Properly designed compressor controls should not only protect
the compressor, they should protect the process.

Centrifugal compressors are common-place in many process plants due to their small
footprint, lower installation cost and low maintenance cost, relative to positive displacement
compressors. 

One significant shortcoming of a centrifugal compressor is its need for surge protection.

Centrifugal compressors raise the pressure of a gas by accelerating the gas in an impeller,
then, by Bernoulli’s principle, slowing the gas down in the diffuser section to convert the
velocity energy into pressure energy. The velocity energy that is imparted to the gas is a
function of the density of the gas and tangential velocity (tip speed) of the impeller.
Therefore, with a given gas and impeller speed, there is a finite pressure ratio
(discharge pressure vs. suction pressure) that can be supported by a centrifugal compressor.

What causes surge?
When the pressure ratio across a centrifugal compressor exceeds the maximum
pressure ratio that can be produced for that gas and impeller speed, the flow will
suddenly reverse through the compressor. The flow reversal allows the discharge
pressure to drop and at the same time will cause the suction pressure to rise, resulting
in a sudden drop in the pressure ratio. The flow reversal essentially relieves the
condition that caused the flow reversal to occur. Within about one second after the flow
reversal started, the flow will start moving forward again through the compressor. If the
condition that caused the high pressure ratio across the compressor in the first place
has not been corrected, the flow reversal will repeat. At this point, the compressor is in
surge.

Preventing surge
To prevent surge from occurring, a minimum flow controller must be implemented.
Unfortunately, a minimum flow controller that is configured to protect against the worst
case surge conditions will severely limit the turndown of the compressor. For processes
that do not use the full capacity of the compressor at all times, a simple minimum flow
controller will waste a lot of driver power. Figure 1 shows the surge limit line for a typical
centrifugal compressor. As the pressure ratio increases, the surge limit flow increases.
For many compressor applications, operating at lower flows also requires less pressure.
To take advantage of this flow/pressure relationship, compressor control engineers
sought techniques that modified the surge limit of the controller to match the actual
surge limit of the compressor under all operating conditions.

Figure 1 - Typical compressor surge limit curve

The earliest implementation of surge control based on surge line prediction that was
invariant to gas conditions was introduced by the Foxboro® Company in the mid-sixties.
Using a single loop pneumatic controller, the surge line was characterized according to
a ΔP vs. h algorithm where ΔP was the differential pressure across the compressor
stage and h was the differential pressure across the suction flow meter. This technique
pioneered by the Foxboro Company was the industry standard for a couple of decades
and as long as the suction pressure was fairly constant, this algorithm and hardware
setup was quite effective for compressor protection allowing significantly better
turndown than was available from a simple minimum flow controller.

The advent of digital anti-surge systems
In the eighties, with the growing popularity of digital controls, other manufactures
introduced digital surge controllers. These controllers used the same algorithm as the
one pioneered by the Foxboro Company twenty years earlier. These new digital controls
used a sloped line that defined the surge line and characterized ΔPc (compressor
differential pressure) to the Y-axis and ΔPo (orifice differential pressure) to the X-axis.
As with the earlier Foxboro pneumatic solution, the digital version worked fine as long
as the suction pressure did not deviate significantly from the design pressure that was
the basis for the surge curve configured in the controller.

These digital anti-surge controllers were quite successful in the eighties and nineties
because they allowed significantly greater turndown without recycling and were good at 
protecting the compressor against surge. A major complaint about these systems was
the abruptness with which they would take action to protect the compressor.
Since an anti-surge controller is usually not actively controlling (when the flow is well to
the right of the surge line), the controller will be wound down until the operating point
approaches the setpoint of the surge controller. The tuning parameters that are
appropriate for this type of controller tend to be “soft,” which leads to a controller that is
slow to react. And since flow signals are noisy, the derivative term of the PID can’t be
used. All these issues result in a protective controller that is slow to react and ultimately
over-reaction when the controller error is severe.

When an upset occurs in a process plant that causes the flow to drop suddenly or the
pressure ratio to rise quickly, these digital anti-surge controllers will open the anti-surge
valve(s) suddenly in order to prevent surge of the compressor. This protects the
compressor, but can create disruptions to the process that are severe enough to trip the
process unit. Some processes can experience damage to process equipment during
such events.

Surge controls are not enough
While it is important to protect a process compressor from surge, it is also important to
protect the process from the shock and disruption of an overly aggressive anti-surge
system response. A properly designed and implemented compressor control system
can not only protect the compressor, it can protect the process, as well.

The key to providing both compressor and process protection is to integrate the
compressor performance control with the anti-surge control. For turbine-driven (or
variable speed) compressors, this means integrating the speed control setpoint with the
anti-surge system. For fixed speed compressors, the positioning of the throttle valves
should be managed by the anti-surge system.

In general, there are very few process changes (aside from the closing of a block valve)
that will drive a compressor into surge at a rate that cannot be gracefully managed by
the surge controller. But a quick drop in speed will send the operating point into the
surge line very quickly.

The most efficient technique for matching the compressor performance to the process
demand is to vary the speed. A performance controller that manages the speed without
coordination with the anti-surge controller can drive the operating point into the surge
controller operating region. An interaction can occur between the performance controller
and the anti-surge controller that results in a low frequency oscillation of the process. A
sudden process change can drive the operating point into the surge limit, resulting in
extreme actions of the surge controller and severe process disruptions.

Protecting the process
Ultimately, the process needs the compressor to always provide a pressure and flow
that matches the demand of the process. A properly integrated compressor control
system not only integrates the performance control and surge control, it incorporates
other control features to soften or eliminate interactions and abrupt controller actions.

Override of surge controller
If the operating point moves to the left of the surge controller setpoint, the normal
proportional and integral (PI) action of the controller will start opening the anti-surge
valve to correct the error (move the operating point back to the right). But during severe
upsets, the tuning constants used in the surge controller might not be aggressive
enough to prevent the operating point from crossing the surge limit line. In this case,
logic is needed that will override the surge controller and start opening the valve open-loop
in response to the magnitude of the controller error.

Adaptive tuning
Aggressive gain is useful for pushing the operating point away from the surge limit line
when the operating point gets close the surge limit. But the aggressive gain can cause
the controller operation to be nervous when operating at the normal setpoint of the
controller. And as the operating point moves away from the control line, a reduced gain
will keep a noisy flow signal from unnecessarily provoking controller action. Likewise, a
high integral term is useful for pushing the operating point back to the setpoint and
helping to drive the anti-surge valve closed more quickly if the process load increases
suddenly.

Anti-surge valve slew rate coordination
An anti-surge controller that can act in 100 msec can get way-ahead of an anti-surge
valve that takes 3 seconds to open. The valve slew rate needs to be incorporated into
the output logic of the controller to assure that the controller does not get too far ahead
of the valve. If the controller gets wound up, by the time the valve catches up, it will
have opened too far and process will be spoiled.

Performance controller de-coupling
By integrating the performance controller with the anti-surge controller, it is possible to
de-couple the performance controller from the speed controller when the operating point
approaches the surge controller setpoint. When the operating point reaches the surge
controller setpoint (control line), the output of the performance controller can be
redirected to open the anti-surge valve in response to a further reduction in the process
demand instead of dropping the speed (or closing the throttling valve further). This
assures that the surge controller does not get provoked by the performance controller. 
This will also give the performance controller the ability to precisely modulate the
compressor output from zero to full load.

Surge anticipation
Algorithms that can anticipate surge or that limit the rate at which the operating point is
allowed to drop can provide the performance controller a way of limiting the demand
that is placed on the surge controller in the first place. To effectively manage the
process demand, this type of feature works best when it is integrated with a
performance controller decoupling feature. The use of this feature will not only reduce
the demand on the surge controller, it will significantly soften the effect of the
compressor control system actions on the process.

Digital controls have provided intelligent protection systems for the problem of
compressor surge that can improve the operating efficiency of the unit. But if the
additional features are not incorporated that provide protection to the process, the gains
in efficiency will be offset by the process disruptions. And worse yet, if the disruptions
cause the operators to distrust the anti-surge system, they are more likely to defeat the
system altogether.

Integration of the compressor performance controls with the anti-surge controls along
with incorporation of good compressor control features will allow the benefits of
improved compressor protection to be realized without sacrificing the smooth operation
of the process.