low-flow conditions. Air is pushed

F A N S

out by the blades and subsequently


pulled back by the high pressure difference between the fan inlet and outlet.
Instead of flowing smoothly through the
fan, the air pulses in and out of the fan
blades. This unstable operation reduces
controllability, increases wear, and can
cause severe acoustical and vibration
problems.
8) The relative cost and efficiency of
fan types can be seen using manufacturers selection programs.
9) This low-frequency noise typically
can be mitigated through the reduction
of system static pressure and the application of lined ductwork.
10) A number of issues associated with
isolating fans in parallel are beyond the
scope of this article. These will be covered
in detail in the projects analysis report
and design guidelines.
11) Fan-system efficiency includes the
fan, motor, drive (belts), and VSD. For a

given operating point, the efficiency of


the motor, drive, and VSD increases as
the load increases. Unless fan efficiency
improves appreciably to offset these
losses, operating one fan generally is
better than operating two. Two fans
running in parallel often operate in surge,
with a fixed pressure setpoint, unlike a
single fan.
12) When box-damper position is not
known, a trim-and-respond algorithm
can be employed for a similar effect.
A promising non-traditional technique
is to directly control fan speed by zone
demand without the use of PID loops.b,c
This technique reduces fan hunting at
low loads.
13) This is true only when fan speed
is controlled with a VSD. If a fan is
controlled using inlet guide vanes or rides
its curve, the savings will be significantly
less, and the fan may enter the surge
region.

REFERENCES

a) Kolderup, E., Hydeman, M., Baker,


M., & Qualmann, R.L. (2002, August).
Measured performance and design guidelines for large commercial HVAC systems.
ACEEE Conference on Energy Efficiency.
b) Hartman, T. (2001, December).
Ultra-efficient cooling with demandbased control. HPAC Engineering, pp.
29-32, 34, 35.
c) Hartman, T. (1995, June). Global
optimization strategies for high-performance controls. Paper presented at ASHRAE
Annual Meeting, San Diego, CA.
d) Hartman, T. (1993, January). Terminal regulated air volume (TRAV)
systems. Paper presented at ASHRAE
Winter Meeting, Chicago, IL.
For HPAC Engineering feature articles
dating back to January 1992, visit
www.hpac.com.

Circle 189
HPAC Engineering May 2003

41

Das könnte Ihnen auch gefallen