Sei sulla pagina 1di 10

Fluids and Solids Handling

Analyzing the Performance


of Pump Networks
Part 2: Improving Pump Efficiency
Jimmy D. Kumana Best energy efficiency practices can
Manuel R. Suarez
Kumana & Associates reduce pump operating costs significantly.
This article reviews the basics and presents a new
way to account for variation in pump efficiency.

P
art 1 (1) of this three-part series reviewed such basics in the piping system, or increase the efficiency by operating
as derivation of the system curve from plant data, the closer to the pump’s best efficiency point (BEP). The two
construction of composite performance curves for most common methods of flowrate control are throttling
pump networks, and how to operate and control pumps in (Figure 1) and recirculation (Figure 2). This control comes at
parallel to avoid surging and cavitation. This second article a cost, however. With throttling, the cost is a higher pressure
discusses how to use variable-frequency drives (VFDs), a drop, while for recirculation, it is a higher flowrate; both of
subset of variable-speed drives (VSDs), and other energy- these require extra power, which ends up heating the fluid.
efficiency measures such as load management (2) to reduce
pump operating costs. It also presents a new way to account Improving energy efficiency
for variation in pump efficiency with changing speed when On average, approximately 40% of the energy supplied
the static head is significant, and uses a natural gas liquid to centrifugal pumps in the CPI is wasted as unrecoverable
(NGL) pipeline example to explain how the concept works. low-grade heat (Table 2). It is significant that only 6% of
Energy requirements (usually electric power) account for energy losses are caused by operational issues, while 34%
nearly 90% of the cost of operating pumps in the chemical are built in at the design stage — that is, they are attribut-
process industries (CPI). Assuming 8,400 hr/yr of continu- able to decisions made during design and the increasingly
ous operation and 72% pump efficiency, the 15-yr cost of common practice of fast-tracking this critical step. When
operating a typical CPI pump would be as shown in Table 1. inadequate time is budgeted for proper engineering analysis,
Increasing energy efficiency clearly offers the greatest design engineers are forced to compensate by oversizing
potential to significantly reduce overall operating costs and pumps, which ensures inefficient operation and can be very
improve profits. Brake horsepower (BHP), the actual horse- costly to undo after production begins.
power delivered to the pump shaft, is: Fortunately, roughly half of the design-related losses

Table 1. Energy accounts for the


largest share of industrial pump costs.
where Q is the flowrate, ΔP is the pressure drop, and η is the Capital Costs (installed, all inclusive) 8.5%
fractional pump efficiency.
Maintenance and Repair, including parts and labor 3.5%
Power consumption can be reduced in three ways —
Energy (electricity at 6¢/kWh) 88.0%
reduce the flow through the pump, reduce the pressure drop

32  www.aiche.org/cep  February 2018  CEP Copyright © 2018 American Institute of Chemical Engineers (AIChE)
Net Flow
to Process Table 2. These five causes
Minimum Flow Bypass Line account for 40% of wasted energy in pumping.
1. Low efficiency due to wrong pump choice 4%
Heat
Exchange 2. Poor installation or maintenance 3%
Storage
FC FC FT 3. Low pump efficiency due to wear 3%
Tank
4. Poor system design (piping, valves, etc.) 10%
FT 5. Poor system control strategy 20%
2,000 40%
Pump Curve
1,600 System w/o CV Table 3. Pumps can be powered by various types of drivers.
Operating Point Driver Type Energy Source
1,200
Head, ft liquid

∆P Loss in Steam Turbine High-pressure steam


800 Control Valve System Dynamic Head exhausting to low pressure
(frictional ∆P) at Target
Flowrate Gas Turbine Fuel
400 Gas Expander Pressure recovery from
Static Head high-presure gas
0
0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 Internal Combustion Engine Fuel
Flow, gpm
Fixed-Speed Motor with Belts Electricity
p Figure 1. In throttling flow control, additional power must be used to Fixed-Speed Motor with Gears Electricity
offset pressure drop.
Fixed-Speed Motor with Electricity
Net Flow Variable-Frequency Drive
to Process
Minimum Flow Bypass Line
the control valve (CV), or needlessly increases recirculation.
FC
Heat
The most effective way to eliminate power loss in indi-
Storage Exchanger vidual pumps is to operate each pump at the speed that exactly
Tank
FY FT matches the minimum process flow and head requirements,
represented by the system curve. This can only be done by
FT varying operating speed. Several options are available for
2,000
equipping a pump with variable-speed capability (Table 3).
Pump Curve
1,600 System w/o CV Compensating for oversized pumps
Required Head A cheap, but not very efficient, way to compensate for
Head, ft liquid

1,200 an oversized pump is to trim its impeller diameter, thereby


reducing delivered head and the control-valve pressure drop.
Operating Point
800 Care must be taken to retain sufficient flexibility to control
Required ∆ (flow)
Flow flow with a throttling valve, by recirculating excess flow
400
Actual Flow back to the suction tank, or by some other means. The cost is
some loss of flow and head capacity.
0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000 The pump’s expected flow and head distribution profile
Flow, gpm
determines whether the best results would be achieved by
p Figure 2. Under recirculation flow control, it is necessary to make up for trimming the impeller or by varying the speed. If head varia-
increased flow in the bypass line. tion is ≤ 5% of design and flow variability is ≤ 10% about
the mean, impeller trimming may be best option. When
can be reduced by retrofits and a relatively simple revamp pump flow and head both vary widely and unpredictably,
of process controls. It is important to recognize that almost however, it is usually best to vary the pump’s operating
no pumping system, whether a single pump or a complex speed continuously.
network, regularly operates at its full original design load. The economic feasibility of retrofitting or replacing an
As discussed in Part 1, part-load operation (also known existing fixed-speed motor with one that has variable-speed
as operation at turndown) imposes a huge energy penalty capability depends mainly on the pump’s flowrate distribu-
(about 20–30% of total power consumption), most of it in tion profile and the local electric utility’s power rates. As

Copyright © 2018 American Institute of Chemical Engineers (AIChE) CEP  February 2018  www.aiche.org/cep  33
Fluids and Solids Handling

FC t Figure 3. Until recently, VFDs were quite expensive, but over the
Variable-frequency
drives (VFDs) must
past 10–15 years, advances in solid-state electronics have
be maintained care- enabled sharp price reductions, especially in the smaller
fully to protect their sizes. The combination of pump and VFD may cost more
sensitive electronic initially than the pump-plus-CV option, but it has lower
components. Outdoor
Rectifier
DC Link
Inverter

M AC installations require lifecycle costs attributable to savings on energy and mainte-


Power protective enclo- nance (Table 4). An integral unit, suitable for clean environ-
Transformer Supply sures and may need ments only, should be considered for new indoor installa-
additional lightning tions or pump replacements of 10 hp or less.
Variable-Frequency Drive protection.
It should be emphasized that, at very low operating
speeds, numerous mechanical and electrical issues can
a rule of thumb, variable-speed operation is most feasible adversely affect VFD performance. Successful VFD applica-
when the average flow is less than 60% of the design value. tion requires a team effort by the electrical and instrumen-
When the pump drive is a motor, VFDs are usually the tation, mechanical, and process engineering departments.
most cost-effective retrofit option. For large new installa- Leaving implementation to a single engineer without cross-
tions (>5,000 hp), a direct-coupled steam or gas turbine functional expertise is a recipe for failure.
drive may be a better choice. If the goal is energy savings
only (rather than superior process control), depending on the Using affinity laws to estimate performance
system flow profile, a two-speed motor can cost less than a The affinity (or fan) laws (Table 5) can be used to
VFD, offering comparable savings at lower capital cost. estimate pump performance at off-design conditions. These
VFDs are generally more desirable than other options, relationships are based on the assumptions that pump
because they also enable exceptionally precise flow control, efficiency is independent of speed N (which is mostly true
with minimal lag time between signal input and actuation. for speeds greater than 50% of the maximum speed) and
A VFD should be thought of as a final control element that impeller diameter D (mostly true for diameters greater than
replaces the function of the control valve without the accom- 80% of the maximum diameter, Dmax), and that the orginal
panying pressure drop. This capability is very important pump design was close to the BEP. As the impeller diameter
for the pumping of slurries. The energy cost savings are is reduced below 0.8Dmax in the same casing, efficiency at
sometimes almost an incidental benefit. Nevertheless, energy the same speed falls off rapidly.
savings is VFDs’ most easily quantifiable benefit, and is Engineers can use the affinity laws with reasonable
commonly used to assess economic payback. confidence to estimate the performance of a pump when
The major drawback of VFDs is that they are solid-state the original impeller is trimmed <15%, e.g., when a 12-in.
electronic devices that must be protected from harsh envi- impeller diameter is trimmed to 11 in. or 10.5 in. (Table 6
ronmental conditions and exposure to dust, heat, moisture, and Figure 4), based on the pump curve provided by the
corrosive gases, etc. If the drives cannot be installed indoors, manufacturer at the time of purchase. Manufacturers’ pump
e.g., inside the motor control center or control room, they curves should be kept safe and legible, along with the origi-
must be enclosed in protective cabinets. Lightning protection nal purchase order documents, in the company library. If any
may also be required for outdoor installations. information is missing, difficulties are likely to arise later,
The principal components of a VFD are the rectifier, the
DC link, and the inverter (Figure 3). Depending on the volt-
Table 5. Use the affinity laws to
age, a transformer may also be required. evaluate pump performance at off-design conditions.

Table 4. A pump with a variable-frequency drive has Constant Constant


lower lifecycle costs than a pump with a control valve. Impeller Diameter Impeller Speed

Without With Capacity


VFD VFD
Pump + Motor Purchase Cost 5 6 Head
Variable-Frequency Drive for Motor 0 3
Installation (piping, valves, etc.) 15 18 Horsepower
Operating Power 55 37
Maintenance (parts and labor) 25 16
Note: Subscript 1 denotes design conditions and subscript 2 denotes the
100 80 new condition.

34  www.aiche.org/cep  February 2018  CEP Copyright © 2018 American Institute of Chemical Engineers (AIChE)
1,800
Table 6. Use the affinity laws to estimate
pump curves for different impeller diameters. 1,600 12-in. dia.
1,400 11-in. dia.
D = 12 in. D = 11 in. D = 10.5 in. 10.5-in. dia.

Head, ft liquid
1,200
Flow, Head, Flow, Head, Flow, Head, 1,000
gpm ft gpm ft gpm ft 800
0 1,500 0 1,260 0 965 600
200 1,496 183 1,257 160 963 400
200
500 1,491 458 1,253 401 959
0
1,000 1,470 917 1,235 802 946 0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000
Flow, gpm
1,500 1,380 1,375 1,160 1,203 888
p Figure 4. The affinity laws allow engineers to quantify possible energy
2,000 1,240 1,833 1,042 1,604 798
efficiency improvements that would be achieved by trimming impellers.
2,500 1,050 2,292 882 2,005 676
3,000 810 2,750 681 2,406 521 2,000 100
1,800 ηB ηA 90
3,500 520 3,208 437 2,807 335
1,600 80
4,000 180 3,667 151 3,208 116 1,400 C 70

Efficiency, %
Head, ft liquid A
1,200 60
especially if the manufacturer has changed the model design 1,000 50
or your original contact at the vendor has left the company. 800 40
B
The use of VFDs enables process engineers to improve 600 30
pump energy efficiency and to achieve a control response 400 20
superior to that provided by other means of flow control. 200 10

When using the affinity laws, it is crucial to remember 0


0 200 400 600 800 1,000 1,200 1,400 1,600
0

that they are only approximations based on the assumption Flow, gpm
of constant pump hydraulic efficiency as speed is reduced,
Pump Curve, 3,600 rpm Pump Curve w/ VFD
which is true only if the static head of the system is negli- System w/o CV Full-Speed Efficiency
gible compared to dynamic head. That is often not the case. Iso-Efficiency Line
When the static head is significant, say >10% of the total,
an efficiency correction must be applied. Over some ranges p Figure 5. The optimum efficiency at reduced speed can be determined
graphically. The first step is determining where the full-speed pump
of flow, the reduced-speed efficiency, η2, will be higher than characteristic curve and the system curve intersect (Point A).
the full-speed efficiency, η1, while over other ranges it may be
lower, depending on whether the full-speed efficiency at the pump curve would intersect Point B at a speed of 2,785 rpm
design condition is to the left or the right of the pump’s BEP. (which is most easily determined graphically, by visual
Moreover, when speed is reduced below 50% of the inspection, using trial-and-error iteration). The challenge
full motor speed, the efficiency at the lower speed must be now is to find the pump efficiency at Point B.
further reduced: According to the affinity laws, the iso-efficiency line
passes through the origin and is a parabola with an exponent
of 2. Draw this line passing through Point B and find its
intersection with the full-speed pump characteristic curve,
where the subscript 1 denotes design (full-speed) conditions. Point C. If the pump curve can be adequately modeled as
The correct method for estimating pump efficiency at a quadratic equation (using parameters a, b, and c in Eq. 3
reduced speed is as follows. Consider the performance and obtained from the vendor’s test data by regression analysis),
system curves shown in Figure 5. Point A is the intersection Point C can be derived algebraically by simultaneously
of the full-speed (3,600 rpm) pump characteristic curve and solving Eq. 3 for the full-speed pump curve and Eq. 4 for the
the system curve when the CV is in the fully open position, iso-efficiency line:
flow is 1,400 gpm, head is 1,383 ft, and efficiency is 82.6%.
Now, let’s assume the actual desired flow is 980 gpm
(Point B), and the required system head is only 870 ft. The
head delivered by a fixed-speed pump would be 1,590 ft
and the CV loss would be 1,590 – 870 = 720 ft. If the pump
were equipped with a VFD, however, the reduced-speed Subtract one from the other to eliminate the system head,

Copyright © 2018 American Institute of Chemical Engineers (AIChE) CEP  February 2018  www.aiche.org/cep  35
Fluids and Solids Handling

Table 7. Calculations for estimating power savings from VFD use.


System Power Consumption, Pump Curve at
Head Full-Speed Required excluding motor loss, hp 2,785 rpm
Flow (Q), Head (H), w/o CV, Efficiency, Speed Efficiency
Throttling CV VFD Savings Q, gpm H, ft
gpm ft of oil ft % (N), rpm at N, rpm
1 1,702 300 0.0 0 0 0 0 0 1 1,019
140 1,718 317 16.2 1,540 35.0 322 28 294 108 1,028
280 1,723 360 30.6 1,670 57.5 342 39 302 217 1,031
420 1,717 424 43.3 1,830 70.1 361 57 305 325 1,028
560 1,701 508 54.2 2,050 76.7 381 83 298 433 1,018
700 1,675 611 63.4 2,270 80.0 401 120 282 542 1,002
840 1,637 732 70.7 2,520 81.6 422 168 253 650 980
980 1,590 870 76.3 2,785 82.3 443 232 211 758 951
1,120 1,531 1,025 80.2 3,050 82.6 464 311 153 866 916
1,260 1,462 1,196 82.3 3,320 82.7 486 408 78 975 875
1,400 1,383 1,383 82.6 3,600 82.6 508 524 –16 1,083 828
Note: Speeds were determined visually on Figure 5 and entered in Table 4 manually.

H, and obtain a quadratic equation in Q that can be solved using the method described here. Table 7 summarizes the
algebraically for the flow QC at Point C, from which the effi- calculation results for this example.
ciency ηC at Point C (which is the same as ηB) can be readily Note that the coordinates of Point B are (980, 870) on
calculated. the system curve and (975, 875) on the reduced-speed pump
For the data depicted in Figure 5, QC = 1,269 gpm, curve. Although these are not exact matches, they are close
HC = 1,458 ft, and ηC = 82.3% enough (<1% error) for engineering and decision-making
In this example, the pump hydraulic efficiency at purposes — and far more accurate than the raw unrecon-
reduced speed, ηB, is only slightly lower than the full-speed ciled plant data that most companies use to monitor pump
efficiency, ηA, but is significantly higher than the full-speed performance.
efficiency of 76.3% at 980 gpm. This may not always be the Also, keep in mind that the VFD itself incurs some
case, however, and will depend on the location of Point A efficiency losses (Figure 6), which must be taken into
relative to the BEP. account when evaluating energy savings relative to fixed-
Most properly selected pumps will have normal oper- speed operation.
ating conditions close to the BEP. Thus, the efficiency
correction at Point B vs. Point A tends to be small and can Estimating VFD energy savings for single pumps
be safely ignored in the first approximation (e.g., for initial For individual pumps, electronic spreadsheets offer
screening purposes). However, it should definitely be taken an easy way to estimate the energy savings that can be
into account when making the final investment decision, obtained with VFDs. Some manufacturers and engineering,
procurement, and construction (EPC) consulting firms may
100
provide such spreadsheet templates for preliminary estimat-
90
80 ing (e.g., Figure 7), but they should be used with caution
Load, % Efficiency, % Load, % Efficiency, %
70 because for simplicity they tend to use embedded generic
Efficiency, %

100 97.0 11.2 90.0


60 80 96.7 8.2 85.0 pump characteristic and efficiency curves. In addition, the
50 65 96.3 6.4 80.0
40 50 95.7 3.5 70.0
freely distributed templates do not adjust for efficiency
30 35 95.2 2.1 60.0 variation with speed, which can introduce large errors in
20 25
16
94.7
93.0
1.6
0
55.0
0
applications in which the static head is significant. Never-
10 theless, they are a good initial screening tool to determine
0
0 10 20 30 40 50 60 70 80 90 100 whether the application warrants further detailed analysis.
Load, % of Design Capacity The methodology for analyzing the performance of
p Figure 6. Running at variable speed during part-load operation involves pumps with VFDs is now illustrated by an example. Con-
some energy losses, which must be accounted for when comparing energy sider the NGL pipeline booster station shown schematically
efficiency with that of fixed-speed operation. in Figure 8. Two identical pumps are installed in parallel,

36  www.aiche.org/cep  February 2018  CEP Copyright © 2018 American Institute of Chemical Engineers (AIChE)
Table 8. Pump and system data for
natural gas liquids (NGL) pipeline booster station.
Hmax (ft) 2,300 Pump maximum head
(at zero flow)
Hn (ft) 1,840 Head (pipe and valve friction
+ static) at intersection of
pump and system curves
Hs (ft) 100 Static head of the system
Hv (ft) 2 Valve friction head at
full-open position (~0)
P (hp) 10,000 Pump power consumption
Qn (gpm) 17,500 Pump flow at full-speed
intersection point
ρ (lb/ft 3 ) 44.8 Liquid density
(ρwater = 62.2 @ 80ºF)
ηm (%) 96 Nominal efficiency of motor
ηp (%) 75.0 Pump efficiency at full-speed
intersection point
ηVSD (%) 98 Nominal efficiency of
variable-speed drive
Time (hr) 8,400 Total operating time per year
Cost (US$) 80.0 Cost of electrical energy
(per MWh)

Table 9. Coefficients for calculating


head or efficiency from Y = a + bQ + cQ2.
Y = head, ft Y = efficiency, %
a 2,300 0
b 0 0.0123

p Figure 7. Commercial software tools use generic pump curves and a generic c –2.0 E–06 –4.57 E–07
efficiency curve. Although they are extremely useful for making initial estimates, they
should never be used as a basis for making investment decisions.

Existing Control System


curve are obtained from the pump manufacturer’s water test
data (which should have been provided when the equipment
was purchased, and which are presumably saved in the com-
Fl pany library). In this example, the data are sufficiently well
PC
correlated by the quadratic equations (Eqs. 3 and 4), with H
in ft of liquid and Q in gpm.
Step 1. From Table 9 and the data in Table 8, calculate
the dynamic head for pipe friction (Hf) at the intersection of
M the system curve and the pump curve: Hf = Hn – Hs – Hv =
1,840 – 100 – 2 = 1,738 ft. Now calculate system head, H,
M for the full range of flows from 0% to 100% divided into
convenient increments, say 10% each:
p Figure 8. At the natural gas liquid (NGL) pipeline in the example, two
pumps are installed in parallel, one in service and one as a spare.

but only one operates at any given time; the other is a spare,
intended for use only in emergencies. Table 8 lists the design where Hs is the static head, ρ is the specific gravity or den-
flowrate and other data. sity in consistent units, and the subscript d refers to an actual
The full-speed head-flow curve and the efficiency-flow operating data point at the desired normal condition (or

Copyright © 2018 American Institute of Chemical Engineers (AIChE) CEP  February 2018  www.aiche.org/cep  37
Fluids and Solids Handling

Table 10. Estimation of power savings at selected flowrates over the entire operating range.
Pump Curve Full-Speed Required Power Consumption,
System Efficiency, Speed (N), Flow at N, Efficiency
Flow, gpm Head, ft Head, ft % rpm gpm at N, % Throttling VFD Savings
1 2,300 100 0.0 839 4,080 42.5 0 0 0
1,750 2,295 130 20.1 955 4,643 47.2 3,785 93 3,692
3,500 2,282 198 37.4 1,181 5,739 55.5 4,044 241 3,803
5,250 2,259 301 51.9 1,456 7,078 64.1 4,327 476 3,851
7,000 2,226 436 63.6 1,752 8,519 71.5 4,641 825 3,816
8,750 2,185 601 72.5 2,058 10,002 77.2 4,994 1,317 3,678
10,500 2,134 795 78.6 2,366 11,503 80.9 5,400 1,994 3,406
12,250 2,075 1,017 81.9 2,676 13,008 82.5 5,877 2,916 2,961
14,000 2,006 1,265 82.4 2,985 14,511 82.1 6,454 4,169 2,285
15,750 1,927 1,540 80.1 3,293 16,009 79.6 7,178 5,886 1,292
17,500 1,840 1,840 75.0 3,600 17,500 75.1 8,131 8,284 0
19,250 1,743 2,165 67.1 3,905 18,984 68.6

design specification if plant data are not available). Notice


that Eq. 5 is not a quadratic, so an algebraic solution is not
possible for the intersection point of the pump curves.
Step 2. For each value of flow, use the affinity laws to cal- where the subscript o denotes the base case (the spreadsheet
culate the pump speed needed to deliver the required system tool, Figure 7, uses n to denote the base case).
head. At the intersection of the full-speed pump curve and the Step 3a. Use the procedure described earlier (with refer-
system curve, when the CV is fully open, Qo = 17,500 gpm, ence to Figure 5) to calculate the iso-efficiency flow, Qiso, at
No = 3,600 gpm, and Ho = 1,840 ft. For each intermediate Point C and the reduced speed N2. Then use the correlation
value of Q and the system head H, the required speed is: coefficients obtained from vendor data (Table 9) to calculate
the hydraulic efficiency of the pump at this flow:

Table 11. Sample of historical NGL flow data needed for each pump.
Flow, gpm % of Design One Step 3b. Alternatively, if the pump
Pump curve cannot be accurately modeled as a
Date 26FI472A 26FI472B Pump A Pump B Pumps
Only
A+B quadratic equation, approximate the cor-
1-Jan-03 0 7,084 0% 40% 20% 40% rected efficiency η2 (3):
2-Jan-03 0 0 0% 0% 0% 0%
3-Jan-03 0 6,815 0% 39% 19% 39%
4-Jan-03 0 0 0% 0% 0% 0% Step 4. For every intermediate value
5-Jan-03 0 0 0% 0% 0% 0% of flow, calculate power consumption; use
6-Jan-03 0 6,670 0% 38% 19% 38% pump head for throttling control (with a
7-Jan-03 0 6,582 0% 38% 19% 38% CV) and system head for the VFD case.
Step 5. Summarize the results, as in
4-Sep-03 5,663 5,667 32% 32% 32% 65%
Table 10.
5-Sep-03 5,744 5,746 33% 33% 33% 65% Step 6. Determine the fluid flow
6-Sep-03 5,437 5,442 31% 31% 31% 62% profile, preferably using actual historical
7-Sep-03 5,952 5,952 34% 34% 34% 68% data over a 1-yr period in increments of
8-Sep-03 5,765 5,768 33% 33% 33% 66% one day (Table 11 and Figure 9), or for bet-
ter accuracy, increments of one shift.
9-Sep-03 5,477 5,479 31% 31% 31% 63%
When approximation is acceptable,
Average 6,589 7,239 37.7% 41.4% 39.5% 79.0%
the amount of effort can be reduced

38  www.aiche.org/cep  February 2018  CEP Copyright © 2018 American Institute of Chemical Engineers (AIChE)
120% 40

35
100%
30
Flow, % of Design

Frequency, %
80%
25

60% 20

15
40%
10
20% 5

0% 0
0 10 20 30 40 50 60 70 80 90 100 110 120
ec an ar y n ug Oc
t
-D -J -M 9-Ma -Ju -A 6- % of Design Flowrate for a Single Pump
10 29 20 28 17

p Figure 9. Use historical data from plant records to calculate the flow p Figure 10. A histogram of the flow profile can be useful for calculating
profile. For a new process or facility, use your best estimates. energy savings on a per-day or per-shift basis.

by creating a histogram of the flow profile for 10 repre- Efficiency and Renewable Energy website (4). In addition,
sentative increments (Figure 10). Although Microsoft’s the calculation of power savings must account for the VFD
Excel now features a built-in histogram-calculation tool, power consumption as an efficiency factor (Figure 6). The
it may be quicker to construct a histogram using the final design employed the cold-switching control scheme
CONDITIONAL COUNT functions (Table 12). shown in Figure 11 (next page).
If the required historical data are not available in the The analysis for a backpressure steam turbine (BPST)
company logs, e.g., for a planned new design, then develop as the VSD (instead of an electric motor with a VFD)
your best estimate of the flow profile. A word of warning, would be similar, except that steam turbine efficiency and
though: Actual plant data never follow the symmetrical net steam costs would be used instead of motor/VFD effi-
normal (Gaussian) distribution that is often seen in some ciency and purchased power cost.
academic simulations, which apply only to random natural Note the deliberate and specific reference to a back­
occurrences; real distributions are always
Table 12. CONDITIONAL COUNT functions can be used
asymmetrical because they are not random.
to construct a histrogram of the flow profile.
Step 7. From the frequency data, esti-
Excel Histogram Function Alternative Development of the Histogram
mate the number of hours per year for each
flow increment and calculate the actual Frequency,
power consumption and cost savings for the Cum Bin Frequency Values Range One Pump
year (Table 13, next page). 0 0 0 Low High Number % Cum
The estimated savings calculated by 8 0.1 8 0 0 10 8 3.2 3.2
this method are $1,058,000 per year. The 8 0.2 0 0.1 10 20 0 0.0 3.2
screening software tool calculated savings
8 0.3 0 0.2 20 30 0 0.0 3.2
of $1,283,539 based on the same data set
26 0.4 18 0.3 30 40 18 7.1 10.3
(Figure 7). That is a 20% difference, which
could be enough to sway the investment 32 0.5 6 0.4 40 50 6 2.4 12.7
decision. 32 0.6 0 0.5 50 60 0 0.0 12.7
This example illustrates the importance 44 0.7 12 0.6 60 70 12 4.8 17.5
of not making investment decisions based 90 0.8 46 0.7 70 80 46 18.3 35.7
on simplistic calculations that are crude
154 0.9 64 0.8 80 90 64 25.4 61.1
approximations at best (e.g., generic pump
and efficiency curves). Often, simplicity 244 1 90 0.9 90 100 90 35.7 96.8
and ease of use are purchased at the cost of 252 1.1 8 1 100 110 8 3.2 100.0
lower accuracy and reliability. 252 1.2 0 1.1 110 120 0 3.2 100.0
Part-load motor efficiencies can be 252 1.3 0 1.2 120 130 0 0.0 100.0
estimated using MotorMaster 4.1, a free
More 0 1.3 130 140 0 0.0 100.0
software tool that can be downloaded from
the U.S. Dept. of Energy’s Office of Energy 252 100

Copyright © 2018 American Institute of Chemical Engineers (AIChE) CEP  February 2018  www.aiche.org/cep  39
Fluids and Solids Handling

Table 13. Final estimation of cost savings based on flow profile. boilers, furnaces, dryers, etc.
Pump Curve Power Consumption Running Time BPST drives are often a more eco-
nomical option for motors of 5 MW or
Flow, Savings,
gpm Head, ft Throttling VFD Savings % hr/yr MWh larger, as long as the plant can make effec-
tive use of the exhaust steam from process
1 2,300 0 0 0 3.2 267 0
heating, and the load is variable. If the
1,750 2,295 3,785 93 3,692 0.0 0 0 motor load is relatively steady and larger
3,500 2,282 4,044 241 3,803 0.0 0 0 than 10 MW, then a gas turbine may be a
5,250 2,259 4,327 476 3,851 7.1 600 1724 more viable option, but only if the exhaust
7,000 2,226 4,641 825 3,816 2.4 200 569 gas from the gas turbine either is used to
generate useful steam in an HRSG or is
8,750 2,185 4,994 1,317 3,678 0.0 0 0
close-coupled to a furnace or boiler.
10,500 2,134 5,400 1,994 3,406 4.8 400 1,016 If the local electric utility offers
12,250 2,075 5,877 2,916 2,961 18.3 1,533 3,387 time-of-use rate schedules, using dual
14,000 2,006 6,454 4,169 2,285 25.4 2,133 3,637 drives on large pumps and compressors
15,750 1,927 7,178 5,886 1,292 35.7 3,000 2,891 (>10 MW) might minimize purchased
energy costs by switching between
17,500 1,840 8,131 8,284 0 3.2 267 0
electrical and thermal energy. Strictly
19,250 1,743 100.0 8,400 13,224
speaking, this is not necessarily an energy
$/MWh 80 efficiency measure, but rather a cost
$/yr 1,058,000 optimization measure similar to arbitrage.

pressure steam turbine. For reasons that are beyond the Estimating energy savings potential
scope of this article, condensing turbines should never be of VFDs for multiple-pump networks
used in any process plant, except in a few rare situations: A network of pumps in series/parallel configuration has
where steam is generated using excess high-temperature an additional degree of freedom for optimizing the load allo-
exothermic heat of reaction from the process or by burning cation among individual pumps. Energy savings are often
a fuel with significant negative cost; in critical duties such comparable to those that can be obtained using VFDs but at
as an emergency boiler feedwater pump; or when the grid substantially lower capital cost.
power supply is unreliable. Several references on pinch This method is known as load management (2), a
analysis (5–7) explain the thermodynamics involved. powerful technique that is woefully underutilized in the
industry. Often, the energy savings obtained through load
Other drive systems management alone are so significant that the incremental
Steam and gas turbines are alternatives to large motors as energy savings obtained by adding VFDs to the network
pump drivers. However, their economics are favorable only cannot justify their incremental capital cost. The applica-
if they can be integrated with the process in cogeneration tion of load management to complex pump networks, both
mode, with the exhaust heat being effectively used in the with and without VFDs, will be covered in Part 3 of this
process to minimize overall fuel consumption in utility-fired article series.
Public drinking-water distribution systems are an excel-
Proposed Controls Revamp (with VFD) lent example of complex pump networks with multiple
sources, multiple delivery points, and intermediate booster
stations (8).
FT
PC A very important point to note is that in a bank of
Switch VFD
parallel pumps (whether identical or not) with the fluid in a
common header going to a single destination, it is not neces-
sary to have a VFD for each pump. A single VFD, after it
M
has been disconnected from the motor that it was controlling
previously, can be switched to any one of the other motors,
M while the rest of the pumps run at fixed speed. This is
known as cold switching, and the lower capital cost greatly
p Figure 11. This cold-switching configuration can be used to reduce improves the return on investment.
the capital costs of the NGL pipeline booster station in the example. However, if the common discharge header branches out,

40  www.aiche.org/cep  February 2018  CEP Copyright © 2018 American Institute of Chemical Engineers (AIChE)
with each branch going to a different location with differ-
ent discharge head requirements, control valves would be Nomenclature
required in each branch line, and the economic benefits a, b, c = coefficients for quadratic equations
would be reduced accordingly. BHP = pump power consumption or brake horsepower
D = impeller diameter
Closing thoughts g = gravitational constant
Variable-frequency drives can be retrofit to fixed-speed H = head of liquid, system head
Hf = dynamic head due to friction
motors (the default case) to provide variable-speed capabil-
H n = head (friction + static) at intersection of pump
ity. They offer a degree of freedom in operation that can be and system curves
exploited to improve overall pumping system efficiency and Hs = static head
reduce energy costs. With the recent advent of powerful and Hv = valve friction head of fully open control valve
cheap solid-state technology and advanced control strategies, N = impeller speed
capital costs for VFDs have dropped substantially, making P = hydraulic power consumption of pump (before
them more economically attractive and increasingly popular motor or VSD efficiency adjustments)
Q = flowrate
with technically progressive companies worldwide. CEP Qiso = hypothetical pump flow at intersection of
full-speed pump curve and reduced-speed
Literature Cited iso-efficiency affinity curve at zero static head
(Point C in Figure 5)
1. Kumana, J., and Suarez, M., “Analyzing the Performance of V = velocity
Pump Networks, Part 1: Basic Theory,” Chemical Engineering Greek Letters
Progress, 114 (1), pp. 34–42 (2018)
ΔP = pressure drop
2. Kumana, J., and A. Aseeri, “Electrical Power Savings in Pump ρ = specific gravity or density (in consistent units)
and Compressor Networks via Load Management,” Proceedings η = efficiency
of 27th Industrial Energy Technology Conference, New Orleans,
ηm = nominal efficiency of motor (typically 93–98%)
LA (May 2005); edited version republished in Saudi Aramco
ηp = pump hydraulic efficiency at full-speed intersec-
Journal of Technology, pp 39–43 (Fall 2005).
tion point (control valve fully open)
3. Coelho, B., and A. Andrade-Campos, “A New Approach for the
ηVSD = nominal efficiency of variable-speed drive (from
Prediction of Speed-Adjusted Pump Efficiency Curves,” Journal
Figure 6)
of Hydraulic Research, 54 (5), pp. 586–593 (2016).
μ = viscosity
4. U.S. Dept. of Energy, “MotorMaster 4.1,” DOE Office of
Subscripts
Energy Efficiency and Renewable Energy, www.energy.gov/eere/
A, B, C = at Point A, B, or C on the head-flow curves
amo/downloads/motormaster-tool.
(Figure 5)
5. Linnhoff, B., et al., “Process Integration for the Efficient Use
d = actual operating data point at the desired normal
of Energy,” Chapter 2, Institution of Chemical Engineers,
(or design) condition
Rugby, U.K. (1994).
max = maximum
6. Kumana, J., and A. Al-Qahtani, “Optimization of Process
n = normal
Topology using Pinch Analysis,” Saudi Aramco Journal of Tech-
o = base case, whether design or normal
nology, pp. 13–23 (Winter 2004).
1 = condition 1 (either full-speed, design, or inlet,
7. Smith, R., “Chemical Process Design and Integration,” John
depending on context)
Wiley & Sons, Ltd., Chichester, U.K., (2005).
2 = condition 2 (reduced-speed, actual, or outlet,
8. Sarbu, I., “A Study of Energy Optimization of Urban Water
depending on context)
Distribution Systems using Potential Elements,” Water, (8) 12,
pp. 593–ff, doi:10.3390/w8120593 (Dec 2016).

JIMMY D. KUMANA (Houston, TX; Email: jkumana@aol.com) has more than MANUEL R. SUAREZ (Email: manuel.r.suarez@gmail.com) has more than 30
35 years of experience working for both manufacturing and engineer- years of experience in process engineering, process documentation,
ing-construction companies. He is the founder of Kumana & Associates, process control and automation, plant operations, logistics of oil stor-
a consulting firm specializing in process integration (pinch analysis) age and transportation, project management, and technical training.
techniques for energy/water optimization, as well as general process His wide experience covers polymers, oil and gas, petrochemicals,
performance troubleshooting/improvement in the full range of chemi- food and beverage, and process equipment fabrication for companies
cal industries. He and his company have been consultants to major in Europe, the Americas, and the Middle East. He has authored or
corporations worldwide, as well as to the U.S. Dept. of Energy, Natural co-authored numerous papers, articles, presentations, and training
Resources Canada, the World Bank, and the United Nations. He has courses on variable-frequency drives as final elements for process con-
authored or co-authored over 70 technical papers and book chapters, trol and variable-speed pumping applications, especially for artificial
and regularly teaches courses on pinch analysis, energy efficiency in lift of oil, and on technical skills training of engineering, operations, and
the process industries, and related subjects. He holds an MS in chemi- maintenance personnel. He earned his BS in chemical engineering from
cal engineering from the Univ. of Cincinnati. the Univ. of Texas at Austin.

Copyright © 2018 American Institute of Chemical Engineers (AIChE) CEP  February 2018  www.aiche.org/cep  41

Potrebbero piacerti anche