Pushing Boundaries: Two Amine Regenerator Case Studies

Jim Critchfield
Retired/Consultant
Joel Cantrell
Bryan Research & Engineering
Christina Daniels
INEOS GAS/SPEC Technology Group

The refinery amine unit regenerator is a significant consumer of energy, and with so much emphasis
currently placed on carbon intensity, there is an increased interest in the performance of regenerators.
The primary function of the amine unit regenerator is to reverse the acid gas absorption reaction, and by
doing so provide adequately-stripped lean solvent for the amine absorber(s). However other often-minor
considerations can also influence regenerator energy requirements.

In this paper we focus on two case studies that highlight regenerator operation. By analyzing the
conditions in the regenerators with the PROMAX®1 amine simulator, we examine performance and
consequences when the systems operate near low-energy and turn-down boundaries.

Both Cases are set in refinery primary treating systems. The first case explores the consequences of a step
change in operating conditions resulting from a refinery crude diet change. The second case reveals effects
of a loss of lean/rich exchanger performance. “Rules of Thumb” for managing reboiler duty are considered
for both cases.

Paper Scope

This paper focuses on refinery primary treating system regenerators. Two specific real-world cases are
discussed.

Energy efficiency is a current emphasis, and in a typical refinery amine system, the regenerator is the major
energy consumer…and therefore a target for conservation. In our carbon-focused reality, refinery energy
advocates seek low-energy modes of operation. For amine regenerators, low energy operation may have
unintended consequences, however. Forewarned is forearmed; we discuss some consequences of low-
energy operation.

The Case studies are analyzed through process simulation to understand the operating conditions, and
comparison is made to operating “rules of thumb” for the Cases. Simulation is also used to extend the
operating conditions to demonstrate the effect of minor process considerations on system performance.

1
ProMax is a registered trademark of Bryan Research and Engineering, LLC
Refiners Operate in a World of Change

Over the last four decades, US refiners have Figure 1 Sulfur In Crude: PADD 3
1.8 3500
responded to shifting crude diets and moving data of U.S.EIA 2023
1.6
3000
environmental targets. Beginning in the mid- 1.4
1980’s, refiners experienced increasing sulfur 1.2
2500

% Sulfur in Crude
content in their average crude diet; that trend

1000 BPD
1 2000

leveled off after the year 2000. In particular, 0.8 1500

PADD 3 (NM, TX, AR, LA, AL, MS) has seen a 0.6
1000
0.4
decline in the concentration of sulfur in the 500
0.2
crude diet, coinciding with the growth in USA
0 0
NGL produced from hydraulic fracturing. The 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025
Year
chilling effect of COVID on the economy – in US Weighted Avg Sulfur PADD 3 Sulfur
particular, travel – hit refineries; creating a US NGL Production, 1000 BPD PADD 3 NGL Production, 1000 BPD

negative “COVID Bobble” that further

Figure 2
decreased sulfur in the crude diet. Data of the
US EIA illustrate these trends in Figure 1.

Starting in the mid 1990’s, US refiners

responded to the increasing demand for sulfur
capacity by investing in sulfur units (Amine,
Claus, and Tail Gas systems). Environmental
pressures also played a role in capacity build; in
many cases acid-gas flaring-avoidance drove
decisions to install spare capacity. Data from
the US EIA in Figure 2 demonstrate that sulfur
charge rates to refineries trended with
installed capacity until the early 2000’s; after
Figure 3 Sulfur Capacity in PADD 3 Refineries
that time, sulfur capacity continued to grow 120
while sulfur charge rates plateaued. Then the Not all sulfur gets to the SRU;
proportion of sulfur captured increases over time
Capacity: anti-flaring (sparing)
sulfur out of fuels
crude flex esp. more coke

COVID Bobble hit; in 2020 charge rates dropped 100

Utilization (for Sulfur, inferred)

further, creating even lower capacity utilization. 80

60 Here
Figure 3 shows inferred capacity utilization of Comes
The
sulfur units in refineries, compared to EIA’s 40 Sulfur
Play with the Toys The Golden Age of The New
Crude Utilization metric. The COVID Bobble hit You've Got Sulfur Investment Normal
20 COVID
the overall refinery industry hard, but it hit data of U.S.EIA 2023 Bobble

refinery sulfur process units harder. The bottom 0

1985 1990 1995 2000 2005 2010 2015 2020 2025
line is that sulfur processing units in refineries Year

Crude Ultilization Inferred Sulfur Utilization - Ratio of Crude Sulfur Charge / Capacity
have had to deal with shifting load utilization,
and for many, turndown operation has become a new normal.

Refinery operations produce hydrocarbon gases containing CO2 as well as H2S; a primary source of CO2 in
fuel gas is the FCC. FCC-generated CO2 is at least partially absorbed in refinery amine units and contributes

2
to acid gas loading. EIA data for FCC charge and capacity imply that CO2 production has been more stable
over time than sulfur charge rates have been. As a result, not only has the total amount of acid gas in the
refinery changed, but so has its average composition.

It’s well-recognized that refinery amine systems become polluted with heat stable amine salts (HSAS); the
strong effect of this pollution on system performance has been a prolific subject of publication. Conversion
units such as FCC are a known source of HSAS pollution. Relatively stable FCC rates suggest HSAS continue
to plague refinery amine units even as sulfur loads change.

The new normal is that refinery sulfur systems must successfully surf throughput *and* composition
changes. This macro environment has implications when applying Rules of Thumb for energy targets; the
effects of rate and composition changes must be understood and incorporated into practical targets for
operating conditions in sulfur units.

The refinery in Case 1 experienced the effect of these trends on amine unit performance: the case study
occurred during this period of new normal, amplified in the COVID Bobble. The sulfur load and the solvent
composition changed during the case study, and effects are discussed later in this paper.

Regeneration Energy

The primary goal of the amine regenerator is to make properly-stripped lean amine so that the refinery’s
H2S absorbers stay on specification. The amine regenerator consumes energy to achieve that goal. There
are three major contributors to energy required in the regenerator:

1. Sensible heat is the heat required to bring the rich solvent (and reflux) up to temperature.
2. Heat of reaction is the heat required to keep the solvent hot as the endothermic evolution of
acid gases occurs.
3. Heat of vaporization (of water) is the heat required to create dilution steam that “pushes” the
acid gases away from the solvent, creating driving force, sending the acid gas up and out of
the column.

The total heat requirement is provided by steam supplied to the reboiler, subject to two minor puts-and-
takes:

4. Preheat of rich amine is the heat recovered from hot, lean amine as it preheats rich amine in
the L/R exchanger. This offsets heat that otherwise would be added to the column.
5. Cooling of the regenerator overhead by heat removed in the condenser happens when cold
reflux returns to the column.

Since the reboiler steam requirement is the largest single component of energy input into the amine
system, it’s the obvious target variable for energy optimization efforts.

3
Energy Targets

The correct amount of energy for the regenerator is that which provides the necessary composition of
lean solvent. In refinery main systems, the sulfide that remains in the lean solvent limits the performance
of the refinery’s gas (and liquid) contactors. So there is incentive to push the lean loading down to very
low values. Excessively deep stripping of lean solvent is energy intensive, and so “over stripping” is often
considered to be a waste of energy.

A balancing act results and refiners usually have targets for both Low and High sulfide in solvent … when
they are able to successfully measure those targets. However the measurement is usually performed
offline, and delayed by the hours (or even days). Operators need practical, real-time targets to keep their
plants running and so it’s often “Rules of Thumb” that are actually applied to adjust energy consumption
in regenerators.

The various Rules of Thumb reflect differences in available information, hardware and operating
philosophy among refineries. Most commonly we encounter the following:

Rules of Thumb: Control Setpoints - The Big Three

The most commonly used Rules of Thumb are Steam-to-Feed, Reflux Ratio, and Overhead Temperature.
These are readily available based on typically installed instrumentation, and are intuitive indications of
heat reaching the top of the regenerator.

Steam-to-Feed (or Steam-to-Lean)

This is a simple ratio of heat input to solvent circulation. The flow rate of steam to the reboiler is divided
by the circulation rate of solvent (typically, measured on the lean). Sensible heat is a portion of the net
heat requirement: the warm lean solvent must be heated to regenerator temperature. So it’s “sensible”
that changed circulation would require different heat input.

The ratio predicts the required heat input only if other conditions in the system are constant. For instance,
if the rich solvent contains more (or less) acid gas loading, the ratio alone doesn’t compensate for that
change. Also if the regenerator is made more (or less) efficient by changes in the chemistry of solvent, the
ratio misses that.

When rich solvent flow meter is used, steam-to-feed ratio is also sensitive to holdup during short-term
level control events; for instance an absorber foaming event that pauses rich solvent flow. In applications
susceptible to such events, the selection of solvent flow meter location (lean, rich) may help to address
that issue.

Reflux Ratio and Stripping Ratio

The hot acid gas at the top of the regenerator is saturated with water. Outside of the regenerator, the acid
gas is cooled and condensed, and the cooled water returns to the top of the regenerator as reflux. The

4
amount of water in the hot gas is a measure of how much heat remains after warming up the solvent,
reversing the reactions and producing acid gas.

There are two flavors of how to exploit that insight for energy control to the regenerator. Since the
flowrate of reflux water is usually measured when it returns to the regenerator, and if the flowrate of acid
gas is also measured, a simple “Reflux Ratio” is available. Because some water vapor is left in the acid gas
downstream of the condenser, the Reflux Ratio changes somewhat depending on the condenser
operation. An improved concept takes the condenser out of the picture, by calculating the water
concentration of the acid gas in the regenerator overhead and correcting the Reflux Ratio for that value.

The strength of the Reflux Ratio or Stripping Ratio is that steam reaching the top of the regenerator
suggests that most of the requirements of the heat-consuming reactions are met, and that the rich solvent
has been preheated to regenerator temperature.

The weakness is that varying composition of the solvent, and the resulting changes in efficiency in the
regenerator aren’t well considered. Another weakness is that acid gas flow meters are not usually
considered “necessary to run” and may not receive attention: the orifice may be in poor condition. The
instrument may not be properly pressure and temperature compensated, and the true meter factor may
not be correct for varying composition of acid gas. Rarely do plants have analyzers on the acid gas stream
to extract the actual values of H2S and CO2molar flow. Usually assumptions are made about that
composition, grounded in past plant tests (or even, design compositions). The easiest assumption is that
acid gas is all H2S + CO2, minus water: typically a correction for the water vapor in acid gas is applied.

The reflux flowrate is a result of controlled level in the Reflux Accumulator Vessel. The success of that
control may be challenged at low reflux rates: the measurement of flowrate may be the average of periods
of low (or zero) flow rate when the controller struggles to maintain level.

Overhead Temperature

The Reflux and Stripping Ratios use measured flowrates (of liquid water and of acid gas) but not all amine
units have those measurements (or trust them if they do). A simpler approach is to infer the water content
of the regenerator overhead by measuring the temperature of the hot acid gas as it leaves the top of the
regenerator. As with the Reflux and Stripping Ratios, the temperature is a measure of how much heat
makes it to the top of the regenerator. A high Overhead Temperature is an indication that sufficient energy
has been supplied to heat the rich solvent and reverse the reactions. Most plants have a temperature
measurement, and many operators simply use this Rule of Thumb of “how hot is enough, how cold is too
little”.

The strength of this approach is simplicity. One weakness is that the boiling point of the solvent varies with
operating pressure. And at low overhead temperatures, the measurement loses sensitivity to process
conditions, leaving the operator without a handle on the heat input. Just as with the other approaches
above, the overhead temperature doesn’t fully consider changes in regenerator efficiency resulting from
solvent composition changes.

5
Because they approximate the same thing – steam that reaches the regenerator overhead -- Overhead
Temperature shares a weakness with Reflux and Stripping Ratios. When hot rich solvent flashes at the feed
tray, that adds heat to the overhead. But that heat “upstairs” doesn’t help much with the dilution steam
needed to strip H2S deeply “downstairs”…at the bottom of the column. In a scenario where a lot of heat is
supplied upstairs, the Overhead Temperature may misguide the steam controller to back off steam.

Rules of Thumb: Control Setpoints – Other Approaches

Temperature(s) Inside the Column

When very low heat rates are applied, the overhead temperature becomes insensitive to heat rate. An
approach to improve control has been to measure temperature lower down in the column, where the
relationship between heat rate and temperature is still sensitive. This approach has the weakness of losing
sensitivity if the ‘sweet spot’ in the relationship temperature/heat input moves away from the point of
measurement.

An approach that has been mentioned is to have multiple thermocouples throughout the column,
sufficient to track changes in the temperature profile. And then to use that profile to control the heat rate.
Application of this technique is uncommon, yet it potentially improves upon a single-point measurement
which is vulnerable to conditions that shift the temperature profile away from the point of measurement.

Gas Analyzer – “Canary in the Coal Mine” Contactor

One practical approach has been to identify a contactor in the refinery that operates with a close-approach
to equilibrium w.r.t. H2S in the treated gas. Ideally, that contactor should operate at low pressure in order
to have higher concentration of H2S in the treated gas at equilibrium. To the extent that the equilibrium-
approach remains in place with shifting conditions in the refinery, an analyzer on this gas stream will
respond to changes in the lean solvent – especially depth of stripping. Gas analyzers are common
technology in refineries, so there is comfort with the application and understanding of its reliability.

Analyzers require periodic calibration and maintenance; that may be difficult to sustain unless the analyzer
is considered a “need to operate” item. One vulnerability is that the approach to the equilibrium condition
– even the equilibrium condition itself -- in the “canary in the coal mine” contactor may change due to
local condition changes (temperature of operation; pressure of operation; liquid/gas ratios; amount and
composition of acid gas removed).

Gas analysis is usually rapid and it directly relates to the lean loading -- when the approach to equilibrium
is close -- providing near real-time indication of over or under stripping. This technique works best with
low-pressure contactors, as they provide more analytical leverage to see H2S values in deeply stripped
solvent. However in a large, integrated refinery, the “Canary” contactor may be quite some distance from
the regenerator. The resulting time-lag in that situation is a potential weakness when dealing with short
term process changes.

6
Control Setpoint Validation

Online, Real Time Measurement of Sulfide in Lean Solvent

This is the holy grail of energy optimization, and over the years technology for this has been launched
several times. Yet we haven’t encountered widespread adoption. Rapid, direct measurement of the sulfide
target makes logical sense from a control viewpoint.

Analyzers that employ optical measurement have the weakness of the dirty, potentially solids-laden
and/or scale-building environment of an amine unit. Other types of techniques have been offered as well.
Justifying this sort of expenditure based solely on steam savings was previously difficult; perhaps the new
emphasis on carbon intensity will foster more widespread adoption in the future.

Offline, Time-Lagged Measurement of Sulfide in Lean Solvent

Usually this measurement involves discreet sample collection, transport from the location of the
regenerator to a lab facility, time in the que with a shared instrument that processes other samples
potentially of different systems. Then reporting and time to react and incorporate the results into
(possibly) adjusted targets for the hands-on control techniques.

The experience with measuring sulfides in lean amine in refinery environments is a checkered one. The
collected amine samples are ‘live’ and subject to loss of sulfide over time – for instance from oxidation
due to air contamination. Also the measurement itself has challenges; it can be a titration to an inflection
endpoint yet multiple inflections may be detected. The instrument often is set to titrate both sulfide and
mercaptide; for the same amount of titrant, these results differ by a factor of 2. Misidentification errors
are common.
Table 1. Summary of Approaches to Guide Reboiler Duty
The most common analysis
Techniqure How to Get it
method (AgNO3) is usually
optimized for a certain mass of Overhead Temperature Measured AG T before condenser
sulfide; deeply stripped and Reflux Ratio Reflux Return Flow Meter / Acid Gas Flow Meter
or or calculate with that:
CONTROL

poorly stripped samples tend Stripping Ratio Moles Water in OVHD / Moles H2S+CO2 in AG
to have less practical accuracy. Steam-to-Feed (or lean) Steam Flow Meter / Lean Solvent Flow Meter
Human errors happen, Temperature Elsewhere (vs OVHD) Thermal wells; skin T’s

including transcription errors. Canary in Coal Mine H2S gas analysis – integrated Low P contactor
(overcirculate; close approach to equilibrium)
VALIDATION

Reset with Lean Sulfide – Measured in Lab Amine sample→Lab: AgNO3 titration (and other);
Even when the samples and Opportunity for human and procedural errors
Reset with Lean Sulfide -- Online Analyzer Various have been attempted; requires specialized
the instrument are accurate, online analyzers (some optical, some chemical)
results can be very noisy – for
good reason. This noise can be a real condition in the regeneration system itself; we will discuss this more
later in the paper. It’s understandable to encounter mistrust of the measurement among operators, who
instead tend to rely on their Rules of Thumb (the real-time controls that they have at hand).
Table 1 summarizes several different approaches to guide reboiler duty.

7
Cases Defined

Case 1. An MEA-Era Refinery: Low Energy and Deep Turndown

Case 1 is a large PADD 3 refinery. The refinery amine system was originally designed to use MEA in its
amine system, and the refinery installed parallel capacity during the “the golden age of sulfur investment”.
Prior to the COVID Bobble, the refinery also converted the heritage-MEA unit to MDEA, a move which was
intended to provide both capacity and lower energy use.

Heritage MEA systems often have some differences compared to modern MDEA designs. As many MEA
systems operated installed thermal reclaimers, those systems tended to avoid swings in concentration of
HSAS. Converting to MDEA meant eliminating the thermal reclaiming function as MDEA does not ‘fit’ MEA-
designed thermal reclaimers. Instead of installing alternative reclaiming technology, the refiner of Case 1
chose to join the many refineries that opt for periodic merchant reclaiming. As a result, the MDEA solvent
system now experiences a periodic high-low cycle in HSAS.

The refinery runs the two amine systems in parallel, with common rich amine feed. The heritage-MEA
regenerator has a single wash stage. The newer regenerator has 5 wash stages.

This refinery has a strong focus on energy conservation metrics – part of the reason for the conversion to
an MDEA-based system. Sulfur complex operators here receive frequent guidance from energy managers
who question (perceived) unnecessary energy consumption; a key variable watched is the sulfide
concentration in the solvent. The refinery had the following target values for depth of stripping:

In the COVID Bobble, turndown operation was the norm; the regenerators were operating below 10% of
system flood. During the data collection period, refinery H2S production decreased in a step change. And
the concentration of heat stable salts did as well. These changes are explored with the help of the ProMax®
process simulator, and “Rules of Thumb” are tested against the results.

The refinery experienced high losses of amine during the COVID Bobble. The refinery conducted a study
to look for sources of losses, and the Regenerator Reflux Purge was a usual suspect.

A minor consideration, sometimes overlooked in refinery amine units, is ammonia management.

Operation at atypical condition influences ammonia management in amine units. Through simulation we
examine how low energy operation affected ammonia management in Case 1’s regenerator.

From the standpoint of the regenerator, the following features are important for the Case:

Case 2. An MDEA Refinery with Heat Exchanger Performance Loss

Case 2 is another integrated refinery amine system that experienced changes in operating conditions. This
refinery had operated for years with MDEA-based solvent, with accepted operating targets for energy
input – their favorite was Overhead Temperature.

8
Over time, the lean/rich heat exchangers began to lose performance. Fouling of the heat transfer surface
was suspected. A study was conducted to determine how much loss of heat transfer could be tolerated
without losing system treating performance.

However the refinery did not measure lean solvent loading, instead relying on offsite analysis by a service
laboratory. As is often the case with offsite measurement, the values were not considered to be completely
reliable due to the potential loss of amine.

Case 1

Rules of Thumb for Reboiler Steam

The data supporting Case 1 were measured during the COVID Bobble; sulfur rates in the refinery were
already low prior to COVID and suddenly decreased further. Amine circulation decreased to <50% of
design. As the regenerator was already at lower energy rates as a result of conversion to MDEA, turndown
further stressed the operation: the column now operated at a very low percent of flood.

Within this turndown operating condition, the amine system confronted the composition changes w.r.t.
HSAS. The COVID Bobble changed sulfur diet to the refinery as well. Table 2 summarizes the swings in
composition experienced during the period:

Table 2: Solvent Compositions in Case 1

HSAS Rich Loading
Composition 1 Start of period Low (Low) Normal
Composition 2 HSAS build High (Low) Normal
Composition 3 Low HSAS, sudden low S diet Low Very Low
Composition 4 Low S diet, HSAS build High Very Low

This refiner uses the Overhead Temperature “Rule

of Thumb” to (via manual control) adjust steam
supply to the reboiler. The operators then sample Figure
Figure44
the lean solvent and the refinery laboratory
measures the resulting concentration of H2S.
Figure 4 shows simulation of that operating curve
in the base case of operation before the COVID
Bobble conditions. At the start of the period, the
concentration of H2S in the lean solvent responds
proportionally to changes in the overhead
temperature. This is the sort of operating line that
makes life easier for plant operators: things
change in an expected, proportional manner.

This plant has defined target values of H2S concentration in the lean solvent. When very low values of H2S
in the lean solvent are measured, action is expected to decrease excess energy consumption in the amine

9
system. Not only are very low values of H2S concentration perceived to be energy-wasteful, they are also
perceived to be corrosion-risky. This refinery set a lower limit of about 50 ppmw H2S in lean solvent.

The upper limit of H2S in lean solvent results from achieving targeted depth of H2S removal in the system’s
lower pressure contactors. That experience translates to about 500 ppmw H2S in the lean solvent. The
“goldilocks’ zone for lean solvent H2S concentration is therefore 50 – 500 ppmw. With the operating line
shown in Figure 5, this can be a happy situation. At the Start of Run condition, a Rule of Thumb of about
205°F delivered on-target H2S.

However the composition changes during the Figure 5

COVID Bobble shift this happy relationship
between lean H2S concentration and regenerator
overhead temperature. Figure 5 shows simulation
results for the additional operating cases. New
operating curves resulted for each composition.
Sadly, one “Rule of Thumb” for Overhead
Temperature does not apply here: for a given
overhead temperature, widely different values of
H2S concentration in the lean solvent result.
Figure 5 shows that the Overhead Temperature
“Rule of Thumb” value required to achieve a targeted concentration of H2S in the lean solvent varied
substantially when the solvent conditions changed. To achieve a 300 ppmw target H2S concentration, the
Overhead Temperature varied from about 183°F to 223°F to achieve a 300 ppmw target H2S concentration.

Different “Rules of Thumb” are often used for

energy input to amine system regenerators. Each
Rule has its proponents; and the merits and debits Figure 6
of each are well debated. One useful thing about
having a process simulation tool (such as
ProMax®) is that it’s possible to test such merits
(and debits) for one’s own situation.

Figures 6 and 7 show simulation results for Steam-

to-Feed and Stripping Ratio Rules of Thumb
(respectively). Unfortunately, as with Overhead
Temperature neither of these two Rules of Thumb
do a good job of handling the variability from
composition changes in Case 1.

10
The key take-away in this: Rules of Thumb are
useful for systems that do not experience solvent
composition changes. For systems that undergo Figure 7
varying conditions, it’s possible to use Rules of
Thumb if one knows the operating line that one
is currently on…but different targets for the Rule
would be needed if conditions change.

The operating lines demonstrate yet another

challenge for operators. The “HSAS Build”
operating lines in the Figures demonstrate a real-
world difficulty in managing H2S concentration in
lean solvent. As the operating line becomes
vertical, small changes in the control variable have large responses.

An example of this challenge is shown in Figure 8, which presents a hypothetical scenario based on the
real-life experience of the unit engineer at the time of the COVID Bobble. HSAS had built to a high-normal
concentration from the Start of Run condition. The refinery’s lab measures (correct) low values of H2S in
the lean solvent. That fact came to the attention of Figure 8
the refinery’s Energy Manager; who asks the sulfur
unit staff: “why are you wasting energy?”. The
operator takes action and cuts reboiler steam. The
regenerator system walks down the operating line
and new solvent samples are analyzed. But
because the operating line is near-vertical at this
point, varying values of H2S in lean solvent are
reported by the lab. It’s human nature to question
sudden, noisy information … that’s what happened
here, too. Until small changes in process
conditions also result in off-spec absorber
performance.

When systems have such a ‘cliff’ in an operating line, it’s usually best to avoid operating at the cliff’s edge.
In this example, only 5% more energy provides some distance from the cliff. When presented with such a
challenge, simulation tools can help one demonstrate that Optimum and Minimum aren’t always the same
thing: a true optimum incorporates non-performance risks when we are able to quantify them.

Amine Loss from Reflux Purge

Another key difference between heritage MEA systems and modern MDEA designs is the premise for
managing amine in the regenerator reflux. Heritage MEA systems often relied upon amine building
significant concentration in the reflux loop, for corrosion avoidance. Refinery MEA systems, like any other
refinery amine system, absorb ammonia when treating the refinery’s ammonia-bearing hydrotreated

11
gases. The absorbed ammonia accumulates in the regenerator overhead, and if not removed from the
regenerator system, ammonia may accumulate to problematic concentration.

In heritage MEA systems, prevention was the bulk of the ammonia strategy: up-front water washing of the
hydrotreated gases removed the large majority of the ammonia, and the MEA systems dealt with only a
small amount of remaining ammonia.

During the “golden age of sulfur investment”, sour water stripper capacity wasn’t always added in concert
with amine and sulfur capacity. Conservation of sour water production was a common strategy to avoid
building more SWS capacity – minimizing hydrotreated-gas water washing helped avoid the need. That
trend meant more ammonia reached amine systems, increasing the importance of managing ammonia in
the amine reflux. Case 1 experienced this
ammonia trend; however the heritage-MEA
system design wasn’t intended to manage such
Figure 9
levels of ammonia. High purge rates in this design
resulted in unintended amine losses; a fact that
was tolerated … until conversion of the amine
system to MDEA.

MDEA is less volatile than MEA, and people expect

to find low losses of amine in reflux. Figure 9
shows simulation results for amine concentration
in the reflux loop in the heritage-MEA design,
compared to the same system operating with the
MDEA-based solvent. The regenerator has only 1 wash stage, and to demonstrate the potential for amine
losses, simulation was done with no purge of reflux. MEA is found at expected concentration in the reflux,
and less MDEA is present (even in this 1 wash stage system). Even though MDEA is indeed less volatile
than MEA, when reflux is purged for ammonia control, the design of Case 1’s reflux loop creates potential
for significant amine losses.

Reflux Loop Operation

Operation at extreme turndown during the COVID Bobble brought amine losses into focus. Amine
replacement is a budget-line-item, and at times of lower cash flow, those items are natural targets for
conservation. The usual potential loss mechanisms were pursued by the refinery – carryover events, filter
changes, solubility loses, etc. But the heritage-MEA system operation presented an additional loss
potential: excessive amine in the reflux loop.

Figure 10 is a conceptual diagram of the regenerator top + reflux loop, showing the single wash stage. The
system originally operated with MEA, which required high reboiler steam rates and therefore made
generous reflux flowrates. But with MDEA operation, turndown of the reflux loop required manual
intervention by the operators. Steam condensate was added to baseload the flowrate of reflux through
the reflux pumps, in order to satisfy their minimum-flow requirement. The flow of steam condensate was
set with a manual valve and monitored with a local flow meter.

12
Water added into the reflux loop merges with the amine system and would ultimately fill the amine system
if not managed; the operators manually purged water to the sour water stripper. The control of that purge
was with a manual valve, monitored with a local FI.

The combined reflux loop flowrate was monitored at the discharge of the reflux pump with a flow
transmitter. No direct measurement of reflux return to the
regenerator exists in this design; during the period of data
collection, reflux return was not known accurately. The
reflux return flow was adjusted to control the water level
in the reflux accumulator.

This reflux loop arrangement was a necessity to adapt to

minimum flow operation, especially at turndown with
MDEA as the solvent. However it is far from ideal for the
stable operation of the regenerator. The reflux return
flowrate naturally varies in a reflux loop, as the level
control system attempts to manage the variability of the
operation. When the reflux return flowrate is a large value,
that variation has less effect on the regenerator. However Figure 10
at turndown, the reflux return variability may cause the
level control valve to close at some points, starving the
wash tray of liquid makeup.

Figures 11 and 12 depict the concept of flow variation when at low reflux return flowrate. Normal flow
variation occurs, and when flowrates are high, that variation is a small percentage of the total … and the
system sees minimal effect.

When reflux return flowrates are very low, the

variation in flowrate becomes significant. If a
minor system upset should occur – for instance,
holdup of solvent in the contactors due to a
foaming event – the reflux loop sees a step change
Figure 11
and that step change may starve the wash stage of
liquid flow. In the examples of Figures 11 and 12,
we show conceptually how that sort of upset
condition might limit wash stage performance and
result in higher, erratic concentration of amine in
the reflux loop.

In application we have observed that type of impact on reflux return loop operation when the LCV
operated at less than 15% open.

13
In the turndown condition of Case 1, simulation demonstrated that reflux return rates were less than 15%
of design rates; therefore we expect that Case 1’s regenerator overhead was vulnerable to this sort of
upset condition.

Reflux Loop Simulation

Case 1 operating data encompass a wide range of Figure 12

conditions: data were collected for operation at high
rich loading and low rich loading, both with
significant concentration of HSAS. We focus on
these data as they were collected when the refinery
was searching for causes of amine loss.

Simulation of the wash tray in this Case comes with

significant challenges.

• The tray operates at severe turndown. Mass transfer and tray hydrodynamic correlations faithfully
produce results in such regions, but one must consider results with a grain of salt. Directionally,
it’s believable that tray effectiveness decreases at extreme turndown, but absolute prediction is
tricky at low percent flood.
• Tray maloperation is *not* simulated, yet this may be a reality in the plant. Weeping or dumping
conditions might exist and those are not simulated.
• Simulation is steady state. Yet the reflux return to the regenerator in this Case is dependent on the
level controller; even normal variation in flow due to level control may be significant when low
reflux rates are experienced.
• Process conditions are not precisely defined, due to the use of linked local flow indication for
condensate makeup and reflux purge.
Plant Data - Amine and Ammonia in Reflux
1200
With those caveats we proceeded to simulate the
plant data. 1000 Figure 13
ppm of Amine or Ammonia

800
Figure 13 summarizes the refinery’s amine-in-reflux
analyses. In the figure, it appears that a trend exists: 600 Higher Rich Loading

higher overhead temperature results in more amine

400
in the reflux. It also appears that the two data sets Lower Rich Loading
lay on the same trend: perhaps one operating line 200

might correlate both sets of data?

0
175 185 195 205 215 225 235
Overhead Temperture °F

14
Simulation explains more about these data (Figure 13). First we simulated the reflux loop with a worst-
case mindset, starting with the higher-rich loading data: what would the reflux look like, if there was no
reflux purging and no wash stage? Figure 14 shows that the resulting concentration of amine in the reflux
was much higher than measured values. Next we added a single wash stage to the simulation, to match
the design of the regenerator. The wash stage appeared to help remove amine from the reflux at higher
overhead temperature, but at the coldest operating Figure 14 Plant Data and Simulation - Amine in Reflux
conditions the wash stage did not decrease the 3000
Magenta: Higher Rich Loading Black: Lower Rich Loading
amine in the reflux loop by much. Lines: Simulations
2500
Circles and Triangles: Process Data No Purge No Wash Tray

Next we included condensate addition and reflux 2000

ppm of Amine
purge in the simulation, using values based on verbal
1500
reports from the operators about their practices with No Purge One Wash Tray

the manual control of that system. We simulated the 1000

system with and without the single wash stage. The

500
results show that the wash stage provides some, but
not much, benefit at moderate operating 0
Condensate Makeup 25 gpm

175 185 195 205 215 225 235

temperature and that benefit decreases as the Overhead Temperture °F

temperature cools.

Simulation answered the question: do the two sets of plant data lay on the same trend? Although the eye
might say yes, the conditions were different enough that the data sets do not in fact connect. Clearly,
solvent conditions matter when it comes to predicting amine concentration in the reflux loop.

We were curious why the wash tray loses effectiveness at lower overhead temperature. These
temperatures correspond to minimum energy input into the system, and the refinery’s focus was on
operating near these conditions.

The answer is perhaps easier to understand if we Figure 15

think of the regenerator column as actually two
separate devices: from the amine feed tray
downwards, the regenerator is a stripping
column. But the wash stage(s) are in fact an
“amine absorber”. Figure 15 shows this concept.
We often use the terminology “amine absorber”
to describe contactors that circulate amine to
scrub H2S from gases. However that contactor is
in fact an H2S absorber. The wash section of a
regenerator column is an absorber that uses
water to scrub amine vapor from the gas
produced in the column below. So, the wash section is an “amine absorber”, and the things that things we
know about absorbers apply:

• They work best when the stages operate in the proper percent of flood.
• They work best when enough liquid is circulated to scrub effectively.

15
 • And they work best when enough stages are in place to do the job.

At cold conditions, the tray operated at low percent of flood. Cold conditions also means little reflux was
available to feed the wash stage. So when the overhead temperature is low, it make sense that the wash
section has lower effectiveness. And with only one Amine in Reflux: OVHD T and Stages (No Purge)
2500
wash stage, limited performance is expected.
Figure 16
2000
No Wash Stage
Figure 16 shows simulation results for the wash

Amine in Reflux, ppmw

section if we were able to vary the number of the 1500

wash stages. Even at cold conditions, more wash

1000
stages results in the sort of lower amine
concentration in reflux. In more modern MDEA- 500 1 Stage
based unit designs, we typically encounter 2-4 2 Stages
3 Stages
wash stages. Even in the case of low liquid rates 0
5 Stages
175 180 185 190 195 200 205 210 215 220 225 230 235 240
and low percent of flood at cold conditions, the Overhead Temperature, °F

“amine absorber” columns with the higher

number of wash stages perform better.

This trend in number of wash stages was confirmed later, after the Case study. Even at turn down; the
refinery measured low values of amine in the reflux from the new regenerator that has 5 wash stages.
Amine in Reflux: OVHD T and Stages + Condensate
Quantifying Amine Losses in the Reflux 180000

160000
Figure 17
The baseload of flow in the reflux loop through 140000

condensate addition made the system operable.

Amine Losses, lb/year

120000

But what was added, was removed as purge. We 100000

included the purge loop operation in the 80000

No Wash Stage
simulations and found that yes, the annual losses 60000

40000
of amine in the reflux were substantial. 1 Stage
2 Stages
20000
3 Stages
5 Stages
0
Figure 17 shows that these losses (for the single 175 180 185 190 195 200 205 210 215 220 225 230 235 240
Overhead Temperature, °F
wash stage, operated in range of 185 – 205°F) were
equivalent to the purchase of an addition 1-2 trucks
per year of MDEA solvent. The shape of the curves in Figures 16 and 17 changes because the same
baseload condensate flowrate was added to all the cases – consistent with how the plant operated.
Dilution is more powerful at low reflux rates.

Losses, What Could Be Done Better?

First, the operability of the reflux loop could be improved. There are three functions of the reflux flowrate.
The minimum flow of the reflux pumps at approximately 20% of the design flowrate determines the
condensate addition rate. There are other approaches to turndown on a centrifugal pump; ideas such as
a spillback loop, impeller change, variable drive controller all could be considered. With better turndown
on the pumps, less condensate could be added (and therefore, less is purged). The pump impeller change

16
might be simplest, yet that comes with a loss of operability at the high range in the pump. Each of these
interventions require physical change in the process and so are more problematic than the makeup/purge
mode of operation.

The constant value of condensate added to the reflux loop often exceeded the minimum required to
operate the reflux pumps. At higher overhead temperature, it would be possible to cut back on the
condensate addition and minimize losses.

Some amount of purging is required for ammonia management. Simulation of the Case 1 conditions sets
that minimum purge at 10% of the pump design flow. In order to maintain water balance, condensate
must be added to replace water lost to the purge for ammonia rejection. As this is a heritage-MEA design,
one might try to argue for improved water wash upstream of the amine unit (at the hydrotreaters which
produce ammonia). To the degree that one succeeds in decreasing the ammonia that reaches the amine
unit, the purge can be decreased … decreasing the amine loss.

The above do not address the issue of variability in level control at low net flowrate through the reflux
loop, and the corresponding impact on the instantaneous reflux return flow rate. Modification to the
control parameters and/or scheme could deal with variability. Alternatives to decrease the risk of starving
wash trays might include switching the level control to the purge valve instead of the reflux return.
However in doing so, attention would still be required to the sufficiency of purge for ammonia control.

Ammonia Breakthrough at Low Energy Input

Ammonia is a headache for amine units, but one that is often ignored. The amine system serves a purpose,
protecting the acid gas from buildup of ammonia. By removing ammonia from the regenerator overhead
with the reflux purge, ammonia is prevented from building up to problematic concentration.

At extremely low energy rates in the regenerator, Ammonia and H2S in Lean Amine: OVHD T, Stages + Condensate
500 100
ammonia distributes down into the regenerator 450
Solids Lines = NH3 in Lean Amine
Dashed Lines = H2S in Lean Amine 90
column, and eventually ‘breaks through’ into the 400 80

Ammonia in Lean Amine, ppmw

lean amine. Simulation demonstrates that at the
H2S in Lean Amine, ppmw

350 70

conditions of Case 1, ammonia breakthrough 300

Figure 18
60

250 50
actually occurs under conditions that still produce
200 40
adequately H2S-stripped lean solvent (Figure 18). 150 30

100 1 Stage 20

The consequences of ammonia in the lean solvent 50 10

depend on the applications in the refinery. In 0 0

0.9 0.92 0.94 0.96 0.98 1
other refineries we have experienced distribution lb/gal steam/feed

of ammonia into treated products, creating off-

specification products at high penalty for the refiner. Case 1’s refiner wasn’t aware of the potential of
ammonia to move around the refinery with lean amine. In general there seems to be low awareness of
this risk of low energy operation.

17
Key Take-Aways

Considering the Big Three “Rules of Thumb”: within a single solvent condition, any Rule of Thumb applied.
However, none were robust enough to guide the Case 1 refinery through the composition changes
experienced.

Despite the high boiling point of MDEA that might lead one to think “there will be no MDEA in reflux”, this
plant design demonstrates that it can happen. Having more wash trays is fortunately more forgiving.

When operating at deep turndown and low energy … things are not normal. In this heritage-MEA design,
deep turndown leads to amine losses due to the need to baseload flow in the reflux loop. A more modern
design, with 2 or more wash trays, would perform better but still be subject to the same trends.

At extremely low reflux rates, reflux return may at times “starve” the column due to reflux drum level
control variability.

At very low energy input to the regenerator, ammonia redistribution into lean amine may occur, even
before H2S stripping is compromised.

Knowledge is power. Demonstrating risks of operation at low energy can inform the conversation around
energy conservation targets, and (hopefully) result in operation away from cliffs and instead at true
(process-reliability-inclusive) optimum conditions.

Case 2

The refinery in Case 2 experienced a loss of performance in the lean/rich heat exchanger, resulting in
cooling of the rich amine. The technical staff sought answers as to how the cooling was affecting the
operation of the regenerator. A study was undertaken to assess the system impact of declining heat
exchange on the system, and to justify the need to clean the exchangers.

When the study was launched, heat exchanger fouling was already progressing, so the “start of run”
condition represented an existing degree of decline in performance. Simulation was used to forecast the
process impact at the end-point temperature expected when one of the two in-series exchangers was
taken out of service for cleaning. The study also included the impact of cleaning both exchangers, resulting
in significantly hotter rich amine.

18
The Plant

Acid Gas
Figure 19 depicts the major
210°F Condenser
elements of the regenerator Figure 19 15
Q_Cond

system. Key to this study is 21 14

200°F 20
that two exchangers exist in 2

15 psig
19
17
15
series, and that the pressure 13
11
Regenerator

in the exchangers, and even Lean/Rich

Exch-1
Lean/Rich
Exch-2
Pipe 9
7
5
downstream of the control 135°F MDEA; 0.3 mol/mol H2S
4
200°F 3
1 17
Rich Amine 3 10 1 Reboiler
valve, is sufficient to limit Typical HSAS content
18
80 psigValve Q_Reb

vapor breakout at most 135°F Q_cool

13
0.81 lb/gal
Cool Lean
261°F
conditions. 100 ppmw H2S Lean Amine
12

Cooler

The regenerator consists of 21 stages, two of which are wash stages. The typical operating conditions are
shown in the figure. Note that there is a long run of pipe to feed the rich amine to the regenerator; the
pressure in the pipe drops as the elevation increases, and hits 15 psig at the point of entry to the feed tray.
The lower pressure near the top of the pipe is not sufficient to prevent vapor breakout at some conditions,
so consideration was given to that risk in the study.

The simulation steps are summarized in Table 3. These steps are labeled to match the annotations on
Figures 20A and 20B.

Table 3. Simulation Cases in Case 2

Condition Rich Temperature Reboiler Duty Manipulated? Lean Solvent H2S
Base Case 200 F Baseline Value At Target
Step 1 170 F Baseline Value Allowed to float
Step 2 170 F Changed as needed At Target
Step 3 230 F Baseline Value Allow to float
Step 4 230 F Changed as needed At Target

Results of the Study

Figures 20A and 20B each show three operating curves for the regenerator; note that the Y-axis is a log
scale. The targeted concentration of H2S in the lean solvent was 100 ppmw; the two figures differ in the
premise of reboiler energy control. Figure 20A demonstrates how the operating curves appear if the
Overhead Temperature guides performance. Figure 20B does the same, except for Steam-to-Feed as the
controlled variable.

The Point B on Figures 20A and 20B is the Base Case for the simulation. The operating curve of the base
case demonstrates that this plant operates near a “cliff”, a point where performance significantly degrades
with a lower energy input.

19
Point 1 represents the location on the operating curve for 170 F rich solvent, if one heat exchanger was
taken out of service but no adjustment was made to the reboiler duty. Note that stripping of H2S is quite
poor.
Figure 20A Figure 20B
10000 1 1
H2S in Lean Solvent (ppmw)

1000

170F
100 2 B 4 4 B 2
200F

3 230F 3

10
170 190 210 230 250 0.5 0.6 0.7 0.8 0.9 1 1.1
Overhead Temperature (F) Steam-to-Feed Ratio (lb/gal)

Point 2 shows the location on the operating curve at 170 F if reboiler duty is adjusted to meet the target
100 ppmw H2S in the lean solvent. Point 3 is on the operating curve that represents both exchangers clean
and at start of run conditions, but strips more deeply than the target. Point 4 is the location that
corresponds to meeting the 100 ppmw H2S target.

The verticality of the operating lines in Figure 20A represent a difficult-to-operate condition, made worse
if one were to premise high lean solvent H2S concentration. Figure 20B also shows near-vertical lines
where it would be difficult to operate.

A horizontal slice of the two figures at 100 ppmw H2S tells us about the change in reboiler duty required
to meet target. Figure 20B is more intuitive at constant circulation because the x-axis is proportional to
reboiler duty. The horizontal slice shows that cooling the rich solvent requires adding more reboiler duty
to reach the stripping target, but heating the solvent does not yield much benefit in decreased reboiler
duty.

Operating with cold rich amine requires an increase of ~8 MMBTU/hr in reboiler duty. Fortunately for this
plant, that duty is within the design of the reboiler. At low heat rates in the reboiler, when H2S stripping is
limited — the conditions create a nearly vertical operating line. Small changes in reboiler duty result in
poorly-stripped solvent: Points 1 and 2 in the figure highlight the difficulty of using Overhead Temperature
to guide operation at such a condition.

Once the steam required in the stripping section of the column is satisfied enough that stripping target is
reached, more heat at the top of the column is of little help. Comparing Points B and 4 in Figure 20B

20
demonstrates that lack of impact. It may seem
Figure 21 Energy Distribution – 100 ppmw H2S Cases
counterintuitive that supplying very hot rich 100
Reboiler
amine to the regenerator does not result in Rich Amine
80
significant decrease in required reboiler duty. The

Energy Added or Removed, MMBTU/hr

Net -2
application of heat to the top of the regenerator 60 Energy
Input
+8
has the most impact on condenser duty; heat 40

applied at that point results in creation of steam 20 +13

above the stripping stages, where H2S is removed Added -13
0
from the solvent. Removed
-4 +11
Condenser
-20

Figure 21 shows the breakdown of contribution to -40

200°F 170°F 230°F
net energy input into the regenerator for the 1 2
Rich Amine Temperature
3

200°F, 170°F, and 230°F simulation cases (all of

which strip to the 100 ppmw target H2S concentration). The loss of rich amine heat at the colder condition
is compensated by less heat removed in the condenser and more heat supplied in the reboiler. At the
hotter condition, the addition of more rich amine heat is balanced by the removal of most of that heat in
the condenser.

It is important to note that a different horizontal slice of Figure 20B at high lean loading, such as with CO2
in Carbon Capture or very lax-performing H2S systems, would give a very different result. Those systems
do not operate stripping stages in pinched conditions and so hotter rich amine would be more beneficial
for reboiler duty than Case 2’s refinery main treating system.

Process Considerations: Flow Regime – Rich Amine

Operating at high rich amine temperature comes Figure 22

100 20
with increasing risk of two-phase flow in the rich 90 18
pipe. Figure 22 shows that above about 215°F, 80 16

Mixture Velocity (ft/s)

vapor breakout causes the velocity to exceed about
Fraction Vapor (%)

70 Volume Fraction Vapor 14

60 12
5-6 ft/sec. The corrosion implication of high-
50 10
velocity flow in carbon steel pipe is a consideration. 40 8
Two-phase flow can also create instability; Figure 30 6
23 shows that at 230°F, the flow regime is well into 20 4
10 Mole Fraction 2
the slug flow regime. The vibration / mechanical Vapor
0 0
considerations can be serious and would require 160 180 200 220 240
Rich Temperature (F)
expert evaluation.

Process Considerations: Reflux Pump

As with Case 1, in this Case the cold conditions in the regenerator overhead result in low reflux flow rates.
The minimum operating condition in the reflux flow loop must be addressed.

21
Process Considerations: Lean Cooler Impact

Although this study focused on the impact of rich amine temperature on the regenerator, note that the
lean amine cooler is directly affected by any change Figure 23
in the lean/rich exchanger. Colder rich amine means
hotter lean amine, and more duty is required in the
lean amine cooler.

Key Take-Aways

Colder rich amine requires more reboiler duty to

compensate. The magnitude of the impact depends
on the regenerator operating conditions.

Specific to the conditions of Case 2:

• Heating up the rich amine above the base case doesn’t result in much benefit for decreased
reboiler duty. Less deeply stripped cases may look different.
• The operating lines for this plant demonstrate “cliffs” and those are wisely avoided.
• Operating with very high rich temperature comes with risk of vapor breakout, and increased pipe
velocity with troublesome/riskier (slugging) flow regime.

Final Thoughts

This paper demonstrates that operating at extremely low energy input comes with complications that are
worth considering in advance. With the emphasis on saving energy for carbon footprint and cost control,
the absolute minimum energy input may look appealing until operability considerations are included.
Demonstrating risks of operation at low energy can inform the conversation around energy targets, and
(hopefully) result in operation away from cliffs and instead at more process-reliability-inclusive conditions.

And when operating at conditions far away from the norm, the common “Rules of Thumb” for managing
reboiler energy input may not adequately guide operations. Such guidelines are a starting point to
understand what is fitting for the conditions that each refinery experiences.

References

Duns, H., Jr. and Ros, N. C. J.: Vertical Flow of Gas and Liquid Mixtures in Wells, Proc. Sixth World Pet.
Congress, Frankfurt (Jun. 19-26, 1963) Section II, Pape 22-PD6

Addington, L. and Ness, C.: An Evaluation of General “Rules of Thumb” in Amine Sweetening Unit Design
and Operation, Gas Processors Association Convention, 2010

22