AN21

U Load Sharing for Parallel

Compressor Networks

PIC — Station Pressure Controller

LSIC — Load-Sharing Controller
UIC — Antisurge Controller
Application Note
FY FT PT PT

SIC LSIC UIC FY

Hot Recycle

A/D Cold Recycle

PIC
PT TT TIC FY CRIC FY

Hot Recycle

SIC LSIC UIC FY

AM FY FT PT PT

SIC — Fuel or Speed Controller

CRIC — Cold Recycle Controller
TIC — Temperature Controller

Figure 1 Simplified P&ID For Compressors Operating In Parallel

When two or more compressors are working in parallel, surge protection

and process efficiency can be maximized by operating them equidistant
from their surge control lines (see ASME paper GT-86-204, available
as CCC Technical Paper 17). This strategy is the basis of our patented
load-sharing algorithms for parallel compressors.
In contrast, conventional load-sharing systems throttle back the compres-
sors one at a time. One might be operating on its surge control line while
the others run flat out. That machine would then be exposed to the full
brunt of any disturbance. Its antisurge valve would often be open even
when other compressors could easily take up the slack. Larger surge pro-
tection margins might also be needed, further eroding process efficiency.
Another common approach is to operate the compressors at the same
flow rate or rotational speed. Unfortunately, even supposedly identical
AN21 (2.1) October 2000
Load Sharing for Parallel Compressor Networks

Analog Input

Station CV
and Load-Sharing Controllers

Controller
to/from all other Antisurge

Capacity
SP
Control

Pressure CRSC
SP Override
Ssel and Lsel CRPOC
Load Balancing
Lavg
CRPOC Lavg Lsel A Ssel CRPOC
A A
CRSC CRSC Analog Inputs

Ssel Ssel
Controllers for same compressor

Capacity Load fA
L L
to/from any other Antisurge

> Control Balancing Mode

S S L S
>
Lavg Lsel CRSC
Ssel RT PI
Load Capacity
(PI+RT)s Balancing Control (PI+RT)s
CRLB + >
CRLB CRLB
PI+RT Load- PI+RT
Loop Sharing Loop Antisurge
Decoupling
+ Decoupling +
Controller Controller

To Performance Control Element To Antisurge Valve

Figure 2 Simplified Diagram Of Parallel Load-sharing Algorithms

compressors will usually have different surge limits, due largely to inter-
nal leakage. As a result, this approach also fails to keep the compressors
equidistant from surge. While this strategy is better than sequential load-
ing, there is still a high likelihood that one compressor will be more apt to
surge or will go into recycle before the others.

CCC Parallel As shown in Figure 1, a CCC load-sharing system for compressors oper-
Load-Sharing ating in parallel consists of one station pressure or flow controller (PIC),
Systems one load-sharing performance controller (LSIC) per compressor, and one
antisurge controller (UIC) per compressor or compressor section. They
use a unique combination of protection and control methods (see Figure
2) to cooperatively maintain the desired pressure or flow while keeping
the compressors the same distance from their surge limits:
• Primary Capacity Control regulates the header flow or pressure.
• Load Balancing keeps the compressors the same distance from
surge to avoid unnecessary recycling or blow-off.
• Pressure Override Control provides accelerated capacity control
when the header pressure or flow exceeds a user-defined limit.
• Recycle Balancing equalizes the recycle rates of the compressors
when recycling is unavoidable.

Page 2 of 6 October 2000

 Load Sharing for Parallel Compressor Networks

Primary The station controller calculates its PID control response from the devia-
Capacity tion of its capacity control variable. Any change in the resulting station
Control control response (CRSC) is transmitted to the load-sharing and antisurge
controllers, which make appropriate changes in their own output signals.
Because this feature is the primary means of controlling the header flow
or pressure, it is called the primary capacity control response. Depending
on the controlled variable, it can also be more specifically referred to as
primary pressure or primary flow control.
When the compressors are recycling and less throughput is needed,
each antisurge controller multiplies the change in the station control
response by its own load-sharing gain and adds the resulting primary
capacity control response to its surge control response. This raises the
recycle rate, thus reducing net flow without hindering surge protection.
Under these conditions, the primary capacity control responses of the
load-sharing performance controllers will be zero.
When the compressors are not recycling or more throughput is needed,
each load-sharing controller multiplies the change in the station control
response by its own load-sharing gain, and adds the resulting primary
capacity control response to its own accumulated integral response.
Under these conditions, the primary capacity control responses of the
antisurge controllers are zero. This increases the compressor speed and
throughput without directly changing the recycle rate.
The overall effect is to throttle the compressor network by manipulating
the most appropriate control elements (speed set points or recycle
valves). This minimizes recycling without compromising surge protection.

Load The load-balancing response is used to keep all of the compressors

Balancing operating the same distance from surge. This is achieved by dynamically
biasing the outputs of the load-sharing controllers:
• Each antisurge controller calculates a deviation that is a measure of
its compressor’s distance from surge. This deviation is zero when the
compressor is operating on its surge control line. A negative devia-
tion implies an unacceptable risk of surge.
• Each load-sharing controller selects the lowest deviation reported by
any of its companion antisurge controllers (that is, that for the section
closest to surge) for use as its capacity control variable.
• The station controller calculates the load-balancing set point by aver-
aging the deviations selected by the load-sharing controllers.
• Finally, each load-sharing controller calculates a PID load-balancing
response to the difference between its selected deviation and station
controller’s average deviation. It then calculates its output signal by
adding the primary capacity and load-balancing responses.
This raises or lowers the flow rate through each compressor until all are
operating the same distance from their surge limits.

AN21 (2.1) Page 3 of 6

Load Sharing for Parallel Compressor Networks

Pressure Our load-sharing control systems also offer a pressure override control
Override (POC) feature that can raise the recycle flow to help control excessively
Control large or rapid changes in the capacity control variable. When controlling
the discharge pressure or net flow, POC is initiated when that variable
rises too high or too fast. When controlling suction pressure, it is initiated
when that variable falls too far or too quickly.
The station controller uses two separate algorithms to calculate the POC
response. One responds to the deviation of the control variable, the other
to its rate of change:
• The one-sided proportional-integral (PI) response counters excessive
deviations of the capacity control variable from the POC set point.
Depending on your application, this set point can be either an abso-
lute threshold or an offset from the capacity control set point.
• The pressure override predictor (POP) counters rapid changes in the
capacity control variable by comparing its instantaneous value (CV)
to a filtered average value (CVavg) and subtracting a user-defined
control threshold (SPPOP). When the objective is to limit the rate of
increase, this deviation is calculated as:
dev POP = ( CV – CV avg – SP POP ) ≥ 0
When the objective is to limit the capacity control variable’s rate of
decrease, this deviation is calculated as:
dev POP = ( CV avg – CV – SP POP ) ≥ 0
If this deviation is positive, which can only occur if the control variable
is changing rapidly in the limited direction, the POC response will
include a term proportional to the predictive deviation:
CR POC = CR PI + CR POP
CR POP = K POP ⋅ dev POP
Otherwise, the predictive response will be zero.
The POC response is sent to the antisurge controllers, which compare it
to their own control responses and use the higher of the two. Thus, this
feature never reduces the recycle rate below that needed to prevent
surge. If the POC response is higher than the surge protection response,
it will increase the recycle flow rate and suction pressure while lowering
the net flow rate and discharge pressure.
When appropriate, the POC response can alternately be used to control
large or rapid deviations of any single-input process variable measured
by the Station Controller.

Page 4 of 6 October 2000

 Load Sharing for Parallel Compressor Networks

Recycle In the event that recycling becomes unavoidable, this feature equalizes
Balancing the recycle rates of the various compressors. When it is enabled, the
Antisurge Controllers monitor each other to determine which of them has
the highest recycle rate.
If the compressors are moving away from surge, the controller with the
highest output will close its valve at an accelerated rate, while the others
slow the rate at which they close their antisurge valves. The furthest open
valve will thus “catch up” with the others.
If the outputs diverge while the compressors are moving toward surge,
the controller with the highest output (presumably closest to surging)
operates normally (so adequate surge protection is not imperilled). The
others then slowly open their valves, thus enabling them to catch up to
the furthest open valve.

Cold-Recycle The load-sharing strategy for parallel compressors is easily adapted to

Control networks that have an overall “cold” recycle loop in addition to a “hot”
recycle loop on each compressor (see Figure 1).
The hot recycle lines are used to protect each individual compressor from
surge due to sudden or short-term disturbances. If sustained recycling
becomes necessary, it is routed through the cold-recycle line, which is
equipped with a heat exchanger to cool the recycle gas. Heat buildup can
thus be avoided without compromising surge protection and without the
expense of multiple heat exchangers.
Our control systems are adapted to such networks by using an antisurge
controller to modulate the cold-recycle valve. That cold-recycle controller
calculates its surge control line DEViation by subtracting its own surge
control margin from the lowest surge control line deviation (DEVmin)
reported by specified companion antisurge controllers:
DEV = DEV min – ( CR SO + CR D ) ⋅ ƒ ( V )
where
CRD = derivative response to the rate at which the operating point
is approaching the surge limit
CRSO = Safety On response to detected surges (which includes a
fixed, initial bias)
Note that this margin, which is usually defined as a constant, is added to
that of the companion controller reporting DEVmin . This assures that such
a cold-recycle control loop will reach its surge control line and begin recy-
cling before any hot-recycle control loop. Except in the case of a rapid or
large disturbance, opening the cold-recycle valve will provide enough
extra flow to prevent surge without opening any hot-recycle valves.

AN21 (2.1) Page 5 of 6

Load Sharing for Parallel Compressor Networks

The impeller and TTC logos, Total Train Control, TTC, Recycle Trip, Safety On, Air Miser, TrainView, and
WOIS are registered trademarks and COMMAND is a trademark of Compressor Controls Corporation.

Page 6 COMPRESSOR CONTROLS CORPORATION

AN21 (2.1) 4725 121st STREET, DES MOINES, IOWA 50323-2316, U.S.A.
October 2000
Tel: (1) 515-270-0857 • Fax: (1) 515-270-1331 • Web: www.cccglobal.com