COMPRESSOR SURGE CONTROL

INTRODUCTION

Implementing microprocessor-based compressor control systems that minimize recycling (or

blow-off) requires consideration for a broad range of operating conditions. The compressor
control system must offer not only efficient and reliable anti-surge control, but must also provide
overload protection, load sharing (in the case of parallel machines), capacity control, process
decoupling and in some cases auto sequencing and other station controls (ESD, SOE, aux.
equipment control, etc.).

The Basic Compressor Control Challenge

The challenge in regulating a centrifugal compressor is to match its throughput to the process
requirements while keeping total flow high enough to prevent surge. The variable natures of the
minimum safe flow rate and the configuration of the compressors often complicate this goal.

Most instrument/control engineers at some time in their careers encountered a control application
for a centrifugal or axial compressor. These applications are generally not well understood.
Analyzing compressor performance characteristics (maps) is usually not a familiar task.
Fundamental information is required in order to implement compressor control systems.

Basic surge and performance control theories have been well documented since the early 1970's.
Since that time there has been conceptionally very little added regarding the phenomenon of
compressor surge, the condition under which it occurs and the methods for preventing it. Some of
today's compressor control companies are, however, presenting the theories of over 20 years ago
(published by Foxboro, Taylor, etc.) as their own "discoveries" and are advertising pending
patents on anti-surge control techniques. Unfortunately, this "black box" theory approach
unnecessarily mystifies the well-established concepts of compressor control.

With respect to control equipment, anti-surge systems have evolved over the years from simple
pneumatic minimum flow controllers, to analog electronic controllers with special algorithms, to
microprocessor-based controllers with special surge control algorithms and the flexibility to
handle a wide variety of compressor control applications.

This paper describes compressor unit control approaches, basic control theories, and addresses
several control considerations required to achieve advanced centrifugal compressor control that
can result in significant energy savings while operating safely over wide varying process
conditions.

--------------------------------------------------------------------------------

TOTAL COMPRESSOR CONTROL

Centrifugal compressors are common dynamic components of continuous processes in many

industries. The compressor, like distillation towers, heaters, reactors, etc., is interactive with all
other components of a process (or pipeline) and cannot be considered an independent system.
Compressors require multi-loop interactive control with the capability of field modification of the

COMPESSOR CONTROL 1
control schemes and parameters in order to adapt to process conditions (existing or new).
Microprocessor-based controllers with a comprehensive function library (including look-up
tables, adaptive tuning, etc.), are ideally suited provided that speed of response is fast enough for
detection of surge spikes and function configuration architecture offers the required flexibility for
a total unit control scheme.

Designing compressor controls that take into account only the machine itself leads to
oversimplification of the requirements and likely poor performances of the overall process
system.

In closely coupled systems, compressor control without dynamic compensation can become
unstable under certain process conditions because of interaction of the recycle valve with
suction/discharge valves and/or driver (turbine, motor). The logical approach to compressor
control must consider any suction dynamics (guide vanes, valves, liquid quench vaporization,
etc.), the driver (motor, steam turbine, gas turbine, etc.), the compressor discharge dynamics
(valves, other booster compressors, etc.) and the recycle valve as an interactive process unit which
must operate safely over as wide a set of conditions (operating window) as possible.

As mentioned before, the design engineer must consider several control constraints:

Limitations in the capacity of the upstream and downstream process that can affect the
compressor. Considering the dynamics of the gas (or air) supply and demand allows corrective
action to be gradual before a hard constraint is reached.
Driver performance. Gas turbine drivers especially have a fairly complex set of dynamic
constraints such as acceleration rate, exhaust temperature limits, fuel flow limits, etc. The turbine
protection controls avoid exceeding the constraints but depending on rate of approach to the limit,
the result may be a total driver shutdown or severe process upset. When the compressor control
considers the turbine driver limits, changes in load can be softened to prevent a limit shutdown.
Compressor surge limits - the objective of an anti-surge controller should not be limited to basic
independent machine protection. The anti-surge control performance as an integral part of the
machine performance control must be considered. Storing real surge points, applying adaptive
control of the recycle valve (variable gain depending on operating region) and compensating for
interaction between surge, overload and process control can significantly expand the operating
window. This allows operation very close to the actual surge lines (4-8%) under all process
conditions. Straight line surge control, even with variable slope, must make allowance for the
poor fit to actual surge points by using a wider margin (15-20%).
In case of sensor failure, machine safety should be maintained by return to minimum flow (Hs)
control with a wide margin of safety, but including loop interaction decoupling. In addition,
continuous sensing of flow with the surge spike detection algorithm, that opens the anti-surge
valve a pre-determined amount in case of momentary reverse flow, should be provided.
Dynamic decoupling of the control strategies that are interactive and have similar time constant,
eliminate regions of instability which not only limit operating range but affect machine safety.
Typically, suction, recycle and discharge valves need to be decoupled to prevent unstable process
oscillations.
Load sharing is normally necessary to prevent uneven distribution of throughput in parallel
machines and to maintain optimum capacity range of multiple machines. Of course, the operator
must always be provided with the capability to disable the load sharing or to select base load
operation.

SURGE CONTROL THEORY

COMPESSOR CONTROL 2
In order to understand the phenomenon of surging from the standpoint of compressor system
variables, the surge characteristics are reviewed and the basic control theory is presented.

Surge Characteristics

A typical family of characteristic curves for a single-stage, variable speed centrifugal compressor
is shown in Fig. 1. These curves depict discharge pressure as a function of mass flow rate.

Assume that the compressor is operating at Point A on the 100% speed curve (mass flow rate
WA) and the external load resistance gradually increases while the speed remains constant, the
flow decreases and the operating point will move to the left along the 100% speed characteristics
curve. When it reaches Point B, the mass flow has decreased to WB and the pressure has
increased from PD,A to PD,B, the maximum pressure the compressor can produce at this speed.
Pont B lies on the surge limit line (the area to the left of this line is the surge zone). At this point
the characteristic curve is practically flat and the operation of the compressor becomes unstable.
This condition, called surging, appears as rapid pulsation in the flow and discharge pressure,
producing high-frequency reversal in the axial thrust on the compressor shaft. The amplitude of
the flow reversal and the frequency of surge cycles vary. In some machines surging can become
severe enough to cause mechanical damage.

The line to the right of the surge limit line represents the surge control line, which is an artificial
line. This line is the boundary of the re-circulation (or blow-off) zone, where re-circulation or
blow-off is used to prevent surge. To the right of the control line is the safe operating zone, where
re-circulation or blow-off is unnecessary. Moving the control line to the right of the surge limit
line and using adaptive controller tuning normally eliminates the possibility of reaching surge
conditions in case of a rapid decrease in flow causes an overshoot to the left of the surge control
line. If the control line is placed too far to the right, the recycle or blow-off valve will be opened
unnecessarily and power wasted. Also, unnecessary opening of the recycle or blow-off valve can
cause inconvenient fluctuations in flow throughput. Consequently, the location of the surge
control line represents a compromise based on actual operating conditions of each individual
system. As a general rule, the surge control line is placed about 7-8% of flow to the right of the
surge limit line. However, there are cases where the compressor must operate closer to the surge
limit line and it is necessary to reduce this margin.
fig.1

Typical Characteristic Curve

COMPESSOR CONTROL 3
Basic Control Theory
diameter (D), the speed (N), and the angular velocity. Davis and Corripio^1 performed a
dimensional analysis using a form of the kinetic energy of a compressor in 1974 to obtain the
following:

(1)

Where Hp = Polytropic Head, (ft-lb)/lb

g = Acceleration due to gravity
ft/sec2
N = Speed, RPM
D = Impeller diameter, ft.
Qs = Suction volume flow rate,
ACFM
Re = Reynolds number
Mn = Mach number

The variation of Mach number (Mn) with velocity is assumed constant providing the gas velocity
does not approach the speed of sound. Reynolds number is assumed to be negligible at high gas
velocities where compressors operate. With these assumptions it is further recognized that the
FAN LAWS are valid along the surge limit line (R.L. Moore^2). Therefore, for a given
compressor geometry the SURGE LIMIT LINE can be reduced to a single point defined by the
following:

(2)

(3)

The surge limit is reduced to a single point

Where
C1,C2 = Constants
Hp = Polytropic head, ft.lb/lb
Qs = Suction flow rate, ACFM
N = Speed, RPM

This is the universal performance curve illustrated graphically by Figure 2.

Surge control theories have been well documented since the early 1970's. The polytropic head
(Hp) ratio to the volumetric suction flow rate squared (Qs^2) forms the basis for the anti-surge
scheme. Equations using the simplified polytropic head (Pr^m'/m') are described in MICON's
Compressor & Turbine Control Algorithms/Function manual. For applications with fairly low
compression ratio and reasonably constant gas composition, adequate anti-surge control can be
achieved by using reduced functions of the polytropic head equations.

COMPESSOR CONTROL 4
The anti-surge control system developed by M.H. White in 1972/3 has been widely discussed
since that time by many others. 4, 5, 6, 7, 8, 9, 10.

White's development of a basic anti-surge control scheme for centrifugal compressors starts with
equations (2) and (3) as do many other references. These equations apply to both, single-stage and
multi-stage (without inter coolers) centrifugal machines.

Speed (N) can be eliminated as a variable by combining equations (2) and (3) to obtain...

(4)

Hp = C3(Qs)^2

Where C3 = C1/(C2)^2

The actual volume flow rate Qs is given by

(5)

Qs = C4(hs/Ds)0.5

Where hs = pressure drop across head-type

flow element at suction, in
inches of H2O.

Ds = Density of gas at the suction lb/ft^3

and

(6)

Where MW = Molecular weight

R = Universal gas constant
Ps, Ts, Zs = Suction pressure, temperature
and compressibility

Equations (5) and (6) are now combined to obtain the following:

(7)

Where C5 = C4 (R)^ 0.5

Q is now eliminated from EQ (4) by using EQ (7) to obtain

(8)

COMPESSOR CONTROL 5
Polytropic head can now be eliminated from EQ (8). This in turn has the advantage of also
eliminating Ts, Ps and MW. Hp is defined as follows:

(9)

Where m' = Polytropic exponent

Where K = Specific Head Ratio

Pe = Polytropic efficiency

An assumption is made now to simplify equation (9). This assumption is that the compression
ratio term [(Pd/Ps)^m'-1.0]/(m') is linear and m' is a constant.

(10)

According to Moore^2 and Chandler6 this assumption is fairly accurate for compressors operating
at low compression ratios and high specific heat ratios. For example, Moore^2 shows the
linearization function of (Rem' + 1.0) to be within approximately 5% for a compressor operating
at a low compression ratio and a specific heat ratio of 1.5.

Combine equations (8), (9) and (10) to obtain...

(11)

In equation (11), the term (Zs + Zd)/2 is Zavg. Little error will result if the term (Zs + Zd)/2(Zs) is
set equal to a constant.

This will give the simple results of:

(12)

hsr = C9 (Pd-Ps)

It is interesting to note that had White^3 not assumed constant molecular weight (MW) and intel
temperature (Ts) that these two variables would have partially canceled during the course of the
algebraic manipulations. This is the reason that White3 arrives at the same results as given by
other references.

It should be realized that the development just covered assumes certain conditions that can lead to
minor errors. These assumptions include the following:

1) The FAN LAWS that establish equations (2) and (3) are accurate.
2) The term [(Pd/Ps)^m,' - 1.0]/(m') is linear.
COMPESSOR CONTROL 6
3) The term (Zs + Zd)/2(Zs) is equal to a constant.

FOR WIDELY VARYING GAS COMPOSITION APPLICATIONS SEE POLYTROPIC HEAD

ALGORITHM IN MICON DESCRIPTION.
Also, field tests on compressors with gases of widely varying densities (more than 35%) have
shown that the basic anti-surge equations (EQ 12, 13, and 14) are not totally self-compensating
for molecular weight and temperature changes. Since such wide variations in density are rare,
(mostly related to large molecular weight changes, because temperature is calculated in absolute
degrees) , one can conclude that temperature and molecular weight corrections are in most cases
not required.

Impact of Large Gas Composition Changes

In some unusual process applications, a compressor may handle gasses of widely varying
composition. Field tests have shown that there is a nonlinear relationship between molecular
weight, MW, and hs, Pd-Ps (orifice pressure differential at suction and differential pressure across
the compressor). This is further complicated by the fact that a change in molecular weight is
accompanied by a change in specific heat ratio, K, and consequently the ratio of (K-1) to K.

The influence on the specific heat ratio (K=Cp/Cv) is often not obvious. Compressor suction K of
course is not affected, but as gas is compressed the K will influence the rate of change of gas
density as shown in Fig. 3.

To compute the effect of drastic gas composition changes on the value of the surge control line,
one must consider the molecular weight impact on the specific heat ratio factor. Since it is not
possible to generalize the effect of these factors, each compressor control application with large
variations in gas composition/density should be examined individually and corrections to the
surge line calculation should be verified by actual field tests.

The above considerations with respect to impact of molecular weight on surge line calculation
relate almost exclusively to compressors, which handle two or more completely different gasses.
An example would be a methane compressor that is also periodically used to purge the system
with nitrogen.

As mentioned before, the self-compensating nature of the orifice pressure differential and
compressor pressure differential variables used in the surge control line equations makes the
requirement of special density (molecular weight and temperature) corrections unnecessary in
most compressor control applications.

Surge Control Line

The optimal position of the surge control line is parallel to the surge limit line. This is
accomplished with a two-dimensional look-up table, a standard function in MICON controllers.
As mentioned before, to minimize re-circulation or blow-off, the surge control line should be set
as close to the surge limit line as possible.

COMPESSOR CONTROL 7
Also critical to the margin setting (% parallel to the surge limit line) of the surge control line is
the effectiveness of the anti-surge system in handling process upsets. This effectiveness depends
on the chosen control strategy and is greatly influenced by the nature of anti-surge system
(parallel/series machines, inlet-guide vanes, etc.), its transmitters, controller and anti-surge valve.

To set the surge control line parallel to the SURGE LINE, a constant is added to the right side of
equation (12) as follows:

(13)
hcl = C9 (Pd-Ps) + A1

Where

hcl = Anti-surge control line

A1 = Constant that establishes how far the control line is from the surge line.

Equation (13) applies when the compression ration (Pd/Ps) is less than 2.5. It has been shown by
others9 that when compression ratio are greater than 2.5 a correction for suction pressure (Ps)
needs to be taken into account as follows:

(14)

hcl = C9(Pd-Ps) + A2(Ps) + A3

Where A2 and A3 = Constants established for the particular application. A3 establishes the surge
CONTROL LINE.

In summary, the theories presented here for surge control line setting, surge characteristics, and
surge control are not new. As stated previously, these theories have been established in the 1970's.
However, the implementation of these theories has been greatly facilitated with the advent of fast
and flexible microprocessor-based controllers such as the MICON S-32, P-200 or RCM-32 units.

Discharge Flow Measurement

Some systems measure flow (DP across the orifice or equal) in the compressor discharge line
without correcting for suction conditions. This does not provide proper surge protection and
should therefore not be applied. In some installations, because of the piping configuration or the
size of the line, it may not be feasible to put the primary flow element in the suction line. This
may also be true when the compressor operates with a very low suction pressure and any
additional drop in the line cannot be tolerated. It may be possible to resolve the problem by using
the inlet eye of the compressor as an orifice. However, it must first be established that this
measurement is a usable/accurate one (i.e. noise-free, repeatable and representative). If this cannot
be done, it is necessary to install the primary flow device in the discharge line and add pressure
and temperature compensation. Note that in some compressor applications the ratio of discharge
to suction temperature may be reasonably constant (remember that the compensation applies to
absolute temperatures) that can eliminate the requirement for temperature transmitters (a constant
can be used to approximate the temperature compensation). And, when the compressor operates at
constant speed, which means that the compression ratio is constant at the surge point, the
instrumentation can be further simplified by applying a pressure constant as compensation.

COMPESSOR CONTROL 8
Inlet-Guide Vanes

Constant-speed centrifugal and axial compressors are frequently equipped with adjustable inlet-
guide vanes to control throughput. In addition, axial compressors may have adjustable stator
blades.

Besides manipulating the compressor throughput, moving these vanes also changes the slope of
the surge limit line (there is a surge line for each vane position). When changes in vane position
produce large variations in the slope of the surge limit line it is usually desirable to measure the
vane position and automatically compensate the anti-surge control algorithm.

Speed of Response Considerations

Surge cycles can occur fast and in some cases surge goes undetected because the applied
instrumentation and controls are too slow to detect the surge phenomenon. Only fast transmitters
and controls can cope with the high frequency of centrifugal and axial compressor transients.
Transmitters should not have a lag of greater than 35 milliseconds and controllers should have a
fast input/loop update time.

Some "smart" transmitters are too slow for surge control applications, others may not be stable at
high frequencies. One should not consider the basic surge phenomenon only. There are pre-surge
oscillations (incipient surge) detection and measurement requirements which are often
incorporated into the surge control strategy.

During field tests, utilizing fast flow transmitters at the suction and discharge side of the
compressors and using high-speed recorders, it has been verified that surge cycles can occur very
fast. These tests showed that the surge cycle rate depends on several factors but is mostly
associated with the size of the machine and its discharge piping configuration. Consideration was
given to upsets such as jammed check valves that are normally located close to the compressor
discharge. The larger the volume, the slower the pre-surge oscillations and the surge cycles. The
flow drops quickly regardless of pressure or compressor speed, from a setpoint level to a reverse
condition in approximately 50-70 milliseconds. Pre-surge oscillations peak-valley times vary
from 60 to 200 milliseconds and surge spike duration has been measured from 150 to 500
milliseconds. Actual surge cycles varied from 400 milliseconds to 30 seconds. These
measurements have confirmed that impending surge and actual surge can produce high-speed
oscillations of forward and reverse flow through the compressor system and that the anti-surge
controller input/loop update time is critical. Considering the measured dynamics, it is
recommended that the anti-surge controller has a min. 10 milliseconds sample rate for pre-surge
oscillation measurement and a loop update time of less than 200 milliseconds. On low volume
compressor applications the loop update time should be decreased to less than 100 milliseconds.

As mentioned before, the recycle or blowoff valve must be fast-acting and should be capable of
handling 100% of design flow rate at a 75% DP across the compressor. The full stroke speed of
the anti-surge valve should be less than 2 seconds, preferably 0.5 to 1.5 seconds (depending on
valve size). This normally requires the use of pneumatic boosters and the utilization of adaptive
gain tuning.

COMPESSOR CONTROL 9
Surge control Considerations

Controlling a centrifugal or axial compressor system with general purpose microprocessor-based

controllers or with a typical distributed control system normally falls short in response time
(better than 150 milliseconds desired) and specific compressor control algorithms (such as non-
linear surge curves, adaptive tuning, oscillation monitor, etc.). These control systems, in many
cases, do not provide adequate protection from surge and often lack the flexibility to properly
compensate for interaction between surge and capacity control.

MICON, a multi-application compressor controller has the required speed of response and
incorporates all functions to provide optimum integrated anti-surge control with process control.
The entire compressor control scheme - capacity control, anti-surge control, load sharing and
startup/shutdown can be implemented in an integrated system. Control interaction problems can
be minimized through dynamic decoupling and the control response deterioration experienced
with most general purpose systems is all but eliminated.

Once transmitters and the anti-surge valve are selected, the proportional band level and reset time
are set based on the parameters of the compressor and its surrounding process. These parameters
include the volume of the system, inertia of the compressor/driver, and many other parameters as
described previously in this paper under "Total Compressor Control". These limitations are
common to all closed control loops with feedback. Thus, modulating control with constant
proportional band setting responds well to normal (slow) upsets; however, without adaptive gain
(or other fast-acting override device) the anti-surge valve cannot be opened quickly enough for
fast upsets such as upstream or downstream equipment failure. Requiring both responses, one for
normal control and the other to cope with rapid upsets, has led to the MICON adaptive gain
method. Thus, for slow upsets, the valve will be controlled by the action of the proportional base
gain plus the reset action; for fast upsets, in which the compressor's operating point drops
significantly below the surge control line, the adaptive gain delivers a control action
corresponding to the pre-determined gain increase, opening the anti-surge valve quickly to
prevent surge. As the operating point of the compressor moves back in the direction of the surge
control line the adaptive gain decreases proportionally until it reaches the base gain setting. The
valve action includes, of course, directional slew rate control which assures that a recycle or
blowoff valve is never closed fast, but can be opened rapidly when needed.

Adaptive gain Tuning

Field tests have confirmed that if the surge control line is set a proper distance from the surge
limit line, then the adaptive gain method performs very well in preventing surge. However, if the
surge control line is by accident (calculation mistake, configuration/data entry error, erroneous
performance data, etc.), set too close to the limit line, the operating point of the compressor can
cross the limit line before the adaptive tuning (or another override signal) takes proper action.
There have even been cases where "old" compressor performance curves (with different process
data) have been mistakenly used to calculate the surge control line and the control line was placed
to the left (in the surge zone) of the surge line. As a result, surge can begin without entering the
normal control zone, causing a rapid flow decrease. With MICON's surge spike detection method,
compressor surge is stopped during the first cycle by rapid opening of the recycle/blowoff valve
and the surge control line is automatically moved a pre-selected amount to the right (a safer
distance from the surge limit line). This is accomplished by sensing the rate and magnitude of
flow decrease. Thus normal process upsets are automatically filtered out. On compressor
shutdown, the spike detection remains active, but the surge control line shift action is disabled.

COMPESSOR CONTROL 10
Incipient Control Backup

In some compressor control applications, incipient surge control is added as a backup algorithm to
the primary and fallback anti-surge control algorithm. This special algorithm can increase the
reliability of the anti-surge control system. Incipient surge could be used as the primary/main anti-
surge control algorithm, however, since the concept depends on high speed, clean flow
measurement that involves high-speed transmitters and special installation consideration, MICON
does not recommend that incipient surge control by itself, (alone) be utilized for compressor anti-
surge control.

Before the compressor reaches the actual surge point, rapid oscillations occur. Compressor field
tests have confirmed this phenomena as an indication of impending surge. However, since this
surge phenomena has special characteristics for each compressor, it is (in practice) not always
easily measured and special signal characterization/ filtering is required.

A very high speed algorithm is required to dissect pre-surge oscillations (10 millisecond sample
rate) into useful data for control purposes.

To prevent high frequency noise from interfering with the pre-surge detector, a special high
frequency filter is used. The effects of low frequency variations caused by normal process
changes and/or operator setpoint changes are isolated by a low frequency cutoff filter adjustable
from 0.2 to 12 HZ.

A high-speed transmitter must be used when implementing incipient surge control techniques.
Fig. 6 shows the trend of an oscillation occurring near actual surge. Control action is initiated
only after Point A has been reached.

The incipient control backup concept has been successfully used in compressor control systems
for many years. It is an exclusive algorithm of MICON. However, it must be emphasized that an
accurate, high-speed flow measurement is required for proper performance of this strategy.

Surge Guard Control Points

EXAMPLE FOR SINGLE CASE COMPRESSOR

The following example, Fig. 7 shows a control block diagram for a motor-driven single case
compressor with decoupled capacity (discharge pressure) control, overload protection and load
sharing. Typical applications involve more complex configurations. Parallel and series machine
configurations require special considerations. Compressors with inlet guide vanes add complexity
(as mentioned before) since the vane position affects the location of the surge limit. Multistage
compressors and compressors with sidestreams are other examples of control challenges.

MICON has developed solutions for virtually all types of compressor configurations. The theories
and concepts described in this Application Note are intended to provide basic information only.

Decoupling of Surge and Process Control

Both, the anti-surge and the throughput controls manipulate the compressor operating point,
usually in opposite directions. These interactions often cause control instability and therefore may
require larger than necessary surge control margins. Effective decoupling of the anti-surge and
throughput (capacity control) loops enables the system to overcome these interactions. Severe
COMPESSOR CONTROL 11
control interactions are found mostly in systems with low discharge and/or suction process
volume (closely coupled systems); systems with large volumes (slow dynamics) are more
forgiving with respect to loop interactions.

Example of Overload Protection Compressor Controls Integration

Consider the following application, (Fig. 7), where the compressor has typical pressure control
and surge control loops. The pressure controller adjusts the butterfly valve (or decreases speed if
it is a turbine driver application) in the closing direction if discharge pressure increases above the
setpoint. The anti-surge controller opens the recycle valve to increase flow in order to prevent
surge. Suppose that the discharge pressure increase is the result of a process upset downstream of
the compressor and that the control reaction is very rapid. The anti-surge controller would
immediately begin to open the recycle valve resulting in a temporary decrease in discharge
pressure. The discharge pressure controller, in an independently configured mode (non-
decoupled), re-opens the suction butterfly valve (or increases the turbine speed) that moves the
operating point of the anti-surge controller away from the surge limit line. An independently
configured (non-decoupled) anti-surge controller would then continue to close the recycle valve
while the discharge pressure increases again, therefore moving the operating point again toward
the surge limit line.

These are incompatible control actions, which usually result in a "hunting" action of the control
loops. The control oscillations can be significant since each controller fights its counterpart. In
non-decoupled loops, this interaction can exist whether the control is on suction pressure,
discharge pressure or flow. To provide control stability there are two configuration/tuning
alternatives. Either de-tune (slow down) the controllers which decreases their effectiveness, or,
provide decoupling between the interacting control loops. The decoupling action normally
consists of a control phase shift of the throughput/ capacity control (in this case discharge
pressure) based on the rate and magnitude of the anti-surge control output. A shift of 0.5 to 3.5
seconds usually provides adequate decoupling, considering that these interaction control problems
are normally associated with low volume process systems.

Another reason for loop decoupling is associated with dead time in the anti-surge loop. These
process response time delays can be caused by the piping configuration of the network between
the compressor suction and the anti-surge valve.

In some compressor systems there is a significant delay, after the recycle valve opens, before the
gas reaches the compressor inlet and the flow through the compressor increases. This dead time is
especially common in turbine driven multi-stream compressors with large suction drums (such as
propylene or ethylene refrigeration compressors). The dead time can be decreased by temporarily
increasing the turbine speed simultaneously with the anti-surge valve opening.

Load Sharing

When centrifugal compressors are operated in parallel, the problem of properly sharing the load is
usually present. The solution is particularly challenging when the compressors have dissimilar
characteristics. But even when the characteristics are supposedly identical, load sharing should be
provided because minor manufacturing differences and piping configuration affect the operating
envelope. An effective system for load sharing can keep compressors at a relative distance from
the surge control line but load sharing is normally activated only once all compressors have
reached the load-sharing enable line. This load sharing strategy assures that all compressors reach
COMPESSOR CONTROL 12
their surge control lines simultaneously. The concept produces the widest range of operation with
no recycle (or blowoff) and the minimum recycle/blowoff to produce any required load. In
addition, the strategy can optimally divide the load whenever the load changes, when a
compressor goes off or on line, or when efficiency changes.

Conventional load sharing strategies employing pressure-to-flow cascade or biasing the output of
a single pressure controller, are usually ineffective. Base loading, another common method, is
also not energy efficient.

Load sharing may be accomplished by the manipulation of a suction valve (as shown on the
illustration), by speed changes (if the compressor is turbine driven), or by guide-vane positioning.
Thus, there are many possible configurations for simultaneous approach to the surge control line.

Since compressors may differ in capacity and characteristics, the absolute distance of the
operating point from the surge control line can be meaningful. For compressors with almost
identical characteristics, it is recommended that the compressors are allowed to operate
unrestricted until the operating point reaches the load-sharing enable line, a line usually set 10%
from the surge control line. For compressors with dissimilar characteristics, it is usually
preferable to set the load share enable line near the normal operation points; or in other words, the
parallel machines operate at equidistance from the surge control line over the complete operation
range.

Each compressor controller's load share algorithm engages once the compressors operating have
crossed the enable line and holds the throughput/capacity control output (suction valve, turbine
speed or guide-vane) in its position until all load sharing compressors have reached the enable
line. Communication between compressor controllers is via a high speed digital peer-to-peer link.
After all compressors have crossed the load-share enable line they are kept equidistance from
their surge limits, thus helping to minimize recycle or blowoff costs. The equal distribution of the
load within the active region (below the load-share enable line) is accomplished by calculating the
average distance from the surge line of all compressors on load sharing and then biasing each
compressor in the proper direction.

Prevention of Driver Overload

Variations in gas (or air) density significantly affect compressor performance. The compressor
driver, a motor or a turbine is sometimes tightly sized and may require load limiting control to
prevent the compressor driver from overload trips during high density demand times. In typical
applications (as shown on the illustration, Fig. 7) the compressor controller overload algorithm
will override the signal of the throughput/capacity controller to modulate a suction valve. In case
of a motor driven unit, the motor power is limited to a setpoint that represents the maximum
power that the motor is allowed to draw. Feedback selection is used to provide a bumpless
transfer between the throughput/capacity control, the load-sharing control, and the overload
control.

For applications that experience fast density changes an additional suction pressure override
control which operates as a pre-act protection may be used. This offers further insurance against
nuisance driver (motor, turbine) trips.

Since speed variation is often the most efficient method for controlling compressor load, turbine
driven machines are commonly in use. These turbines represent a piece of complex machinery by
themselves and incorporate safety systems (exhaust, temp. limits, etc.) to eliminate, or at least

COMPESSOR CONTROL 13
minimize the possibility of damage due to overload or in the event of other unusual operating
conditions.

Series or Parallel Machine Surge Control

When two or more compressors are connected in series, using surge control on each individual
machine provides the best protection. This, however, requires a re-circulation or blowoff valve for
each compressor. In many designs (especially in large pipeline applications) the individual
recycle control architecture may not be very practical due to common cooler systems and
therefore one common "cold" re-circulation valve is often provided. Duration of reverse flow and
thrust bearing load exposure during compressor trips should be examined for common valve or
individual machine configurations with consideration given to automatic on/off "hot" recycle
control. In special designs, each machine may require complete "hot" recycle control
(manipulating, not on/off) in addition to the "cold" recycle control.

In some series compressor configurations, two compressors are driven by the same prime mover.
These machines are sometimes considered as a single unit and the installation usually includes
only one anti-surge valve.

When two or more compressors are working in parallel, the greatest surge protection and
efficiency results when both units operate equidistant from their respective surge control lines
after the operating points have crossed the loadsharing enable lines (this has been described
previously). It is, of course, assumed that each compressor is provided with its own
recycle/blowoff valve. In some parallel applications (such as pipeline stations with common
cooler arrangement), each machine may have a "hot" recycle valve to provide rapid and short-
term protection, while one common "cold" recycle valve provides the normal anti-surge control.

Startup/Shutdown Features

Both the anti-surge and the throughput/ capacity controls often include additional algorithms to
provide for startup and shutdown mode of the compressor in accordance with the compressor
manufacturer's specifications and in compliance with the process requirements.

Typical Startup/shutdown features are:

I. Startup and Purge Sequences. The recycle valve is kept at either 100% or 0%, as required for
shutdown or purge. After the compressor is started, opening of the valve is determined by the
anti-surge controller.
II. Startup Sequence. After the compressor is started the control maintains the minimum turbine
speed (for turbine driven units) for a pre-selected time delay and then ramps the governor to a
position required by the system.

III. Startup Sequence. Startup ramp by the control is aborted when either the check valve into the
system is open (typically sensed by a differential pressure switch across the checkvalve) or upon
reaching a preset pressure.

IV. Shutdown. On compressor shutdown, the recycle or blowoff valve is opened 100%.

V. Shutdown. On compressor shutdown, the throughput/capacity control ramps down the output
to prepare for future startup. The controller is also transferred to auto mode.

COMPESSOR CONTROL 14
FIELD TESTING OF SURGE POINTS

A request to test the compressor's surge limits in order to determine the true surge control line is
understandably unpopular because of the perceived risk of damage to the machine. However, the
MICON anti-surge control system minimizes the danger by its quick open recycle/blowoff valve
action upon detection of the first surge cycle.

Field testing of surge points is important since the actual compressor surge limits usually differ
significantly from the limits indicated by the compressor map. The surge line shown on a
manufacturer's compressor map results from an analysis of the compressor design and is not
normally verified by testing. The map data is usually conservative. Also, piping arrangements on
the compressor suction side can account for significant surge point discrepancies.
Accommodating these inaccuracies often requires unnecessarily wide surge margins, which can
result in excessive recycling or blowoff.

Therefore, a strong economic incentive exists to establish and confirm the surge line by
experimental field testing.

To calibrate the surge limit line, one or several surge tests have to be carried out. For variable
speed compressors (turbine driven units) the surge test should be taken at several speed settings in
order to establish a reasonably complete surge line. Test points are then linearly interpolated by
MICON's look-up table function to generate a precise surge control line.

The distance between the surge limits and the surge control line is examined and selected so that
the severest process disturbances (such as fastest possible closing of the discharge or suction line)
will not cause surge. Since adaptive tuning is utilized in the control algorithm and the anti-surge
valve opens at high speed, in most cases the margin can be set between 4% and 8% of flowrate.

CONTROLLER RELIABILITY AND SECURITY

Reliability and security are key considerations of the design and application of compressor control
systems. MICON hardware is built using advanced solid state circuitry - subjected to stringent
quality control inspection and testing- provided with on-line self diagnostics and optionally
employing redundant hardware.

The redundant controller system provides for automatic bumpless switching to the backup
controller if certain error(s) are detected in the primary controller.

High control system reliability however,is not only a function of hardware redundancy, fault
tolerant compressor control strategy configuration is equally as important. In case of
sensor/transmitter failure, machine safety should be maintained. Fallback functions should be
configured so the recycle/blowoff valve does not necessarily have to be opened, instead the
controller reverts automatically to a simpler strategy (typically h). Although these fallback
strategies generally decrease the compressor efficiency and the operating window at low capacity
conditions, significant energy savings is still obtained since without fallback controls many field
instrument malfunctions would result in complete opening of the recycle/blowoff valve or
compressor shutdown.

On typical surge control applications minimum flow (DP across suction element) and incipient
surge can be selected as fallback strategies. Surge spike override should always be used as an
additional safety measure. Additionally, on motor driven compressors, the current signal
(approximation of BHP) can be used in a fallback algorithm.
COMPESSOR CONTROL 15
SUMMARY

Although many of the advantages of the MICON system in compressor control are obvious after
reading this Application Note, others are not and are often worth another review.

Total compressor control, not just anti-surge control. MICON accommodates anti-surge control,
overload protection, load sharing and capacity control in one or two dedicated units. Control
interaction problems are practically eliminated through dynamic decoupling.
Control strategy flexibility. MICON has the flexibility to change control strategies or include new
strategies in the field. Over 100 control, arithmetic and logic functions are built into the MICON
controller, including look-up tables, oscillation monitor and adaptive tuning.
Fault-tolerant control configuration. MICON can easily cope with the configuration complexity of
automatic fall-back strategies (such as minimum h or incipient surge control).
Automatic control backup. Completely automatic hardware redundancy is available as a standard
option.
Communication interfaces. Each controller is equipped with two communication interfaces - a
peer-to-peer link and a vertical communication link. In addition, a MODBUS interface is offered.
Load sharing features. For parallel or special series compressor configuration MICON includes
advanced load sharing techniques.
In summary, this application note lists the most common considerations which should be taken
into account during design and start-up of a compressor control system. MICON has been
successfully implementing compressor controls since 1978. MICON has the know-how and
equipment to furnish a good solution regardless of how simple or complex the control challenge
may be.
Return to MICON's main menu

Return to MICON's table of contents menu

COMPESSOR CONTROL 16